ShuriZma
/
pcsx2
mirror of https://github.com/PCSX2/pcsx2.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

GSdx: Format

This commit is contained in:
TellowKrinkle 2021-04-12 04:31:30 -05:00 committed by tellowkrinkle
parent fafbb3cc63
commit ae1bc651d6
128 changed files with 11631 additions and 9678 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

105
plugins/GSdx/GS.cpp
View File

@ -161,7 +161,8 @@ EXPORT_C GSclose()
{ {
gsopen_done = false; gsopen_done = false;
if(s_gs == NULL) return; if (s_gs == NULL)
return;
s_gs->ResetDevice(); s_gs->ResetDevice();
@ -223,7 +224,8 @@ static int _GSopen(void** dsp, const char* title, GSRendererType renderer, int t
#if defined(__unix__) #if defined(__unix__)
// Note: EGL code use GLX otherwise maybe it could be also compatible with Windows // Note: EGL code use GLX otherwise maybe it could be also compatible with Windows
// Yes OpenGL code isn't complicated enough ! // Yes OpenGL code isn't complicated enough !
switch (GSWndEGL::SelectPlatform()) { switch (GSWndEGL::SelectPlatform())
{
#if GS_EGL_X11 #if GS_EGL_X11
case EGL_PLATFORM_X11_KHR: case EGL_PLATFORM_X11_KHR:
wnds.push_back(std::make_shared<GSWndEGL_X11>()); wnds.push_back(std::make_shared<GSWndEGL_X11>());
@ -389,7 +391,8 @@ static int _GSopen(void** dsp, const char* title, GSRendererType renderer, int t
return -1; return -1;
} }
if (renderer == GSRendererType::OGL_HW && theApp.GetConfigI("debug_glsl_shader") == 2) { if (renderer == GSRendererType::OGL_HW && theApp.GetConfigI("debug_glsl_shader") == 2)
{
printf("GSdx: test OpenGL shader. Please wait...\n\n"); printf("GSdx: test OpenGL shader. Please wait...\n\n");
static_cast<GSDeviceOGL*>(s_gs->m_dev)->SelfShaderTest(); static_cast<GSDeviceOGL*>(s_gs->m_dev)->SelfShaderTest();
printf("\nGSdx: test OpenGL shader done. It will now exit\n"); printf("\nGSdx: test OpenGL shader done. It will now exit\n");
@ -401,12 +404,14 @@ static int _GSopen(void** dsp, const char* title, GSRendererType renderer, int t
EXPORT_C_(void) GSosdLog(const char* utf8, uint32 color) EXPORT_C_(void) GSosdLog(const char* utf8, uint32 color)
{ {
if(s_gs && s_gs->m_dev) s_gs->m_dev->m_osd.Log(utf8); if (s_gs && s_gs->m_dev)
s_gs->m_dev->m_osd.Log(utf8);
} }
EXPORT_C_(void) GSosdMonitor(const char* key, const char* value, uint32 color) EXPORT_C_(void) GSosdMonitor(const char* key, const char* value, uint32 color)
{ {
if(s_gs && s_gs->m_dev) s_gs->m_dev->m_osd.Monitor(key, value); if (s_gs && s_gs->m_dev)
s_gs->m_dev->m_osd.Monitor(key, value);
} }
EXPORT_C_(int) GSopen2(void** dsp, uint32 flags) EXPORT_C_(int) GSopen2(void** dsp, uint32 flags)
@ -430,9 +435,7 @@ EXPORT_C_(int) GSopen2(void** dsp, uint32 flags)
case GSRendererType::OGL_SW: case GSRendererType::OGL_SW:
#ifdef _WIN32 #ifdef _WIN32
{ {
const auto config_renderer = static_cast<GSRendererType>( const auto config_renderer = static_cast<GSRendererType>(theApp.GetConfigI("Renderer"));
theApp.GetConfigI("Renderer")
);
if (current_renderer == config_renderer) if (current_renderer == config_renderer)
current_renderer = GSUtil::GetBestRenderer(); current_renderer = GSUtil::GetBestRenderer();
@ -739,7 +742,8 @@ EXPORT_C GSconfigure()
{ {
try try
{ {
if(!GSUtil::CheckSSE()) return; if (!GSUtil::CheckSSE())
return;
theApp.Init(); theApp.Init();
@ -759,7 +763,8 @@ EXPORT_C GSconfigure()
// We can convince it that touching that pool would be unsafe by running all GTK calls within a CFRunLoop // We can convince it that touching that pool would be unsafe by running all GTK calls within a CFRunLoop
// (Blocks submitted to the main queue by dispatch_async are run by its CFRunLoop) // (Blocks submitted to the main queue by dispatch_async are run by its CFRunLoop)
dispatch_async(dispatch_get_main_queue(), ^{ dispatch_async(dispatch_get_main_queue(), ^{
if (RunLinuxDialog()) { if (RunLinuxDialog())
{
theApp.ReloadConfig(); theApp.ReloadConfig();
// Force a reload of the gs state // Force a reload of the gs state
theApp.SetCurrentRendererType(GSRendererType::Undefined); theApp.SetCurrentRendererType(GSRendererType::Undefined);
@ -767,15 +772,16 @@ EXPORT_C GSconfigure()
}); });
#else #else
if (RunLinuxDialog()) { if (RunLinuxDialog())
{
theApp.ReloadConfig(); theApp.ReloadConfig();
// Force a reload of the gs state // Force a reload of the gs state
theApp.SetCurrentRendererType(GSRendererType::Undefined); theApp.SetCurrentRendererType(GSRendererType::Undefined);
} }
#endif #endif
}
} catch (GSDXRecoverableError) catch (GSDXRecoverableError)
{ {
} }
} }
@ -802,7 +808,8 @@ EXPORT_C GSirqCallback(void (*irq)())
} }
} }
void pt(const char* str){ void pt(const char* str)
{
struct tm* current; struct tm* current;
time_t now; time_t now;
@ -814,12 +821,14 @@ void pt(const char* str){
EXPORT_C_(bool) GSsetupRecording(std::string& filename) EXPORT_C_(bool) GSsetupRecording(std::string& filename)
{ {
if (s_gs == NULL) { if (s_gs == NULL)
{
printf("GSdx: no s_gs for recording\n"); printf("GSdx: no s_gs for recording\n");
return false; return false;
} }
#if defined(__unix__) || defined(__APPLE__) #if defined(__unix__) || defined(__APPLE__)
if (!theApp.GetConfigB("capture_enabled")) { if (!theApp.GetConfigB("capture_enabled"))
{
printf("GSdx: Recording is disabled\n"); printf("GSdx: Recording is disabled\n");
return false; return false;
} }
@ -914,7 +923,8 @@ public:
: m_console(NULL) : m_console(NULL)
, m_title(title) , m_title(title)
{ {
if(open) Open(); if (open)
Open();
} }
Console::~Console() Console::~Console()
@ -983,10 +993,15 @@ EXPORT_C GSReplay(HWND hwnd, HINSTANCE hinst, LPSTR lpszCmdLine, int nCmdShow)
char* start = lpszCmdLine; char* start = lpszCmdLine;
char* end = NULL; char* end = NULL;
long n = strtol(lpszCmdLine, &end, 10); long n = strtol(lpszCmdLine, &end, 10);
if(end > start) {renderer = static_cast<GSRendererType>(n); lpszCmdLine = end;} if (end > start)
{
renderer = static_cast<GSRendererType>(n);
lpszCmdLine = end;
}
} }
while(*lpszCmdLine == ' ') lpszCmdLine++; while (*lpszCmdLine == ' ')
lpszCmdLine++;
::SetPriorityClass(::GetCurrentProcess(), HIGH_PRIORITY_CLASS); ::SetPriorityClass(::GetCurrentProcess(), HIGH_PRIORITY_CLASS);
@ -1027,17 +1042,24 @@ EXPORT_C GSReplay(HWND hwnd, HINSTANCE hinst, LPSTR lpszCmdLine, int nCmdShow)
GSvsync(1); GSvsync(1);
struct Packet {uint8 type, param; uint32 size, addr; std::vector<uint8> buff;}; struct Packet
{
uint8 type, param;
uint32 size, addr;
std::vector<uint8> buff;
};
auto read_packet = [&file](uint8 type) { auto read_packet = [&file](uint8 type) {
Packet p; Packet p;
p.type = type; p.type = type;
switch(p.type) { switch (p.type)
{
case 0: case 0:
file->Read(&p.param, 1); file->Read(&p.param, 1);
file->Read(&p.size, 4); file->Read(&p.size, 4);
switch(p.param) { switch (p.param)
{
case 0: case 0:
p.buff.resize(0x4000); p.buff.resize(0x4000);
p.addr = 0x4000 - p.size; p.addr = 0x4000 - p.size;
@ -1138,7 +1160,8 @@ EXPORT_C GSBenchmark(HWND hwnd, HINSTANCE hinst, LPSTR lpszCmdLine, int nCmdShow
uint8* ptr = (uint8*)_aligned_malloc(1024 * 1024 * 4, 32); uint8* ptr = (uint8*)_aligned_malloc(1024 * 1024 * 4, 32);
for(int i = 0; i < 1024 * 1024 * 4; i++) ptr[i] = (uint8)i; for (int i = 0; i < 1024 * 1024 * 4; i++)
ptr[i] = (uint8)i;
// //
@ -1275,7 +1298,8 @@ EXPORT_C GSBenchmark(HWND hwnd, HINSTANCE hinst, LPSTR lpszCmdLine, int nCmdShow
uint8* ptr = (uint8*)_aligned_malloc(1024 * 1024 * 4, 32); uint8* ptr = (uint8*)_aligned_malloc(1024 * 1024 * 4, 32);
for(int i = 0; i < 1024 * 1024 * 4; i++) ptr[i] = (uint8)i; for (int i = 0; i < 1024 * 1024 * 4; i++)
ptr[i] = (uint8)i;
const GSLocalMemory::psm_t& psm = GSLocalMemory::m_psm[PSM_PSMCT32]; const GSLocalMemory::psm_t& psm = GSLocalMemory::m_psm[PSM_PSMCT32];
@ -1344,7 +1368,12 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
return; return;
} }
struct Packet {uint8 type, param; uint32 size, addr; std::vector<uint8> buff;}; struct Packet
{
uint8 type, param;
uint32 size, addr;
std::vector<uint8> buff;
};
std::list<Packet*> packets; std::list<Packet*> packets;
std::vector<uint8> buff; std::vector<uint8> buff;
@ -1356,7 +1385,8 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
int finished = theApp.GetConfigI("linux_replay"); int finished = theApp.GetConfigI("linux_replay");
bool repack_dump = (finished < 0); bool repack_dump = (finished < 0);
if (theApp.GetConfigI("dump")) { if (theApp.GetConfigI("dump"))
{
fprintf(stderr, "Dump is enabled. Replay will be disabled\n"); fprintf(stderr, "Dump is enabled. Replay will be disabled\n");
finished = 1; finished = 1;
} }
@ -1365,11 +1395,13 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
void* hWnd = NULL; void* hWnd = NULL;
int err = _GSopen((void**)&hWnd, "", m_renderer); int err = _GSopen((void**)&hWnd, "", m_renderer);
if (err != 0) { if (err != 0)
{
fprintf(stderr, "Error failed to GSopen\n"); fprintf(stderr, "Error failed to GSopen\n");
return; return;
} }
if (s_gs->m_wnd == NULL) return; if (s_gs->m_wnd == NULL)
return;
{ // Read .gs content { // Read .gs content
std::string f(lpszCmdLine); std::string f(lpszCmdLine);
@ -1492,7 +1524,8 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
case 2: case 2:
if(buff.size() < p->size) buff.resize(p->size); if (buff.size() < p->size)
buff.resize(p->size);
GSreadFIFO2(&buff[0], p->size / 16); GSreadFIFO2(&buff[0], p->size / 16);
@ -1506,11 +1539,16 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
} }
} }
if (finished >= 200) { if (finished >= 200)
{
; // Nop for Nvidia Profiler ; // Nop for Nvidia Profiler
} else if (finished > 90) { }
else if (finished > 90)
{
sleep(1); sleep(1);
} else { }
else
{
finished--; finished--;
} }
} }
@ -1522,8 +1560,7 @@ EXPORT_C GSReplay(char* lpszCmdLine, int renderer)
fprintf(stderr, "memory bandwith. T: %f KB/f. V: %f KB/f. U: %f KB/f\n", fprintf(stderr, "memory bandwith. T: %f KB/f. V: %f KB/f. U: %f KB/f\n",
(float)g_real_texture_upload_byte / (float)total_frame_nb, (float)g_real_texture_upload_byte / (float)total_frame_nb,
(float)g_vertex_upload_byte / (float)total_frame_nb, (float)g_vertex_upload_byte / (float)total_frame_nb,
(float)g_uniform_upload_byte/(float)total_frame_nb (float)g_uniform_upload_byte / (float)total_frame_nb);
);
#endif #endif
for (auto i = packets.begin(); i != packets.end(); i++) for (auto i = packets.begin(); i != packets.end(); i++)

177
plugins/GSdx/GS.h
View File

@ -152,7 +152,7 @@ enum GIF_FLG
GIF_FLG_PACKED = 0, GIF_FLG_PACKED = 0,
GIF_FLG_REGLIST = 1, GIF_FLG_REGLIST = 1,
GIF_FLG_IMAGE = 2, GIF_FLG_IMAGE = 2,
GIF_FLG_IMAGE2 = 3 GIF_FLG_IMAGE2 = 3,
}; };
enum GS_PSM enum GS_PSM
@ -263,14 +263,14 @@ union name \
bool operator == (const union name& r) const {return ((GSVector4i)r).eq(*this);} \ bool operator == (const union name& r) const {return ((GSVector4i)r).eq(*this);} \
bool operator != (const union name& r) const {return !((GSVector4i)r).eq(*this);} \ bool operator != (const union name& r) const {return !((GSVector4i)r).eq(*this);} \
operator GSVector4i() const {return GSVector4i::loadl(this);} \ operator GSVector4i() const {return GSVector4i::loadl(this);} \
struct { \ struct {
#define REG128(name) \ #define REG128(name) \
union name \ union name \
{ \ { \
uint64 u64[2]; \ uint64 u64[2]; \
uint32 u32[4]; \ uint32 u32[4]; \
struct { \ struct {
#define REG32_(prefix, name) REG32(prefix##name) #define REG32_(prefix, name) REG32(prefix##name)
#define REG64_(prefix, name) REG64(prefix##name) #define REG64_(prefix, name) REG64(prefix##name)
@ -282,20 +282,20 @@ union name \
#define REG32_SET(name) \ #define REG32_SET(name) \
union name \ union name \
{ \ { \
uint32 u32; \ uint32 u32;
#define REG64_SET(name) \ #define REG64_SET(name) \
union name \ union name \
{ \ { \
uint64 u64; \ uint64 u64; \
uint32 u32[2]; \ uint32 u32[2];
#define REG128_SET(name) \ #define REG128_SET(name) \
union name \ union name \
{ \ { \
__m128i m128; \ __m128i m128; \
uint64 u64[2]; \ uint64 u64[2]; \
uint32 u32[4]; \ uint32 u32[4];
#define REG_SET_END }; #define REG_SET_END };
@ -840,8 +840,10 @@ REG_END2
{ {
if (TBW < 2) if (TBW < 2)
{ {
if(PSM == PSM_PSMT8) return TW > 7 || TH > 6; if (PSM == PSM_PSMT8)
if(PSM == PSM_PSMT4) return TW > 7 || TH > 7; return TW > 7 || TH > 6;
if (PSM == PSM_PSMT4)
return TW > 7 || TH > 7;
} }
// The recast of TBW seems useless but it avoid tons of warning from GCC... // The recast of TBW seems useless but it avoid tons of warning from GCC...
@ -1121,7 +1123,13 @@ struct alignas(32) GIFPath
uint32 type; uint32 type;
GSVector4i regs; GSVector4i regs;
enum {TYPE_UNKNOWN, TYPE_ADONLY, TYPE_STQRGBAXYZF2, TYPE_STQRGBAXYZ2}; enum
{
TYPE_UNKNOWN,
TYPE_ADONLY,
TYPE_STQRGBAXYZF2,
TYPE_STQRGBAXYZ2
};
__forceinline void SetTag(const void* mem) __forceinline void SetTag(const void* mem)
{ {
@ -1137,7 +1145,8 @@ struct alignas(32) GIFPath
nloop = a & 0x7fff; nloop = a & 0x7fff;
if(nloop == 0) return; if (nloop == 0)
return;
GSVector4i v = GSVector4i::loadl(&src->REGS); // REGS not stored to tag.REGS, only into this->regs, restored before saving the state though GSVector4i v = GSVector4i::loadl(&src->REGS); // REGS not stored to tag.REGS, only into this->regs, restored before saving the state though
@ -1157,30 +1166,59 @@ struct alignas(32) GIFPath
{ {
switch (nreg) switch (nreg)
{ {
case 1: break; case 1:
case 2: break; break;
case 2:
break;
case 3: case 3:
if(regs.u32[0] == 0x00040102) type = TYPE_STQRGBAXYZF2; // many games, TODO: formats mixed with NOPs (xeno2: 040f010f02, 04010f020f, mgs3: 04010f0f02, 0401020f0f, 04010f020f) // many games, TODO: formats mixed with NOPs (xeno2: 040f010f02, 04010f020f, mgs3: 04010f0f02, 0401020f0f, 04010f020f)
if(regs.u32[0] == 0x00050102) type = TYPE_STQRGBAXYZ2; // GoW (has other crazy formats, like ...030503050103) if (regs.u32[0] == 0x00040102)
type = TYPE_STQRGBAXYZF2;
// GoW (has other crazy formats, like ...030503050103)
if (regs.u32[0] == 0x00050102)
type = TYPE_STQRGBAXYZ2;
// TODO: common types with UV instead // TODO: common types with UV instead
break; break;
case 4: break; case 4:
case 5: break; break;
case 6: break; case 5:
case 7: break; break;
case 8: break; case 6:
break;
case 7:
break;
case 8:
break;
case 9: case 9:
if(regs.u32[0] == 0x02040102 && regs.u32[1] == 0x01020401 && regs.u32[2] == 0x00000004) {type = TYPE_STQRGBAXYZF2; nreg = 3; nloop *= 3;} // ffx // ffx
if (regs.u32[0] == 0x02040102 && regs.u32[1] == 0x01020401 && regs.u32[2] == 0x00000004)
{
type = TYPE_STQRGBAXYZF2;
nreg = 3;
nloop *= 3;
}
break;
case 10:
break;
case 11:
break; break;
case 10: break;
case 11: break;
case 12: case 12:
if(regs.u32[0] == 0x02040102 && regs.u32[1] == 0x01020401 && regs.u32[2] == 0x04010204) {type = TYPE_STQRGBAXYZF2; nreg = 3; nloop *= 4;} // dq8 (not many, mostly 040102) // dq8 (not many, mostly 040102)
if (regs.u32[0] == 0x02040102 && regs.u32[1] == 0x01020401 && regs.u32[2] == 0x04010204)
{
type = TYPE_STQRGBAXYZF2;
nreg = 3;
nloop *= 4;
}
break;
case 13:
break;
case 14:
break;
case 15:
break;
case 16:
break; break;
case 13: break;
case 14: break;
case 15: break;
case 16: break;
default: default:
__assume(0); __assume(0);
} }
@ -1234,7 +1272,8 @@ struct GSPrivRegSet
uint64 _pad6; uint64 _pad6;
GSRegSYNCV SYNCV; GSRegSYNCV SYNCV;
uint64 _pad7; uint64 _pad7;
struct { struct
{
GSRegDISPFB DISPFB; GSRegDISPFB DISPFB;
uint64 _pad1; uint64 _pad1;
GSRegDISPLAY DISPLAY; GSRegDISPLAY DISPLAY;
@ -1276,10 +1315,13 @@ struct GSPrivRegSet
{ {
for (int i = 0; i < 2; i++) for (int i = 0; i < 2; i++)
{ {
if (!fp) return; if (!fp)
return;
if(i == 0 && !PMODE.EN1) continue; if (i == 0 && !PMODE.EN1)
if(i == 1 && !PMODE.EN2) continue; continue;
if (i == 1 && !PMODE.EN2)
continue;
fprintf(fp, "DISPFB[%d] BP=%05x BW=%u PSM=%u DBX=%u DBY=%u\n", fprintf(fp, "DISPFB[%d] BP=%05x BW=%u PSM=%u DBX=%u DBY=%u\n",
i, i,
@ -1287,8 +1329,7 @@ struct GSPrivRegSet
DISP[i].DISPFB.FBW, DISP[i].DISPFB.FBW,
DISP[i].DISPFB.PSM, DISP[i].DISPFB.PSM,
DISP[i].DISPFB.DBX, DISP[i].DISPFB.DBX,
DISP[i].DISPFB.DBY DISP[i].DISPFB.DBY);
);
fprintf(fp, "DISPLAY[%d] DX=%u DY=%u DW=%u DH=%u MAGH=%u MAGV=%u\n", fprintf(fp, "DISPLAY[%d] DX=%u DY=%u DW=%u DH=%u MAGH=%u MAGV=%u\n",
i, i,
@ -1297,8 +1338,7 @@ struct GSPrivRegSet
DISP[i].DISPLAY.DW, DISP[i].DISPLAY.DW,
DISP[i].DISPLAY.DH, DISP[i].DISPLAY.DH,
DISP[i].DISPLAY.MAGH, DISP[i].DISPLAY.MAGH,
DISP[i].DISPLAY.MAGV DISP[i].DISPLAY.MAGV);
);
} }
fprintf(fp, "PMODE EN1=%u EN2=%u CRTMD=%u MMOD=%u AMOD=%u SLBG=%u ALP=%u\n", fprintf(fp, "PMODE EN1=%u EN2=%u CRTMD=%u MMOD=%u AMOD=%u SLBG=%u ALP=%u\n",
@ -1308,8 +1348,7 @@ struct GSPrivRegSet
PMODE.MMOD, PMODE.MMOD,
PMODE.AMOD, PMODE.AMOD,
PMODE.SLBG, PMODE.SLBG,
PMODE.ALP PMODE.ALP);
);
fprintf(fp, "SMODE1 CLKSEL=%u CMOD=%u EX=%u GCONT=%u LC=%u NVCK=%u PCK2=%u PEHS=%u PEVS=%u PHS=%u PRST=%u PVS=%u RC=%u SINT=%u SLCK=%u SLCK2=%u SPML=%u T1248=%u VCKSEL=%u VHP=%u XPCK=%u\n", fprintf(fp, "SMODE1 CLKSEL=%u CMOD=%u EX=%u GCONT=%u LC=%u NVCK=%u PCK2=%u PEHS=%u PEVS=%u PHS=%u PRST=%u PVS=%u RC=%u SINT=%u SLCK=%u SLCK2=%u SPML=%u T1248=%u VCKSEL=%u VHP=%u XPCK=%u\n",
SMODE1.CLKSEL, SMODE1.CLKSEL,
@ -1332,29 +1371,24 @@ struct GSPrivRegSet
SMODE1.T1248, SMODE1.T1248,
SMODE1.VCKSEL, SMODE1.VCKSEL,
SMODE1.VHP, SMODE1.VHP,
SMODE1.XPCK SMODE1.XPCK);
);
fprintf(fp, "SMODE2 INT=%u FFMD=%u DPMS=%u\n", fprintf(fp, "SMODE2 INT=%u FFMD=%u DPMS=%u\n",
SMODE2.INT, SMODE2.INT,
SMODE2.FFMD, SMODE2.FFMD,
SMODE2.DPMS SMODE2.DPMS);
);
fprintf(fp, "SRFSH %08x_%08x\n", fprintf(fp, "SRFSH %08x_%08x\n",
SRFSH.u32[0], SRFSH.u32[0],
SRFSH.u32[1] SRFSH.u32[1]);
);
fprintf(fp, "SYNCH1 %08x_%08x\n", fprintf(fp, "SYNCH1 %08x_%08x\n",
SYNCH1.u32[0], SYNCH1.u32[0],
SYNCH1.u32[1] SYNCH1.u32[1]);
);
fprintf(fp, "SYNCH2 %08x_%08x\n", fprintf(fp, "SYNCH2 %08x_%08x\n",
SYNCH2.u32[0], SYNCH2.u32[0],
SYNCH2.u32[1] SYNCH2.u32[1]);
);
fprintf(fp, "SYNCV VBP=%u VBPE=%u VDP=%u VFP=%u VFPE=%u VS=%u\n", fprintf(fp, "SYNCV VBP=%u VBPE=%u VDP=%u VFP=%u VFPE=%u VS=%u\n",
SYNCV.VBP, SYNCV.VBP,
@ -1362,28 +1396,23 @@ struct GSPrivRegSet
SYNCV.VDP, SYNCV.VDP,
SYNCV.VFP, SYNCV.VFP,
SYNCV.VFPE, SYNCV.VFPE,
SYNCV.VS SYNCV.VS);
);
fprintf(fp, "CSR %08x_%08x\n", fprintf(fp, "CSR %08x_%08x\n",
CSR.u32[0], CSR.u32[0],
CSR.u32[1] CSR.u32[1]);
);
fprintf(fp, "BGCOLOR B=%u G=%u R=%u\n", fprintf(fp, "BGCOLOR B=%u G=%u R=%u\n",
BGCOLOR.B, BGCOLOR.B,
BGCOLOR.G, BGCOLOR.G,
BGCOLOR.R BGCOLOR.R);
);
fprintf(fp, "EXTBUF BP=0x%x BW=%u FBIN=%u WFFMD=%u EMODA=%u EMODC=%u WDX=%u WDY=%u\n", fprintf(fp, "EXTBUF BP=0x%x BW=%u FBIN=%u WFFMD=%u EMODA=%u EMODC=%u WDX=%u WDY=%u\n",
EXTBUF.EXBP, EXTBUF.EXBW, EXTBUF.FBIN, EXTBUF.WFFMD, EXTBUF.EXBP, EXTBUF.EXBW, EXTBUF.FBIN, EXTBUF.WFFMD,
EXTBUF.EMODA, EXTBUF.EMODC, EXTBUF.WDX, EXTBUF.WDY EXTBUF.EMODA, EXTBUF.EMODC, EXTBUF.WDX, EXTBUF.WDY);
);
fprintf(fp, "EXTDATA SX=%u SY=%u SMPH=%u SMPV=%u WW=%u WH=%u\n", fprintf(fp, "EXTDATA SX=%u SY=%u SMPH=%u SMPV=%u WW=%u WH=%u\n",
EXTDATA.SX, EXTDATA.SY, EXTDATA.SMPH, EXTDATA.SMPV, EXTDATA.WW, EXTDATA.WH EXTDATA.SX, EXTDATA.SY, EXTDATA.SMPH, EXTDATA.SMPV, EXTDATA.WW, EXTDATA.WH);
);
fprintf(fp, "EXTWRITE EN=%u\n", EXTWRITE.WRITE); fprintf(fp, "EXTWRITE EN=%u\n", EXTWRITE.WRITE);
} }
@ -1391,7 +1420,8 @@ struct GSPrivRegSet
void Dump(const std::string& filename) void Dump(const std::string& filename)
{ {
FILE* fp = fopen(filename.c_str(), "wt"); FILE* fp = fopen(filename.c_str(), "wt");
if (fp) { if (fp)
{
Dump(fp); Dump(fp);
fclose(fp); fclose(fp);
} }
@ -1400,13 +1430,34 @@ struct GSPrivRegSet
#pragma pack(pop) #pragma pack(pop)
enum {KEYPRESS=1, KEYRELEASE=2}; enum
struct GSKeyEventData {uint32 key, type;}; {
KEYPRESS = 1,
KEYRELEASE = 2
};
struct GSKeyEventData
{
uint32 key, type;
};
enum {FREEZE_LOAD=0, FREEZE_SAVE=1, FREEZE_SIZE=2}; enum
struct GSFreezeData {int size; uint8* data;}; {
FREEZE_LOAD = 0,
FREEZE_SAVE = 1,
FREEZE_SIZE = 2
};
struct GSFreezeData
{
int size;
uint8* data;
};
enum stateType {ST_WRITE, ST_TRANSFER, ST_VSYNC}; enum stateType
{
ST_WRITE,
ST_TRANSFER,
ST_VSYNC
};
enum class GSVideoMode : uint8 enum class GSVideoMode : uint8
{ {

3
plugins/GSdx/GSAlignedClass.h
View File

@ -21,7 +21,8 @@
#pragma once #pragma once
template<int i> class GSAlignedClass template <int i>
class GSAlignedClass
{ {
public: public:
GSAlignedClass() {} GSAlignedClass() {}

69
plugins/GSdx/GSBlock.h
View File

@ -53,7 +53,8 @@ class GSBlock
static const GSVector4i m_uw8hmask3; static const GSVector4i m_uw8hmask3;
public: public:
template<int i, int alignment, uint32 mask> __forceinline static void WriteColumn32(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int i, int alignment, uint32 mask>
__forceinline static void WriteColumn32(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
const uint8* RESTRICT s0 = &src[srcpitch * 0]; const uint8* RESTRICT s0 = &src[srcpitch * 0];
const uint8* RESTRICT s1 = &src[srcpitch * 1]; const uint8* RESTRICT s1 = &src[srcpitch * 1];
@ -177,7 +178,8 @@ public:
#endif #endif
} }
template<int i, int alignment> __forceinline static void WriteColumn16(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int i, int alignment>
__forceinline static void WriteColumn16(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
const uint8* RESTRICT s0 = &src[srcpitch * 0]; const uint8* RESTRICT s0 = &src[srcpitch * 0];
const uint8* RESTRICT s1 = &src[srcpitch * 1]; const uint8* RESTRICT s1 = &src[srcpitch * 1];
@ -250,7 +252,8 @@ public:
#endif #endif
} }
template<int i, int alignment> __forceinline static void WriteColumn8(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int i, int alignment>
__forceinline static void WriteColumn8(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
// TODO: read unaligned as WriteColumn32 does and try saving a few shuffles // TODO: read unaligned as WriteColumn32 does and try saving a few shuffles
@ -312,7 +315,8 @@ public:
#endif #endif
} }
template<int i, int alignment> __forceinline static void WriteColumn4(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int i, int alignment>
__forceinline static void WriteColumn4(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
//printf("WriteColumn4\n"); //printf("WriteColumn4\n");
@ -347,7 +351,8 @@ public:
((GSVector4i*)dst)[i * 4 + 3] = v3; ((GSVector4i*)dst)[i * 4 + 3] = v3;
} }
template<int alignment, uint32 mask> static void WriteColumn32(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment, uint32 mask>
static void WriteColumn32(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
switch ((y >> 1) & 3) switch ((y >> 1) & 3)
{ {
@ -359,7 +364,8 @@ public:
} }
} }
template<int alignment> static void WriteColumn16(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteColumn16(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
switch ((y >> 1) & 3) switch ((y >> 1) & 3)
{ {
@ -371,7 +377,8 @@ public:
} }
} }
template<int alignment> static void WriteColumn8(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteColumn8(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
switch ((y >> 2) & 3) switch ((y >> 2) & 3)
{ {
@ -383,7 +390,8 @@ public:
} }
} }
template<int alignment> static void WriteColumn4(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteColumn4(int y, uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
switch ((y >> 2) & 3) switch ((y >> 2) & 3)
{ {
@ -395,7 +403,8 @@ public:
} }
} }
template<int alignment, uint32 mask> static void WriteBlock32(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment, uint32 mask>
static void WriteBlock32(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
WriteColumn32<0, alignment, mask>(dst, src, srcpitch); WriteColumn32<0, alignment, mask>(dst, src, srcpitch);
src += srcpitch * 2; src += srcpitch * 2;
@ -406,7 +415,8 @@ public:
WriteColumn32<3, alignment, mask>(dst, src, srcpitch); WriteColumn32<3, alignment, mask>(dst, src, srcpitch);
} }
template<int alignment> static void WriteBlock16(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteBlock16(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
WriteColumn16<0, alignment>(dst, src, srcpitch); WriteColumn16<0, alignment>(dst, src, srcpitch);
src += srcpitch * 2; src += srcpitch * 2;
@ -417,7 +427,8 @@ public:
WriteColumn16<3, alignment>(dst, src, srcpitch); WriteColumn16<3, alignment>(dst, src, srcpitch);
} }
template<int alignment> static void WriteBlock8(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteBlock8(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
WriteColumn8<0, alignment>(dst, src, srcpitch); WriteColumn8<0, alignment>(dst, src, srcpitch);
src += srcpitch * 4; src += srcpitch * 4;
@ -428,7 +439,8 @@ public:
WriteColumn8<3, alignment>(dst, src, srcpitch); WriteColumn8<3, alignment>(dst, src, srcpitch);
} }
template<int alignment> static void WriteBlock4(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch) template <int alignment>
static void WriteBlock4(uint8* RESTRICT dst, const uint8* RESTRICT src, int srcpitch)
{ {
WriteColumn4<0, alignment>(dst, src, srcpitch); WriteColumn4<0, alignment>(dst, src, srcpitch);
src += srcpitch * 4; src += srcpitch * 4;
@ -439,7 +451,8 @@ public:
WriteColumn4<3, alignment>(dst, src, srcpitch); WriteColumn4<3, alignment>(dst, src, srcpitch);
} }
template<int i> __forceinline static void ReadColumn32(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch) template <int i>
__forceinline static void ReadColumn32(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch)
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -476,7 +489,8 @@ public:
#endif #endif
} }
template<int i> __forceinline static void ReadColumn16(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch) template <int i>
__forceinline static void ReadColumn16(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch)
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -517,7 +531,8 @@ public:
#endif #endif
} }
template<int i> __forceinline static void ReadColumn8(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch) template <int i>
__forceinline static void ReadColumn8(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch)
{ {
//for(int j = 0; j < 64; j++) ((uint8*)src)[j] = (uint8)j; //for(int j = 0; j < 64; j++) ((uint8*)src)[j] = (uint8)j;
@ -583,7 +598,8 @@ public:
#endif #endif
} }
template<int i> __forceinline static void ReadColumn4(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch) template <int i>
__forceinline static void ReadColumn4(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch)
{ {
//printf("ReadColumn4\n"); //printf("ReadColumn4\n");
@ -990,17 +1006,20 @@ public:
#endif #endif
} }
template<bool AEM, class V> __forceinline static V Expand24to32(const V& c, const V& TA0) template <bool AEM, class V>
__forceinline static V Expand24to32(const V& c, const V& TA0)
{ {
return c | (AEM ? TA0.andnot(c == V::zero()) : TA0); // TA0 & (c != GSVector4i::zero()) return c | (AEM ? TA0.andnot(c == V::zero()) : TA0); // TA0 & (c != GSVector4i::zero())
} }
template<bool AEM, class V> __forceinline static V Expand16to32(const V& c, const V& TA0, const V& TA1) template <bool AEM, class V>
__forceinline static V Expand16to32(const V& c, const V& TA0, const V& TA1)
{ {
return ((c & m_rxxx) << 3) | ((c & m_xgxx) << 6) | ((c & m_xxbx) << 9) | (AEM ? TA0.blend8(TA1, c.sra16(15)).andnot(c == V::zero()) : TA0.blend(TA1, c.sra16(15))); return ((c & m_rxxx) << 3) | ((c & m_xgxx) << 6) | ((c & m_xxbx) << 9) | (AEM ? TA0.blend8(TA1, c.sra16(15)).andnot(c == V::zero()) : TA0.blend(TA1, c.sra16(15)));
} }
template<bool AEM> static void ExpandBlock24(const uint32* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA) template <bool AEM>
static void ExpandBlock24(const uint32* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA)
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -1047,7 +1066,8 @@ public:
#endif #endif
} }
template<bool AEM> static void ExpandBlock16(const uint16* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA) // do not inline, uses too many xmm regs template <bool AEM>
static void ExpandBlock16(const uint16* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA) // do not inline, uses too many xmm regs
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -1361,7 +1381,8 @@ public:
{ {
uint8* s = (uint8*)src; uint8* s = (uint8*)src;
for (int j = 0; j < 8; j++, s += srcpitch) for (int j = 0; j < 8; j++, s += srcpitch)
for(int i = 0; i < 4; i++) s[i] = (columnTable32[j][i*2] & 0x0f) | (columnTable32[j][i*2+1] << 4); for (int i = 0; i < 4; i++)
s[i] = (columnTable32[j][i * 2] & 0x0f) | (columnTable32[j][i * 2 + 1] << 4);
} }
GSVector4i v4, v5, v6; GSVector4i v4, v5, v6;
@ -1536,7 +1557,8 @@ public:
#endif #endif
} }
template<bool AEM> __forceinline static void ReadAndExpandBlock24(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA) template <bool AEM>
__forceinline static void ReadAndExpandBlock24(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA)
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -1612,7 +1634,8 @@ public:
#endif #endif
} }
template<bool AEM> __forceinline static void ReadAndExpandBlock16(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA) template <bool AEM>
__forceinline static void ReadAndExpandBlock16(const uint8* RESTRICT src, uint8* RESTRICT dst, int dstpitch, const GIFRegTEXA& TEXA)
{ {
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501

60
plugins/GSdx/GSCapture.cpp
View File

@ -26,33 +26,45 @@
#ifdef _WIN32 #ifdef _WIN32
class CPinInfo : public PIN_INFO { class CPinInfo : public PIN_INFO
{
public: public:
CPinInfo() { pFilter = NULL; } CPinInfo() { pFilter = NULL; }
~CPinInfo() { if (pFilter) pFilter->Release(); } ~CPinInfo()
{
if (pFilter)
pFilter->Release();
}
}; };
class CFilterInfo : public FILTER_INFO { class CFilterInfo : public FILTER_INFO
{
public: public:
CFilterInfo() { pGraph = NULL; } CFilterInfo() { pGraph = NULL; }
~CFilterInfo() { if (pGraph) pGraph->Release(); } ~CFilterInfo()
{
if (pGraph)
pGraph->Release();
}
}; };
#define BeginEnumFilters(pFilterGraph, pEnumFilters, pBaseFilter) \ #define BeginEnumFilters(pFilterGraph, pEnumFilters, pBaseFilter) \
{CComPtr<IEnumFilters> pEnumFilters; \ { \
CComPtr<IEnumFilters> pEnumFilters; \
if(pFilterGraph && SUCCEEDED(pFilterGraph->EnumFilters(&pEnumFilters))) \ if(pFilterGraph && SUCCEEDED(pFilterGraph->EnumFilters(&pEnumFilters))) \
{ \ { \
for(CComPtr<IBaseFilter> pBaseFilter; S_OK == pEnumFilters->Next(1, &pBaseFilter, 0); pBaseFilter = NULL) \ for(CComPtr<IBaseFilter> pBaseFilter; S_OK == pEnumFilters->Next(1, &pBaseFilter, 0); pBaseFilter = NULL) \
{ \ {
#define EndEnumFilters }}} #define EndEnumFilters }}}
#define BeginEnumPins(pBaseFilter, pEnumPins, pPin) \ #define BeginEnumPins(pBaseFilter, pEnumPins, pPin) \
{CComPtr<IEnumPins> pEnumPins; \ { \
CComPtr<IEnumPins> pEnumPins; \
if(pBaseFilter && SUCCEEDED(pBaseFilter->EnumPins(&pEnumPins))) \ if(pBaseFilter && SUCCEEDED(pBaseFilter->EnumPins(&pEnumPins))) \
{ \ { \
for(CComPtr<IPin> pPin; S_OK == pEnumPins->Next(1, &pPin, 0); pPin = NULL) \ for(CComPtr<IPin> pPin; S_OK == pEnumPins->Next(1, &pPin, 0); pPin = NULL) \
{ \ {
#define EndEnumPins }}} #define EndEnumPins }}}
@ -76,9 +88,9 @@ GSSource : public CBaseFilter, private CCritSec, public IGSSource
STDMETHODIMP NonDelegatingQueryInterface(REFIID riid, void** ppv) STDMETHODIMP NonDelegatingQueryInterface(REFIID riid, void** ppv)
{ {
return return riid == __uuidof(IGSSource)
riid == __uuidof(IGSSource) ? GetInterface((IGSSource*)this, ppv) : ? GetInterface((IGSSource*)this, ppv)
__super::NonDelegatingQueryInterface(riid, ppv); : __super::NonDelegatingQueryInterface(riid, ppv);
} }
class GSSourceOutputPin : public CBaseOutputPin class GSSourceOutputPin : public CBaseOutputPin
@ -126,8 +138,10 @@ GSSource : public CBaseFilter, private CCritSec, public IGSSource
vih.bmiHeader.biSizeImage = m_size.x * m_size.y * 4; vih.bmiHeader.biSizeImage = m_size.x * m_size.y * 4;
mt.SetFormat((uint8*)&vih, sizeof(vih)); mt.SetFormat((uint8*)&vih, sizeof(vih));
if(colorspace == 1) m_mts.insert(m_mts.begin(), mt); if (colorspace == 1)
else m_mts.push_back(mt); m_mts.insert(m_mts.begin(), mt);
else
m_mts.push_back(mt);
} }
HRESULT GSSourceOutputPin::DecideBufferSize(IMemAllocator* pAlloc, ALLOCATOR_PROPERTIES* pProperties) HRESULT GSSourceOutputPin::DecideBufferSize(IMemAllocator* pAlloc, ALLOCATOR_PROPERTIES* pProperties)
@ -173,8 +187,10 @@ GSSource : public CBaseFilter, private CCritSec, public IGSSource
{ {
CheckPointer(pmt, E_POINTER); CheckPointer(pmt, E_POINTER);
if(i < 0) return E_INVALIDARG; if (i < 0)
if(i > 1) return VFW_S_NO_MORE_ITEMS; return E_INVALIDARG;
if (i > 1)
return VFW_S_NO_MORE_ITEMS;
*pmt = m_mts[i]; *pmt = m_mts[i];
@ -195,7 +211,6 @@ GSSource : public CBaseFilter, private CCritSec, public IGSSource
GSSourceOutputPin* m_output; GSSourceOutputPin* m_output;
public: public:
GSSource(int w, int h, float fps, IUnknown* pUnk, HRESULT& hr, int colorspace) GSSource(int w, int h, float fps, IUnknown* pUnk, HRESULT& hr, int colorspace)
: CBaseFilter("GSSource", pUnk, this, __uuidof(this), &hr) : CBaseFilter("GSSource", pUnk, this, __uuidof(this), &hr)
, m_output(NULL) , m_output(NULL)
@ -364,7 +379,8 @@ public:
static IPin* GetFirstPin(IBaseFilter* pBF, PIN_DIRECTION dir) static IPin* GetFirstPin(IBaseFilter* pBF, PIN_DIRECTION dir)
{ {
if(!pBF) return(NULL); if (!pBF)
return nullptr;
BeginEnumPins(pBF, pEP, pPin) BeginEnumPins(pBF, pEP, pPin)
{ {
@ -374,12 +390,12 @@ static IPin* GetFirstPin(IBaseFilter* pBF, PIN_DIRECTION dir)
{ {
IPin* pRet = pPin.Detach(); IPin* pRet = pPin.Detach();
pRet->Release(); pRet->Release();
return(pRet); return pRet;
} }
} }
EndEnumPins EndEnumPins
return(NULL); return nullptr;
} }
#endif #endif
@ -470,8 +486,7 @@ bool GSCapture::BeginCapture(float fps, GSVector2i recommendedResolution, float
} }
else else
{ {
if(FAILED(hr = m_graph->AddFilter(m_src, L"Source")) if (FAILED(hr = m_graph->AddFilter(m_src, L"Source")) || FAILED(hr = m_graph->AddFilter(dlg.m_enc, L"Encoder")))
|| FAILED(hr = m_graph->AddFilter(dlg.m_enc, L"Encoder")))
{ {
return false; return false;
} }
@ -519,7 +534,8 @@ bool GSCapture::BeginCapture(float fps, GSVector2i recommendedResolution, float
m_size.x = theApp.GetConfigI("CaptureWidth"); m_size.x = theApp.GetConfigI("CaptureWidth");
m_size.y = theApp.GetConfigI("CaptureHeight"); m_size.y = theApp.GetConfigI("CaptureHeight");
for(int i = 0; i < m_threads; i++) { for (int i = 0; i < m_threads; i++)
{
m_workers.push_back(std::unique_ptr<GSPng::Worker>(new GSPng::Worker(&GSPng::Process))); m_workers.push_back(std::unique_ptr<GSPng::Worker>(new GSPng::Worker(&GSPng::Process)));
} }

9
plugins/GSdx/GSClut.h
View File

@ -71,9 +71,12 @@ class alignas(32) GSClut : public GSAlignedClass<32>
void WriteCLUT16S_I8_CSM1(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); void WriteCLUT16S_I8_CSM1(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);
void WriteCLUT16S_I4_CSM1(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); void WriteCLUT16S_I4_CSM1(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);
template<int n> void WriteCLUT32_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); template <int n>
template<int n> void WriteCLUT16_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); void WriteCLUT32_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);
template<int n> void WriteCLUT16S_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); template <int n>
void WriteCLUT16_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);
template <int n>
void WriteCLUT16S_CSM2(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);
void WriteCLUT_NULL(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT); void WriteCLUT_NULL(const GIFRegTEX0& TEX0, const GIFRegTEXCLUT& TEXCLUT);

9
plugins/GSdx/GSCrc.cpp
View File

@ -544,10 +544,11 @@ CRC::Game CRC::Lookup(uint32 crc)
int crcDups = 0; int crcDups = 0;
for (size_t i = 0; i < countof(m_games); i++) for (size_t i = 0; i < countof(m_games); i++)
{ {
if( !IsCrcExcluded( exclusions, m_games[i].crc ) ){ if (!IsCrcExcluded(exclusions, m_games[i].crc))
if(m_map[m_games[i].crc]){ {
printf("[FIXME] GSdx: Duplicate CRC: 0x%08X: (game-id/region-id) %d/%d overrides %d/%d\n" if (m_map[m_games[i].crc])
, m_games[i].crc, m_games[i].title, m_games[i].region, m_map[m_games[i].crc]->title, m_map[m_games[i].crc]->region); {
printf("[FIXME] GSdx: Duplicate CRC: 0x%08X: (game-id/region-id) %d/%d overrides %d/%d\n", m_games[i].crc, m_games[i].title, m_games[i].region, m_map[m_games[i].crc]->title, m_map[m_games[i].crc]->region);
crcDups++; crcDups++;
} }

24
plugins/GSdx/GSDrawingContext.cpp
View File

@ -31,22 +31,29 @@ static int findmax(int tl, int br, int limit, int wm, int minuv, int maxuv)
if (wm == CLAMP_CLAMP) if (wm == CLAMP_CLAMP)
{ {
if(uv > limit) uv = limit; if (uv > limit)
uv = limit;
} }
else if (wm == CLAMP_REPEAT) else if (wm == CLAMP_REPEAT)
{ {
if(tl < 0) uv = limit; // wrap around if (tl < 0)
else if(uv > limit) uv = limit; uv = limit; // wrap around
else if (uv > limit)
uv = limit;
} }
else if (wm == CLAMP_REGION_CLAMP) else if (wm == CLAMP_REGION_CLAMP)
{ {
if(uv < minuv) uv = minuv; if (uv < minuv)
if(uv > maxuv) uv = maxuv; uv = minuv;
if (uv > maxuv)
uv = maxuv;
} }
else if (wm == CLAMP_REGION_REPEAT) else if (wm == CLAMP_REGION_REPEAT)
{ {
if(tl < 0) uv = minuv | maxuv; // wrap around, just use (any & mask) | fix if (tl < 0)
else uv = std::min(uv, minuv) | maxuv; // (any & mask) cannot be larger than mask, select br if that is smaller (not br & mask because there might be a larger value between tl and br when &'ed with the mask) uv = minuv | maxuv; // wrap around, just use (any & mask) | fix
else
uv = std::min(uv, minuv) | maxuv; // (any & mask) cannot be larger than mask, select br if that is smaller (not br & mask because there might be a larger value between tl and br when &'ed with the mask)
} }
return uv; return uv;
@ -74,7 +81,8 @@ static int extend(int uv, int size)
GIFRegTEX0 GSDrawingContext::GetSizeFixedTEX0(const GSVector4& st, bool linear, bool mipmap) GIFRegTEX0 GSDrawingContext::GetSizeFixedTEX0(const GSVector4& st, bool linear, bool mipmap)
{ {
if(mipmap) return TEX0; // no mipmaping allowed if (mipmap)
return TEX0; // no mipmaping allowed
// find the optimal value for TW/TH by analyzing vertex trace and clamping values, extending only for region modes where uv may be outside // find the optimal value for TW/TH by analyzing vertex trace and clamping values, extending only for region modes where uv may be outside

3
plugins/GSdx/GSDrawingContext.h
View File

@ -197,7 +197,8 @@ public:
{ {
// Append on purpose so env + context are merged into a single file // Append on purpose so env + context are merged into a single file
FILE* fp = fopen(filename.c_str(), "at"); FILE* fp = fopen(filename.c_str(), "at");
if (!fp) return; if (!fp)
return;
fprintf(fp, "XYOFFSET\n" fprintf(fp, "XYOFFSET\n"
"\tX:%u\n" "\tX:%u\n"

8
plugins/GSdx/GSDrawingEnvironment.h
View File

@ -88,7 +88,8 @@ public:
void Dump(const std::string& filename) void Dump(const std::string& filename)
{ {
FILE* fp = fopen(filename.c_str(), "wt"); FILE* fp = fopen(filename.c_str(), "wt");
if (!fp) return; if (!fp)
return;
fprintf(fp, "PRIM\n" fprintf(fp, "PRIM\n"
"\tPRIM:%u\n" "\tPRIM:%u\n"
@ -182,8 +183,8 @@ public:
, BITBLTBUF.SBW, BITBLTBUF.SBP, BITBLTBUF.SPSM, BITBLTBUF.DBW, BITBLTBUF.DPSM, BITBLTBUF.DBP); , BITBLTBUF.SBW, BITBLTBUF.SBP, BITBLTBUF.SPSM, BITBLTBUF.DBW, BITBLTBUF.DPSM, BITBLTBUF.DBP);
fprintf(fp, "TRXDIR\n" fprintf(fp, "TRXDIR\n"
"\tXDIR:%u\n\n" "\tXDIR:%u\n\n",
, TRXDIR.XDIR); TRXDIR.XDIR);
fprintf(fp, "TRXPOS\n" fprintf(fp, "TRXPOS\n"
"\tDIRY:%u\n" "\tDIRY:%u\n"
@ -201,5 +202,4 @@ public:
fclose(fp); fclose(fp);
} }
}; };

9
plugins/GSdx/GSDump.cpp
View File

@ -122,7 +122,8 @@ GSDumpXz::GSDumpXz(const std::string& fn, uint32 crc, const GSFreezeData& fd, co
{ {
m_strm = LZMA_STREAM_INIT; m_strm = LZMA_STREAM_INIT;
lzma_ret ret = lzma_easy_encoder(&m_strm, 6 /*level*/, LZMA_CHECK_CRC64); lzma_ret ret = lzma_easy_encoder(&m_strm, 6 /*level*/, LZMA_CHECK_CRC64);
if (ret != LZMA_OK) { if (ret != LZMA_OK)
{
fprintf(stderr, "GSDumpXz: Error initializing LZMA encoder ! (error code %u)\n", ret); fprintf(stderr, "GSDumpXz: Error initializing LZMA encoder ! (error code %u)\n", ret);
return; return;
} }
@ -176,13 +177,15 @@ void GSDumpXz::Flush()
void GSDumpXz::Compress(lzma_action action, lzma_ret expected_status) void GSDumpXz::Compress(lzma_action action, lzma_ret expected_status)
{ {
std::vector<uint8> out_buff(1024 * 1024); std::vector<uint8> out_buff(1024 * 1024);
do { do
{
m_strm.next_out = out_buff.data(); m_strm.next_out = out_buff.data();
m_strm.avail_out = out_buff.size(); m_strm.avail_out = out_buff.size();
lzma_ret ret = lzma_code(&m_strm, action); lzma_ret ret = lzma_code(&m_strm, action);
if (ret != expected_status) { if (ret != expected_status)
{
fprintf(stderr, "GSDumpXz: Error %d\n", (int)ret); fprintf(stderr, "GSDumpXz: Error %d\n", (int)ret);
return; return;
} }

169
plugins/GSdx/GSLocalMemory.cpp
View File

@ -85,7 +85,8 @@ GSLocalMemory::GSLocalMemory()
: m_clut(this) : m_clut(this)
{ {
m_use_fifo_alloc = theApp.GetConfigB("UserHacks") && theApp.GetConfigB("wrap_gs_mem"); m_use_fifo_alloc = theApp.GetConfigB("UserHacks") && theApp.GetConfigB("wrap_gs_mem");
switch (theApp.GetCurrentRendererType()) { switch (theApp.GetCurrentRendererType())
{
case GSRendererType::OGL_SW: case GSRendererType::OGL_SW:
m_use_fifo_alloc = true; m_use_fifo_alloc = true;
break; break;
@ -240,7 +241,8 @@ GSLocalMemory::GSLocalMemory()
m_psm[i].pal = 0; m_psm[i].pal = 0;
m_psm[i].bs = GSVector2i(8, 8); m_psm[i].bs = GSVector2i(8, 8);
m_psm[i].pgs = GSVector2i(64, 32); m_psm[i].pgs = GSVector2i(64, 32);
for(int j = 0; j < 8; j++) m_psm[i].rowOffset[j] = rowOffset32; for (int j = 0; j < 8; j++)
m_psm[i].rowOffset[j] = rowOffset32;
m_psm[i].blockOffset = blockOffset32; m_psm[i].blockOffset = blockOffset32;
m_psm[i].msk = 0xff; m_psm[i].msk = 0xff;
m_psm[i].depth = 0; m_psm[i].depth = 0;
@ -435,7 +437,8 @@ GSLocalMemory::GSLocalMemory()
m_psm[PSM_PSMT8].pal = m_psm[PSM_PSMT8H].pal = 256; m_psm[PSM_PSMT8].pal = m_psm[PSM_PSMT8H].pal = 256;
m_psm[PSM_PSMT4].pal = m_psm[PSM_PSMT4HL].pal = m_psm[PSM_PSMT4HH].pal = 16; m_psm[PSM_PSMT4].pal = m_psm[PSM_PSMT4HL].pal = m_psm[PSM_PSMT4HH].pal = 16;
for(size_t i = 0; i < countof(m_psm); i++) m_psm[i].fmt = 3; for (size_t i = 0; i < countof(m_psm); i++)
m_psm[i].fmt = 3;
m_psm[PSM_PSMCT32].fmt = m_psm[PSM_PSMZ32].fmt = 0; m_psm[PSM_PSMCT32].fmt = m_psm[PSM_PSMZ32].fmt = 0;
m_psm[PSM_PSMCT24].fmt = m_psm[PSM_PSMZ24].fmt = 1; m_psm[PSM_PSMCT24].fmt = m_psm[PSM_PSMZ24].fmt = 1;
m_psm[PSM_PSMCT16].fmt = m_psm[PSM_PSMZ16].fmt = 2; m_psm[PSM_PSMCT16].fmt = m_psm[PSM_PSMZ16].fmt = 2;
@ -493,9 +496,12 @@ GSLocalMemory::~GSLocalMemory()
else else
vmfree(m_vm8, m_vmsize * 4); vmfree(m_vm8, m_vmsize * 4);
for(auto &i : m_omap) delete i.second; for (auto& i : m_omap)
for(auto &i : m_pomap) _aligned_free(i.second); delete i.second;
for(auto &i : m_po4map) _aligned_free(i.second); for (auto& i : m_pomap)
_aligned_free(i.second);
for (auto& i : m_po4map)
_aligned_free(i.second);
for (auto& i : m_p2tmap) for (auto& i : m_p2tmap)
{ {
@ -853,12 +859,14 @@ void GSLocalMemory::WriteImageTopBottom(int l, int r, int y, int h, const uint8*
break; break;
case PSM_PSMT8: case PSM_PSMT8:
GSBlock::ReadColumn8(y, dst, buff, 16); GSBlock::ReadColumn8(y, dst, buff, 16);
for(int i = 0, j = y2; i < h2; i++, j++) memcpy(&buff[j * 16], &src[i * srcpitch + x], 16); for (int i = 0, j = y2; i < h2; i++, j++)
memcpy(&buff[j * 16], &src[i * srcpitch + x], 16);
GSBlock::WriteColumn8<32>(y, dst, buff, 16); GSBlock::WriteColumn8<32>(y, dst, buff, 16);
break; break;
case PSM_PSMT4: case PSM_PSMT4:
GSBlock::ReadColumn4(y, dst, buff, 16); GSBlock::ReadColumn4(y, dst, buff, 16);
for(int i = 0, j = y2; i < h2; i++, j++) memcpy(&buff[j * 16], &src[i * srcpitch + (x >> 1)], 16); for (int i = 0, j = y2; i < h2; i++, j++)
memcpy(&buff[j * 16], &src[i * srcpitch + (x >> 1)], 16);
GSBlock::WriteColumn4<32>(y, dst, buff, 16); GSBlock::WriteColumn4<32>(y, dst, buff, 16);
break; break;
// TODO // TODO
@ -940,12 +948,14 @@ void GSLocalMemory::WriteImageTopBottom(int l, int r, int y, int h, const uint8*
break; break;
case PSM_PSMT8: case PSM_PSMT8:
GSBlock::ReadColumn8(y, dst, buff, 16); GSBlock::ReadColumn8(y, dst, buff, 16);
for(int i = 0; i < h; i++) memcpy(&buff[i * 16], &src[i * srcpitch + x], 16); for (int i = 0; i < h; i++)
memcpy(&buff[i * 16], &src[i * srcpitch + x], 16);
GSBlock::WriteColumn8<32>(y, dst, buff, 16); GSBlock::WriteColumn8<32>(y, dst, buff, 16);
break; break;
case PSM_PSMT4: case PSM_PSMT4:
GSBlock::ReadColumn4(y, dst, buff, 16); GSBlock::ReadColumn4(y, dst, buff, 16);
for(int i = 0; i < h; i++) memcpy(&buff[i * 16], &src[i * srcpitch + (x >> 1)], 16); for (int i = 0; i < h; i++)
memcpy(&buff[i * 16], &src[i * srcpitch + (x >> 1)], 16);
GSBlock::WriteColumn4<32>(y, dst, buff, 16); GSBlock::WriteColumn4<32>(y, dst, buff, 16);
break; break;
// TODO // TODO
@ -959,7 +969,8 @@ void GSLocalMemory::WriteImageTopBottom(int l, int r, int y, int h, const uint8*
template <int psm, int bsx, int bsy, int trbpp> template <int psm, int bsx, int bsy, int trbpp>
void GSLocalMemory::WriteImage(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
int l = (int)TRXPOS.DSAX; int l = (int)TRXPOS.DSAX;
int r = l + (int)TRXREG.RRW; int r = l + (int)TRXREG.RRW;
@ -1076,7 +1087,8 @@ static bool IsTopLeftAligned(int dsax, int tx, int ty, int bw, int bh)
void GSLocalMemory::WriteImage24(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage24(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
uint32 bp = BITBLTBUF.DBP; uint32 bp = BITBLTBUF.DBP;
uint32 bw = BITBLTBUF.DBW; uint32 bw = BITBLTBUF.DBW;
@ -1110,7 +1122,8 @@ void GSLocalMemory::WriteImage24(int& tx, int& ty, const uint8* src, int len, GI
void GSLocalMemory::WriteImage8H(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage8H(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
uint32 bp = BITBLTBUF.DBP; uint32 bp = BITBLTBUF.DBP;
uint32 bw = BITBLTBUF.DBW; uint32 bw = BITBLTBUF.DBW;
@ -1144,7 +1157,8 @@ void GSLocalMemory::WriteImage8H(int& tx, int& ty, const uint8* src, int len, GI
void GSLocalMemory::WriteImage4HL(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage4HL(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
uint32 bp = BITBLTBUF.DBP; uint32 bp = BITBLTBUF.DBP;
uint32 bw = BITBLTBUF.DBW; uint32 bw = BITBLTBUF.DBW;
@ -1178,7 +1192,8 @@ void GSLocalMemory::WriteImage4HL(int& tx, int& ty, const uint8* src, int len, G
void GSLocalMemory::WriteImage4HH(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage4HH(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
uint32 bp = BITBLTBUF.DBP; uint32 bp = BITBLTBUF.DBP;
uint32 bw = BITBLTBUF.DBW; uint32 bw = BITBLTBUF.DBW;
@ -1212,7 +1227,8 @@ void GSLocalMemory::WriteImage4HH(int& tx, int& ty, const uint8* src, int len, G
void GSLocalMemory::WriteImage24Z(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImage24Z(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(TRXREG.RRW == 0) return; if (TRXREG.RRW == 0)
return;
uint32 bp = BITBLTBUF.DBP; uint32 bp = BITBLTBUF.DBP;
uint32 bw = BITBLTBUF.DBW; uint32 bw = BITBLTBUF.DBW;
@ -1246,7 +1262,8 @@ void GSLocalMemory::WriteImage24Z(int& tx, int& ty, const uint8* src, int len, G
void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG)
{ {
if(len <= 0) return; if (len <= 0)
return;
const uint8* pb = (uint8*)src; const uint8* pb = (uint8*)src;
const uint16* pw = (uint16*)src; const uint16* pw = (uint16*)src;
@ -1278,7 +1295,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel32(addr + offset[x], *pd); WritePixel32(addr + offset[x], *pd);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1298,7 +1319,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel24(addr + offset[x], *(uint32*)pb); WritePixel24(addr + offset[x], *(uint32*)pb);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1320,7 +1345,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel16(addr + offset[x], *pw); WritePixel16(addr + offset[x], *pw);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1337,7 +1366,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel8(addr + offset[x], *pb); WritePixel8(addr + offset[x], *pb);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1355,7 +1388,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel4(addr + offset[x + 1], *pb >> 4); WritePixel4(addr + offset[x + 1], *pb >> 4);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1372,7 +1409,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel8H(addr + offset[x], *pb); WritePixel8H(addr + offset[x], *pb);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1390,7 +1431,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel4HL(addr + offset[x + 1], *pb >> 4); WritePixel4HL(addr + offset[x + 1], *pb >> 4);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1408,7 +1453,11 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
WritePixel4HH(addr + offset[x + 1], *pb >> 4); WritePixel4HH(addr + offset[x + 1], *pb >> 4);
} }
if(x >= ex) {x = sx; y++;} if (x >= ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1422,7 +1471,8 @@ void GSLocalMemory::WriteImageX(int& tx, int& ty, const uint8* src, int len, GIF
void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) const void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITBLTBUF& BITBLTBUF, GIFRegTRXPOS& TRXPOS, GIFRegTRXREG& TRXREG) const
{ {
if(len <= 0) return; if (len <= 0)
return;
uint8* RESTRICT pb = (uint8*)dst; uint8* RESTRICT pb = (uint8*)dst;
uint16* RESTRICT pw = (uint16*)dst; uint16* RESTRICT pw = (uint16*)dst;
@ -1467,7 +1517,8 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
GSVector4i::store<false>(&pd[0], GSVector4i::load(&ps[off + 0], &ps[off + 4])); GSVector4i::store<false>(&pd[0], GSVector4i::load(&ps[off + 0], &ps[off + 4]));
GSVector4i::store<false>(&pd[4], GSVector4i::load(&ps[off + 8], &ps[off + 12])); GSVector4i::store<false>(&pd[4], GSVector4i::load(&ps[off + 8], &ps[off + 12]));
for(int i = 0; i < 8; i++) ASSERT(pd[i] == ps[offset[x + i]]); for (int i = 0; i < 8; i++)
ASSERT(pd[i] == ps[offset[x + i]]);
} }
for (; len > 0 && x < ex; len--, x++, pd++) for (; len > 0 && x < ex; len--, x++, pd++)
@ -1475,7 +1526,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pd = ps[offset[x]]; *pd = ps[offset[x]];
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1499,7 +1554,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
pb[2] = (uint8)(c >> 16); pb[2] = (uint8)(c >> 16);
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1529,7 +1588,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pw = ps[offset[x]]; *pw = ps[offset[x]];
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1554,7 +1617,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pb = ps[offset[x]]; *pb = ps[offset[x]];
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1571,7 +1638,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pb = (uint8)(ReadPixel4(addr + offset[x + 0]) | (ReadPixel4(addr + offset[x + 1]) << 4)); *pb = (uint8)(ReadPixel4(addr + offset[x + 0]) | (ReadPixel4(addr + offset[x + 1]) << 4));
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1596,7 +1667,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pb = (uint8)(ps[offset[x]] >> 24); *pb = (uint8)(ps[offset[x]] >> 24);
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1616,7 +1691,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pb = (uint8)(c0 | c1); *pb = (uint8)(c0 | c1);
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1636,7 +1715,11 @@ void GSLocalMemory::ReadImageX(int& tx, int& ty, uint8* dst, int len, GIFRegBITB
*pb = (uint8)(c0 | c1); *pb = (uint8)(c0 | c1);
} }
if(x == ex) {x = sx; y++;} if (x == ex)
{
x = sx;
y++;
}
} }
break; break;
@ -1687,10 +1770,12 @@ void GSLocalMemory::ReadTextureGPU24(const GSOffset* RESTRICT off, const GSVecto
// Convert packed RGB scanline to 32 bits RGBA // Convert packed RGB scanline to 32 bits RGBA
ASSERT(dstpitch >= r.width() * 4); ASSERT(dstpitch >= r.width() * 4);
for(int y = r.top; y < r.bottom; y ++) { for (int y = r.top; y < r.bottom; y++)
{
uint8* line = dst + y * dstpitch; uint8* line = dst + y * dstpitch;
for(int x = r.right; x >= r.left; x--) { for (int x = r.right; x >= r.left; x--)
{
*(uint32*)&line[x * 4] = *(uint32*)&line[x * 3] & 0xFFFFFF; *(uint32*)&line[x * 4] = *(uint32*)&line[x * 3] & 0xFFFFFF;
} }
} }
@ -1852,9 +1937,7 @@ void GSLocalMemory::ReadTexture(const GSOffset* RESTRICT off, const GSVector4i&
readTexel rt = psm.rt; readTexel rt = psm.rt;
readTexture rtx = psm.rtx; readTexture rtx = psm.rtx;
if(r.width() < psm.bs.x || r.height() < psm.bs.y if (r.width() < psm.bs.x || r.height() < psm.bs.y || (r.left & (psm.bs.x - 1)) || (r.top & (psm.bs.y - 1)) || (r.right & (psm.bs.x - 1)) || (r.bottom & (psm.bs.y - 1)))
|| (r.left & (psm.bs.x - 1)) || (r.top & (psm.bs.y - 1))
|| (r.right & (psm.bs.x - 1)) || (r.bottom & (psm.bs.y - 1)))
{ {
GIFRegTEX0 TEX0; GIFRegTEX0 TEX0;
@ -1870,7 +1953,8 @@ void GSLocalMemory::ReadTexture(const GSOffset* RESTRICT off, const GSVector4i&
{ {
// TODO: expand r to block size, read into temp buffer // TODO: expand r to block size, read into temp buffer
if(!aligned) printf("unaligned memory pointer passed to ReadTexture\n"); if (!aligned)
printf("unaligned memory pointer passed to ReadTexture\n");
for (int y = r.top; y < r.bottom; y++, dst += dstpitch) for (int y = r.top; y < r.bottom; y++, dst += dstpitch)
{ {
@ -2085,7 +2169,8 @@ uint32* GSOffset::GetPages(const GSVector4i& rect, uint32* pages, GSVector4i* bb
GSVector4i r = rect.ralign<Align_Outside>(bs); GSVector4i r = rect.ralign<Align_Outside>(bs);
if(bbox != NULL) *bbox = r; if (bbox != NULL)
*bbox = r;
// worst case: // worst case:
// bp page-aligned: (w * h) / (64 * 32) // bp page-aligned: (w * h) / (64 * 32)

12
plugins/GSdx/GSLocalMemory.h
View File

@ -164,7 +164,10 @@ protected:
__forceinline static uint32 Expand16To32(uint16 c, const GIFRegTEXA& TEXA) __forceinline static uint32 Expand16To32(uint16 c, const GIFRegTEXA& TEXA)
{ {
return (((c & 0x8000) ? TEXA.TA1 : (!TEXA.AEM | c) ? TEXA.TA0 : 0) << 24) | ((c & 0x7c00) << 9) | ((c & 0x03e0) << 6) | ((c & 0x001f) << 3); return (((c & 0x8000) ? TEXA.TA1 : (!TEXA.AEM | c) ? TEXA.TA0 : 0) << 24)
| ((c & 0x7c00) << 9)
| ((c & 0x03e0) << 6)
| ((c & 0x001f) << 3);
} }
// TODO // TODO
@ -557,7 +560,8 @@ public:
__forceinline void WritePixel4(uint32 addr, uint32 c) __forceinline void WritePixel4(uint32 addr, uint32 c)
{ {
int shift = (addr & 1) << 2; addr >>= 1; int shift = (addr & 1) << 2;
addr >>= 1;
m_vm8[addr] = (uint8)((m_vm8[addr] & (0xf0 >> shift)) | ((c & 0x0f) << shift)); m_vm8[addr] = (uint8)((m_vm8[addr] & (0xf0 >> shift)) | ((c & 0x0f) << shift));
} }
@ -913,10 +917,10 @@ public:
// //
template<typename T> void ReadTexture(const GSOffset* RESTRICT off, const GSVector4i& r, uint8* dst, int dstpitch, const GIFRegTEXA& TEXA); template <typename T>
void ReadTexture(const GSOffset* RESTRICT off, const GSVector4i& r, uint8* dst, int dstpitch, const GIFRegTEXA& TEXA);
// //
void SaveBMP(const std::string& fn, uint32 bp, uint32 bw, uint32 psm, int w, int h); void SaveBMP(const std::string& fn, uint32 bp, uint32 bw, uint32 psm, int w, int h);
}; };

76
plugins/GSdx/GSLzma.cpp
View File

@ -21,32 +21,36 @@
#include "stdafx.h" #include "stdafx.h"
#include "GSLzma.h" #include "GSLzma.h"
GSDumpFile::GSDumpFile(char* filename, const char* repack_filename) { GSDumpFile::GSDumpFile(char* filename, const char* repack_filename)
{
m_fp = fopen(filename, "rb"); m_fp = fopen(filename, "rb");
if (m_fp == nullptr) { if (m_fp == nullptr)
{
fprintf(stderr, "failed to open %s\n", filename); fprintf(stderr, "failed to open %s\n", filename);
throw "BAD"; // Just exit the program throw "BAD"; // Just exit the program
} }
m_repack_fp = nullptr; m_repack_fp = nullptr;
if (repack_filename) { if (repack_filename)
{
m_repack_fp = fopen(repack_filename, "wb"); m_repack_fp = fopen(repack_filename, "wb");
if (m_repack_fp == nullptr) if (m_repack_fp == nullptr)
fprintf(stderr, "failed to open %s for repack\n", repack_filename); fprintf(stderr, "failed to open %s for repack\n", repack_filename);
} }
} }
void GSDumpFile::Repack(void* ptr, size_t size) { void GSDumpFile::Repack(void* ptr, size_t size)
{
if (m_repack_fp == nullptr) if (m_repack_fp == nullptr)
return; return;
size_t ret = fwrite(ptr, 1, size, m_repack_fp); size_t ret = fwrite(ptr, 1, size, m_repack_fp);
if (ret != size) if (ret != size)
fprintf(stderr, "Failed to repack\n"); fprintf(stderr, "Failed to repack\n");
} }
GSDumpFile::~GSDumpFile() { GSDumpFile::~GSDumpFile()
{
if (m_fp) if (m_fp)
fclose(m_fp); fclose(m_fp);
if (m_repack_fp) if (m_repack_fp)
@ -54,13 +58,16 @@ GSDumpFile::~GSDumpFile() {
} }
/******************************************************************/ /******************************************************************/
GSDumpLzma::GSDumpLzma(char* filename, const char* repack_filename) : GSDumpFile(filename, repack_filename) { GSDumpLzma::GSDumpLzma(char* filename, const char* repack_filename)
: GSDumpFile(filename, repack_filename)
{
memset(&m_strm, 0, sizeof(lzma_stream)); memset(&m_strm, 0, sizeof(lzma_stream));
lzma_ret ret = lzma_stream_decoder(&m_strm, UINT32_MAX, 0); lzma_ret ret = lzma_stream_decoder(&m_strm, UINT32_MAX, 0);
if (ret != LZMA_OK) { if (ret != LZMA_OK)
{
fprintf(stderr, "Error initializing the decoder! (error code %u)\n", ret); fprintf(stderr, "Error initializing the decoder! (error code %u)\n", ret);
throw "BAD"; // Just exit the program throw "BAD"; // Just exit the program
} }
@ -78,18 +85,21 @@ GSDumpLzma::GSDumpLzma(char* filename, const char* repack_filename) : GSDumpFile
m_strm.next_out = m_area; m_strm.next_out = m_area;
} }
void GSDumpLzma::Decompress() { void GSDumpLzma::Decompress()
{
lzma_action action = LZMA_RUN; lzma_action action = LZMA_RUN;
m_strm.next_out = m_area; m_strm.next_out = m_area;
m_strm.avail_out = m_buff_size; m_strm.avail_out = m_buff_size;
// Nothing left in the input buffer. Read data from the file // Nothing left in the input buffer. Read data from the file
if (m_strm.avail_in == 0 && !feof(m_fp)) { if (m_strm.avail_in == 0 && !feof(m_fp))
{
m_strm.next_in = m_inbuf; m_strm.next_in = m_inbuf;
m_strm.avail_in = fread(m_inbuf, 1, BUFSIZ, m_fp); m_strm.avail_in = fread(m_inbuf, 1, BUFSIZ, m_fp);
if (ferror(m_fp)) { if (ferror(m_fp))
{
fprintf(stderr, "Read error: %s\n", strerror(errno)); fprintf(stderr, "Read error: %s\n", strerror(errno));
throw "BAD"; // Just exit the program throw "BAD"; // Just exit the program
} }
@ -97,10 +107,12 @@ void GSDumpLzma::Decompress() {
lzma_ret ret = lzma_code(&m_strm, action); lzma_ret ret = lzma_code(&m_strm, action);
if (ret != LZMA_OK) { if (ret != LZMA_OK)
{
if (ret == LZMA_STREAM_END) if (ret == LZMA_STREAM_END)
fprintf(stderr, "LZMA decoder finished without error\n\n"); fprintf(stderr, "LZMA decoder finished without error\n\n");
else { else
{
fprintf(stderr, "Decoder error: (error code %u)\n", ret); fprintf(stderr, "Decoder error: (error code %u)\n", ret);
throw "BAD"; // Just exit the program throw "BAD"; // Just exit the program
} }
@ -110,16 +122,20 @@ void GSDumpLzma::Decompress() {
m_avail = m_buff_size - m_strm.avail_out; m_avail = m_buff_size - m_strm.avail_out;
} }
bool GSDumpLzma::IsEof() { bool GSDumpLzma::IsEof()
{
return feof(m_fp) && m_avail == 0 && m_strm.avail_in == 0; return feof(m_fp) && m_avail == 0 && m_strm.avail_in == 0;
} }
bool GSDumpLzma::Read(void* ptr, size_t size) { bool GSDumpLzma::Read(void* ptr, size_t size)
{
size_t off = 0; size_t off = 0;
uint8_t* dst = (uint8_t*)ptr; uint8_t* dst = (uint8_t*)ptr;
size_t full_size = size; size_t full_size = size;
while (size && !IsEof()) { while (size && !IsEof())
if (m_avail == 0) { {
if (m_avail == 0)
{
Decompress(); Decompress();
} }
@ -131,7 +147,8 @@ bool GSDumpLzma::Read(void* ptr, size_t size) {
off += l; off += l;
} }
if (size == 0) { if (size == 0)
{
Repack(ptr, full_size); Repack(ptr, full_size);
return true; return true;
} }
@ -139,7 +156,8 @@ bool GSDumpLzma::Read(void* ptr, size_t size) {
return false; return false;
} }
GSDumpLzma::~GSDumpLzma() { GSDumpLzma::~GSDumpLzma()
{
lzma_end(&m_strm); lzma_end(&m_strm);
if (m_inbuf) if (m_inbuf)
@ -150,26 +168,32 @@ GSDumpLzma::~GSDumpLzma() {
/******************************************************************/ /******************************************************************/
GSDumpRaw::GSDumpRaw(char* filename, const char* repack_filename) : GSDumpFile(filename, repack_filename) { GSDumpRaw::GSDumpRaw(char* filename, const char* repack_filename)
: GSDumpFile(filename, repack_filename)
{
m_buff_size = 0; m_buff_size = 0;
m_area = NULL; m_area = nullptr;
m_inbuf = NULL; m_inbuf = nullptr;
m_avail = 0; m_avail = 0;
m_start = 0; m_start = 0;
} }
bool GSDumpRaw::IsEof() { bool GSDumpRaw::IsEof()
{
return !!feof(m_fp); return !!feof(m_fp);
} }
bool GSDumpRaw::Read(void* ptr, size_t size) { bool GSDumpRaw::Read(void* ptr, size_t size)
{
size_t ret = fread(ptr, 1, size, m_fp); size_t ret = fread(ptr, 1, size, m_fp);
if (ret != size && ferror(m_fp)) { if (ret != size && ferror(m_fp))
{
fprintf(stderr, "GSDumpRaw:: Read error (%zu/%zu)\n", ret, size); fprintf(stderr, "GSDumpRaw:: Read error (%zu/%zu)\n", ret, size);
throw "BAD"; // Just exit the program throw "BAD"; // Just exit the program
} }
if (ret == size) { if (ret == size)
{
Repack(ptr, size); Repack(ptr, size);
return true; return true;
} }

13
plugins/GSdx/GSLzma.h
View File

@ -20,7 +20,8 @@
#include <lzma.h> #include <lzma.h>
class GSDumpFile { class GSDumpFile
{
FILE* m_repack_fp; FILE* m_repack_fp;
protected: protected:
@ -36,8 +37,8 @@ class GSDumpFile {
virtual ~GSDumpFile(); virtual ~GSDumpFile();
}; };
class GSDumpLzma : public GSDumpFile { class GSDumpLzma : public GSDumpFile
{
lzma_stream m_strm; lzma_stream m_strm;
size_t m_buff_size; size_t m_buff_size;
@ -50,7 +51,6 @@ class GSDumpLzma : public GSDumpFile {
void Decompress(); void Decompress();
public: public:
GSDumpLzma(char* filename, const char* repack_filename); GSDumpLzma(char* filename, const char* repack_filename);
virtual ~GSDumpLzma(); virtual ~GSDumpLzma();
@ -58,8 +58,8 @@ class GSDumpLzma : public GSDumpFile {
bool Read(void* ptr, size_t size) final; bool Read(void* ptr, size_t size) final;
}; };
class GSDumpRaw : public GSDumpFile { class GSDumpRaw : public GSDumpFile
{
size_t m_buff_size; size_t m_buff_size;
uint8_t* m_area; uint8_t* m_area;
uint8_t* m_inbuf; uint8_t* m_inbuf;
@ -68,7 +68,6 @@ class GSDumpRaw : public GSDumpFile {
size_t m_start; size_t m_start;
public: public:
GSDumpRaw(char* filename, const char* repack_filename); GSDumpRaw(char* filename, const char* repack_filename);
virtual ~GSDumpRaw() = default; virtual ~GSDumpRaw() = default;

15
plugins/GSdx/GSPerfMon.h
View File

@ -35,7 +35,14 @@ public:
enum counter_t enum counter_t
{ {
Frame, Prim, Draw, Swizzle, Unswizzle, Fillrate, Quad, SyncPoint, Frame,
Prim,
Draw,
Swizzle,
Unswizzle,
Fillrate,
Quad,
SyncPoint,
CounterLast, CounterLast,
}; };
@ -70,6 +77,10 @@ class GSPerfMonAutoTimer
int m_timer; int m_timer;
public: public:
GSPerfMonAutoTimer(GSPerfMon* pm, int timer = GSPerfMon::Main) {m_timer = timer; (m_pm = pm)->Start(m_timer);} GSPerfMonAutoTimer(GSPerfMon* pm, int timer = GSPerfMon::Main)
{
m_timer = timer;
(m_pm = pm)->Start(m_timer);
}
~GSPerfMonAutoTimer() { m_pm->Stop(m_timer); } ~GSPerfMonAutoTimer() { m_pm->Stop(m_timer); }
}; };

18
plugins/GSdx/GSPng.cpp
View File

@ -23,7 +23,8 @@
#include <zlib.h> #include <zlib.h>
#include <png.h> #include <png.h>
struct { struct
{
int type; int type;
int bytes_per_pixel_in; int bytes_per_pixel_in;
int bytes_per_pixel_out; int bytes_per_pixel_out;
@ -39,7 +40,8 @@ struct {
{PNG_COLOR_TYPE_GRAY, 4, 2, 16, {"_R32I_lsb.png", "_R32I_msb.png"}}, // R32I_PNG {PNG_COLOR_TYPE_GRAY, 4, 2, 16, {"_R32I_lsb.png", "_R32I_msb.png"}}, // R32I_PNG
}; };
namespace GSPng { namespace GSPng
{
bool SaveFile(const std::string& file, const Format fmt, const uint8* const image, bool SaveFile(const std::string& file, const Format fmt, const uint8* const image,
uint8* const row, const int width, const int height, const int pitch, uint8* const row, const int width, const int height, const int pitch,
@ -60,7 +62,8 @@ namespace GSPng {
png_infop info_ptr = nullptr; png_infop info_ptr = nullptr;
bool success; bool success;
try { try
{
if (png_ptr == nullptr) if (png_ptr == nullptr)
throw GSDXRecoverableError(); throw GSDXRecoverableError();
@ -82,7 +85,8 @@ namespace GSPng {
if (rb_swapped && type != PNG_COLOR_TYPE_GRAY) if (rb_swapped && type != PNG_COLOR_TYPE_GRAY)
png_set_bgr(png_ptr); png_set_bgr(png_ptr);
for (int y = 0; y < height; ++y) { for (int y = 0; y < height; ++y)
{
for (int x = 0; x < width; ++x) for (int x = 0; x < width; ++x)
for (int i = 0; i < bytes_per_pixel_out; ++i) for (int i = 0; i < bytes_per_pixel_out; ++i)
row[bytes_per_pixel_out * x + i] = image[y * pitch + bytes_per_pixel_in * x + i + offset]; row[bytes_per_pixel_out * x + i] = image[y * pitch + bytes_per_pixel_in * x + i + offset];
@ -91,7 +95,9 @@ namespace GSPng {
png_write_end(png_ptr, nullptr); png_write_end(png_ptr, nullptr);
success = true; success = true;
} catch (GSDXRecoverableError&) { }
catch (GSDXRecoverableError&)
{
fprintf(stderr, "Failed to write image %s\n", file.c_str()); fprintf(stderr, "Failed to write image %s\n", file.c_str());
success = false; success = false;
@ -148,4 +154,4 @@ namespace GSPng {
Save(item->m_fmt, item->m_file, item->m_image, item->m_w, item->m_h, item->m_pitch, item->m_compression); Save(item->m_fmt, item->m_file, item->m_image, item->m_w, item->m_h, item->m_pitch, item->m_compression);
} }
} } // namespace GSPng

8
plugins/GSdx/GSPng.h
View File

@ -22,8 +22,10 @@
#include "GSThread_CXX11.h" #include "GSThread_CXX11.h"
namespace GSPng { namespace GSPng
enum Format { {
enum Format
{
START = 0, START = 0,
RGBA_PNG = 0, RGBA_PNG = 0,
RGB_PNG, RGB_PNG,
@ -55,4 +57,4 @@ namespace GSPng {
void Process(std::shared_ptr<Transaction>& item); void Process(std::shared_ptr<Transaction>& item);
using Worker = GSJobQueue<std::shared_ptr<Transaction>, 16>; using Worker = GSJobQueue<std::shared_ptr<Transaction>, 16>;
} } // namespace GSPng

7
plugins/GSdx/GSState.h
View File

@ -274,11 +274,14 @@ public:
template<int index> void Transfer(const uint8* mem, uint32 size); template<int index> void Transfer(const uint8* mem, uint32 size);
int Freeze(GSFreezeData* fd, bool sizeonly); int Freeze(GSFreezeData* fd, bool sizeonly);
int Defrost(const GSFreezeData* fd); int Defrost(const GSFreezeData* fd);
void GetLastTag(uint32* tag) {*tag = m_path3hack; m_path3hack = 0;} void GetLastTag(uint32* tag)
{
*tag = m_path3hack;
m_path3hack = 0;
}
virtual void SetGameCRC(uint32 crc, int options); virtual void SetGameCRC(uint32 crc, int options);
void SetFrameSkip(int skip); void SetFrameSkip(int skip);
void SetRegsMem(uint8* basemem); void SetRegsMem(uint8* basemem);
void SetIrqCallback(void (*irq)()); void SetIrqCallback(void (*irq)());
void SetMultithreaded(bool mt = true); void SetMultithreaded(bool mt = true);
}; };

24
plugins/GSdx/GSThread_CXX11.h
View File

@ -24,7 +24,8 @@
#include "GSdx.h" #include "GSdx.h"
#include "Utilities/boost_spsc_queue.hpp" #include "Utilities/boost_spsc_queue.hpp"
template<class T, int CAPACITY> class GSJobQueue final template <class T, int CAPACITY>
class GSJobQueue final
{ {
private: private:
std::thread m_thread; std::thread m_thread;
@ -37,12 +38,15 @@ private:
std::condition_variable m_empty; std::condition_variable m_empty;
std::condition_variable m_notempty; std::condition_variable m_notempty;
void ThreadProc() { void ThreadProc()
{
std::unique_lock<std::mutex> l(m_lock); std::unique_lock<std::mutex> l(m_lock);
while (true) { while (true)
{
while (m_queue.empty()) { while (m_queue.empty())
{
if (m_exit) if (m_exit)
return; return;
@ -64,9 +68,9 @@ private:
} }
public: public:
GSJobQueue(std::function<void(T&)> func) : GSJobQueue(std::function<void(T&)> func)
m_func(func), : m_func(func)
m_exit(false) , m_exit(false)
{ {
m_thread = std::thread(&GSJobQueue::ThreadProc, this); m_thread = std::thread(&GSJobQueue::ThreadProc, this);
} }
@ -87,7 +91,8 @@ public:
return m_queue.empty(); return m_queue.empty();
} }
void Push(const T& item) { void Push(const T& item)
{
while (!m_queue.push(item)) while (!m_queue.push(item))
std::this_thread::yield(); std::this_thread::yield();
@ -109,7 +114,8 @@ public:
assert(IsEmpty()); assert(IsEmpty());
} }
void operator() (T& item) { void operator()(T& item)
{
m_func(item); m_func(item);
} }
}; };

18
plugins/GSdx/GSUtil.cpp
View File

@ -211,7 +211,8 @@ bool GSUtil::CheckSSE()
{ {
bool status = true; bool status = true;
struct ISA { struct ISA
{
Xbyak::util::Cpu::Type type; Xbyak::util::Cpu::Type type;
const char* name; const char* name;
}; };
@ -228,8 +229,10 @@ bool GSUtil::CheckSSE()
#endif #endif
}; };
for (size_t i = 0; i < countof(checks); i++) { for (size_t i = 0; i < countof(checks); i++)
if(!g_cpu.has(checks[i].type)) { {
if (!g_cpu.has(checks[i].type))
{
fprintf(stderr, "This CPU does not support %s\n", checks[i].name); fprintf(stderr, "This CPU does not support %s\n", checks[i].name);
status = false; status = false;
@ -323,9 +326,11 @@ GSRendererType GSUtil::GetBestRenderer()
#ifdef _WIN32 #ifdef _WIN32
void GSmkdir(const wchar_t* dir) void GSmkdir(const wchar_t* dir)
{ {
if (!CreateDirectory(dir, nullptr)) { if (!CreateDirectory(dir, nullptr))
{
DWORD errorID = ::GetLastError(); DWORD errorID = ::GetLastError();
if (errorID != ERROR_ALREADY_EXISTS) { if (errorID != ERROR_ALREADY_EXISTS)
{
fprintf(stderr, "Failed to create directory: %ls error %u\n", dir, errorID); fprintf(stderr, "Failed to create directory: %ls error %u\n", dir, errorID);
} }
} }
@ -355,7 +360,8 @@ std::string GStempdir()
const char* psm_str(int psm) const char* psm_str(int psm)
{ {
switch(psm) { switch (psm)
{
// Normal color // Normal color
case PSM_PSMCT32: return "C_32"; case PSM_PSMCT32: return "C_32";
case PSM_PSMCT24: return "C_24"; case PSM_PSMCT24: return "C_24";

6
plugins/GSdx/GSVector.cpp
View File

@ -177,14 +177,16 @@ GSVector4i GSVector4i::fit(int arx, int ary) const
{ {
w = h * arx / ary; w = h * arx / ary;
r.left = (r.left + r.right - w) >> 1; r.left = (r.left + r.right - w) >> 1;
if(r.left & 1) r.left++; if (r.left & 1)
r.left++;
r.right = r.left + w; r.right = r.left + w;
} }
else else
{ {
h = w * ary / arx; h = w * ary / arx;
r.top = (r.top + r.bottom - h) >> 1; r.top = (r.top + r.bottom - h) >> 1;
if(r.top & 1) r.top++; if (r.top & 1)
r.top++;
r.bottom = r.top + h; r.bottom = r.top + h;
} }

3
plugins/GSdx/GSVector.h
View File

@ -41,7 +41,8 @@ enum Round_Mode
#pragma pack(push, 1) #pragma pack(push, 1)
template<class T> class GSVector2T template <class T>
class GSVector2T
{ {
public: public:
union union

33
plugins/GSdx/GSVector4.h
View File

@ -248,7 +248,8 @@ public:
return (v + v) - (v * v) * *this; return (v + v) - (v * v) * *this;
} }
template<int mode> __forceinline GSVector4 round() const template <int mode>
__forceinline GSVector4 round() const
{ {
return GSVector4(_mm_round_ps(m, mode)); return GSVector4(_mm_round_ps(m, mode));
} }
@ -396,7 +397,8 @@ public:
return GSVector4(_mm_hsub_ps(m, v.m)); return GSVector4(_mm_hsub_ps(m, v.m));
} }
template<int i> __forceinline GSVector4 dp(const GSVector4& v) const template <int i>
__forceinline GSVector4 dp(const GSVector4& v) const
{ {
return GSVector4(_mm_dp_ps(m, v.m, i)); return GSVector4(_mm_dp_ps(m, v.m, i));
} }
@ -431,7 +433,8 @@ public:
return GSVector4(_mm_max_ps(m, a)); return GSVector4(_mm_max_ps(m, a));
} }
template<int mask> __forceinline GSVector4 blend32(const GSVector4& a) const template <int mask>
__forceinline GSVector4 blend32(const GSVector4& a) const
{ {
return GSVector4(_mm_blend_ps(m, a, mask)); return GSVector4(_mm_blend_ps(m, a, mask));
} }
@ -506,7 +509,8 @@ public:
return v.blend32(*this, *this == *this); return v.blend32(*this, *this == *this);
} }
template<int src, int dst> __forceinline GSVector4 insert32(const GSVector4& v) const template <int src, int dst>
__forceinline GSVector4 insert32(const GSVector4& v) const
{ {
// TODO: use blendps when src == dst // TODO: use blendps when src == dst
@ -582,12 +586,14 @@ GSVector.h:2973:15: error: shadows template parm 'int i'
template<int i> __forceinline int extract32() const template<int i> __forceinline int extract32() const
#endif #endif
template<int index> __forceinline int extract32() const template <int index>
__forceinline int extract32() const
{ {
return _mm_extract_ps(m, index); return _mm_extract_ps(m, index);
} }
#else #else
template<int i> __forceinline int extract32() const template <int i>
__forceinline int extract32() const
{ {
return _mm_extract_ps(m, i); return _mm_extract_ps(m, i);
} }
@ -630,7 +636,8 @@ GSVector.h:2973:15: error: shadows template parm 'int i'
return GSVector4(v) + (m_x4f800000 & GSVector4::cast(v.sra32(31))); return GSVector4(v) + (m_x4f800000 & GSVector4::cast(v.sra32(31)));
} }
template<bool aligned> __forceinline static GSVector4 load(const void* p) template <bool aligned>
__forceinline static GSVector4 load(const void* p)
{ {
return GSVector4(aligned ? _mm_load_ps((const float*)p) : _mm_loadu_ps((const float*)p)); return GSVector4(aligned ? _mm_load_ps((const float*)p) : _mm_loadu_ps((const float*)p));
} }
@ -650,10 +657,13 @@ GSVector.h:2973:15: error: shadows template parm 'int i'
_mm_storeh_pd((double*)p, _mm_castps_pd(v.m)); _mm_storeh_pd((double*)p, _mm_castps_pd(v.m));
} }
template<bool aligned> __forceinline static void store(void* p, const GSVector4& v) template <bool aligned>
__forceinline static void store(void* p, const GSVector4& v)
{ {
if(aligned) _mm_store_ps((float*)p, v.m); if (aligned)
else _mm_storeu_ps((float*)p, v.m); _mm_store_ps((float*)p, v.m);
else
_mm_storeu_ps((float*)p, v.m);
} }
__forceinline static void store(float* p, const GSVector4& v) __forceinline static void store(float* p, const GSVector4& v)
@ -710,7 +720,8 @@ GSVector.h:2973:15: error: shadows template parm 'int i'
b = v2.h2l(v0); b = v2.h2l(v0);
c = v1.l2h(v3); c = v1.l2h(v3);
d = v3.h2l(v1); d = v3.h2l(v1);
*/ } */
}
__forceinline GSVector4 operator-() const __forceinline GSVector4 operator-() const
{ {

190
plugins/GSdx/GSVector4i.h
View File

@ -70,7 +70,8 @@ public:
__m128i m; __m128i m;
}; };
__forceinline constexpr GSVector4i(): x(0), y(0), z(0), w(0) __forceinline constexpr GSVector4i()
: x(0), y(0), z(0), w(0)
{ {
} }
@ -237,7 +238,8 @@ public:
return sat_i32(a); return sat_i32(a);
} }
template<int mode> __forceinline GSVector4i ralign(const GSVector2i& a) const template <int mode>
__forceinline GSVector4i ralign(const GSVector2i& a) const
{ {
// a must be 1 << n // a must be 1 << n
@ -422,14 +424,16 @@ public:
return GSVector4i(_mm_blendv_epi8(m, a, mask)); return GSVector4i(_mm_blendv_epi8(m, a, mask));
} }
template<int mask> __forceinline GSVector4i blend16(const GSVector4i& a) const template <int mask>
__forceinline GSVector4i blend16(const GSVector4i& a) const
{ {
return GSVector4i(_mm_blend_epi16(m, a, mask)); return GSVector4i(_mm_blend_epi16(m, a, mask));
} }
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
template<int mask> __forceinline GSVector4i blend32(const GSVector4i& v) const template <int mask>
__forceinline GSVector4i blend32(const GSVector4i& v) const
{ {
return GSVector4i(_mm_blend_epi32(m, v.m, mask)); return GSVector4i(_mm_blend_epi32(m, v.m, mask));
} }
@ -660,17 +664,20 @@ public:
return GSVector4i(_mm_cvtepu32_epi64(m)); return GSVector4i(_mm_cvtepu32_epi64(m));
} }
template<int i> __forceinline GSVector4i srl() const template <int i>
__forceinline GSVector4i srl() const
{ {
return GSVector4i(_mm_srli_si128(m, i)); return GSVector4i(_mm_srli_si128(m, i));
} }
template<int i> __forceinline GSVector4i srl(const GSVector4i& v) template <int i>
__forceinline GSVector4i srl(const GSVector4i& v)
{ {
return GSVector4i(_mm_alignr_epi8(v.m, m, i)); return GSVector4i(_mm_alignr_epi8(v.m, m, i));
} }
template<int i> __forceinline GSVector4i sll() const template <int i>
__forceinline GSVector4i sll() const
{ {
return GSVector4i(_mm_slli_si128(m, i)); return GSVector4i(_mm_slli_si128(m, i));
} }
@ -874,21 +881,24 @@ public:
return GSVector4i(_mm_madd_epi16(m, v.m)); return GSVector4i(_mm_madd_epi16(m, v.m));
} }
template<int shift> __forceinline GSVector4i lerp16(const GSVector4i& a, const GSVector4i& f) const template <int shift>
__forceinline GSVector4i lerp16(const GSVector4i& a, const GSVector4i& f) const
{ {
// (a - this) * f << shift + this // (a - this) * f << shift + this
return add16(a.sub16(*this).modulate16<shift>(f)); return add16(a.sub16(*this).modulate16<shift>(f));
} }
template<int shift> __forceinline static GSVector4i lerp16(const GSVector4i& a, const GSVector4i& b, const GSVector4i& c) template <int shift>
__forceinline static GSVector4i lerp16(const GSVector4i& a, const GSVector4i& b, const GSVector4i& c)
{ {
// (a - b) * c << shift // (a - b) * c << shift
return a.sub16(b).modulate16<shift>(c); return a.sub16(b).modulate16<shift>(c);
} }
template<int shift> __forceinline static GSVector4i lerp16(const GSVector4i& a, const GSVector4i& b, const GSVector4i& c, const GSVector4i& d) template <int shift>
__forceinline static GSVector4i lerp16(const GSVector4i& a, const GSVector4i& b, const GSVector4i& c, const GSVector4i& d)
{ {
// (a - b) * c << shift + d // (a - b) * c << shift + d
@ -902,7 +912,8 @@ public:
return add16(a.sub16(*this).mul16l(f).sra16(4)); return add16(a.sub16(*this).mul16l(f).sra16(4));
} }
template<int shift> __forceinline GSVector4i modulate16(const GSVector4i& f) const template <int shift>
__forceinline GSVector4i modulate16(const GSVector4i& f) const
{ {
// a * f << shift // a * f << shift
@ -1007,56 +1018,66 @@ public:
return _mm_testz_si128(m, m) != 0; return _mm_testz_si128(m, m) != 0;
} }
template<int i> __forceinline GSVector4i insert8(int a) const template <int i>
__forceinline GSVector4i insert8(int a) const
{ {
return GSVector4i(_mm_insert_epi8(m, a, i)); return GSVector4i(_mm_insert_epi8(m, a, i));
} }
template<int i> __forceinline int extract8() const template <int i>
__forceinline int extract8() const
{ {
return _mm_extract_epi8(m, i); return _mm_extract_epi8(m, i);
} }
template<int i> __forceinline GSVector4i insert16(int a) const template <int i>
__forceinline GSVector4i insert16(int a) const
{ {
return GSVector4i(_mm_insert_epi16(m, a, i)); return GSVector4i(_mm_insert_epi16(m, a, i));
} }
template<int i> __forceinline int extract16() const template <int i>
__forceinline int extract16() const
{ {
return _mm_extract_epi16(m, i); return _mm_extract_epi16(m, i);
} }
template<int i> __forceinline GSVector4i insert32(int a) const template <int i>
__forceinline GSVector4i insert32(int a) const
{ {
return GSVector4i(_mm_insert_epi32(m, a, i)); return GSVector4i(_mm_insert_epi32(m, a, i));
} }
template<int i> __forceinline int extract32() const template <int i>
__forceinline int extract32() const
{ {
if(i == 0) return GSVector4i::store(*this); if (i == 0)
return GSVector4i::store(*this);
return _mm_extract_epi32(m, i); return _mm_extract_epi32(m, i);
} }
#ifdef _M_AMD64 #ifdef _M_AMD64
template<int i> __forceinline GSVector4i insert64(int64 a) const template <int i>
__forceinline GSVector4i insert64(int64 a) const
{ {
return GSVector4i(_mm_insert_epi64(m, a, i)); return GSVector4i(_mm_insert_epi64(m, a, i));
} }
template<int i> __forceinline int64 extract64() const template <int i>
__forceinline int64 extract64() const
{ {
if(i == 0) return GSVector4i::storeq(*this); if (i == 0)
return GSVector4i::storeq(*this);
return _mm_extract_epi64(m, i); return _mm_extract_epi64(m, i);
} }
#endif #endif
template <int src, class T>
template<int src, class T> __forceinline GSVector4i gather8_4(const T* ptr) const __forceinline GSVector4i gather8_4(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1080,7 +1101,8 @@ public:
return v; return v;
} }
template<class T> __forceinline GSVector4i gather8_8(const T* ptr) const template <class T>
__forceinline GSVector4i gather8_8(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1104,7 +1126,8 @@ public:
return v; return v;
} }
template<int dst, class T> __forceinline GSVector4i gather8_16(const T* ptr, const GSVector4i& a) const template <int dst, class T>
__forceinline GSVector4i gather8_16(const T* ptr, const GSVector4i& a) const
{ {
GSVector4i v = a; GSVector4i v = a;
@ -1120,7 +1143,8 @@ public:
return v; return v;
} }
template<int dst, class T> __forceinline GSVector4i gather8_32(const T* ptr, const GSVector4i& a) const template <int dst, class T>
__forceinline GSVector4i gather8_32(const T* ptr, const GSVector4i& a) const
{ {
GSVector4i v = a; GSVector4i v = a;
@ -1132,7 +1156,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather16_4(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather16_4(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1148,7 +1173,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather16_8(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather16_8(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1164,7 +1190,8 @@ public:
return v; return v;
} }
template<class T>__forceinline GSVector4i gather16_16(const T* ptr) const template <class T>
__forceinline GSVector4i gather16_16(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1180,7 +1207,8 @@ public:
return v; return v;
} }
template<class T1, class T2>__forceinline GSVector4i gather16_16(const T1* ptr1, const T2* ptr2) const template <class T1, class T2>
__forceinline GSVector4i gather16_16(const T1* ptr1, const T2* ptr2) const
{ {
GSVector4i v; GSVector4i v;
@ -1196,7 +1224,8 @@ public:
return v; return v;
} }
template<int dst, class T> __forceinline GSVector4i gather16_32(const T* ptr, const GSVector4i& a) const template <int dst, class T>
__forceinline GSVector4i gather16_32(const T* ptr, const GSVector4i& a) const
{ {
GSVector4i v = a; GSVector4i v = a;
@ -1208,7 +1237,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather32_4(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather32_4(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1219,7 +1249,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather32_8(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather32_8(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1231,7 +1262,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather32_16(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather32_16(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1243,7 +1275,8 @@ public:
return v; return v;
} }
template<class T> __forceinline GSVector4i gather32_32(const T* ptr) const template <class T>
__forceinline GSVector4i gather32_32(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1255,7 +1288,8 @@ public:
return v; return v;
} }
template<class T1, class T2> __forceinline GSVector4i gather32_32(const T1* ptr1, const T2* ptr2) const template <class T1, class T2>
__forceinline GSVector4i gather32_32(const T1* ptr1, const T2* ptr2) const
{ {
GSVector4i v; GSVector4i v;
@ -1269,7 +1303,8 @@ public:
#if defined(_M_AMD64) #if defined(_M_AMD64)
template<int src, class T> __forceinline GSVector4i gather64_4(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_4(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1279,7 +1314,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_8(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_8(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1289,7 +1325,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_16(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_16(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1299,7 +1336,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_32(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_32(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1309,7 +1347,8 @@ public:
return v; return v;
} }
template<class T> __forceinline GSVector4i gather64_64(const T* ptr) const template <class T>
__forceinline GSVector4i gather64_64(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1321,7 +1360,8 @@ public:
#else #else
template<int src, class T> __forceinline GSVector4i gather64_4(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_4(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1330,7 +1370,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_8(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_8(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1339,7 +1380,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_16(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_16(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1348,7 +1390,8 @@ public:
return v; return v;
} }
template<int src, class T> __forceinline GSVector4i gather64_32(const T* ptr) const template <int src, class T>
__forceinline GSVector4i gather64_32(const T* ptr) const
{ {
GSVector4i v; GSVector4i v;
@ -1359,7 +1402,8 @@ public:
#endif #endif
template<class T> __forceinline void gather8_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather8_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather8_4<0>(ptr); dst[0] = gather8_4<0>(ptr);
dst[1] = gather8_4<8>(ptr); dst[1] = gather8_4<8>(ptr);
@ -1370,7 +1414,8 @@ public:
dst[0] = gather8_8<>(ptr); dst[0] = gather8_8<>(ptr);
} }
template<class T> __forceinline void gather16_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather16_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather16_4<0>(ptr); dst[0] = gather16_4<0>(ptr);
dst[1] = gather16_4<4>(ptr); dst[1] = gather16_4<4>(ptr);
@ -1378,18 +1423,21 @@ public:
dst[3] = gather16_4<12>(ptr); dst[3] = gather16_4<12>(ptr);
} }
template<class T> __forceinline void gather16_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather16_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather16_8<0>(ptr); dst[0] = gather16_8<0>(ptr);
dst[1] = gather16_8<8>(ptr); dst[1] = gather16_8<8>(ptr);
} }
template<class T> __forceinline void gather16_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather16_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather16_16<>(ptr); dst[0] = gather16_16<>(ptr);
} }
template<class T> __forceinline void gather32_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather32_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather32_4<0>(ptr); dst[0] = gather32_4<0>(ptr);
dst[1] = gather32_4<2>(ptr); dst[1] = gather32_4<2>(ptr);
@ -1401,7 +1449,8 @@ public:
dst[7] = gather32_4<14>(ptr); dst[7] = gather32_4<14>(ptr);
} }
template<class T> __forceinline void gather32_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather32_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather32_8<0>(ptr); dst[0] = gather32_8<0>(ptr);
dst[1] = gather32_8<4>(ptr); dst[1] = gather32_8<4>(ptr);
@ -1409,18 +1458,21 @@ public:
dst[3] = gather32_8<12>(ptr); dst[3] = gather32_8<12>(ptr);
} }
template<class T> __forceinline void gather32_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather32_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather32_16<0>(ptr); dst[0] = gather32_16<0>(ptr);
dst[1] = gather32_16<4>(ptr); dst[1] = gather32_16<4>(ptr);
} }
template<class T> __forceinline void gather32_32(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather32_32(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather32_32<>(ptr); dst[0] = gather32_32<>(ptr);
} }
template<class T> __forceinline void gather64_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather64_4(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather64_4<0>(ptr); dst[0] = gather64_4<0>(ptr);
dst[1] = gather64_4<1>(ptr); dst[1] = gather64_4<1>(ptr);
@ -1440,7 +1492,8 @@ public:
dst[15] = gather64_4<15>(ptr); dst[15] = gather64_4<15>(ptr);
} }
template<class T> __forceinline void gather64_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather64_8(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather64_8<0>(ptr); dst[0] = gather64_8<0>(ptr);
dst[1] = gather64_8<2>(ptr); dst[1] = gather64_8<2>(ptr);
@ -1452,7 +1505,8 @@ public:
dst[7] = gather64_8<14>(ptr); dst[7] = gather64_8<14>(ptr);
} }
template<class T> __forceinline void gather64_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather64_16(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather64_16<0>(ptr); dst[0] = gather64_16<0>(ptr);
dst[1] = gather64_16<2>(ptr); dst[1] = gather64_16<2>(ptr);
@ -1460,7 +1514,8 @@ public:
dst[3] = gather64_16<8>(ptr); dst[3] = gather64_16<8>(ptr);
} }
template<class T> __forceinline void gather64_32(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather64_32(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather64_32<0>(ptr); dst[0] = gather64_32<0>(ptr);
dst[1] = gather64_32<2>(ptr); dst[1] = gather64_32<2>(ptr);
@ -1468,7 +1523,8 @@ public:
#ifdef _M_AMD64 #ifdef _M_AMD64
template<class T> __forceinline void gather64_64(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const template <class T>
__forceinline void gather64_64(const T* RESTRICT ptr, GSVector4i* RESTRICT dst) const
{ {
dst[0] = gather64_64<>(ptr); dst[0] = gather64_64<>(ptr);
} }
@ -1508,7 +1564,8 @@ public:
return GSVector4i(_mm_unpacklo_epi64(lo, hi)); return GSVector4i(_mm_unpacklo_epi64(lo, hi));
} }
*/ */
template<bool aligned> __forceinline static GSVector4i load(const void* p) template <bool aligned>
__forceinline static GSVector4i load(const void* p)
{ {
return GSVector4i(aligned ? _mm_load_si128((__m128i*)p) : _mm_loadu_si128((__m128i*)p)); return GSVector4i(aligned ? _mm_load_si128((__m128i*)p) : _mm_loadu_si128((__m128i*)p));
} }
@ -1548,10 +1605,13 @@ public:
GSVector4i::storeh(ph, v); GSVector4i::storeh(ph, v);
} }
template<bool aligned> __forceinline static void store(void* p, const GSVector4i& v) template <bool aligned>
__forceinline static void store(void* p, const GSVector4i& v)
{ {
if(aligned) _mm_store_si128((__m128i*)p, v.m); if (aligned)
else _mm_storeu_si128((__m128i*)p, v.m); _mm_store_si128((__m128i*)p, v.m);
else
_mm_storeu_si128((__m128i*)p, v.m);
} }
__forceinline static int store(const GSVector4i& v) __forceinline static int store(const GSVector4i& v)
@ -1573,7 +1633,8 @@ public:
const GSVector4i* s = (const GSVector4i*)src; const GSVector4i* s = (const GSVector4i*)src;
GSVector4i* d = (GSVector4i*)dst; GSVector4i* d = (GSVector4i*)dst;
if(size == 0) return; if (size == 0)
return;
size_t i = 0; size_t i = 0;
size_t j = size >> 6; size_t j = size >> 6;
@ -1588,7 +1649,8 @@ public:
size &= 63; size &= 63;
if(size == 0) return; if (size == 0)
return;
memcpy(d, s, size); memcpy(d, s, size);
} }

36
plugins/GSdx/GSVector8.h
View File

@ -251,7 +251,8 @@ public:
return (v + v) - (v * v) * *this; return (v + v) - (v * v) * *this;
} }
template<int mode> __forceinline GSVector8 round() const template <int mode>
__forceinline GSVector8 round() const
{ {
return GSVector8(_mm256_round_ps(m, mode)); return GSVector8(_mm256_round_ps(m, mode));
} }
@ -396,7 +397,8 @@ public:
return GSVector8(_mm256_hsub_ps(m, v.m)); return GSVector8(_mm256_hsub_ps(m, v.m));
} }
template<int i> __forceinline GSVector8 dp(const GSVector8& v) const template <int i>
__forceinline GSVector8 dp(const GSVector8& v) const
{ {
return GSVector8(_mm256_dp_ps(m, v.m, i)); return GSVector8(_mm256_dp_ps(m, v.m, i));
} }
@ -431,7 +433,8 @@ public:
return GSVector8(_mm256_max_ps(m, a)); return GSVector8(_mm256_max_ps(m, a));
} }
template<int mask> __forceinline GSVector8 blend32(const GSVector8& a) const template <int mask>
__forceinline GSVector8 blend32(const GSVector8& a) const
{ {
return GSVector8(_mm256_blend_ps(m, a, mask)); return GSVector8(_mm256_blend_ps(m, a, mask));
} }
@ -496,7 +499,8 @@ public:
return v.blend32(*this, *this == *this); return v.blend32(*this, *this == *this);
} }
template<int src, int dst> __forceinline GSVector8 insert32(const GSVector8& v) const template <int src, int dst>
__forceinline GSVector8 insert32(const GSVector8& v) const
{ {
// TODO: use blendps when src == dst // TODO: use blendps when src == dst
@ -551,25 +555,29 @@ public:
return *this; return *this;
} }
template<int i> __forceinline int extract32() const template <int i>
__forceinline int extract32() const
{ {
ASSERT(i < 8); ASSERT(i < 8);
return extract<i / 4>().template extract32<i & 3>(); return extract<i / 4>().template extract32<i & 3>();
} }
template<int i> __forceinline GSVector8 insert(__m128 m) const template <int i>
__forceinline GSVector8 insert(__m128 m) const
{ {
ASSERT(i < 2); ASSERT(i < 2);
return GSVector8(_mm256_insertf128_ps(this->m, m, i)); return GSVector8(_mm256_insertf128_ps(this->m, m, i));
} }
template<int i> __forceinline GSVector4 extract() const template <int i>
__forceinline GSVector4 extract() const
{ {
ASSERT(i < 2); ASSERT(i < 2);
if(i == 0) return GSVector4(_mm256_castps256_ps128(m)); if (i == 0)
return GSVector4(_mm256_castps256_ps128(m));
return GSVector4(_mm256_extractf128_ps(m, i)); return GSVector4(_mm256_extractf128_ps(m, i));
} }
@ -606,7 +614,8 @@ public:
return loadh(ph, loadl(pl)); return loadh(ph, loadl(pl));
} }
template<bool aligned> __forceinline static GSVector8 load(const void* p) template <bool aligned>
__forceinline static GSVector8 load(const void* p)
{ {
return GSVector8(aligned ? _mm256_load_ps((const float*)p) : _mm256_loadu_ps((const float*)p)); return GSVector8(aligned ? _mm256_load_ps((const float*)p) : _mm256_loadu_ps((const float*)p));
} }
@ -623,10 +632,13 @@ public:
_mm_store_ps((float*)p, _mm256_extractf128_ps(v.m, 1)); _mm_store_ps((float*)p, _mm256_extractf128_ps(v.m, 1));
} }
template<bool aligned> __forceinline static void store(void* p, const GSVector8& v) template <bool aligned>
__forceinline static void store(void* p, const GSVector8& v)
{ {
if(aligned) _mm256_store_ps((float*)p, v.m); if (aligned)
else _mm256_storeu_ps((float*)p, v.m); _mm256_store_ps((float*)p, v.m);
else
_mm256_storeu_ps((float*)p, v.m);
} }
// //

72
plugins/GSdx/GSVector8i.h
View File

@ -316,7 +316,8 @@ public:
return GSVector8i(_mm256_blendv_epi8(m, a, mask)); return GSVector8i(_mm256_blendv_epi8(m, a, mask));
} }
template<int mask> __forceinline GSVector8i blend16(const GSVector8i& a) const template <int mask>
__forceinline GSVector8i blend16(const GSVector8i& a) const
{ {
return GSVector8i(_mm256_blend_epi16(m, a, mask)); return GSVector8i(_mm256_blend_epi16(m, a, mask));
} }
@ -582,17 +583,20 @@ public:
// //
template<int i> __forceinline GSVector8i srl() const template <int i>
__forceinline GSVector8i srl() const
{ {
return GSVector8i(_mm256_srli_si256(m, i)); return GSVector8i(_mm256_srli_si256(m, i));
} }
template<int i> __forceinline GSVector8i srl(const GSVector8i& v) template <int i>
__forceinline GSVector8i srl(const GSVector8i& v)
{ {
return GSVector8i(_mm256_alignr_epi8(v.m, m, i)); return GSVector8i(_mm256_alignr_epi8(v.m, m, i));
} }
template<int i> __forceinline GSVector8i sll() const template <int i>
__forceinline GSVector8i sll() const
{ {
return GSVector8i(_mm256_slli_si256(m, i)); return GSVector8i(_mm256_slli_si256(m, i));
//return GSVector8i(_mm256_slli_si128(m, i)); //return GSVector8i(_mm256_slli_si128(m, i));
@ -848,21 +852,24 @@ public:
return GSVector8i(_mm256_madd_epi16(m, v.m)); return GSVector8i(_mm256_madd_epi16(m, v.m));
} }
template<int shift> __forceinline GSVector8i lerp16(const GSVector8i& a, const GSVector8i& f) const template <int shift>
__forceinline GSVector8i lerp16(const GSVector8i& a, const GSVector8i& f) const
{ {
// (a - this) * f << shift + this // (a - this) * f << shift + this
return add16(a.sub16(*this).modulate16<shift>(f)); return add16(a.sub16(*this).modulate16<shift>(f));
} }
template<int shift> __forceinline static GSVector8i lerp16(const GSVector8i& a, const GSVector8i& b, const GSVector8i& c) template <int shift>
__forceinline static GSVector8i lerp16(const GSVector8i& a, const GSVector8i& b, const GSVector8i& c)
{ {
// (a - b) * c << shift // (a - b) * c << shift
return a.sub16(b).modulate16<shift>(c); return a.sub16(b).modulate16<shift>(c);
} }
template<int shift> __forceinline static GSVector8i lerp16(const GSVector8i& a, const GSVector8i& b, const GSVector8i& c, const GSVector8i& d) template <int shift>
__forceinline static GSVector8i lerp16(const GSVector8i& a, const GSVector8i& b, const GSVector8i& c, const GSVector8i& d)
{ {
// (a - b) * c << shift + d // (a - b) * c << shift + d
@ -876,7 +883,8 @@ public:
return add16(a.sub16(*this).mul16l(f).sra16(4)); return add16(a.sub16(*this).mul16l(f).sra16(4));
} }
template<int shift> __forceinline GSVector8i modulate16(const GSVector8i& f) const template <int shift>
__forceinline GSVector8i modulate16(const GSVector8i& f) const
{ {
// a * f << shift // a * f << shift
@ -977,7 +985,8 @@ public:
// TODO: extract/insert // TODO: extract/insert
template<int i> __forceinline int extract8() const template <int i>
__forceinline int extract8() const
{ {
ASSERT(i < 32); ASSERT(i < 32);
@ -986,7 +995,8 @@ public:
return v.extract8<i & 15>(); return v.extract8<i & 15>();
} }
template<int i> __forceinline int extract16() const template <int i>
__forceinline int extract16() const
{ {
ASSERT(i < 16); ASSERT(i < 16);
@ -995,27 +1005,32 @@ public:
return v.extract16<i & 8>(); return v.extract16<i & 8>();
} }
template<int i> __forceinline int extract32() const template <int i>
__forceinline int extract32() const
{ {
ASSERT(i < 8); ASSERT(i < 8);
GSVector4i v = extract<i / 4>(); GSVector4i v = extract<i / 4>();
if((i & 3) == 0) return GSVector4i::store(v); if ((i & 3) == 0)
return GSVector4i::store(v);
return v.extract32<i & 3>(); return v.extract32<i & 3>();
} }
template<int i> __forceinline GSVector4i extract() const template <int i>
__forceinline GSVector4i extract() const
{ {
ASSERT(i < 2); ASSERT(i < 2);
if(i == 0) return GSVector4i(_mm256_castsi256_si128(m)); if (i == 0)
return GSVector4i(_mm256_castsi256_si128(m));
return GSVector4i(_mm256_extracti128_si256(m, i)); return GSVector4i(_mm256_extracti128_si256(m, i));
} }
template<int i> __forceinline GSVector8i insert(__m128i m) const template <int i>
__forceinline GSVector8i insert(__m128i m) const
{ {
ASSERT(i < 2); ASSERT(i < 2);
@ -1024,7 +1039,8 @@ public:
// TODO: gather // TODO: gather
template<class T> __forceinline GSVector8i gather32_32(const T* ptr) const template <class T>
__forceinline GSVector8i gather32_32(const T* ptr) const
{ {
GSVector4i v0; GSVector4i v0;
GSVector4i v1; GSVector4i v1;
@ -1060,7 +1076,8 @@ public:
return GSVector8i(_mm256_i32gather_epi32((const int*)ptr, m, 4)); return GSVector8i(_mm256_i32gather_epi32((const int*)ptr, m, 4));
} }
template<class T1, class T2> __forceinline GSVector8i gather32_32(const T1* ptr1, const T2* ptr2) const template <class T1, class T2>
__forceinline GSVector8i gather32_32(const T1* ptr1, const T2* ptr2) const
{ {
GSVector4i v0; GSVector4i v0;
GSVector4i v1; GSVector4i v1;
@ -1091,7 +1108,8 @@ public:
return gather32_32<uint32>(ptr1).gather32_32<uint32>(ptr2); return gather32_32<uint32>(ptr1).gather32_32<uint32>(ptr2);
} }
template<class T> __forceinline void gather32_32(const T* RESTRICT ptr, GSVector8i* RESTRICT dst) const template <class T>
__forceinline void gather32_32(const T* RESTRICT ptr, GSVector8i* RESTRICT dst) const
{ {
dst[0] = gather32_32<>(ptr); dst[0] = gather32_32<>(ptr);
} }
@ -1144,7 +1162,8 @@ public:
// return GSVector8i(l).insert<1>(h); // return GSVector8i(l).insert<1>(h);
} }
template<bool aligned> __forceinline static GSVector8i load(const void* p) template <bool aligned>
__forceinline static GSVector8i load(const void* p)
{ {
return GSVector8i(aligned ? _mm256_load_si256((__m256i*)p) : _mm256_loadu_si256((__m256i*)p)); return GSVector8i(aligned ? _mm256_load_si256((__m256i*)p) : _mm256_loadu_si256((__m256i*)p));
} }
@ -1184,10 +1203,13 @@ public:
GSVector8i::storeh(ph, v); GSVector8i::storeh(ph, v);
} }
template<bool aligned> __forceinline static void store(void* p, const GSVector8i& v) template <bool aligned>
__forceinline static void store(void* p, const GSVector8i& v)
{ {
if(aligned) _mm256_store_si256((__m256i*)p, v.m); if (aligned)
else _mm256_storeu_si256((__m256i*)p, v.m); _mm256_store_si256((__m256i*)p, v.m);
else
_mm256_storeu_si256((__m256i*)p, v.m);
} }
__forceinline static int store(const GSVector8i& v) __forceinline static int store(const GSVector8i& v)
@ -1209,7 +1231,8 @@ public:
const GSVector8i* s = (const GSVector8i*)src; const GSVector8i* s = (const GSVector8i*)src;
GSVector8i* d = (GSVector8i*)dst; GSVector8i* d = (GSVector8i*)dst;
if(size == 0) return; if (size == 0)
return;
size_t i = 0; size_t i = 0;
size_t j = size >> 7; size_t j = size >> 7;
@ -1224,7 +1247,8 @@ public:
size &= 127; size &= 127;
if(size == 0) return; if (size == 0)
return;
memcpy(d, s, size); memcpy(d, s, size);
} }

62
plugins/GSdx/GSdx.cpp
View File

@ -47,11 +47,14 @@ bool GSdxApp::LoadResource(int id, std::vector<char>& buff, const wchar_t* type)
{ {
buff.clear(); buff.clear();
HRSRC hRsrc = FindResource((HMODULE)s_hModule, MAKEINTRESOURCE(id), type != NULL ? type : (LPWSTR)RT_RCDATA); HRSRC hRsrc = FindResource((HMODULE)s_hModule, MAKEINTRESOURCE(id), type != NULL ? type : (LPWSTR)RT_RCDATA);
if(!hRsrc) return false; if (!hRsrc)
return false;
HGLOBAL hGlobal = ::LoadResource((HMODULE)s_hModule, hRsrc); HGLOBAL hGlobal = ::LoadResource((HMODULE)s_hModule, hRsrc);
if(!hGlobal) return false; if (!hGlobal)
return false;
DWORD size = SizeofResource((HMODULE)s_hModule, hRsrc); DWORD size = SizeofResource((HMODULE)s_hModule, hRsrc);
if(!size) return false; if (!size)
return false;
// On Linux resources are always NULL terminated // On Linux resources are always NULL terminated
// Add + 1 on size to do the same for compatibility sake (required by GSDeviceOGL) // Add + 1 on size to do the same for compatibility sake (required by GSDeviceOGL)
buff.resize(size + 1); buff.resize(size + 1);
@ -66,7 +69,8 @@ bool GSdxApp::LoadResource(int id, std::vector<char>& buff, const wchar_t* type)
bool GSdxApp::LoadResource(int id, std::vector<char>& buff, const char* type) bool GSdxApp::LoadResource(int id, std::vector<char>& buff, const char* type)
{ {
std::string path; std::string path;
switch (id) { switch (id)
{
case IDR_COMMON_GLSL: case IDR_COMMON_GLSL:
path = "/GSdx/res/glsl/common_header.glsl"; path = "/GSdx/res/glsl/common_header.glsl";
break; break;
@ -104,7 +108,8 @@ bool GSdxApp::LoadResource(int id, std::vector<char>& buff, const char* type)
size_t size = 0; size_t size = 0;
const void* data = g_bytes_get_data(bytes, &size); const void* data = g_bytes_get_data(bytes, &size);
if (data == nullptr || size == 0) { if (data == nullptr || size == 0)
{
printf("Failed to get data for resource: %d\n", id); printf("Failed to get data for resource: %d\n", id);
return false; return false;
} }
@ -125,11 +130,13 @@ size_t GSdxApp::GetIniString(const char* lpAppName, const char* lpKeyName, const
std::string key(lpKeyName); std::string key(lpKeyName);
std::string value = m_configuration_map[key]; std::string value = m_configuration_map[key];
if (value.empty()) { if (value.empty())
{
// save the value for futur call // save the value for futur call
m_configuration_map[key] = std::string(lpDefault); m_configuration_map[key] = std::string(lpDefault);
strcpy(lpReturnedString, lpDefault); strcpy(lpReturnedString, lpDefault);
} else }
else
strcpy(lpReturnedString, value.c_str()); strcpy(lpReturnedString, value.c_str());
return 0; return 0;
@ -146,16 +153,19 @@ bool GSdxApp::WriteIniString(const char* lpAppName, const char* lpKeyName, const
// Save config to a file // Save config to a file
FILE* f = px_fopen(lpFileName, "w"); FILE* f = px_fopen(lpFileName, "w");
if (f == NULL) return false; // FIXME print a nice message if (f == NULL)
return false; // FIXME print a nice message
// Maintain compatibility with GSDumpGUI/old Windows ini. // Maintain compatibility with GSDumpGUI/old Windows ini.
#ifdef _WIN32 #ifdef _WIN32
fprintf(f, "[Settings]\n"); fprintf(f, "[Settings]\n");
#endif #endif
for (const auto& entry : m_configuration_map) { for (const auto& entry : m_configuration_map)
{
// Do not save the inifile key which is not an option // Do not save the inifile key which is not an option
if (entry.first.compare("inifile") == 0) continue; if (entry.first.compare("inifile") == 0)
continue;
// Only keep option that have a default value (allow to purge old option of the GSdx.ini) // Only keep option that have a default value (allow to purge old option of the GSdx.ini)
if (!entry.second.empty() && m_default_configuration.find(entry.first) != m_default_configuration.end()) if (!entry.second.empty() && m_default_configuration.find(entry.first) != m_default_configuration.end())
@ -171,11 +181,13 @@ int GSdxApp::GetIniInt(const char* lpAppName, const char* lpKeyName, int nDefaul
BuildConfigurationMap(lpFileName); BuildConfigurationMap(lpFileName);
std::string value = m_configuration_map[std::string(lpKeyName)]; std::string value = m_configuration_map[std::string(lpKeyName)];
if (value.empty()) { if (value.empty())
{
// save the value for futur call // save the value for futur call
SetConfig(lpKeyName, nDefault); SetConfig(lpKeyName, nDefault);
return nDefault; return nDefault;
} else }
else
return atoi(value.c_str()); return atoi(value.c_str());
} }
@ -425,10 +437,12 @@ void GSdxApp::Init()
void GSdxApp::ReloadConfig() void GSdxApp::ReloadConfig()
{ {
if (m_configuration_map.empty()) return; if (m_configuration_map.empty())
return;
auto file = m_configuration_map.find("inifile"); auto file = m_configuration_map.find("inifile");
if (file == m_configuration_map.end()) return; if (file == m_configuration_map.end())
return;
// A map was built so reload it // A map was built so reload it
std::string filename = file->second; std::string filename = file->second;
@ -440,7 +454,8 @@ void GSdxApp::BuildConfigurationMap(const char* lpFileName)
{ {
// Check if the map was already built // Check if the map was already built
std::string inifile_value(lpFileName); std::string inifile_value(lpFileName);
if ( inifile_value.compare(m_configuration_map["inifile"]) == 0 ) return; if (inifile_value.compare(m_configuration_map["inifile"]) == 0)
return;
m_configuration_map["inifile"] = inifile_value; m_configuration_map["inifile"] = inifile_value;
// Load config from file // Load config from file
@ -453,7 +468,8 @@ void GSdxApp::BuildConfigurationMap(const char* lpFileName)
return; return;
std::string line; std::string line;
while (std::getline(file, line)) { while (std::getline(file, line))
{
const auto separator = line.find('='); const auto separator = line.find('=');
if (separator == std::string::npos) if (separator == std::string::npos)
continue; continue;
@ -507,9 +523,12 @@ std::string GSdxApp::GetConfigS(const char* entry)
char buff[4096] = {0}; char buff[4096] = {0};
auto def = m_default_configuration.find(entry); auto def = m_default_configuration.find(entry);
if (def != m_default_configuration.end()) { if (def != m_default_configuration.end())
{
GetIniString(m_section.c_str(), entry, def->second.c_str(), buff, countof(buff), m_ini.c_str()); GetIniString(m_section.c_str(), entry, def->second.c_str(), buff, countof(buff), m_ini.c_str());
} else { }
else
{
fprintf(stderr, "Option %s doesn't have a default value\n", entry); fprintf(stderr, "Option %s doesn't have a default value\n", entry);
GetIniString(m_section.c_str(), entry, "", buff, countof(buff), m_ini.c_str()); GetIniString(m_section.c_str(), entry, "", buff, countof(buff), m_ini.c_str());
} }
@ -526,9 +545,12 @@ int GSdxApp::GetConfigI(const char* entry)
{ {
auto def = m_default_configuration.find(entry); auto def = m_default_configuration.find(entry);
if (def != m_default_configuration.end()) { if (def != m_default_configuration.end())
{
return GetIniInt(m_section.c_str(), entry, std::stoi(def->second), m_ini.c_str()); return GetIniInt(m_section.c_str(), entry, std::stoi(def->second), m_ini.c_str());
} else { }
else
{
fprintf(stderr, "Option %s doesn't have a default value\n", entry); fprintf(stderr, "Option %s doesn't have a default value\n", entry);
return GetIniInt(m_section.c_str(), entry, 0, m_ini.c_str()); return GetIniInt(m_section.c_str(), entry, 0, m_ini.c_str());
} }

22
plugins/GSdx/GSdx.h
View File

@ -39,7 +39,10 @@ public:
void* GetModuleHandlePtr(); void* GetModuleHandlePtr();
#ifdef _WIN32 #ifdef _WIN32
HMODULE GetModuleHandle() {return (HMODULE)GetModuleHandlePtr();} HMODULE GetModuleHandle()
{
return (HMODULE)GetModuleHandlePtr();
}
#endif #endif
void BuildConfigurationMap(const char* lpFileName); void BuildConfigurationMap(const char* lpFileName);
@ -59,7 +62,10 @@ public:
void SetConfig(const char* entry, int value); void SetConfig(const char* entry, int value);
// Avoid issue with overloading // Avoid issue with overloading
template <typename T> template <typename T>
T GetConfigT(const char* entry) { return static_cast<T>(GetConfigI(entry)); } T GetConfigT(const char* entry)
{
return static_cast<T>(GetConfigI(entry));
}
int GetConfigI(const char* entry); int GetConfigI(const char* entry);
bool GetConfigB(const char* entry); bool GetConfigB(const char* entry);
std::string GetConfigS(const char* entry); std::string GetConfigS(const char* entry);
@ -87,8 +93,14 @@ public:
std::vector<GSSetting> m_gs_tv_shaders; std::vector<GSSetting> m_gs_tv_shaders;
}; };
struct GSDXError {}; struct GSDXError
struct GSDXRecoverableError : GSDXError {}; {
struct GSDXErrorGlVertexArrayTooSmall : GSDXError {}; };
struct GSDXRecoverableError : GSDXError
{
};
struct GSDXErrorGlVertexArrayTooSmall : GSDXError
{
};
extern GSdxApp theApp; extern GSdxApp theApp;

12
plugins/GSdx/Renderers/Common/GSDevice.cpp
View File

@ -42,7 +42,8 @@ GSDevice::GSDevice()
GSDevice::~GSDevice() GSDevice::~GSDevice()
{ {
for(auto t : m_pool) delete t; for (auto t : m_pool)
delete t;
delete m_backbuffer; delete m_backbuffer;
delete m_merge; delete m_merge;
@ -60,7 +61,8 @@ bool GSDevice::Create(const std::shared_ptr<GSWnd>& wnd)
bool GSDevice::Reset(int w, int h) bool GSDevice::Reset(int w, int h)
{ {
for(auto t : m_pool) delete t; for (auto t : m_pool)
delete t;
m_pool.clear(); m_pool.clear();
@ -364,7 +366,11 @@ void GSDevice::ShadeBoost()
bool GSDevice::ResizeTexture(GSTexture** t, int type, int w, int h) bool GSDevice::ResizeTexture(GSTexture** t, int type, int w, int h)
{ {
if(t == NULL) {ASSERT(0); return false;} if (t == NULL)
{
ASSERT(0);
return false;
}
GSTexture* t2 = *t; GSTexture* t2 = *t;

22
plugins/GSdx/Renderers/Common/GSDevice.h
View File

@ -124,7 +124,10 @@ enum HWBlendFlags
}; };
// Determines the HW blend function for DX11/OGL // Determines the HW blend function for DX11/OGL
struct HWBlend { uint16 flags, op, src, dst; }; struct HWBlend
{
uint16 flags, op, src, dst;
};
class GSDevice : public GSAlignedClass<32> class GSDevice : public GSAlignedClass<32>
{ {
@ -157,8 +160,14 @@ protected:
GSTexture* m_blend; GSTexture* m_blend;
GSTexture* m_target_tmp; GSTexture* m_target_tmp;
GSTexture* m_current; GSTexture* m_current;
struct {size_t stride, start, count, limit;} m_vertex; struct
struct {size_t start, count, limit;} m_index; {
size_t stride, start, count, limit;
} m_vertex;
struct
{
size_t start, count, limit;
} m_index;
unsigned int m_frame; // for ageing the pool unsigned int m_frame; // for ageing the pool
bool m_linear_present; bool m_linear_present;
@ -180,7 +189,12 @@ public:
void Recycle(GSTexture* t); void Recycle(GSTexture* t);
enum {Windowed, Fullscreen, DontCare}; enum
{
Windowed,
Fullscreen,
DontCare
};
virtual bool Create(const std::shared_ptr<GSWnd>& wnd); virtual bool Create(const std::shared_ptr<GSWnd>& wnd);
virtual bool Reset(int w, int h); virtual bool Reset(int w, int h);

118
plugins/GSdx/Renderers/Common/GSFastList.h
View File

@ -23,7 +23,8 @@
#pragma once #pragma once
template <class T> template <class T>
struct Element { struct Element
{
T data; T data;
uint16 next_index; uint16 next_index;
uint16 prev_index; uint16 prev_index;
@ -33,8 +34,10 @@ template <class T>
class FastListIterator; class FastListIterator;
template <class T> template <class T>
class FastList { class FastList
{
friend class FastListIterator<T>; friend class FastListIterator<T>;
private: private:
// The index of the first element of the list is m_buffer[0].next_index // The index of the first element of the list is m_buffer[0].next_index
// The first Element<T> of the list has prev_index equal to 0 // The first Element<T> of the list has prev_index equal to 0
@ -56,16 +59,19 @@ private:
uint16* m_free_indexes_stack; uint16* m_free_indexes_stack;
public: public:
__forceinline FastList() { __forceinline FastList()
{
m_buffer = nullptr; m_buffer = nullptr;
clear(); clear();
} }
__forceinline ~FastList() { __forceinline ~FastList()
{
_aligned_free(m_buffer); _aligned_free(m_buffer);
} }
void clear() { void clear()
{
// Initialize m_capacity to 4 so we avoid to Grow() on initial insertions // Initialize m_capacity to 4 so we avoid to Grow() on initial insertions
// The code doesn't break if this value is changed with anything from 1 to USHRT_MAX // The code doesn't break if this value is changed with anything from 1 to USHRT_MAX
m_capacity = 4; m_capacity = 4;
@ -83,14 +89,17 @@ public:
m_free_indexes_stack_top = 0; m_free_indexes_stack_top = 0;
// m_buffer index 0 is reserved for auxiliary element // m_buffer index 0 is reserved for auxiliary element
for (uint16 i = 0; i < m_capacity - 1; i++) { for (uint16 i = 0; i < m_capacity - 1; i++)
{
m_free_indexes_stack[i] = i + 1; m_free_indexes_stack[i] = i + 1;
} }
} }
// Insert the element in front of the list and return its position in m_buffer // Insert the element in front of the list and return its position in m_buffer
__forceinline uint16 InsertFront(const T& data) { __forceinline uint16 InsertFront(const T& data)
if (Full()) { {
if (Full())
{
Grow(); Grow();
} }
@ -101,81 +110,99 @@ public:
return free_index; return free_index;
} }
__forceinline void push_front(const T& data) { __forceinline void push_front(const T& data)
{
InsertFront(data); InsertFront(data);
} }
__forceinline const T& back() const { __forceinline const T& back() const
{
return m_buffer[LastIndex()].data; return m_buffer[LastIndex()].data;
} }
__forceinline void pop_back() { __forceinline void pop_back()
{
EraseIndex(LastIndex()); EraseIndex(LastIndex());
} }
__forceinline uint16 size() const { __forceinline uint16 size() const
{
return m_free_indexes_stack_top; return m_free_indexes_stack_top;
} }
__forceinline bool empty() const { __forceinline bool empty() const
{
return size() == 0; return size() == 0;
} }
__forceinline void EraseIndex(const uint16 index) { __forceinline void EraseIndex(const uint16 index)
{
ListRemove(index); ListRemove(index);
m_free_indexes_stack[--m_free_indexes_stack_top] = index; m_free_indexes_stack[--m_free_indexes_stack_top] = index;
} }
__forceinline void MoveFront(const uint16 index) { __forceinline void MoveFront(const uint16 index)
if (FirstIndex() != index) { {
if (FirstIndex() != index)
{
ListRemove(index); ListRemove(index);
ListInsertFront(index); ListInsertFront(index);
} }
} }
__forceinline const FastListIterator<T> begin() const { __forceinline const FastListIterator<T> begin() const
{
return FastListIterator<T>(this, FirstIndex()); return FastListIterator<T>(this, FirstIndex());
} }
__forceinline const FastListIterator<T> end() const { __forceinline const FastListIterator<T> end() const
{
return FastListIterator<T>(this, 0); return FastListIterator<T>(this, 0);
} }
__forceinline FastListIterator<T> erase(FastListIterator<T> i) { __forceinline FastListIterator<T> erase(FastListIterator<T> i)
{
EraseIndex(i.Index()); EraseIndex(i.Index());
return ++i; return ++i;
} }
private: private:
// Accessed by FastListIterator<T> using class friendship // Accessed by FastListIterator<T> using class friendship
__forceinline const T& Data(const uint16 index) const { __forceinline const T& Data(const uint16 index) const
{
return m_buffer[index].data; return m_buffer[index].data;
} }
// Accessed by FastListIterator<T> using class friendship // Accessed by FastListIterator<T> using class friendship
__forceinline uint16 NextIndex(const uint16 index) const { __forceinline uint16 NextIndex(const uint16 index) const
{
return m_buffer[index].next_index; return m_buffer[index].next_index;
} }
// Accessed by FastListIterator<T> using class friendship // Accessed by FastListIterator<T> using class friendship
__forceinline uint16 PrevIndex(const uint16 index) const { __forceinline uint16 PrevIndex(const uint16 index) const
{
return m_buffer[index].prev_index; return m_buffer[index].prev_index;
} }
__forceinline uint16 FirstIndex() const { __forceinline uint16 FirstIndex() const
{
return m_buffer[0].next_index; return m_buffer[0].next_index;
} }
__forceinline uint16 LastIndex() const { __forceinline uint16 LastIndex() const
{
return m_buffer[0].prev_index; return m_buffer[0].prev_index;
} }
__forceinline bool Full() const { __forceinline bool Full() const
{
// The minus one is due to the presence of the auxiliary element // The minus one is due to the presence of the auxiliary element
return size() == m_capacity - 1; return size() == m_capacity - 1;
} }
__forceinline void ListInsertFront(const uint16 index) { __forceinline void ListInsertFront(const uint16 index)
{
// Update prev / next indexes to add m_buffer[index] to the chain // Update prev / next indexes to add m_buffer[index] to the chain
Element<T>& head = m_buffer[0]; Element<T>& head = m_buffer[0];
m_buffer[index].prev_index = 0; m_buffer[index].prev_index = 0;
@ -184,15 +211,18 @@ private:
head.next_index = index; head.next_index = index;
} }
__forceinline void ListRemove(const uint16 index) { __forceinline void ListRemove(const uint16 index)
{
// Update prev / next indexes to remove m_buffer[index] from the chain // Update prev / next indexes to remove m_buffer[index] from the chain
const Element<T>& to_remove = m_buffer[index]; const Element<T>& to_remove = m_buffer[index];
m_buffer[to_remove.prev_index].next_index = to_remove.next_index; m_buffer[to_remove.prev_index].next_index = to_remove.next_index;
m_buffer[to_remove.next_index].prev_index = to_remove.prev_index; m_buffer[to_remove.next_index].prev_index = to_remove.prev_index;
} }
void Grow() { void Grow()
if (m_capacity == USHRT_MAX) { {
if (m_capacity == USHRT_MAX)
{
throw std::runtime_error("FastList size maxed out at USHRT_MAX (65535) elements, cannot grow futhermore."); throw std::runtime_error("FastList size maxed out at USHRT_MAX (65535) elements, cannot grow futhermore.");
} }
@ -210,7 +240,8 @@ private:
m_free_indexes_stack = new_free_indexes_stack; m_free_indexes_stack = new_free_indexes_stack;
// Initialize the additional space in the stack // Initialize the additional space in the stack
for (uint16 i = m_capacity - 1; i < new_capacity - 1; i++) { for (uint16 i = m_capacity - 1; i < new_capacity - 1; i++)
{
m_free_indexes_stack[i] = i + 1; m_free_indexes_stack[i] = i + 1;
} }
@ -228,50 +259,59 @@ private:
uint16 m_index; uint16 m_index;
public: public:
__forceinline FastListIterator(const FastList<T>* fastlist, const uint16 index) { __forceinline FastListIterator(const FastList<T>* fastlist, const uint16 index)
{
m_fastlist = fastlist; m_fastlist = fastlist;
m_index = index; m_index = index;
} }
__forceinline bool operator!=(const FastListIterator<T>& other) const { __forceinline bool operator!=(const FastListIterator<T>& other) const
{
return (m_index != other.m_index); return (m_index != other.m_index);
} }
__forceinline bool operator==(const FastListIterator<T>& other) const { __forceinline bool operator==(const FastListIterator<T>& other) const
{
return (m_index == other.m_index); return (m_index == other.m_index);
} }
// Prefix increment // Prefix increment
__forceinline const FastListIterator<T>& operator++() { __forceinline const FastListIterator<T>& operator++()
{
m_index = m_fastlist->NextIndex(m_index); m_index = m_fastlist->NextIndex(m_index);
return *this; return *this;
} }
// Postfix increment // Postfix increment
__forceinline const FastListIterator<T> operator++(int) { __forceinline const FastListIterator<T> operator++(int)
{
FastListIterator<T> copy(*this); FastListIterator<T> copy(*this);
++(*this); ++(*this);
return copy; return copy;
} }
// Prefix decrement // Prefix decrement
__forceinline const FastListIterator<T>& operator--() { __forceinline const FastListIterator<T>& operator--()
{
m_index = m_fastlist->PrevIndex(m_index); m_index = m_fastlist->PrevIndex(m_index);
return *this; return *this;
} }
// Postfix decrement // Postfix decrement
__forceinline const FastListIterator<T> operator--(int) { __forceinline const FastListIterator<T> operator--(int)
{
FastListIterator<T> copy(*this); FastListIterator<T> copy(*this);
--(*this); --(*this);
return copy; return copy;
} }
__forceinline const T& operator*() const { __forceinline const T& operator*() const
{
return m_fastlist->Data(m_index); return m_fastlist->Data(m_index);
} }
__forceinline uint16 Index() const { __forceinline uint16 Index() const
{
return m_index; return m_index;
} }
}; };

6
plugins/GSdx/Renderers/Common/GSFunctionMap.h
View File

@ -28,7 +28,8 @@
#include "Renderers/SW/GSScanlineEnvironment.h" #include "Renderers/SW/GSScanlineEnvironment.h"
template<class KEY, class VALUE> class GSFunctionMap template <class KEY, class VALUE>
class GSFunctionMap
{ {
protected: protected:
struct ActivePtr struct ActivePtr
@ -53,7 +54,8 @@ public:
virtual ~GSFunctionMap() virtual ~GSFunctionMap()
{ {
for(auto &i : m_map_active) delete i.second; for (auto& i : m_map_active)
delete i.second;
} }
VALUE operator[](KEY key) VALUE operator[](KEY key)

201
plugins/GSdx/Renderers/Common/GSOsdManager.cpp
View File

@ -26,14 +26,17 @@
#include "resource.h" #include "resource.h"
#endif #endif
void GSOsdManager::LoadFont() { void GSOsdManager::LoadFont()
{
FT_Error error = FT_New_Face(m_library, theApp.GetConfigS("osd_fontname").c_str(), 0, &m_face); FT_Error error = FT_New_Face(m_library, theApp.GetConfigS("osd_fontname").c_str(), 0, &m_face);
if (error) { if (error)
{
FT_Error error_load_res = 1; FT_Error error_load_res = 1;
if (theApp.LoadResource(IDR_FONT_ROBOTO, resource_data_buffer)) if (theApp.LoadResource(IDR_FONT_ROBOTO, resource_data_buffer))
error_load_res = FT_New_Memory_Face(m_library, (const FT_Byte*)resource_data_buffer.data(), resource_data_buffer.size(), 0, &m_face); error_load_res = FT_New_Memory_Face(m_library, (const FT_Byte*)resource_data_buffer.data(), resource_data_buffer.size(), 0, &m_face);
if (error_load_res) { if (error_load_res)
{
m_face = NULL; m_face = NULL;
fprintf(stderr, "Failed to init freetype face from external and internal resource\n"); fprintf(stderr, "Failed to init freetype face from external and internal resource\n");
if (error == FT_Err_Unknown_File_Format) if (error == FT_Err_Unknown_File_Format)
@ -45,11 +48,14 @@ void GSOsdManager::LoadFont() {
LoadSize(); LoadSize();
} }
void GSOsdManager::LoadSize() { void GSOsdManager::LoadSize()
if (!m_face) return; {
if (!m_face)
return;
FT_Error error = FT_Set_Pixel_Sizes(m_face, 0, m_size);; FT_Error error = FT_Set_Pixel_Sizes(m_face, 0, m_size);
if (error) { if (error)
{
fprintf(stderr, "Failed to init the face size\n"); fprintf(stderr, "Failed to init the face size\n");
return; return;
} }
@ -60,7 +66,8 @@ void GSOsdManager::LoadSize() {
m_atlas_h = m_size + 10; // another random guess m_atlas_h = m_size + 10; // another random guess
} }
GSOsdManager::GSOsdManager() : m_atlas_h(0) GSOsdManager::GSOsdManager()
: m_atlas_h(0)
, m_atlas_w(0) , m_atlas_w(0)
, m_max_width(0) , m_max_width(0)
, m_onscreen_messages(0) , m_onscreen_messages(0)
@ -79,7 +86,8 @@ GSOsdManager::GSOsdManager() : m_atlas_h(0)
m_color = r | (g << 8) | (b << 16) | (255 << 24); m_color = r | (g << 8) | (b << 16) | (255 << 24);
if (FT_Init_FreeType(&m_library)) { if (FT_Init_FreeType(&m_library))
{
m_face = NULL; m_face = NULL;
fprintf(stderr, "Failed to init the freetype library\n"); fprintf(stderr, "Failed to init the freetype library\n");
return; return;
@ -91,24 +99,30 @@ GSOsdManager::GSOsdManager() : m_atlas_h(0)
AddGlyph(' '); AddGlyph(' ');
} }
GSOsdManager::~GSOsdManager() { GSOsdManager::~GSOsdManager()
{
FT_Done_FreeType(m_library); FT_Done_FreeType(m_library);
} }
GSVector2i GSOsdManager::get_texture_font_size() { GSVector2i GSOsdManager::get_texture_font_size()
{
return GSVector2i(m_atlas_w, m_atlas_h); return GSVector2i(m_atlas_w, m_atlas_h);
} }
void GSOsdManager::upload_texture_atlas(GSTexture* t) { void GSOsdManager::upload_texture_atlas(GSTexture* t)
if (!m_face) return; {
if (!m_face)
return;
if (m_char_info.size() > 96) // we only reserved space for this many glyphs if (m_char_info.size() > 96) // we only reserved space for this many glyphs
fprintf(stderr, "More than 96 glyphs needed for OSD"); fprintf(stderr, "More than 96 glyphs needed for OSD");
// This can be sped up a bit by only uploading new glyphs // This can be sped up a bit by only uploading new glyphs
int x = 0; int x = 0;
for(auto &pair : m_char_info) { for (auto& pair : m_char_info)
if(FT_Load_Char(m_face, pair.first, FT_LOAD_RENDER)) { {
if (FT_Load_Char(m_face, pair.first, FT_LOAD_RENDER))
{
fprintf(stderr, "failed to load char U%d\n", (int)pair.first); fprintf(stderr, "failed to load char U%d\n", (int)pair.first);
continue; continue;
} }
@ -127,7 +141,8 @@ void GSOsdManager::upload_texture_atlas(GSTexture* t) {
if (r.width()) if (r.width())
t->Update(r, m_face->glyph->bitmap.buffer, m_face->glyph->bitmap.pitch); t->Update(r, m_face->glyph->bitmap.buffer, m_face->glyph->bitmap.pitch);
if (r.width() > m_max_width) m_max_width = r.width(); if (r.width() > m_max_width)
m_max_width = r.width();
pair.second.tx = (float)x / m_atlas_w; pair.second.tx = (float)x / m_atlas_w;
pair.second.ty = (float)pair.second.bh / m_atlas_h; pair.second.ty = (float)pair.second.bh / m_atlas_h;
@ -142,35 +157,50 @@ void GSOsdManager::upload_texture_atlas(GSTexture* t) {
#if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4) #if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4)
/* This is dumb in that it doesn't check for malformed UTF8. This function /* This is dumb in that it doesn't check for malformed UTF8. This function
* is not expected to operate on user input, but only on compiled in strings */ * is not expected to operate on user input, but only on compiled in strings */
void dumb_utf8_to_utf32(const char *utf8, char32_t *utf32, unsigned size) { void dumb_utf8_to_utf32(const char* utf8, char32_t* utf32, unsigned size)
while(*utf8 && --size) { {
if((*utf8 & 0xF1) == 0xF0) { while (*utf8 && --size)
{
if ((*utf8 & 0xF1) == 0xF0)
{
*utf32++ = (utf8[0] & 0x07) << 18 | (utf8[1] & 0x3F) << 12 | (utf8[2] & 0x3F) << 6 | utf8[3] & 0x3F; *utf32++ = (utf8[0] & 0x07) << 18 | (utf8[1] & 0x3F) << 12 | (utf8[2] & 0x3F) << 6 | utf8[3] & 0x3F;
utf8 += 4; utf8 += 4;
} else if((*utf8 & 0xF0) == 0xE0) { }
else if ((*utf8 & 0xF0) == 0xE0)
{
*utf32++ = (utf8[0] & 0x0F) << 12 | (utf8[1] & 0x3F) << 6 | utf8[2] & 0x3F; *utf32++ = (utf8[0] & 0x0F) << 12 | (utf8[1] & 0x3F) << 6 | utf8[2] & 0x3F;
utf8 += 3; utf8 += 3;
} else if((*utf8 & 0xE0) == 0xC0) { }
else if ((*utf8 & 0xE0) == 0xC0)
{
*utf32++ = (utf8[0] & 0x1F) << 6 | utf8[1] & 0x3F; *utf32++ = (utf8[0] & 0x1F) << 6 | utf8[1] & 0x3F;
utf8 += 2; utf8 += 2;
} else if((*utf8 & 0x80) == 0x00) { }
else if ((*utf8 & 0x80) == 0x00)
{
*utf32++ = utf8[0] & 0x7F; *utf32++ = utf8[0] & 0x7F;
utf8 += 1; utf8 += 1;
} }
} }
if(size) *utf32 = *utf8; // Copy NUL char if (size)
*utf32 = *utf8; // Copy NUL char
} }
#endif #endif
void GSOsdManager::AddGlyph(char32_t codepoint) { void GSOsdManager::AddGlyph(char32_t codepoint)
if (!m_face) return; {
if(m_char_info.count(codepoint) == 0) { if (!m_face)
return;
if (m_char_info.count(codepoint) == 0)
{
m_texture_dirty = true; m_texture_dirty = true;
m_char_info[codepoint]; // add it m_char_info[codepoint]; // add it
if(FT_HAS_KERNING(m_face)) { if (FT_HAS_KERNING(m_face))
{
FT_UInt new_glyph = FT_Get_Char_Index(m_face, codepoint); FT_UInt new_glyph = FT_Get_Char_Index(m_face, codepoint);
for(auto pair : m_char_info) { for (auto pair : m_char_info)
{
FT_Vector delta; FT_Vector delta;
FT_UInt glyph_index = FT_Get_Char_Index(m_face, pair.first); FT_UInt glyph_index = FT_Get_Char_Index(m_face, pair.first);
@ -181,14 +211,16 @@ void GSOsdManager::AddGlyph(char32_t codepoint) {
} }
} }
void GSOsdManager::Log(const char *utf8) { void GSOsdManager::Log(const char* utf8)
{
if (!m_log_enabled) if (!m_log_enabled)
return; return;
#if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4) #if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4)
char32_t buffer[256]; char32_t buffer[256];
dumb_utf8_to_utf32(utf8, buffer, countof(buffer)); dumb_utf8_to_utf32(utf8, buffer, countof(buffer));
for(char32_t* c = buffer; *c; ++c) AddGlyph(*c); for (char32_t* c = buffer; *c; ++c)
AddGlyph(*c);
#else #else
#if _MSC_VER == 1900 #if _MSC_VER == 1900
std::wstring_convert<std::codecvt_utf8<unsigned int>, unsigned int> conv; std::wstring_convert<std::codecvt_utf8<unsigned int>, unsigned int> conv;
@ -196,24 +228,28 @@ void GSOsdManager::Log(const char *utf8) {
std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> conv; std::wstring_convert<std::codecvt_utf8<char32_t>, char32_t> conv;
#endif #endif
std::u32string buffer = conv.from_bytes(utf8); std::u32string buffer = conv.from_bytes(utf8);
for(auto const &c : buffer) AddGlyph(c); for (auto const& c : buffer)
AddGlyph(c);
#endif #endif
m_onscreen_messages++; m_onscreen_messages++;
m_log.push_back(log_info{buffer, std::chrono::system_clock::time_point()}); m_log.push_back(log_info{buffer, std::chrono::system_clock::time_point()});
} }
void GSOsdManager::Monitor(const char *key, const char *value) { void GSOsdManager::Monitor(const char* key, const char* value)
{
if (!m_monitor_enabled) if (!m_monitor_enabled)
return; return;
if(value && *value) { if (value && *value)
{
#if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4) #if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4)
char32_t buffer[256], vbuffer[256]; char32_t buffer[256], vbuffer[256];
dumb_utf8_to_utf32(key, buffer, countof(buffer)); dumb_utf8_to_utf32(key, buffer, countof(buffer));
dumb_utf8_to_utf32(value, vbuffer, countof(vbuffer)); dumb_utf8_to_utf32(value, vbuffer, countof(vbuffer));
for(char32_t* c = buffer; *c; ++c) AddGlyph(*c); for (char32_t* c = buffer; *c; ++c)
for(char32_t* c = vbuffer; *c; ++c) AddGlyph(*c); AddGlyph(*c);
for (char32_t* c = vbuffer; *c; ++c)
AddGlyph(*c);
#else #else
#if _MSC_VER == 1900 #if _MSC_VER == 1900
std::wstring_convert<std::codecvt_utf8<unsigned int>, unsigned int> conv; std::wstring_convert<std::codecvt_utf8<unsigned int>, unsigned int> conv;
@ -222,11 +258,15 @@ void GSOsdManager::Monitor(const char *key, const char *value) {
#endif #endif
std::u32string buffer = conv.from_bytes(key); std::u32string buffer = conv.from_bytes(key);
std::u32string vbuffer = conv.from_bytes(value); std::u32string vbuffer = conv.from_bytes(value);
for(auto const &c : buffer) AddGlyph(c); for (auto const& c : buffer)
for(auto const &c : vbuffer) AddGlyph(c); AddGlyph(c);
for (auto const& c : vbuffer)
AddGlyph(c);
#endif #endif
m_monitor[buffer] = vbuffer; m_monitor[buffer] = vbuffer;
} else { }
else
{
#if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4) #if __GNUC__ < 5 || (__GNUC__ == 5 && __GNUC_MINOR__ < 4)
char32_t buffer[256]; char32_t buffer[256];
dumb_utf8_to_utf32(key, buffer, countof(buffer)); dumb_utf8_to_utf32(key, buffer, countof(buffer));
@ -242,7 +282,8 @@ void GSOsdManager::Monitor(const char *key, const char *value) {
} }
} }
void GSOsdManager::RenderGlyph(GSVertexPT1* dst, const glyph_info g, float x, float y, uint32 color) { void GSOsdManager::RenderGlyph(GSVertexPT1* dst, const glyph_info g, float x, float y, uint32 color)
{
float x2 = x + g.bl * (2.0f / m_real_size.x); float x2 = x + g.bl * (2.0f / m_real_size.x);
float y2 = -y - g.bt * (2.0f / m_real_size.y); float y2 = -y - g.bt * (2.0f / m_real_size.y);
float w = g.bw * (2.0f / m_real_size.x); float w = g.bw * (2.0f / m_real_size.x);
@ -274,10 +315,13 @@ void GSOsdManager::RenderGlyph(GSVertexPT1* dst, const glyph_info g, float x, fl
++dst; ++dst;
} }
void GSOsdManager::RenderString(GSVertexPT1* dst, const std::u32string msg, float x, float y, uint32 color) { void GSOsdManager::RenderString(GSVertexPT1* dst, const std::u32string msg, float x, float y, uint32 color)
{
char32_t p = 0; char32_t p = 0;
for(const auto & c : msg) { for (const auto& c : msg)
if(p) { {
if (p)
{
x += m_kern_info[std::make_pair(p, c)] * (2.0f / m_real_size.x); x += m_kern_info[std::make_pair(p, c)] * (2.0f / m_real_size.x);
} }
@ -293,36 +337,47 @@ void GSOsdManager::RenderString(GSVertexPT1* dst, const std::u32string msg, floa
} }
} }
size_t GSOsdManager::Size() { size_t GSOsdManager::Size()
{
size_t sum = 0; size_t sum = 0;
if(m_log_enabled) { if (m_log_enabled)
{
float offset = 0; float offset = 0;
for(auto it = m_log.begin(); it != m_log.end(); ++it) { for (auto it = m_log.begin(); it != m_log.end(); ++it)
{
float y = 1 - ((m_size + 2) * (it - m_log.begin() + 1)) * (2.0f / m_real_size.y); float y = 1 - ((m_size + 2) * (it - m_log.begin() + 1)) * (2.0f / m_real_size.y);
if(y + offset < -1) break; if (y + offset < -1)
break;
std::chrono::duration<float> elapsed; std::chrono::duration<float> elapsed;
if(it->OnScreen.time_since_epoch().count() == 0) { if (it->OnScreen.time_since_epoch().count() == 0)
{
elapsed = std::chrono::seconds(0); elapsed = std::chrono::seconds(0);
} else { }
else
{
elapsed = std::chrono::system_clock::now() - it->OnScreen; elapsed = std::chrono::system_clock::now() - it->OnScreen;
if(elapsed > std::chrono::seconds(m_log_timeout) || m_onscreen_messages > m_max_onscreen_messages) { if (elapsed > std::chrono::seconds(m_log_timeout) || m_onscreen_messages > m_max_onscreen_messages)
{
continue; continue;
} }
} }
float ratio = (elapsed - std::chrono::seconds(m_log_timeout / 2)).count() / std::chrono::seconds(m_log_timeout / 2).count(); float ratio = (elapsed - std::chrono::seconds(m_log_timeout / 2)).count() / std::chrono::seconds(m_log_timeout / 2).count();
ratio = ratio > 1.0f ? 1.0f : ratio < 0.0f ? 0.0f : ratio; ratio = ratio > 1.0f ? 1.0f : ratio < 0.0f ? 0.0f :
ratio;
y += offset += ((m_size + 2) * (2.0f / m_real_size.y)) * ratio; y += offset += ((m_size + 2) * (2.0f / m_real_size.y)) * ratio;
sum += it->msg.size(); sum += it->msg.size();
} }
} }
if(m_monitor_enabled) { if (m_monitor_enabled)
for(const auto &pair : m_monitor) { {
for (const auto& pair : m_monitor)
{
sum += pair.first.size(); sum += pair.first.size();
sum += pair.second.size(); sum += pair.second.size();
} }
@ -331,12 +386,15 @@ size_t GSOsdManager::Size() {
return sum * 6; return sum * 6;
} }
float GSOsdManager::StringSize(const std::u32string msg) { float GSOsdManager::StringSize(const std::u32string msg)
{
char32_t p = 0; char32_t p = 0;
float x = 0.0; float x = 0.0;
for(auto c : msg) { for (auto c : msg)
if(p) { {
if (p)
{
x += m_kern_info[std::make_pair(p, c)] * (2.0f / m_real_size.x); x += m_kern_info[std::make_pair(p, c)] * (2.0f / m_real_size.x);
} }
@ -349,30 +407,36 @@ float GSOsdManager::StringSize(const std::u32string msg) {
return x; return x;
} }
size_t GSOsdManager::GeneratePrimitives(GSVertexPT1* dst, size_t count) { size_t GSOsdManager::GeneratePrimitives(GSVertexPT1* dst, size_t count)
{
size_t drawn = 0; size_t drawn = 0;
float opacity = m_opacity * 0.01f; float opacity = m_opacity * 0.01f;
if(m_log_enabled) { if (m_log_enabled)
{
float offset = 0; float offset = 0;
for(auto it = m_log.begin(); it != m_log.end();) { for (auto it = m_log.begin(); it != m_log.end();)
{
float x = -1 + 8 * (2.0f / m_real_size.x); float x = -1 + 8 * (2.0f / m_real_size.x);
float y = 1 - ((m_size + 2) * (it - m_log.begin() + 1)) * (2.0f / m_real_size.y); float y = 1 - ((m_size + 2) * (it - m_log.begin() + 1)) * (2.0f / m_real_size.y);
if(y + offset < -1) break; if (y + offset < -1)
break;
if (it->OnScreen.time_since_epoch().count() == 0) if (it->OnScreen.time_since_epoch().count() == 0)
it->OnScreen = std::chrono::system_clock::now(); it->OnScreen = std::chrono::system_clock::now();
std::chrono::duration<float> elapsed = std::chrono::system_clock::now() - it->OnScreen; std::chrono::duration<float> elapsed = std::chrono::system_clock::now() - it->OnScreen;
if(elapsed > std::chrono::seconds(m_log_timeout) || m_onscreen_messages > m_max_onscreen_messages) { if (elapsed > std::chrono::seconds(m_log_timeout) || m_onscreen_messages > m_max_onscreen_messages)
{
m_onscreen_messages--; m_onscreen_messages--;
it = m_log.erase(it); it = m_log.erase(it);
continue; continue;
} }
if(it->msg.size() * 6 > count - drawn) break; if (it->msg.size() * 6 > count - drawn)
break;
float ratio = (elapsed - std::chrono::seconds(m_log_timeout / 2)).count() / std::chrono::seconds(m_log_timeout / 2).count(); float ratio = (elapsed - std::chrono::seconds(m_log_timeout / 2)).count() / std::chrono::seconds(m_log_timeout / 2).count();
ratio = ratio > 1.0f ? 1.0f : ratio < 0.0f ? 0.0f : ratio; ratio = ratio > 1.0f ? 1.0f : ratio < 0.0f ? 0.0f : ratio;
@ -387,13 +451,15 @@ size_t GSOsdManager::GeneratePrimitives(GSVertexPT1* dst, size_t count) {
} }
} }
if(m_monitor_enabled) { if (m_monitor_enabled)
{
// pair.first is the key and second is the value and color // pair.first is the key and second is the value and color
// Since the monitor is right justified, but we render from left to right // Since the monitor is right justified, but we render from left to right
// we need to find the longest string // we need to find the longest string
float first_max = 0.0, second_max = 0.0; float first_max = 0.0, second_max = 0.0;
for(const auto &pair : m_monitor) { for (const auto& pair : m_monitor)
{
float first_len = StringSize(pair.first); float first_len = StringSize(pair.first);
float second_len = StringSize(pair.second); float second_len = StringSize(pair.second);
@ -402,8 +468,10 @@ size_t GSOsdManager::GeneratePrimitives(GSVertexPT1* dst, size_t count) {
} }
size_t line = 1; size_t line = 1;
for(const auto &pair : m_monitor) { for (const auto& pair : m_monitor)
if((pair.first.size() + pair.second.size()) * 6 > count - drawn) break; {
if ((pair.first.size() + pair.second.size()) * 6 > count - drawn)
break;
// Calculate where to start rendering from by taking the right most position 1.0 // Calculate where to start rendering from by taking the right most position 1.0
// and subtracting (going left) 8 scaled pixels for a margin, then subtracting // and subtracting (going left) 8 scaled pixels for a margin, then subtracting
@ -432,4 +500,3 @@ size_t GSOsdManager::GeneratePrimitives(GSVertexPT1* dst, size_t count) {
return drawn; return drawn;
} }

11
plugins/GSdx/Renderers/Common/GSOsdManager.h
View File

@ -28,8 +28,10 @@
#include <ft2build.h> #include <ft2build.h>
#include FT_FREETYPE_H #include FT_FREETYPE_H
class GSOsdManager { class GSOsdManager
struct glyph_info { {
struct glyph_info
{
int32 ax; // advance.x int32 ax; // advance.x
int32 ay; // advance.y int32 ay; // advance.y
@ -56,7 +58,8 @@ class GSOsdManager {
int32 m_max_width; int32 m_max_width;
int32 m_onscreen_messages; int32 m_onscreen_messages;
struct log_info { struct log_info
{
std::u32string msg; std::u32string msg;
std::chrono::system_clock::time_point OnScreen; std::chrono::system_clock::time_point OnScreen;
}; };
@ -77,7 +80,6 @@ class GSOsdManager {
int m_max_onscreen_messages; int m_max_onscreen_messages;
public: public:
GSOsdManager(); GSOsdManager();
~GSOsdManager(); ~GSOsdManager();
@ -97,6 +99,5 @@ class GSOsdManager {
size_t GeneratePrimitives(GSVertexPT1* dst, size_t count); size_t GeneratePrimitives(GSVertexPT1* dst, size_t count);
private: private:
std::vector<char> resource_data_buffer; std::vector<char> resource_data_buffer;
}; };

22
plugins/GSdx/Renderers/Common/GSRenderer.cpp
View File

@ -80,7 +80,8 @@ bool GSRenderer::CreateDevice(GSDevice* dev)
void GSRenderer::ResetDevice() void GSRenderer::ResetDevice()
{ {
if(m_dev) m_dev->Reset(1, 1); if (m_dev)
m_dev->Reset(1, 1);
} }
bool GSRenderer::Merge(int field) bool GSRenderer::Merge(int field)
@ -187,9 +188,12 @@ bool GSRenderer::Merge(int field)
} }
else else
{ {
if(en[0]) tex[0] = GetOutput(0, y_offset[0]); if (en[0])
if(en[1]) tex[1] = GetOutput(1, y_offset[1]); tex[0] = GetOutput(0, y_offset[0]);
if(feedback_merge) tex[2] = GetFeedbackOutput(); if (en[1])
tex[1] = GetOutput(1, y_offset[1]);
if (feedback_merge)
tex[2] = GetFeedbackOutput();
} }
GSVector4 src[2]; GSVector4 src[2];
@ -198,7 +202,8 @@ bool GSRenderer::Merge(int field)
for (int i = 0; i < 2; i++) for (int i = 0; i < 2; i++)
{ {
if(!en[i] || !tex[i]) continue; if (!en[i] || !tex[i])
continue;
GSVector4i r = fr[i]; GSVector4i r = fr[i];
GSVector4 scale = GSVector4(tex[i]->GetScale()).xyxy(); GSVector4 scale = GSVector4(tex[i]->GetScale()).xyxy();
@ -309,7 +314,8 @@ void GSRenderer::SetVSync(int vsync)
{ {
m_vsync = vsync; m_vsync = vsync;
if(m_dev) m_dev->SetVSync(m_vsync); if (m_dev)
m_dev->SetVSync(m_vsync);
} }
void GSRenderer::VSync(int field) void GSRenderer::VSync(int field)
@ -370,8 +376,7 @@ void GSRenderer::VSync(int field)
(int)m_perfmon.Get(GSPerfMon::Draw), (int)m_perfmon.Get(GSPerfMon::Draw),
m_perfmon.CPU(), m_perfmon.CPU(),
m_perfmon.Get(GSPerfMon::Swizzle) / 1024, m_perfmon.Get(GSPerfMon::Swizzle) / 1024,
m_perfmon.Get(GSPerfMon::Unswizzle) / 1024 m_perfmon.Get(GSPerfMon::Unswizzle) / 1024);
);
double fillrate = m_perfmon.Get(GSPerfMon::Fillrate); double fillrate = m_perfmon.Get(GSPerfMon::Fillrate);
@ -617,7 +622,6 @@ void GSRenderer::KeyEvent(GSKeyEventData* e)
printf("GSdx: Dithering is now %s.\n", dither_msg[m_dithering]); printf("GSdx: Dithering is now %s.\n", dither_msg[m_dithering]);
return; return;
} }
} }
} }

25
plugins/GSdx/Renderers/Common/GSTexture.h
View File

@ -35,15 +35,32 @@ protected:
bool m_sparse; bool m_sparse;
public: public:
struct GSMap {uint8* bits; int pitch;}; struct GSMap
{
uint8* bits;
int pitch;
};
enum {RenderTarget = 1, DepthStencil, Texture, Offscreen, Backbuffer, SparseRenderTarget, SparseDepthStencil}; enum
{
RenderTarget = 1,
DepthStencil,
Texture,
Offscreen,
Backbuffer,
SparseRenderTarget,
SparseDepthStencil
};
public: public:
GSTexture(); GSTexture();
virtual ~GSTexture() {} virtual ~GSTexture() {}
virtual operator bool() {ASSERT(0); return false;} virtual operator bool()
{
ASSERT(0);
return false;
}
virtual bool Update(const GSVector4i& r, const void* data, int pitch, int layer = 0) = 0; virtual bool Update(const GSVector4i& r, const void* data, int pitch, int layer = 0) = 0;
virtual bool Map(GSMap& m, const GSVector4i* r = NULL, int layer = 0) = 0; virtual bool Map(GSMap& m, const GSVector4i* r = NULL, int layer = 0) = 0;
@ -62,7 +79,7 @@ public:
int GetType() const { return m_type; } int GetType() const { return m_type; }
int GetFormat() const { return m_format; } int GetFormat() const { return m_format; }
virtual void CommitPages(const GSVector2i& region, bool commit) {}; virtual void CommitPages(const GSVector2i& region, bool commit) {}
void CommitRegion(const GSVector2i& region); void CommitRegion(const GSVector2i& region);
void Commit(); void Commit();
void Uncommit(); void Uncommit();

17
plugins/GSdx/Renderers/Common/GSVertex.h
View File

@ -50,11 +50,22 @@ struct alignas(32) GSVertex
GSVertex() = default; // Warning object is potentially used in hot path GSVertex() = default; // Warning object is potentially used in hot path
#if _M_SSE >= 0x500 #if _M_SSE >= 0x500
GSVertex(const GSVertex& v) {mx = v.mx;} GSVertex(const GSVertex& v)
{
mx = v.mx;
}
void operator=(const GSVertex& v) { mx = v.mx; } void operator=(const GSVertex& v) { mx = v.mx; }
#else #else
GSVertex(const GSVertex& v) {m[0] = v.m[0]; m[1] = v.m[1];} GSVertex(const GSVertex& v)
void operator = (const GSVertex& v) {m[0] = v.m[0]; m[1] = v.m[1];} {
m[0] = v.m[0];
m[1] = v.m[1];
}
void operator=(const GSVertex& v)
{
m[0] = v.m[0];
m[1] = v.m[1];
}
#endif #endif
}; };

3
plugins/GSdx/Renderers/Common/GSVertexList.h
View File

@ -21,7 +21,8 @@
#pragma once #pragma once
template <class Vertex> class GSVertexList template <class Vertex>
class GSVertexList
{ {
void* m_base; void* m_base;
Vertex* m_v[3]; Vertex* m_v[3];

33
plugins/GSdx/Renderers/Common/GSVertexTrace.cpp
View File

@ -68,7 +68,8 @@ void GSVertexTrace::Update(const void* vertex, const uint32* index, int v_count,
// Potential float overflow detected. Better uses the slower division instead // Potential float overflow detected. Better uses the slower division instead
// Note: If Q is too big, 1/Q will end up as 0. 1e30 is a random number // Note: If Q is too big, 1/Q will end up as 0. 1e30 is a random number
// that feel big enough. // that feel big enough.
if (!fst && !m_accurate_stq && m_min.t.z > 1e30) { if (!fst && !m_accurate_stq && m_min.t.z > 1e30)
{
fprintf(stderr, "Vertex Trace: float overflow detected ! min %e max %e\n", m_min.t.z, m_max.t.z); fprintf(stderr, "Vertex Trace: float overflow detected ! min %e max %e\n", m_min.t.z, m_max.t.z);
m_accurate_stq = true; m_accurate_stq = true;
(this->*m_fmm[m_accurate_stq][color][fst][tme][iip][primclass])(vertex, index, i_count); (this->*m_fmm[m_accurate_stq][color][fst][tme][iip][primclass])(vertex, index, i_count);
@ -79,7 +80,8 @@ void GSVertexTrace::Update(const void* vertex, const uint32* index, int v_count,
m_alpha.valid = false; m_alpha.valid = false;
// I'm not sure of the cost. In doubt let's do it only when depth is enabled // I'm not sure of the cost. In doubt let's do it only when depth is enabled
if(m_state->m_context->TEST.ZTE == 1 && m_state->m_context->TEST.ZTST > ZTST_ALWAYS) { if (m_state->m_context->TEST.ZTE == 1 && m_state->m_context->TEST.ZTST > ZTST_ALWAYS)
{
CorrectDepthTrace(vertex, v_count); CorrectDepthTrace(vertex, v_count);
} }
@ -104,7 +106,12 @@ void GSVertexTrace::Update(const void* vertex, const uint32* index, int v_count,
GSVector4::storel(&m_lod, m_max.t.uph(m_min.t).log2(3).neg() * (float)(1 << TEX1.L) + K); GSVector4::storel(&m_lod, m_max.t.uph(m_min.t).log2(3).neg() * (float)(1 << TEX1.L) + K);
if(m_lod.x > m_lod.y) {float tmp = m_lod.x; m_lod.x = m_lod.y; m_lod.y = tmp;} if (m_lod.x > m_lod.y)
{
float tmp = m_lod.x;
m_lod.x = m_lod.y;
m_lod.y = tmp;
}
} }
else else
{ {
@ -518,21 +525,29 @@ void GSVertexTrace::CorrectDepthTrace(const void* vertex, int count)
uint32 z = v[0].XYZ.Z; uint32 z = v[0].XYZ.Z;
// ought to check only 1/2 for sprite // ought to check only 1/2 for sprite
if (z & 1) { if (z & 1)
{
// Check that first bit is always 1 // Check that first bit is always 1
for (int i = 0; i < count; i++) { for (int i = 0; i < count; i++)
{
z &= v[i].XYZ.Z; z &= v[i].XYZ.Z;
} }
} else { }
else
{
// Check that first bit is always 0 // Check that first bit is always 0
for (int i = 0; i < count; i++) { for (int i = 0; i < count; i++)
{
z |= v[i].XYZ.Z; z |= v[i].XYZ.Z;
} }
} }
if (z == v[0].XYZ.Z) { if (z == v[0].XYZ.Z)
{
m_eq.z = 1; m_eq.z = 1;
} else { }
else
{
m_eq.z = 0; m_eq.z = 0;
} }
} }

12
plugins/GSdx/Renderers/Common/GSVertexTrace.h
View File

@ -34,8 +34,16 @@ class alignas(32) GSVertexTrace : public GSAlignedClass<32>
BiFiltering m_force_filter; BiFiltering m_force_filter;
public: public:
struct Vertex {GSVector4i c; GSVector4 p, t;}; struct Vertex
struct VertexAlpha {int min, max; bool valid;}; {
GSVector4i c;
GSVector4 p, t;
};
struct VertexAlpha
{
int min, max;
bool valid;
};
bool m_accurate_stq; bool m_accurate_stq;
protected: protected:

55
plugins/GSdx/Renderers/DX11/GSDevice11.cpp
View File

@ -160,8 +160,7 @@ bool GSDevice11::Create(const std::shared_ptr<GSWnd> &wnd)
const HRESULT result = D3D11CreateDevice( const HRESULT result = D3D11CreateDevice(
adapter, driver_type, nullptr, flags, adapter, driver_type, nullptr, flags,
supported_levels.data(), supported_levels.size(), supported_levels.data(), supported_levels.size(),
D3D11_SDK_VERSION, &m_dev, &level, &m_ctx D3D11_SDK_VERSION, &m_dev, &level, &m_ctx);
);
if (FAILED(result)) if (FAILED(result))
{ {
@ -189,8 +188,7 @@ bool GSDevice11::Create(const std::shared_ptr<GSWnd> &wnd)
const HRESULT result = m_factory->CreateSwapChainForHwnd( const HRESULT result = m_factory->CreateSwapChainForHwnd(
m_dev, reinterpret_cast<HWND>(m_wnd->GetHandle()), m_dev, reinterpret_cast<HWND>(m_wnd->GetHandle()),
&swapchain_description, nullptr, nullptr, &m_swapchain &swapchain_description, nullptr, nullptr, &m_swapchain);
);
if (FAILED(result)) if (FAILED(result))
{ {
@ -208,7 +206,8 @@ bool GSDevice11::Create(const std::shared_ptr<GSWnd> &wnd)
else else
m_d3d_texsize = D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION; m_d3d_texsize = D3D10_REQ_TEXTURE2D_U_OR_V_DIMENSION;
{ // HACK: check nVIDIA {
// HACK: check nVIDIA
// Note: It can cause issues on several games such as SOTC, Fatal Frame, plus it adds border offset. // Note: It can cause issues on several games such as SOTC, Fatal Frame, plus it adds border offset.
bool disable_safe_features = theApp.GetConfigB("UserHacks") && theApp.GetConfigB("UserHacks_Disable_Safe_Features"); bool disable_safe_features = theApp.GetConfigB("UserHacks") && theApp.GetConfigB("UserHacks_Disable_Safe_Features");
m_hack_topleft_offset = (m_upscale_multiplier != 1 && nvidia_vendor && !disable_safe_features) ? -0.01f : 0.0f; m_hack_topleft_offset = (m_upscale_multiplier != 1 && nvidia_vendor && !disable_safe_features) ? -0.01f : 0.0f;
@ -435,8 +434,7 @@ bool GSDevice11::Create(const std::shared_ptr<GSWnd> &wnd)
GSVector2i tex_font = m_osd.get_texture_font_size(); GSVector2i tex_font = m_osd.get_texture_font_size();
m_font = std::unique_ptr<GSTexture>( m_font = std::unique_ptr<GSTexture>(
CreateSurface(GSTexture::Texture, tex_font.x, tex_font.y, DXGI_FORMAT_R8_UNORM) CreateSurface(GSTexture::Texture, tex_font.x, tex_font.y, DXGI_FORMAT_R8_UNORM));
);
return true; return true;
} }
@ -553,13 +551,15 @@ void GSDevice11::DrawIndexedPrimitive(int offset, int count)
void GSDevice11::ClearRenderTarget(GSTexture* t, const GSVector4& c) void GSDevice11::ClearRenderTarget(GSTexture* t, const GSVector4& c)
{ {
if (!t) return; if (!t)
return;
m_ctx->ClearRenderTargetView(*(GSTexture11*)t, c.v); m_ctx->ClearRenderTargetView(*(GSTexture11*)t, c.v);
} }
void GSDevice11::ClearRenderTarget(GSTexture* t, uint32 c) void GSDevice11::ClearRenderTarget(GSTexture* t, uint32 c)
{ {
if (!t) return; if (!t)
return;
GSVector4 color = GSVector4::rgba32(c) * (1.0f / 255); GSVector4 color = GSVector4::rgba32(c) * (1.0f / 255);
m_ctx->ClearRenderTargetView(*(GSTexture11*)t, color.v); m_ctx->ClearRenderTargetView(*(GSTexture11*)t, color.v);
@ -567,13 +567,15 @@ void GSDevice11::ClearRenderTarget(GSTexture* t, uint32 c)
void GSDevice11::ClearDepth(GSTexture* t) void GSDevice11::ClearDepth(GSTexture* t)
{ {
if (!t) return; if (!t)
return;
m_ctx->ClearDepthStencilView(*(GSTexture11*)t, D3D11_CLEAR_DEPTH, 0.0f, 0); m_ctx->ClearDepthStencilView(*(GSTexture11*)t, D3D11_CLEAR_DEPTH, 0.0f, 0);
} }
void GSDevice11::ClearStencil(GSTexture* t, uint8 c) void GSDevice11::ClearStencil(GSTexture* t, uint8 c)
{ {
if (!t) return; if (!t)
return;
m_ctx->ClearDepthStencilView(*(GSTexture11*)t, D3D11_CLEAR_STENCIL, 0, c); m_ctx->ClearDepthStencilView(*(GSTexture11*)t, D3D11_CLEAR_STENCIL, 0, c);
} }
@ -758,8 +760,8 @@ void GSDevice11::StretchRect(GSTexture* sTex, const GSVector4& sRect, GSTexture*
return; return;
} }
bool draw_in_depth = (ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT32] || ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT24] || bool draw_in_depth = (ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT32] || ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT24]
ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT16] || ps == m_convert.ps[ShaderConvert_RGB5A1_TO_FLOAT16]); || ps == m_convert.ps[ShaderConvert_RGBA8_TO_FLOAT16] || ps == m_convert.ps[ShaderConvert_RGB5A1_TO_FLOAT16]);
BeginScene(); BeginScene();
@ -837,7 +839,8 @@ void GSDevice11::RenderOsd(GSTexture* dt)
OMSetBlendState(m_merge.bs, 0); OMSetBlendState(m_merge.bs, 0);
OMSetRenderTargets(dt, NULL); OMSetRenderTargets(dt, NULL);
if(m_osd.m_texture_dirty) { if (m_osd.m_texture_dirty)
{
m_osd.upload_texture_atlas(m_font.get()); m_osd.upload_texture_atlas(m_font.get());
} }
@ -911,7 +914,8 @@ void GSDevice11::InitExternalFX()
{ {
if (!ExShader_Compiled) if (!ExShader_Compiled)
{ {
try { try
{
std::string config_name(theApp.GetConfigS("shaderfx_conf")); std::string config_name(theApp.GetConfigS("shaderfx_conf"));
std::ifstream fconfig(config_name); std::ifstream fconfig(config_name);
std::stringstream shader; std::stringstream shader;
@ -935,7 +939,8 @@ void GSDevice11::InitExternalFX()
fprintf(stderr, "GSdx: External shader '%s' not loaded and will be disabled!\n", shader_name.c_str()); fprintf(stderr, "GSdx: External shader '%s' not loaded and will be disabled!\n", shader_name.c_str());
} }
} }
catch (GSDXRecoverableError) { catch (GSDXRecoverableError)
{
printf("GSdx: failed to compile external post-processing shader. \n"); printf("GSdx: failed to compile external post-processing shader. \n");
} }
ExShader_Compiled = true; ExShader_Compiled = true;
@ -968,13 +973,15 @@ void GSDevice11::InitFXAA()
{ {
if (!FXAA_Compiled) if (!FXAA_Compiled)
{ {
try { try
{
std::vector<char> shader; std::vector<char> shader;
theApp.LoadResource(IDR_FXAA_FX, shader); theApp.LoadResource(IDR_FXAA_FX, shader);
ShaderMacro sm(m_shader.model); ShaderMacro sm(m_shader.model);
CreateShader(shader, "fxaa.fx", nullptr, "ps_main", sm.GetPtr(), &m_fxaa.ps); CreateShader(shader, "fxaa.fx", nullptr, "ps_main", sm.GetPtr(), &m_fxaa.ps);
} }
catch (GSDXRecoverableError) { catch (GSDXRecoverableError)
{
printf("GSdx: failed to compile fxaa shader.\n"); printf("GSdx: failed to compile fxaa shader.\n");
} }
FXAA_Compiled = true; FXAA_Compiled = true;
@ -1102,7 +1109,8 @@ bool GSDevice11::IAMapVertexBuffer(void** vertex, size_t stride, size_t count)
hr = m_dev->CreateBuffer(&bd, NULL, &m_vb); hr = m_dev->CreateBuffer(&bd, NULL, &m_vb);
if(FAILED(hr)) return false; if (FAILED(hr))
return false;
} }
D3D11_MAP type = D3D11_MAP_WRITE_NO_OVERWRITE; D3D11_MAP type = D3D11_MAP_WRITE_NO_OVERWRITE;
@ -1178,7 +1186,8 @@ void GSDevice11::IASetIndexBuffer(const void* index, size_t count)
hr = m_dev->CreateBuffer(&bd, NULL, &m_ib); hr = m_dev->CreateBuffer(&bd, NULL, &m_ib);
if(FAILED(hr)) return; if (FAILED(hr))
return;
} }
D3D11_MAP type = D3D11_MAP_WRITE_NO_OVERWRITE; D3D11_MAP type = D3D11_MAP_WRITE_NO_OVERWRITE;
@ -1283,7 +1292,8 @@ void GSDevice11::PSSetShaderResource(int i, GSTexture* sr)
{ {
ID3D11ShaderResourceView* srv = NULL; ID3D11ShaderResourceView* srv = NULL;
if(sr) srv = *(GSTexture11*)sr; if (sr)
srv = *(GSTexture11*)sr;
PSSetShaderResourceView(i, srv, sr); PSSetShaderResourceView(i, srv, sr);
} }
@ -1494,8 +1504,7 @@ void GSDevice11::CompileShader(const std::vector<char>& source, const char* fn,
const HRESULT hr = D3DCompile( const HRESULT hr = D3DCompile(
source.data(), source.size(), fn, macro, source.data(), source.size(), fn, macro,
include, entry, shader_model.c_str(), include, entry, shader_model.c_str(),
flags, 0, shader, &error flags, 0, shader, &error);
);
if (error) if (error)
fprintf(stderr, "%s\n", (const char*)error->GetBufferPointer()); fprintf(stderr, "%s\n", (const char*)error->GetBufferPointer());

60
plugins/GSdx/Renderers/DX11/GSDevice11.h
View File

@ -89,8 +89,14 @@ public:
operator uint32() const { return key; } operator uint32() const { return key; }
VSSelector() : key(0) {} VSSelector()
VSSelector(uint32 k) : key(k) {} : key(0)
{
}
VSSelector(uint32 k)
: key(k)
{
}
}; };
struct alignas(32) PSConstantBuffer struct alignas(32) PSConstantBuffer
@ -192,8 +198,14 @@ public:
operator uint32() { return key; } operator uint32() { return key; }
GSSelector() : key(0) {} GSSelector()
GSSelector(uint32 k) : key(k) {} : key(0)
{
}
GSSelector(uint32 k)
: key(k)
{
}
}; };
struct PSSelector struct PSSelector
@ -260,7 +272,10 @@ public:
operator uint64() { return key; } operator uint64() { return key; }
PSSelector() : key(0) {} PSSelector()
: key(0)
{
}
}; };
struct PSSamplerSelector struct PSSamplerSelector
@ -279,7 +294,10 @@ public:
operator uint32() { return key & 0x7; } operator uint32() { return key & 0x7; }
PSSamplerSelector() : key(0) {} PSSamplerSelector()
: key(0)
{
}
}; };
struct OMDepthStencilSelector struct OMDepthStencilSelector
@ -300,7 +318,10 @@ public:
operator uint32() { return key & 0x3f; } operator uint32() { return key & 0x3f; }
OMDepthStencilSelector() : key(0) {} OMDepthStencilSelector()
: key(0)
{
}
}; };
struct OMBlendSelector struct OMBlendSelector
@ -331,7 +352,10 @@ public:
operator uint32() { return key & 0x1fff; } operator uint32() { return key & 0x1fff; }
OMBlendSelector() : key(0) {} OMBlendSelector()
: key(0)
{
}
}; };
#pragma pack(pop) #pragma pack(pop)
@ -341,13 +365,21 @@ public:
struct mcstr struct mcstr
{ {
const char *name, *def; const char *name, *def;
mcstr(const char* n, const char* d) : name(n), def(d) {} mcstr(const char* n, const char* d)
: name(n)
, def(d)
{
}
}; };
struct mstring struct mstring
{ {
std::string name, def; std::string name, def;
mstring(const char* n, std::string d) : name(n), def(d) {} mstring(const char* n, std::string d)
: name(n)
, def(d)
{
}
}; };
std::vector<mstring> mlist; std::vector<mstring> mlist;
@ -494,7 +526,12 @@ private:
std::unique_ptr<GSTexture> m_font; std::unique_ptr<GSTexture> m_font;
protected: protected:
struct {D3D_FEATURE_LEVEL level; std::string model, vs, gs, ps, cs;} m_shader; struct
{
D3D_FEATURE_LEVEL level;
std::string model, vs, gs, ps, cs;
} m_shader;
public: public:
GSDevice11(); GSDevice11();
virtual ~GSDevice11() {} virtual ~GSDevice11() {}
@ -568,4 +605,3 @@ public:
void CompileShader(const std::vector<char>& source, const char* fn, ID3DInclude* include, const char* entry, D3D_SHADER_MACRO* macro, ID3DBlob** shader, std::string shader_model); void CompileShader(const std::vector<char>& source, const char* fn, ID3DInclude* include, const char* entry, D3D_SHADER_MACRO* macro, ID3DBlob** shader, std::string shader_model);
}; };

12
plugins/GSdx/Renderers/DX11/GSRendererDX11.cpp
View File

@ -102,7 +102,8 @@ void GSRendererDX11::SetupIA(const float& sx, const float& sy)
for (unsigned int i = 0; i < m_vertex.next; i++) for (unsigned int i = 0; i < m_vertex.next; i++)
{ {
if (PRIM->TME && PRIM->FST) d[i].UV &= 0x3FEF3FEF; if (PRIM->TME && PRIM->FST)
d[i].UV &= 0x3FEF3FEF;
} }
} }
@ -401,7 +402,6 @@ void GSRendererDX11::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache:
// fprintf(stderr, "%d: Green channel (wrong mask) (fbmask %x)\n", s_n, m_context->FRAME.FBMSK >> 24); // fprintf(stderr, "%d: Green channel (wrong mask) (fbmask %x)\n", s_n, m_context->FRAME.FBMSK >> 24);
m_ps_sel.channel = ChannelFetch_GREEN; m_ps_sel.channel = ChannelFetch_GREEN;
} }
} }
else if (green) else if (green)
{ {
@ -491,7 +491,8 @@ void GSRendererDX11::EmulateBlending()
case ACC_BLEND_BASIC_D3D11: case ACC_BLEND_BASIC_D3D11:
sw_blending |= accumulation_blend || blend_non_recursive; sw_blending |= accumulation_blend || blend_non_recursive;
[[fallthrough]]; [[fallthrough]];
default: break; default:
break;
} }
// Color clip // Color clip
@ -532,7 +533,8 @@ void GSRendererDX11::EmulateBlending()
{ {
m_om_bsel.accu_blend = 1; m_om_bsel.accu_blend = 1;
if (ALPHA.A == 2) { if (ALPHA.A == 2)
{
// The blend unit does a reverse subtraction so it means // The blend unit does a reverse subtraction so it means
// the shader must output a positive value. // the shader must output a positive value.
// Replace 0 - Cs by Cs - 0 // Replace 0 - Cs by Cs - 0
@ -619,7 +621,6 @@ void GSRendererDX11::EmulateTextureSampler(const GSTextureCache::Source* tex)
GSVector4 half_offset = RealignTargetTextureCoordinate(tex); GSVector4 half_offset = RealignTargetTextureCoordinate(tex);
vs_cb.Texture_Scale_Offset.z = half_offset.x; vs_cb.Texture_Scale_Offset.z = half_offset.x;
vs_cb.Texture_Scale_Offset.w = half_offset.y; vs_cb.Texture_Scale_Offset.w = half_offset.y;
} }
else if (tex->m_target) else if (tex->m_target)
{ {
@ -684,7 +685,6 @@ void GSRendererDX11::EmulateTextureSampler(const GSTextureCache::Source* tex)
// Note 4 bits indexes are converted to 8 bits // Note 4 bits indexes are converted to 8 bits
m_ps_sel.fmt = 3 << 2; m_ps_sel.fmt = 3 << 2;
} }
else else
{ {

3
plugins/GSdx/Renderers/DX11/GSRendererDX11.h
View File

@ -27,7 +27,8 @@
class GSRendererDX11 final : public GSRendererHW class GSRendererDX11 final : public GSRendererHW
{ {
enum ACC_BLEND_D3D11 { enum ACC_BLEND_D3D11
{
ACC_BLEND_NONE_D3D11 = 0, ACC_BLEND_NONE_D3D11 = 0,
ACC_BLEND_BASIC_D3D11 = 1, ACC_BLEND_BASIC_D3D11 = 1,
ACC_BLEND_MEDIUM_D3D11 = 2, ACC_BLEND_MEDIUM_D3D11 = 2,

12
plugins/GSdx/Renderers/DX11/GSTexture11.cpp
View File

@ -36,10 +36,14 @@ GSTexture11::GSTexture11(ID3D11Texture2D* texture)
m_size.x = (int)m_desc.Width; m_size.x = (int)m_desc.Width;
m_size.y = (int)m_desc.Height; m_size.y = (int)m_desc.Height;
if(m_desc.BindFlags & D3D11_BIND_RENDER_TARGET) m_type = RenderTarget; if (m_desc.BindFlags & D3D11_BIND_RENDER_TARGET)
else if(m_desc.BindFlags & D3D11_BIND_DEPTH_STENCIL) m_type = DepthStencil; m_type = RenderTarget;
else if(m_desc.BindFlags & D3D11_BIND_SHADER_RESOURCE) m_type = Texture; else if (m_desc.BindFlags & D3D11_BIND_DEPTH_STENCIL)
else if(m_desc.Usage == D3D11_USAGE_STAGING) m_type = Offscreen; m_type = DepthStencil;
else if (m_desc.BindFlags & D3D11_BIND_SHADER_RESOURCE)
m_type = Texture;
else if (m_desc.Usage == D3D11_USAGE_STAGING)
m_type = Offscreen;
m_format = (int)m_desc.Format; m_format = (int)m_desc.Format;

6
plugins/GSdx/Renderers/DX11/GSTextureCache11.cpp
View File

@ -121,13 +121,15 @@ void GSTextureCache11::Read(Source* t, const GSVector4i& r)
const GIFRegTEX0& TEX0 = t->m_TEX0; const GIFRegTEX0& TEX0 = t->m_TEX0;
if (GSTexture* offscreen = m_renderer->m_dev->CreateOffscreen(r.width(), r.height())) { if (GSTexture* offscreen = m_renderer->m_dev->CreateOffscreen(r.width(), r.height()))
{
m_renderer->m_dev->CopyRect(t->m_texture, offscreen, r); m_renderer->m_dev->CopyRect(t->m_texture, offscreen, r);
GSTexture::GSMap m; GSTexture::GSMap m;
GSVector4i r_offscreen(0, 0, r.width(), r.height()); GSVector4i r_offscreen(0, 0, r.width(), r.height());
if (offscreen->Map(m, &r_offscreen)) { if (offscreen->Map(m, &r_offscreen))
{
GSOffset* off = m_renderer->m_mem.GetOffset(TEX0.TBP0, TEX0.TBW, TEX0.PSM); GSOffset* off = m_renderer->m_mem.GetOffset(TEX0.TBP0, TEX0.TBW, TEX0.PSM);
m_renderer->m_mem.WritePixel32(m.bits, m.pitch, off, r); m_renderer->m_mem.WritePixel32(m.bits, m.pitch, off, r);

12
plugins/GSdx/Renderers/DX11/GSTextureFX11.cpp
View File

@ -38,7 +38,8 @@ bool GSDevice11::CreateTextureFX()
hr = m_dev->CreateBuffer(&bd, NULL, &m_vs_cb); hr = m_dev->CreateBuffer(&bd, NULL, &m_vs_cb);
if(FAILED(hr)) return false; if (FAILED(hr))
return false;
memset(&bd, 0, sizeof(bd)); memset(&bd, 0, sizeof(bd));
@ -48,7 +49,8 @@ bool GSDevice11::CreateTextureFX()
hr = m_dev->CreateBuffer(&bd, NULL, &m_gs_cb); hr = m_dev->CreateBuffer(&bd, NULL, &m_gs_cb);
if (FAILED(hr)) return false; if (FAILED(hr))
return false;
memset(&bd, 0, sizeof(bd)); memset(&bd, 0, sizeof(bd));
@ -58,7 +60,8 @@ bool GSDevice11::CreateTextureFX()
hr = m_dev->CreateBuffer(&bd, NULL, &m_ps_cb); hr = m_dev->CreateBuffer(&bd, NULL, &m_ps_cb);
if(FAILED(hr)) return false; if (FAILED(hr))
return false;
D3D11_SAMPLER_DESC sd; D3D11_SAMPLER_DESC sd;
@ -75,7 +78,8 @@ bool GSDevice11::CreateTextureFX()
hr = m_dev->CreateSamplerState(&sd, &m_palette_ss); hr = m_dev->CreateSamplerState(&sd, &m_palette_ss);
if(FAILED(hr)) return false; if (FAILED(hr))
return false;
// create layout // create layout

327
plugins/GSdx/Renderers/HW/GSRendererHW.cpp
View File

@ -44,7 +44,8 @@ GSRendererHW::GSRendererHW(GSTextureCache* tc)
m_conservative_framebuffer = theApp.GetConfigB("conservative_framebuffer"); m_conservative_framebuffer = theApp.GetConfigB("conservative_framebuffer");
m_accurate_date = theApp.GetConfigB("accurate_date"); m_accurate_date = theApp.GetConfigB("accurate_date");
if (theApp.GetConfigB("UserHacks")) { if (theApp.GetConfigB("UserHacks"))
{
m_userhacks_enabled_gs_mem_clear = !theApp.GetConfigB("UserHacks_Disable_Safe_Features"); m_userhacks_enabled_gs_mem_clear = !theApp.GetConfigB("UserHacks_Disable_Safe_Features");
m_userHacks_enabled_unscale_ptln = !theApp.GetConfigB("UserHacks_Disable_Safe_Features"); m_userHacks_enabled_unscale_ptln = !theApp.GetConfigB("UserHacks_Disable_Safe_Features");
m_userhacks_align_sprite_X = theApp.GetConfigB("UserHacks_align_sprite_X"); m_userhacks_align_sprite_X = theApp.GetConfigB("UserHacks_align_sprite_X");
@ -55,7 +56,9 @@ GSRendererHW::GSRendererHW(GSTextureCache* tc)
m_userhacks_tcoffset_x = theApp.GetConfigI("UserHacks_TCOffsetX") / -1000.0f; m_userhacks_tcoffset_x = theApp.GetConfigI("UserHacks_TCOffsetX") / -1000.0f;
m_userhacks_tcoffset_y = theApp.GetConfigI("UserHacks_TCOffsetY") / -1000.0f; m_userhacks_tcoffset_y = theApp.GetConfigI("UserHacks_TCOffsetY") / -1000.0f;
m_userhacks_tcoffset = m_userhacks_tcoffset_x < 0.0f || m_userhacks_tcoffset_y < 0.0f; m_userhacks_tcoffset = m_userhacks_tcoffset_x < 0.0f || m_userhacks_tcoffset_y < 0.0f;
} else { }
else
{
m_userhacks_enabled_gs_mem_clear = true; m_userhacks_enabled_gs_mem_clear = true;
m_userHacks_enabled_unscale_ptln = true; m_userHacks_enabled_unscale_ptln = true;
m_userhacks_align_sprite_X = false; m_userhacks_align_sprite_X = false;
@ -65,12 +68,14 @@ GSRendererHW::GSRendererHW(GSTextureCache* tc)
m_userHacks_HPO = 0; m_userHacks_HPO = 0;
} }
if (!m_upscale_multiplier) { //Custom Resolution if (!m_upscale_multiplier) // Custom Resolution
{
m_custom_width = m_width = theApp.GetConfigI("resx"); m_custom_width = m_width = theApp.GetConfigI("resx");
m_custom_height = m_height = theApp.GetConfigI("resy"); m_custom_height = m_height = theApp.GetConfigI("resy");
} }
if (m_upscale_multiplier == 1) { // hacks are only needed for upscaling issues. if (m_upscale_multiplier == 1) // hacks are only needed for upscaling issues.
{
m_userhacks_round_sprite_offset = 0; m_userhacks_round_sprite_offset = 0;
m_userhacks_align_sprite_X = false; m_userhacks_align_sprite_X = false;
m_userHacks_merge_sprite = false; m_userHacks_merge_sprite = false;
@ -326,7 +331,8 @@ GSTexture* GSRendererHW::GetOutput(int i, int& y_offset)
t = rt->m_texture; t = rt->m_texture;
int delta = TEX0.TBP0 - rt->m_TEX0.TBP0; int delta = TEX0.TBP0 - rt->m_TEX0.TBP0;
if (delta > 0 && DISPFB.FBW != 0) { if (delta > 0 && DISPFB.FBW != 0)
{
int pages = delta >> 5u; int pages = delta >> 5u;
int y_pages = pages / DISPFB.FBW; int y_pages = pages / DISPFB.FBW;
y_offset = y_pages * GSLocalMemory::m_psm[DISPFB.PSM].pgs.y; y_offset = y_pages * GSLocalMemory::m_psm[DISPFB.PSM].pgs.y;
@ -398,7 +404,8 @@ void GSRendererHW::Lines2Sprites()
v0.XYZ.Z = v1.XYZ.Z; v0.XYZ.Z = v1.XYZ.Z;
v0.FOG = v1.FOG; v0.FOG = v1.FOG;
if (PRIM->TME && !PRIM->FST) { if (PRIM->TME && !PRIM->FST)
{
GSVector4 st0 = GSVector4::loadl(&v0.ST.u64); GSVector4 st0 = GSVector4::loadl(&v0.ST.u64);
GSVector4 st1 = GSVector4::loadl(&v1.ST.u64); GSVector4 st1 = GSVector4::loadl(&v1.ST.u64);
GSVector4 Q = GSVector4(v1.RGBAQ.Q, v1.RGBAQ.Q, v1.RGBAQ.Q, v1.RGBAQ.Q); GSVector4 Q = GSVector4(v1.RGBAQ.Q, v1.RGBAQ.Q, v1.RGBAQ.Q, v1.RGBAQ.Q);
@ -508,7 +515,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
read_ba = (tex_pos > 112 && tex_pos < 144); read_ba = (tex_pos > 112 && tex_pos < 144);
bool half_bottom = false; bool half_bottom = false;
switch (m_userhacks_ts_half_bottom) { switch (m_userhacks_ts_half_bottom)
{
case 0: case 0:
// Force Disabled. // Force Disabled.
// Force Disabled will help games such as Xenosaga. // Force Disabled will help games such as Xenosaga.
@ -541,10 +549,12 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
break; break;
} }
if (PRIM->FST) { if (PRIM->FST)
{
GL_INS("First vertex is P: %d => %d T: %d => %d", v[0].XYZ.X, v[1].XYZ.X, v[0].U, v[1].U); GL_INS("First vertex is P: %d => %d T: %d => %d", v[0].XYZ.X, v[1].XYZ.X, v[0].U, v[1].U);
for(size_t i = 0; i < count; i += 2) { for (size_t i = 0; i < count; i += 2)
{
if (write_ba) if (write_ba)
v[i].XYZ.X -= 128u; v[i].XYZ.X -= 128u;
else else
@ -555,7 +565,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
else else
v[i+1].U += 128u; v[i+1].U += 128u;
if (!half_bottom){ if (!half_bottom)
{
// Height is too big (2x). // Height is too big (2x).
int tex_offset = v[i].V & 0xF; int tex_offset = v[i].V & 0xF;
GSVector4i offset(o.OFY, tex_offset, o.OFY, tex_offset); GSVector4i offset(o.OFY, tex_offset, o.OFY, tex_offset);
@ -569,11 +580,14 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
v[i + 1].V = (uint16)tmp.w; v[i + 1].V = (uint16)tmp.w;
} }
} }
} else { }
else
{
const float offset_8pix = 8.0f / tw; const float offset_8pix = 8.0f / tw;
GL_INS("First vertex is P: %d => %d T: %f => %f (offset %f)", v[0].XYZ.X, v[1].XYZ.X, v[0].ST.S, v[1].ST.S, offset_8pix); GL_INS("First vertex is P: %d => %d T: %f => %f (offset %f)", v[0].XYZ.X, v[1].XYZ.X, v[0].ST.S, v[1].ST.S, offset_8pix);
for(size_t i = 0; i < count; i += 2) { for (size_t i = 0; i < count; i += 2)
{
if (write_ba) if (write_ba)
v[i].XYZ.X -= 128u; v[i].XYZ.X -= 128u;
else else
@ -584,7 +598,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
else else
v[i+1].ST.S += offset_8pix; v[i+1].ST.S += offset_8pix;
if (!half_bottom) { if (!half_bottom)
{
// Height is too big (2x). // Height is too big (2x).
GSVector4i offset(o.OFY, o.OFY); GSVector4i offset(o.OFY, o.OFY);
@ -606,7 +621,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
else else
m_vt.m_max.p.x += 8.0f; m_vt.m_max.p.x += 8.0f;
if (!half_bottom) { if (!half_bottom)
{
float delta_Y = m_vt.m_max.p.y - m_vt.m_min.p.y; float delta_Y = m_vt.m_max.p.y - m_vt.m_min.p.y;
m_vt.m_max.p.y -= delta_Y / 2.0f; m_vt.m_max.p.y -= delta_Y / 2.0f;
} }
@ -616,7 +632,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
else else
m_vt.m_max.t.x += 8.0f; m_vt.m_max.t.x += 8.0f;
if (!half_bottom) { if (!half_bottom)
{
float delta_T = m_vt.m_max.t.y - m_vt.m_min.t.y; float delta_T = m_vt.m_max.t.y - m_vt.m_min.t.y;
m_vt.m_max.t.y -= delta_T / 2.0f; m_vt.m_max.t.y -= delta_T / 2.0f;
} }
@ -624,7 +641,8 @@ void GSRendererHW::ConvertSpriteTextureShuffle(bool& write_ba, bool& read_ba)
GSVector4 GSRendererHW::RealignTargetTextureCoordinate(const GSTextureCache::Source* tex) GSVector4 GSRendererHW::RealignTargetTextureCoordinate(const GSTextureCache::Source* tex)
{ {
if (m_userHacks_HPO <= 1 || GetUpscaleMultiplier() == 1) return GSVector4(0.0f); if (m_userHacks_HPO <= 1 || GetUpscaleMultiplier() == 1)
return GSVector4(0.0f);
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
const GSVector2& scale = tex->m_texture->GetScale(); const GSVector2& scale = tex->m_texture->GetScale();
@ -635,27 +653,41 @@ GSVector4 GSRendererHW::RealignTargetTextureCoordinate(const GSTextureCache::Sou
// FIXME Let's start with something wrong same mess on X and Y // FIXME Let's start with something wrong same mess on X and Y
// FIXME Maybe it will be enough to check linear // FIXME Maybe it will be enough to check linear
if (PRIM->FST) { if (PRIM->FST)
{
if (m_userHacks_HPO == 3) { if (m_userHacks_HPO == 3)
if (!linear && t_position == 8) { {
half_offset.x = 8; if (!linear && t_position == 8)
half_offset.y = 8; {
} else if (linear && t_position == 16) {
half_offset.x = 16;
half_offset.y = 16;
} else if (m_vt.m_min.p.x == -0.5f) {
half_offset.x = 8; half_offset.x = 8;
half_offset.y = 8; half_offset.y = 8;
} }
} else { else if (linear && t_position == 16)
if (!linear && t_position == 8) { {
half_offset.x = 16;
half_offset.y = 16;
}
else if (m_vt.m_min.p.x == -0.5f)
{
half_offset.x = 8;
half_offset.y = 8;
}
}
else
{
if (!linear && t_position == 8)
{
half_offset.x = 8 - 8 / scale.x; half_offset.x = 8 - 8 / scale.x;
half_offset.y = 8 - 8 / scale.y; half_offset.y = 8 - 8 / scale.y;
} else if (linear && t_position == 16) { }
else if (linear && t_position == 16)
{
half_offset.x = 16 - 16 / scale.x; half_offset.x = 16 - 16 / scale.x;
half_offset.y = 16 - 16 / scale.y; half_offset.y = 16 - 16 / scale.y;
} else if (m_vt.m_min.p.x == -0.5f) { }
else if (m_vt.m_min.p.x == -0.5f)
{
half_offset.x = 8; half_offset.x = 8;
half_offset.y = 8; half_offset.y = 8;
} }
@ -663,8 +695,9 @@ GSVector4 GSRendererHW::RealignTargetTextureCoordinate(const GSTextureCache::Sou
GL_INS("offset detected %f,%f t_pos %d (linear %d, scale %f)", GL_INS("offset detected %f,%f t_pos %d (linear %d, scale %f)",
half_offset.x, half_offset.y, t_position, linear, scale.x); half_offset.x, half_offset.y, t_position, linear, scale.x);
}
} else if (m_vt.m_eq.q) { else if (m_vt.m_eq.q)
{
float tw = (float)(1 << m_context->TEX0.TW); float tw = (float)(1 << m_context->TEX0.TW);
float th = (float)(1 << m_context->TEX0.TH); float th = (float)(1 << m_context->TEX0.TH);
float q = v[0].RGBAQ.Q; float q = v[0].RGBAQ.Q;
@ -675,7 +708,6 @@ GSVector4 GSRendererHW::RealignTargetTextureCoordinate(const GSTextureCache::Sou
GL_INS("ST offset detected %f,%f (linear %d, scale %f)", GL_INS("ST offset detected %f,%f (linear %d, scale %f)",
half_offset.x, half_offset.y, linear, scale.x); half_offset.x, half_offset.y, linear, scale.x);
} }
return half_offset; return half_offset;
@ -692,8 +724,10 @@ GSVector4i GSRendererHW::ComputeBoundingBox(const GSVector2& rtscale, const GSVe
void GSRendererHW::MergeSprite(GSTextureCache::Source* tex) void GSRendererHW::MergeSprite(GSTextureCache::Source* tex)
{ {
// Upscaling hack to avoid various line/grid issues // Upscaling hack to avoid various line/grid issues
if (m_userHacks_merge_sprite && tex && tex->m_target && (m_vt.m_primclass == GS_SPRITE_CLASS)) { if (m_userHacks_merge_sprite && tex && tex->m_target && (m_vt.m_primclass == GS_SPRITE_CLASS))
if (PRIM->FST && GSLocalMemory::m_psm[tex->m_TEX0.PSM].fmt < 2 && ((m_vt.m_eq.value & 0xCFFFF) == 0xCFFFF)) { {
if (PRIM->FST && GSLocalMemory::m_psm[tex->m_TEX0.PSM].fmt < 2 && ((m_vt.m_eq.value & 0xCFFFF) == 0xCFFFF))
{
// Ideally the hack ought to be enabled in a true paving mode only. I don't know how to do it accurately // Ideally the hack ought to be enabled in a true paving mode only. I don't know how to do it accurately
// neither in a fast way. So instead let's just take the hypothesis that all sprites must have the same // neither in a fast way. So instead let's just take the hypothesis that all sprites must have the same
@ -704,10 +738,12 @@ void GSRendererHW::MergeSprite(GSTextureCache::Source* tex)
// SSE optimization: shuffle m[1] to have (4*32 bits) X, Y, U, V // SSE optimization: shuffle m[1] to have (4*32 bits) X, Y, U, V
int first_dpX = v[1].XYZ.X - v[0].XYZ.X; int first_dpX = v[1].XYZ.X - v[0].XYZ.X;
int first_dpU = v[1].U - v[0].U; int first_dpU = v[1].U - v[0].U;
for (size_t i = 0; i < m_vertex.next; i += 2) { for (size_t i = 0; i < m_vertex.next; i += 2)
{
int dpX = v[i + 1].XYZ.X - v[i].XYZ.X; int dpX = v[i + 1].XYZ.X - v[i].XYZ.X;
int dpU = v[i + 1].U - v[i].U; int dpU = v[i + 1].U - v[i].U;
if (dpX != first_dpX || dpU != first_dpU) { if (dpX != first_dpX || dpU != first_dpU)
{
is_paving = false; is_paving = false;
break; break;
} }
@ -720,7 +756,8 @@ void GSRendererHW::MergeSprite(GSTextureCache::Source* tex)
GL_INS("PP SAMPLER: Dp %f %f Dt %f %f. Is blit %d, is paving %d, count %d", delta_p.x, delta_p.y, delta_t.x, delta_t.y, is_blit, is_paving, m_vertex.tail); GL_INS("PP SAMPLER: Dp %f %f Dt %f %f. Is blit %d, is paving %d, count %d", delta_p.x, delta_p.y, delta_t.x, delta_t.y, is_blit, is_paving, m_vertex.tail);
#endif #endif
if (is_paving) { if (is_paving)
{
// Replace all sprite with a single fullscreen sprite. // Replace all sprite with a single fullscreen sprite.
GSVertex* s = &m_vertex.buff[0]; GSVertex* s = &m_vertex.buff[0];
@ -752,7 +789,8 @@ void GSRendererHW::InvalidateLocalMem(const GIFRegBITBLTBUF& BITBLTBUF, const GS
{ {
// printf("[%d] InvalidateLocalMem %d,%d - %d,%d %05x (%d)\n", (int)m_perfmon.GetFrame(), r.left, r.top, r.right, r.bottom, (int)BITBLTBUF.SBP, (int)BITBLTBUF.SPSM); // printf("[%d] InvalidateLocalMem %d,%d - %d,%d %05x (%d)\n", (int)m_perfmon.GetFrame(), r.left, r.top, r.right, r.bottom, (int)BITBLTBUF.SBP, (int)BITBLTBUF.SPSM);
if(clut) return; // FIXME if (clut)
return; // FIXME
m_tc->InvalidateLocalMem(m_mem.GetOffset(BITBLTBUF.SBP, BITBLTBUF.SBW, BITBLTBUF.SPSM), r); m_tc->InvalidateLocalMem(m_mem.GetOffset(BITBLTBUF.SBP, BITBLTBUF.SBW, BITBLTBUF.SPSM), r);
} }
@ -1035,7 +1073,8 @@ void GSRendererHW::RoundSpriteOffset()
size_t count = m_vertex.next; size_t count = m_vertex.next;
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
for(size_t i = 0; i < count; i += 2) { for (size_t i = 0; i < count; i += 2)
{
// Performance note: if it had any impact on perf, someone would port it to SSE (AKA GSVector) // Performance note: if it had any impact on perf, someone would port it to SSE (AKA GSVector)
// Compute the coordinate of first and last texels (in native with a linear filtering) // Compute the coordinate of first and last texels (in native with a linear filtering)
@ -1048,7 +1087,8 @@ void GSRendererHW::RoundSpriteOffset()
uint16 tx0 = Interpolate_UV(ax0, v[i].U, v[i + 1].U); uint16 tx0 = Interpolate_UV(ax0, v[i].U, v[i + 1].U);
uint16 tx1 = Interpolate_UV(ax1, v[i].U, v[i + 1].U); uint16 tx1 = Interpolate_UV(ax1, v[i].U, v[i + 1].U);
#ifdef DEBUG_U #ifdef DEBUG_U
if (debug) { if (debug)
{
fprintf(stderr, "u0:%d and u1:%d\n", v[i].U, v[i + 1].U); fprintf(stderr, "u0:%d and u1:%d\n", v[i].U, v[i + 1].U);
fprintf(stderr, "a0:%f and a1:%f\n", ax0, ax1); fprintf(stderr, "a0:%f and a1:%f\n", ax0, ax1);
fprintf(stderr, "t0:%d and t1:%d\n", tx0, tx1); fprintf(stderr, "t0:%d and t1:%d\n", tx0, tx1);
@ -1064,7 +1104,8 @@ void GSRendererHW::RoundSpriteOffset()
uint16 ty0 = Interpolate_UV(ay0, v[i].V, v[i + 1].V); uint16 ty0 = Interpolate_UV(ay0, v[i].V, v[i + 1].V);
uint16 ty1 = Interpolate_UV(ay1, v[i].V, v[i + 1].V); uint16 ty1 = Interpolate_UV(ay1, v[i].V, v[i + 1].V);
#ifdef DEBUG_V #ifdef DEBUG_V
if (debug) { if (debug)
{
fprintf(stderr, "v0:%d and v1:%d\n", v[i].V, v[i + 1].V); fprintf(stderr, "v0:%d and v1:%d\n", v[i].V, v[i + 1].V);
fprintf(stderr, "a0:%f and a1:%f\n", ay0, ay1); fprintf(stderr, "a0:%f and a1:%f\n", ay0, ay1);
fprintf(stderr, "t0:%d and t1:%d\n", ty0, ty1); fprintf(stderr, "t0:%d and t1:%d\n", ty0, ty1);
@ -1088,33 +1129,47 @@ void GSRendererHW::RoundSpriteOffset()
// of interpolation migth trigger a discard (with alpha testing) // of interpolation migth trigger a discard (with alpha testing)
// Let's use something simple that correct really bad case (for a couple of 2D games). // Let's use something simple that correct really bad case (for a couple of 2D games).
// I hope it won't create too much glitches. // I hope it won't create too much glitches.
if (linear) { if (linear)
{
int Lu = v[i + 1].U - v[i].U; int Lu = v[i + 1].U - v[i].U;
// Note 32 is based on taisho-mononoke // Note 32 is based on taisho-mononoke
if ((Lu > 0) && (Lu <= (Lx+32))) { if ((Lu > 0) && (Lu <= (Lx + 32)))
{
v[i + 1].U -= 8; v[i + 1].U -= 8;
} }
} else { }
if (tx0 <= tx1) { else
{
if (tx0 <= tx1)
{
v[i].U = tx0; v[i].U = tx0;
v[i + 1].U = tx1 + 16; v[i + 1].U = tx1 + 16;
} else { }
else
{
v[i].U = tx0 + 15; v[i].U = tx0 + 15;
v[i + 1].U = tx1; v[i + 1].U = tx1;
} }
} }
#endif #endif
#if 1 #if 1
if (linear) { if (linear)
{
int Lv = v[i + 1].V - v[i].V; int Lv = v[i + 1].V - v[i].V;
if ((Lv > 0) && (Lv <= (Ly+32))) { if ((Lv > 0) && (Lv <= (Ly + 32)))
{
v[i + 1].V -= 8; v[i + 1].V -= 8;
} }
} else { }
if (ty0 <= ty1) { else
{
if (ty0 <= ty1)
{
v[i].V = ty0; v[i].V = ty0;
v[i + 1].V = ty1 + 16; v[i + 1].V = ty1 + 16;
} else { }
else
{
v[i].V = ty0 + 15; v[i].V = ty0 + 15;
v[i + 1].V = ty1; v[i + 1].V = ty1;
} }
@ -1129,13 +1184,13 @@ void GSRendererHW::RoundSpriteOffset()
if (debug) if (debug)
fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].V, v[i + 1].V); fprintf(stderr, "GREP_AFTER %d => %d\n\n", v[i].V, v[i + 1].V);
#endif #endif
} }
} }
void GSRendererHW::Draw() void GSRendererHW::Draw()
{ {
if(m_dev->IsLost() || IsBadFrame()) { if (m_dev->IsLost() || IsBadFrame())
{
GL_INS("Warning skipping a draw call (%d)", s_n); GL_INS("Warning skipping a draw call (%d)", s_n);
return; return;
} }
@ -1187,22 +1242,31 @@ void GSRendererHW::Draw()
const GSVector4 delta_p = m_vt.m_max.p - m_vt.m_min.p; const GSVector4 delta_p = m_vt.m_max.p - m_vt.m_min.p;
bool single_page = (delta_p.x <= 64.0f) && (delta_p.y <= 64.0f); bool single_page = (delta_p.x <= 64.0f) && (delta_p.y <= 64.0f);
if (m_channel_shuffle) { if (m_channel_shuffle)
{
m_channel_shuffle = draw_sprite_tex && (m_context->TEX0.PSM == PSM_PSMT8) && single_page; m_channel_shuffle = draw_sprite_tex && (m_context->TEX0.PSM == PSM_PSMT8) && single_page;
if (m_channel_shuffle) { if (m_channel_shuffle)
{
GL_CACHE("Channel shuffle effect detected SKIP"); GL_CACHE("Channel shuffle effect detected SKIP");
return; return;
} }
} else if (draw_sprite_tex && m_context->FRAME.Block() == m_context->TEX0.TBP0) { }
else if (draw_sprite_tex && m_context->FRAME.Block() == m_context->TEX0.TBP0)
{
// Special post-processing effect // Special post-processing effect
if ((m_context->TEX0.PSM == PSM_PSMT8) && single_page) { if ((m_context->TEX0.PSM == PSM_PSMT8) && single_page)
{
GL_INS("Channel shuffle effect detected"); GL_INS("Channel shuffle effect detected");
m_channel_shuffle = true; m_channel_shuffle = true;
} else { }
else
{
GL_DBG("Special post-processing effect not supported"); GL_DBG("Special post-processing effect not supported");
m_channel_shuffle = false; m_channel_shuffle = false;
} }
} else { }
else
{
m_channel_shuffle = false; m_channel_shuffle = false;
} }
@ -1214,7 +1278,8 @@ void GSRendererHW::Draw()
GSTextureCache::Target* rt = NULL; GSTextureCache::Target* rt = NULL;
GSTexture* rt_tex = NULL; GSTexture* rt_tex = NULL;
if (!no_rt) { if (!no_rt)
{
rt = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::RenderTarget, true, fm); rt = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::RenderTarget, true, fm);
rt_tex = rt->m_texture; rt_tex = rt->m_texture;
} }
@ -1225,7 +1290,8 @@ void GSRendererHW::Draw()
GSTextureCache::Target* ds = NULL; GSTextureCache::Target* ds = NULL;
GSTexture* ds_tex = NULL; GSTexture* ds_tex = NULL;
if (!no_ds) { if (!no_ds)
{
ds = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::DepthStencil, context->DepthWrite()); ds = m_tc->LookupTarget(TEX0, m_width, m_height, GSTextureCache::DepthStencil, context->DepthWrite());
ds_tex = ds->m_texture; ds_tex = ds->m_texture;
} }
@ -1240,33 +1306,42 @@ void GSRendererHW::Draw()
m_lod = GSVector2i(0, 0); m_lod = GSVector2i(0, 0);
// Code from the SW renderer // Code from the SW renderer
if (IsMipMapActive()) { if (IsMipMapActive())
{
int interpolation = (context->TEX1.MMIN & 1) + 1; // 1: round, 2: tri int interpolation = (context->TEX1.MMIN & 1) + 1; // 1: round, 2: tri
int k = (m_context->TEX1.K + 8) >> 4; int k = (m_context->TEX1.K + 8) >> 4;
int lcm = m_context->TEX1.LCM; int lcm = m_context->TEX1.LCM;
if ((int)m_vt.m_lod.x >= mxl) { if ((int)m_vt.m_lod.x >= mxl)
{
k = mxl; // set lod to max level k = mxl; // set lod to max level
lcm = 1; // constant lod lcm = 1; // constant lod
} }
if (PRIM->FST) { if (PRIM->FST)
{
ASSERT(lcm == 1); ASSERT(lcm == 1);
ASSERT(((m_vt.m_min.t.uph(m_vt.m_max.t) == GSVector4::zero()).mask() & 3) == 3); // ratchet and clank (menu) ASSERT(((m_vt.m_min.t.uph(m_vt.m_max.t) == GSVector4::zero()).mask() & 3) == 3); // ratchet and clank (menu)
lcm = 1; lcm = 1;
} }
if (lcm == 1) { if (lcm == 1)
{
m_lod.x = std::max<int>(k, 0); m_lod.x = std::max<int>(k, 0);
m_lod.y = m_lod.x; m_lod.y = m_lod.x;
} else { }
else
{
// Not constant but who care ! // Not constant but who care !
if (interpolation == 2) { if (interpolation == 2)
{
// Mipmap Linear. Both layers are sampled, only take the big one // Mipmap Linear. Both layers are sampled, only take the big one
m_lod.x = std::max<int>((int)floor(m_vt.m_lod.x), 0); m_lod.x = std::max<int>((int)floor(m_vt.m_lod.x), 0);
} else { }
else
{
// On GS lod is a fixed float number 7:4 (4 bit for the frac part) // On GS lod is a fixed float number 7:4 (4 bit for the frac part)
#if 0 #if 0
m_lod.x = std::max<int>((int)round(m_vt.m_lod.x + 0.0625), 0); m_lod.x = std::max<int>((int)round(m_vt.m_lod.x + 0.0625), 0);
@ -1294,13 +1369,16 @@ void GSRendererHW::Draw()
MIP_CLAMP.MAXU >>= m_lod.x; MIP_CLAMP.MAXU >>= m_lod.x;
MIP_CLAMP.MAXV >>= m_lod.x; MIP_CLAMP.MAXV >>= m_lod.x;
for (int i = 0; i < m_lod.x; i++) { for (int i = 0; i < m_lod.x; i++)
{
m_vt.m_min.t *= 0.5f; m_vt.m_min.t *= 0.5f;
m_vt.m_max.t *= 0.5f; m_vt.m_max.t *= 0.5f;
} }
GL_CACHE("Mipmap LOD %d %d (%f %f) new size %dx%d (K %d L %u)", m_lod.x, m_lod.y, m_vt.m_lod.x, m_vt.m_lod.y, 1 << TEX0.TW, 1 << TEX0.TH, m_context->TEX1.K, m_context->TEX1.L); GL_CACHE("Mipmap LOD %d %d (%f %f) new size %dx%d (K %d L %u)", m_lod.x, m_lod.y, m_vt.m_lod.x, m_vt.m_lod.y, 1 << TEX0.TW, 1 << TEX0.TH, m_context->TEX1.K, m_context->TEX1.L);
} else { }
else
{
TEX0 = GetTex0Layer(0); TEX0 = GetTex0Layer(0);
} }
@ -1313,12 +1391,14 @@ void GSRendererHW::Draw()
m_src = tex_psm.depth ? m_tc->LookupDepthSource(TEX0, env.TEXA, r) : m_tc->LookupSource(TEX0, env.TEXA, r); m_src = tex_psm.depth ? m_tc->LookupDepthSource(TEX0, env.TEXA, r) : m_tc->LookupSource(TEX0, env.TEXA, r);
// Round 2 // Round 2
if (IsMipMapActive() && m_mipmap == 2 && !tex_psm.depth) { if (IsMipMapActive() && m_mipmap == 2 && !tex_psm.depth)
{
// Upload remaining texture layers // Upload remaining texture layers
GSVector4 tmin = m_vt.m_min.t; GSVector4 tmin = m_vt.m_min.t;
GSVector4 tmax = m_vt.m_max.t; GSVector4 tmax = m_vt.m_max.t;
for (int layer = m_lod.x + 1; layer <= m_lod.y; layer++) { for (int layer = m_lod.x + 1; layer <= m_lod.y; layer++)
{
const GIFRegTEX0& MIP_TEX0 = GetTex0Layer(layer); const GIFRegTEX0& MIP_TEX0 = GetTex0Layer(layer);
m_context->offset.tex = m_mem.GetOffset(MIP_TEX0.TBP0, MIP_TEX0.TBW, MIP_TEX0.PSM); m_context->offset.tex = m_mem.GetOffset(MIP_TEX0.TBP0, MIP_TEX0.TBW, MIP_TEX0.PSM);
@ -1348,7 +1428,8 @@ void GSRendererHW::Draw()
&& draw_sprite_tex && m_src->m_32_bits_fmt; && draw_sprite_tex && m_src->m_32_bits_fmt;
// Okami mustn't call this code // Okami mustn't call this code
if (m_texture_shuffle && m_vertex.next < 3 && PRIM->FST && (m_context->FRAME.FBMSK == 0)) { if (m_texture_shuffle && m_vertex.next < 3 && PRIM->FST && (m_context->FRAME.FBMSK == 0))
{
// Avious dubious call to m_texture_shuffle on 16 bits games // Avious dubious call to m_texture_shuffle on 16 bits games
// The pattern is severals column of 8 pixels. A single sprite // The pattern is severals column of 8 pixels. A single sprite
// smell fishy but a big sprite is wrong. // smell fishy but a big sprite is wrong.
@ -1373,14 +1454,18 @@ void GSRendererHW::Draw()
// Texture shuffle is not yet supported with strange clamp mode // Texture shuffle is not yet supported with strange clamp mode
ASSERT(!m_texture_shuffle || (context->CLAMP.WMS < 3 && context->CLAMP.WMT < 3)); ASSERT(!m_texture_shuffle || (context->CLAMP.WMS < 3 && context->CLAMP.WMT < 3));
if (m_src->m_target && m_context->TEX0.PSM == PSM_PSMT8 && single_page && draw_sprite_tex) { if (m_src->m_target && m_context->TEX0.PSM == PSM_PSMT8 && single_page && draw_sprite_tex)
{
GL_INS("Channel shuffle effect detected (2nd shot)"); GL_INS("Channel shuffle effect detected (2nd shot)");
m_channel_shuffle = true; m_channel_shuffle = true;
} else { }
else
{
m_channel_shuffle = false; m_channel_shuffle = false;
} }
} }
if (rt) { if (rt)
{
// Be sure texture shuffle detection is properly propagated // Be sure texture shuffle detection is properly propagated
// Otherwise set or clear the flag (Code in texture cache only set the flag) // Otherwise set or clear the flag (Code in texture cache only set the flag)
// Note: it is important to clear the flag when RT is used as a real 16 bits target. // Note: it is important to clear the flag when RT is used as a real 16 bits target.
@ -1393,7 +1478,8 @@ void GSRendererHW::Draw()
std::string s; std::string s;
if (s_n >= s_saven) { if (s_n >= s_saven)
{
// Dump Register state // Dump Register state
s = format("%05d_context.txt", s_n); s = format("%05d_context.txt", s_n);
@ -1434,7 +1520,6 @@ void GSRendererHW::Draw()
if (ds_tex) if (ds_tex)
ds_tex->Save(m_dump_root + s); ds_tex->Save(m_dump_root + s);
} }
} }
// The rectangle of the draw // The rectangle of the draw
@ -1446,12 +1531,14 @@ void GSRendererHW::Draw()
return; return;
} }
if (!OI_BlitFMV(rt, m_src, m_r)) { if (!OI_BlitFMV(rt, m_src, m_r))
{
GL_INS("Warning skipping a draw call (%d)", s_n); GL_INS("Warning skipping a draw call (%d)", s_n);
return; return;
} }
if (m_userhacks_enabled_gs_mem_clear) { if (m_userhacks_enabled_gs_mem_clear)
{
// Constant Direct Write without texture/test/blending (aka a GS mem clear) // Constant Direct Write without texture/test/blending (aka a GS mem clear)
if ((m_vt.m_primclass == GS_SPRITE_CLASS) && !PRIM->TME // Direct write if ((m_vt.m_primclass == GS_SPRITE_CLASS) && !PRIM->TME // Direct write
&& (!PRIM->ABE || m_context->ALPHA.IsOpaque()) // No transparency && (!PRIM->ABE || m_context->ALPHA.IsOpaque()) // No transparency
@ -1470,23 +1557,27 @@ void GSRendererHW::Draw()
// A couple of hack to avoid upscaling issue. So far it seems to impacts mostly sprite // A couple of hack to avoid upscaling issue. So far it seems to impacts mostly sprite
// Note: first hack corrects both position and texture coordinate // Note: first hack corrects both position and texture coordinate
// Note: second hack corrects only the texture coordinate // Note: second hack corrects only the texture coordinate
if ((m_upscale_multiplier > 1) && (m_vt.m_primclass == GS_SPRITE_CLASS)) { if ((m_upscale_multiplier > 1) && (m_vt.m_primclass == GS_SPRITE_CLASS))
{
size_t count = m_vertex.next; size_t count = m_vertex.next;
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
// Hack to avoid vertical black line in various games (ace combat/tekken) // Hack to avoid vertical black line in various games (ace combat/tekken)
if (m_userhacks_align_sprite_X) { if (m_userhacks_align_sprite_X)
{
// Note for performance reason I do the check only once on the first // Note for performance reason I do the check only once on the first
// primitive // primitive
int win_position = v[1].XYZ.X - context->XYOFFSET.OFX; int win_position = v[1].XYZ.X - context->XYOFFSET.OFX;
const bool unaligned_position = ((win_position & 0xF) == 8); const bool unaligned_position = ((win_position & 0xF) == 8);
const bool unaligned_texture = ((v[1].U & 0xF) == 0) && PRIM->FST; // I'm not sure this check is useful const bool unaligned_texture = ((v[1].U & 0xF) == 0) && PRIM->FST; // I'm not sure this check is useful
const bool hole_in_vertex = (count < 4) || (v[1].XYZ.X != v[2].XYZ.X); const bool hole_in_vertex = (count < 4) || (v[1].XYZ.X != v[2].XYZ.X);
if (hole_in_vertex && unaligned_position && (unaligned_texture || !PRIM->FST)) { if (hole_in_vertex && unaligned_position && (unaligned_texture || !PRIM->FST))
{
// Normaly vertex are aligned on full pixels and texture in half // Normaly vertex are aligned on full pixels and texture in half
// pixels. Let's extend the coverage of an half-pixel to avoid // pixels. Let's extend the coverage of an half-pixel to avoid
// hole after upscaling // hole after upscaling
for(size_t i = 0; i < count; i += 2) { for (size_t i = 0; i < count; i += 2)
{
v[i + 1].XYZ.X += 8; v[i + 1].XYZ.X += 8;
// I really don't know if it is a good idea. Neither what to do for !PRIM->FST // I really don't know if it is a good idea. Neither what to do for !PRIM->FST
if (unaligned_texture) if (unaligned_texture)
@ -1496,14 +1587,18 @@ void GSRendererHW::Draw()
} }
// Noting to do if no texture is sampled // Noting to do if no texture is sampled
if (PRIM->FST && draw_sprite_tex) { if (PRIM->FST && draw_sprite_tex)
if ((m_userhacks_round_sprite_offset > 1) || (m_userhacks_round_sprite_offset == 1 && !m_vt.IsLinear())) { {
if ((m_userhacks_round_sprite_offset > 1) || (m_userhacks_round_sprite_offset == 1 && !m_vt.IsLinear()))
{
if (m_vt.IsLinear()) if (m_vt.IsLinear())
RoundSpriteOffset<true>(); RoundSpriteOffset<true>();
else else
RoundSpriteOffset<false>(); RoundSpriteOffset<false>();
} }
} else { }
else
{
; // vertical line in Yakuza (note check m_userhacks_align_sprite_X behavior) ; // vertical line in Yakuza (note check m_userhacks_align_sprite_X behavior)
} }
} }
@ -1522,10 +1617,12 @@ void GSRendererHW::Draw()
// Help to detect rendering outside of the framebuffer // Help to detect rendering outside of the framebuffer
#if _DEBUG #if _DEBUG
if (m_upscale_multiplier * m_r.z > m_width) { if (m_upscale_multiplier * m_r.z > m_width)
{
GL_INS("ERROR: RT width is too small only %d but require %d", m_width, m_upscale_multiplier * m_r.z); GL_INS("ERROR: RT width is too small only %d but require %d", m_width, m_upscale_multiplier * m_r.z);
} }
if (m_upscale_multiplier * m_r.w > m_height) { if (m_upscale_multiplier * m_r.w > m_height)
{
GL_INS("ERROR: RT height is too small only %d but require %d", m_height, m_upscale_multiplier * m_r.w); GL_INS("ERROR: RT height is too small only %d but require %d", m_height, m_upscale_multiplier * m_r.w);
} }
#endif #endif
@ -1629,7 +1726,8 @@ void GSRendererHW::Hacks::SetGameCRC(const CRC::Game& game)
m_oo = m_oo_map[hash]; m_oo = m_oo_map[hash];
m_cu = m_cu_map[hash]; m_cu = m_cu_map[hash];
if (game.flags & CRC::PointListPalette) { if (game.flags & CRC::PointListPalette)
{
ASSERT(m_oi == NULL); ASSERT(m_oi == NULL);
m_oi = &GSRendererHW::OI_PointListPalette; m_oi = &GSRendererHW::OI_PointListPalette;
@ -1644,7 +1742,8 @@ void GSRendererHW::OI_DoubleHalfClear(GSTexture* rt, GSTexture* ds)
// Note gs mem clear must be tested before calling this function // Note gs mem clear must be tested before calling this function
// Limit further to unmask Z write // Limit further to unmask Z write
if (!m_context->ZBUF.ZMSK && rt && ds) { if (!m_context->ZBUF.ZMSK && rt && ds)
{
const GSVertex* v = &m_vertex.buff[0]; const GSVertex* v = &m_vertex.buff[0];
const GSLocalMemory::psm_t& frame_psm = GSLocalMemory::m_psm[m_context->FRAME.PSM]; const GSLocalMemory::psm_t& frame_psm = GSLocalMemory::m_psm[m_context->FRAME.PSM];
//const GSLocalMemory::psm_t& depth_psm = GSLocalMemory::m_psm[m_context->ZBUF.PSM]; //const GSLocalMemory::psm_t& depth_psm = GSLocalMemory::m_psm[m_context->ZBUF.PSM];
@ -1665,16 +1764,20 @@ void GSRendererHW::OI_DoubleHalfClear(GSTexture* rt, GSTexture* ds)
// Frame and depth pointer can be inverted // Frame and depth pointer can be inverted
uint32 base; uint32 base;
uint32 half; uint32 half;
if (m_context->FRAME.FBP > m_context->ZBUF.ZBP) { if (m_context->FRAME.FBP > m_context->ZBUF.ZBP)
{
base = m_context->ZBUF.ZBP; base = m_context->ZBUF.ZBP;
half = m_context->FRAME.FBP; half = m_context->FRAME.FBP;
} else { }
else
{
base = m_context->FRAME.FBP; base = m_context->FRAME.FBP;
half = m_context->ZBUF.ZBP; half = m_context->ZBUF.ZBP;
} }
// If both buffers are side by side we can expect a fast clear in on-going // If both buffers are side by side we can expect a fast clear in on-going
if (half <= (base + written_pages)) { if (half <= (base + written_pages))
{
uint32 color = v[1].RGBAQ.u32[0]; uint32 color = v[1].RGBAQ.u32[0];
bool clear_depth = (m_context->FRAME.FBP > m_context->ZBUF.ZBP); bool clear_depth = (m_context->FRAME.FBP > m_context->ZBUF.ZBP);
@ -1686,11 +1789,14 @@ void GSRendererHW::OI_DoubleHalfClear(GSTexture* rt, GSTexture* ds)
GSVector4i commitRect = ComputeBoundingBox(t->GetScale(), t->GetSize()); GSVector4i commitRect = ComputeBoundingBox(t->GetScale(), t->GetSize());
t->CommitRegion(GSVector2i(commitRect.z, 2 * commitRect.w)); t->CommitRegion(GSVector2i(commitRect.z, 2 * commitRect.w));
if (clear_depth) { if (clear_depth)
{
// Only pure clear are supported for depth // Only pure clear are supported for depth
ASSERT(color == 0); ASSERT(color == 0);
m_dev->ClearDepth(t); m_dev->ClearDepth(t);
} else { }
else
{
m_dev->ClearRenderTarget(t, color); m_dev->ClearRenderTarget(t, color);
} }
} }
@ -1703,7 +1809,8 @@ void GSRendererHW::OI_GsMemClear()
// Note gs mem clear must be tested before calling this function // Note gs mem clear must be tested before calling this function
// Limit it further to a full screen 0 write // Limit it further to a full screen 0 write
if ((m_vertex.next == 2) && m_vt.m_min.c.eq(GSVector4i(0))) { if ((m_vertex.next == 2) && m_vt.m_min.c.eq(GSVector4i(0)))
{
GSOffset* off = m_context->offset.fb; GSOffset* off = m_context->offset.fb;
GSVector4i r = GSVector4i(m_vt.m_min.p.xyxy(m_vt.m_max.p)).rintersect(GSVector4i(m_context->scissor.in)); GSVector4i r = GSVector4i(m_vt.m_min.p.xyxy(m_vt.m_max.p)).rintersect(GSVector4i(m_context->scissor.in));
// Limit the hack to a single fullscreen clear. Some games might use severals column to clear a screen // Limit the hack to a single fullscreen clear. Some games might use severals column to clear a screen
@ -1717,7 +1824,8 @@ void GSRendererHW::OI_GsMemClear()
// FIXME: loop can likely be optimized with AVX/SSE. Pixels aren't // FIXME: loop can likely be optimized with AVX/SSE. Pixels aren't
// linear but the value will be done for all pixels of a block. // linear but the value will be done for all pixels of a block.
// FIXME: maybe we could limit the write to the top and bottom row page. // FIXME: maybe we could limit the write to the top and bottom row page.
if (format == 0) { if (format == 0)
{
// Based on WritePixel32 // Based on WritePixel32
for (int y = r.top; y < r.bottom; y++) for (int y = r.top; y < r.bottom; y++)
{ {
@ -1729,7 +1837,9 @@ void GSRendererHW::OI_GsMemClear()
d[col[x]] = 0; // Here the constant color d[col[x]] = 0; // Here the constant color
} }
} }
} else if (format == 1) { }
else if (format == 1)
{
// Based on WritePixel24 // Based on WritePixel24
for (int y = r.top; y < r.bottom; y++) for (int y = r.top; y < r.bottom; y++)
{ {
@ -1741,7 +1851,9 @@ void GSRendererHW::OI_GsMemClear()
d[col[x]] &= 0xff000000; // Clear the color d[col[x]] &= 0xff000000; // Clear the color
} }
} }
} else if (format == 2) { }
else if (format == 2)
{
; // Hack is used for FMV which are likely 24/32 bits. Let's keep the for reference ; // Hack is used for FMV which are likely 24/32 bits. Let's keep the for reference
#if 0 #if 0
// Based on WritePixel16 // Based on WritePixel16
@ -1762,7 +1874,8 @@ void GSRendererHW::OI_GsMemClear()
bool GSRendererHW::OI_BlitFMV(GSTextureCache::Target* _rt, GSTextureCache::Source* tex, const GSVector4i& r_draw) bool GSRendererHW::OI_BlitFMV(GSTextureCache::Target* _rt, GSTextureCache::Source* tex, const GSVector4i& r_draw)
{ {
if (r_draw.w > 1024 && (m_vt.m_primclass == GS_SPRITE_CLASS) && (m_vertex.next == 2) && PRIM->TME && !PRIM->ABE && tex && !tex->m_target && m_context->TEX0.TBW > 0) { if (r_draw.w > 1024 && (m_vt.m_primclass == GS_SPRITE_CLASS) && (m_vertex.next == 2) && PRIM->TME && !PRIM->ABE && tex && !tex->m_target && m_context->TEX0.TBW > 0)
{
GL_PUSH("OI_BlitFMV"); GL_PUSH("OI_BlitFMV");
GL_INS("OI_BlitFMV"); GL_INS("OI_BlitFMV");
@ -1800,7 +1913,8 @@ bool GSRendererHW::OI_BlitFMV(GSTextureCache::Target* _rt, GSTextureCache::Sourc
// Do the blit. With a Copy mess to avoid issue with limited API (dx) // Do the blit. With a Copy mess to avoid issue with limited API (dx)
// m_dev->StretchRect(tex->m_texture, sRect, tex->m_texture, dRect); // m_dev->StretchRect(tex->m_texture, sRect, tex->m_texture, dRect);
GSVector4i r_full(0, 0, tw, th); GSVector4i r_full(0, 0, tw, th);
if (GSTexture* rt = m_dev->CreateRenderTarget(tw, th)) { if (GSTexture* rt = m_dev->CreateRenderTarget(tw, th))
{
m_dev->CopyRect(tex->m_texture, rt, r_full); m_dev->CopyRect(tex->m_texture, rt, r_full);
m_dev->StretchRect(tex->m_texture, sRect, rt, dRect); m_dev->StretchRect(tex->m_texture, sRect, rt, dRect);
@ -1891,7 +2005,8 @@ bool GSRendererHW::OI_FFXII(GSTexture* rt, GSTexture* ds, GSTextureCache::Source
{ {
// incoming pixels are stored in columns, one column is 16x512, total res 448x512 or 448x454 // incoming pixels are stored in columns, one column is 16x512, total res 448x512 or 448x454
if(!video) video = new uint32[512 * 512]; if (!video)
video = new uint32[512 * 512];
int ox = m_context->XYOFFSET.OFX - 8; int ox = m_context->XYOFFSET.OFX - 8;
int oy = m_context->XYOFFSET.OFY - 8; int oy = m_context->XYOFFSET.OFY - 8;
@ -1903,7 +2018,8 @@ bool GSRendererHW::OI_FFXII(GSTexture* rt, GSTexture* ds, GSTextureCache::Source
int x = (v->XYZ.X - ox) >> 4; int x = (v->XYZ.X - ox) >> 4;
int y = (v->XYZ.Y - oy) >> 4; int y = (v->XYZ.Y - oy) >> 4;
if (x < 0 || x >= 448 || y < 0 || y >= (int)lines) return false; // le sigh if (x < 0 || x >= 448 || y < 0 || y >= (int)lines)
return false; // le sigh
video[(y << 8) + (y << 7) + (y << 6) + x] = v->RGBAQ.u32[0]; video[(y << 8) + (y << 7) + (y << 6) + x] = v->RGBAQ.u32[0];
} }
@ -2194,7 +2310,8 @@ bool GSRendererHW::OI_ArTonelico2(GSTexture* rt, GSTexture* ds, GSTextureCache::
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
if (m_vertex.next == 2 && !PRIM->TME && m_context->FRAME.FBW == 10 && v->XYZ.Z == 0 && m_context->TEST.ZTST == ZTST_ALWAYS) { if (m_vertex.next == 2 && !PRIM->TME && m_context->FRAME.FBW == 10 && v->XYZ.Z == 0 && m_context->TEST.ZTST == ZTST_ALWAYS)
{
GL_INS("OI_ArTonelico2"); GL_INS("OI_ArTonelico2");
if (ds) if (ds)
ds->Commit(); // Don't bother to save few MB for a single game ds->Commit(); // Don't bother to save few MB for a single game

14
plugins/GSdx/Renderers/HW/GSRendererHW.h
View File

@ -73,7 +73,8 @@ private:
class Hacks class Hacks
{ {
template<class T> class HackEntry template <class T>
class HackEntry
{ {
public: public:
CRC::Title title; CRC::Title title;
@ -88,7 +89,8 @@ private:
} }
}; };
template<class T> class FunctionMap : public GSFunctionMap<uint32, T> template <class T>
class FunctionMap : public GSFunctionMap<uint32, T>
{ {
std::list<HackEntry<T>>& m_tbl; std::list<HackEntry<T>>& m_tbl;
@ -109,7 +111,10 @@ private:
} }
public: public:
FunctionMap(std::list<HackEntry<T> >& tbl) : m_tbl(tbl) {} FunctionMap(std::list<HackEntry<T>>& tbl)
: m_tbl(tbl)
{
}
}; };
std::list<HackEntry<OI_Ptr>> m_oi_list; std::list<HackEntry<OI_Ptr>> m_oi_list;
@ -139,7 +144,8 @@ private:
void SwSpriteRender(); void SwSpriteRender();
bool CanUseSwSpriteRender(bool allow_64x64_sprite); bool CanUseSwSpriteRender(bool allow_64x64_sprite);
template <bool linear> void RoundSpriteOffset(); template <bool linear>
void RoundSpriteOffset();
protected: protected:
GSTextureCache* m_tc; GSTextureCache* m_tc;

394
plugins/GSdx/Renderers/HW/GSTextureCache.cpp
View File

@ -31,7 +31,8 @@ GSTextureCache::GSTextureCache(GSRenderer* r)
: m_renderer(r) : m_renderer(r)
, m_palette_map(r) , m_palette_map(r)
{ {
if (theApp.GetConfigB("UserHacks")) { if (theApp.GetConfigB("UserHacks"))
{
UserHacks_HalfPixelOffset = theApp.GetConfigI("UserHacks_HalfPixelOffset") == 1; UserHacks_HalfPixelOffset = theApp.GetConfigI("UserHacks_HalfPixelOffset") == 1;
m_preload_frame = theApp.GetConfigB("preload_frame_with_gs_data"); m_preload_frame = theApp.GetConfigB("preload_frame_with_gs_data");
m_disable_partial_invalidation = theApp.GetConfigB("UserHacks_DisablePartialInvalidation"); m_disable_partial_invalidation = theApp.GetConfigB("UserHacks_DisablePartialInvalidation");
@ -39,7 +40,9 @@ GSTextureCache::GSTextureCache(GSRenderer* r)
m_cpu_fb_conversion = theApp.GetConfigB("UserHacks_CPU_FB_Conversion"); m_cpu_fb_conversion = theApp.GetConfigB("UserHacks_CPU_FB_Conversion");
m_texture_inside_rt = theApp.GetConfigB("UserHacks_TextureInsideRt"); m_texture_inside_rt = theApp.GetConfigB("UserHacks_TextureInsideRt");
m_wrap_gs_mem = theApp.GetConfigB("wrap_gs_mem"); m_wrap_gs_mem = theApp.GetConfigB("wrap_gs_mem");
} else { }
else
{
UserHacks_HalfPixelOffset = false; UserHacks_HalfPixelOffset = false;
m_preload_frame = false; m_preload_frame = false;
m_disable_partial_invalidation = false; m_disable_partial_invalidation = false;
@ -77,7 +80,8 @@ void GSTextureCache::RemovePartial()
for (int type = 0; type < 2; type++) for (int type = 0; type < 2; type++)
{ {
for (auto t : m_dst[type]) delete t; for (auto t : m_dst[type])
delete t;
m_dst[type].clear(); m_dst[type].clear();
} }
@ -89,7 +93,8 @@ void GSTextureCache::RemoveAll()
for (int type = 0; type < 2; type++) for (int type = 0; type < 2; type++)
{ {
for (auto t : m_dst[type]) delete t; for (auto t : m_dst[type])
delete t;
m_dst[type].clear(); m_dst[type].clear();
} }
@ -99,13 +104,17 @@ void GSTextureCache::RemoveAll()
GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0, const GIFRegTEXA& TEXA, const GSVector4i& r, bool palette) GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0, const GIFRegTEXA& TEXA, const GSVector4i& r, bool palette)
{ {
if (!m_can_convert_depth) { if (!m_can_convert_depth)
{
GL_CACHE("LookupDepthSource not supported (0x%x, F:0x%x)", TEX0.TBP0, TEX0.PSM); GL_CACHE("LookupDepthSource not supported (0x%x, F:0x%x)", TEX0.TBP0, TEX0.PSM);
if (m_renderer->m_game.title == CRC::JackieChanAdv || m_renderer->m_game.title == CRC::SVCChaos) { if (m_renderer->m_game.title == CRC::JackieChanAdv || m_renderer->m_game.title == CRC::SVCChaos)
{
// JackieChan and SVCChaos cause regressions when skipping the draw calls when depth is disabled/not supported. // JackieChan and SVCChaos cause regressions when skipping the draw calls when depth is disabled/not supported.
// This way we make sure there are no regressions on D3D as well. // This way we make sure there are no regressions on D3D as well.
return LookupSource(TEX0, TEXA, r); return LookupSource(TEX0, TEXA, r);
} else { }
else
{
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
} }
@ -119,24 +128,30 @@ GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0
uint32 bp = TEX0.TBP0; uint32 bp = TEX0.TBP0;
uint32 psm = TEX0.PSM; uint32 psm = TEX0.PSM;
for(auto t : m_dst[DepthStencil]) { for (auto t : m_dst[DepthStencil])
{
if (t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM)) if (t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM))
{ {
ASSERT(GSLocalMemory::m_psm[t->m_TEX0.PSM].depth); ASSERT(GSLocalMemory::m_psm[t->m_TEX0.PSM].depth);
if (t->m_age == 0) { if (t->m_age == 0)
{
// Perfect Match // Perfect Match
dst = t; dst = t;
break; break;
} else if (t->m_age == 1) { }
else if (t->m_age == 1)
{
// Better than nothing (Full Spectrum Warrior) // Better than nothing (Full Spectrum Warrior)
dst = t; dst = t;
} }
} }
} }
if (!dst) { if (!dst)
{
// Retry on the render target (Silent Hill 4) // Retry on the render target (Silent Hill 4)
for(auto t : m_dst[RenderTarget]) { for (auto t : m_dst[RenderTarget])
{
// FIXME: do I need to allow m_age == 1 as a potential match (as DepthStencil) ??? // FIXME: do I need to allow m_age == 1 as a potential match (as DepthStencil) ???
if (!t->m_age && t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM)) if (!t->m_age && t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM))
{ {
@ -147,7 +162,8 @@ GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0
} }
} }
if (dst) { if (dst)
{
GL_CACHE("TC depth: dst %s hit: %d (0x%x, %s)", to_string(dst->m_type), GL_CACHE("TC depth: dst %s hit: %d (0x%x, %s)", to_string(dst->m_type),
dst->m_texture ? dst->m_texture->GetID() : 0, dst->m_texture ? dst->m_texture->GetID() : 0,
TEX0.TBP0, psm_str(psm)); TEX0.TBP0, psm_str(psm));
@ -167,12 +183,15 @@ GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0
// texture cache list. It means that a new Source is created everytime we need it. // texture cache list. It means that a new Source is created everytime we need it.
// If it is too expensive, one could cut memory allocation in Source constructor for this // If it is too expensive, one could cut memory allocation in Source constructor for this
// use case. // use case.
if (palette) { if (palette)
{
AttachPaletteToSource(src, psm_s.pal, true); AttachPaletteToSource(src, psm_s.pal, true);
} }
m_src.m_surfaces.insert(src); m_src.m_surfaces.insert(src);
} else { }
else
{
GL_CACHE("TC depth: ERROR miss (0x%x, %s)", TEX0.TBP0, psm_str(psm)); GL_CACHE("TC depth: ERROR miss (0x%x, %s)", TEX0.TBP0, psm_str(psm));
// Possible ? In this case we could call LookupSource // Possible ? In this case we could call LookupSource
// Or just put a basic texture // Or just put a basic texture
@ -182,10 +201,13 @@ GSTextureCache::Source* GSTextureCache::LookupDepthSource(const GIFRegTEX0& TEX0
// Note: might worth to check previous frame // Note: might worth to check previous frame
// Note: otherwise return NULL and skip the draw // Note: otherwise return NULL and skip the draw
if (m_renderer->m_game.title == CRC::JackieChanAdv || m_renderer->m_game.title == CRC::SVCChaos) { if (m_renderer->m_game.title == CRC::JackieChanAdv || m_renderer->m_game.title == CRC::SVCChaos)
{
// JackieChan and SVCChaos cause regressions when skipping the draw calls so we reuse the old code for these two. // JackieChan and SVCChaos cause regressions when skipping the draw calls so we reuse the old code for these two.
return LookupSource(TEX0, TEXA, r); return LookupSource(TEX0, TEXA, r);
} else { }
else
{
// Full Spectrum Warrior: first draw call of cut-scene rendering // Full Spectrum Warrior: first draw call of cut-scene rendering
// The game tries to emulate a texture shuffle with an old depth buffer // The game tries to emulate a texture shuffle with an old depth buffer
// (don't exists yet for us due to the cache) // (don't exists yet for us due to the cache)
@ -222,7 +244,8 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
continue; continue;
// Target are converted (AEM & palette) on the fly by the GPU. They don't need extra check // Target are converted (AEM & palette) on the fly by the GPU. They don't need extra check
if (!s->m_target) { if (!s->m_target)
{
// We request a palette texture (psm_s.pal). If the texture was // We request a palette texture (psm_s.pal). If the texture was
// converted by the CPU (!s->m_palette), we need to ensure // converted by the CPU (!s->m_palette), we need to ensure
// palette content is the same. // palette content is the same.
@ -269,8 +292,10 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
bool texture_inside_rt = ShallSearchTextureInsideRt(); bool texture_inside_rt = ShallSearchTextureInsideRt();
for(auto t : m_dst[RenderTarget]) { for (auto t : m_dst[RenderTarget])
if(t->m_used && t->m_dirty.empty()) { {
if (t->m_used && t->m_dirty.empty())
{
// Typical bug (MGS3 blue cloud): // Typical bug (MGS3 blue cloud):
// 1/ RT used as 32 bits => alpha channel written // 1/ RT used as 32 bits => alpha channel written
// 2/ RT used as 24 bits => no update of alpha channel // 2/ RT used as 24 bits => no update of alpha channel
@ -280,7 +305,8 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
// Solution: consider the RT as 32 bits if the alpha was used in the past // Solution: consider the RT as 32 bits if the alpha was used in the past
uint32 t_psm = (t->m_dirty_alpha) ? t->m_TEX0.PSM & ~0x1 : t->m_TEX0.PSM; uint32 t_psm = (t->m_dirty_alpha) ? t->m_TEX0.PSM & ~0x1 : t->m_TEX0.PSM;
if (GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t_psm)) { if (GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t_psm))
{
// It is a complex to convert the code in shader. As a reference, let's do it on the CPU, it will be slow but // It is a complex to convert the code in shader. As a reference, let's do it on the CPU, it will be slow but
// 1/ it just works :) // 1/ it just works :)
// 2/ even with upscaling // 2/ even with upscaling
@ -295,8 +321,9 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
dst = t; dst = t;
break; break;
}
} else if ((t->m_TEX0.TBW >= 16) && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0 + t->m_TEX0.TBW * 0x10, t->m_TEX0.PSM)) { else if ((t->m_TEX0.TBW >= 16) && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0 + t->m_TEX0.TBW * 0x10, t->m_TEX0.PSM))
{
// Detect half of the render target (fix snow engine game) // Detect half of the render target (fix snow engine game)
// Target Page (8KB) have always a width of 64 pixels // Target Page (8KB) have always a width of 64 pixels
// Half of the Target is TBW/2 pages * 8KB / (1 block * 256B) = 0x10 // Half of the Target is TBW/2 pages * 8KB / (1 block * 256B) = 0x10
@ -304,8 +331,9 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
dst = t; dst = t;
break; break;
}
} else if (texture_inside_rt && psm == PSM_PSMCT32 && t->m_TEX0.PSM == psm && t->m_TEX0.TBP0 < bp && t->m_end_block >= bp) { else if (texture_inside_rt && psm == PSM_PSMCT32 && t->m_TEX0.PSM == psm && t->m_TEX0.TBP0 < bp && t->m_end_block >= bp)
{
// Only PSMCT32 to limit false hits // Only PSMCT32 to limit false hits
// Check if it is possible to hit with valid <x,y> offset on the given Target // Check if it is possible to hit with valid <x,y> offset on the given Target
@ -404,23 +432,28 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
// //
// Sigh... They don't help us. // Sigh... They don't help us.
if (dst == NULL && m_can_convert_depth) { if (dst == NULL && m_can_convert_depth)
{
// Let's try a trick to avoid to use wrongly a depth buffer // Let's try a trick to avoid to use wrongly a depth buffer
// Unfortunately, I don't have any Arc the Lad testcase // Unfortunately, I don't have any Arc the Lad testcase
// //
// 1/ Check only current frame, I guess it is only used as a postprocessing effect // 1/ Check only current frame, I guess it is only used as a postprocessing effect
for(auto t : m_dst[DepthStencil]) { for (auto t : m_dst[DepthStencil])
{
if (!t->m_age && t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM)) if (!t->m_age && t->m_used && t->m_dirty.empty() && GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM))
{ {
GL_INS("TC: Warning depth format read as color format. Pixels will be scrambled"); GL_INS("TC: Warning depth format read as color format. Pixels will be scrambled");
// Let's fetch a depth format texture. Rational, it will avoid the texture allocation and the // Let's fetch a depth format texture. Rational, it will avoid the texture allocation and the
// rescaling of the current function. // rescaling of the current function.
if (psm_s.bpp > 8) { if (psm_s.bpp > 8)
{
GIFRegTEX0 depth_TEX0; GIFRegTEX0 depth_TEX0;
depth_TEX0.u32[0] = TEX0.u32[0] | (0x30u << 20u); depth_TEX0.u32[0] = TEX0.u32[0] | (0x30u << 20u);
depth_TEX0.u32[1] = TEX0.u32[1]; depth_TEX0.u32[1] = TEX0.u32[1];
return LookupDepthSource(depth_TEX0, TEXA, r); return LookupDepthSource(depth_TEX0, TEXA, r);
} else { }
else
{
return LookupDepthSource(TEX0, TEXA, r, true); return LookupDepthSource(TEX0, TEXA, r, true);
} }
} }
@ -433,25 +466,30 @@ GSTextureCache::Source* GSTextureCache::LookupSource(const GIFRegTEX0& TEX0, con
if (src == NULL) if (src == NULL)
{ {
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG
if (dst) { if (dst)
{
GL_CACHE("TC: dst %s hit (%s): %d (0x%x, %s)", to_string(dst->m_type), half_right ? "half" : "full", GL_CACHE("TC: dst %s hit (%s): %d (0x%x, %s)", to_string(dst->m_type), half_right ? "half" : "full",
dst->m_texture ? dst->m_texture->GetID() : 0, dst->m_texture ? dst->m_texture->GetID() : 0,
TEX0.TBP0, psm_str(TEX0.PSM)); TEX0.TBP0, psm_str(TEX0.PSM));
} else { }
else
{
GL_CACHE("TC: src miss (0x%x, 0x%x, %s)", TEX0.TBP0, psm_s.pal > 0 ? TEX0.CBP : 0, psm_str(TEX0.PSM)); GL_CACHE("TC: src miss (0x%x, 0x%x, %s)", TEX0.TBP0, psm_s.pal > 0 ? TEX0.CBP : 0, psm_str(TEX0.PSM));
} }
#endif #endif
src = CreateSource(TEX0, TEXA, dst, half_right, x_offset, y_offset); src = CreateSource(TEX0, TEXA, dst, half_right, x_offset, y_offset);
new_source = true; new_source = true;
}
} else { else
{
GL_CACHE("TC: src hit: %d (0x%x, 0x%x, %s)", GL_CACHE("TC: src hit: %d (0x%x, 0x%x, %s)",
src->m_texture ? src->m_texture->GetID() : 0, src->m_texture ? src->m_texture->GetID() : 0,
TEX0.TBP0, psm_s.pal > 0 ? TEX0.CBP : 0, TEX0.TBP0, psm_s.pal > 0 ? TEX0.CBP : 0,
psm_str(TEX0.PSM)); psm_str(TEX0.PSM));
} }
if (src->m_palette && !new_source && !src->ClutMatch({ clut, psm_s.pal })) { if (src->m_palette && !new_source && !src->ClutMatch({clut, psm_s.pal}))
{
AttachPaletteToSource(src, psm_s.pal, true); AttachPaletteToSource(src, psm_s.pal, true);
} }
@ -497,7 +535,8 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
Target* dst = NULL; Target* dst = NULL;
auto& list = m_dst[type]; auto& list = m_dst[type];
for(auto i = list.begin(); i != list.end(); ++i) { for (auto i = list.begin(); i != list.end(); ++i)
{
Target* t = *i; Target* t = *i;
if (bp == t->m_TEX0.TBP0) if (bp == t->m_TEX0.TBP0)
@ -513,32 +552,40 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
} }
} }
if (dst) { if (dst)
{
GL_CACHE("TC: Lookup Target(%s) %dx%d, hit: %d (0x%x, %s)", to_string(type), w, h, dst->m_texture->GetID(), bp, psm_str(TEX0.PSM)); GL_CACHE("TC: Lookup Target(%s) %dx%d, hit: %d (0x%x, %s)", to_string(type), w, h, dst->m_texture->GetID(), bp, psm_str(TEX0.PSM));
dst->Update(); dst->Update();
dst->m_dirty_alpha |= (psm_s.trbpp == 32 && (fbmask & 0xFF000000) != 0xFF000000) || (psm_s.trbpp == 16); dst->m_dirty_alpha |= (psm_s.trbpp == 32 && (fbmask & 0xFF000000) != 0xFF000000) || (psm_s.trbpp == 16);
}
} else if (m_can_convert_depth) { else if (m_can_convert_depth)
{
int rev_type = (type == DepthStencil) ? RenderTarget : DepthStencil; int rev_type = (type == DepthStencil) ? RenderTarget : DepthStencil;
// Depth stencil/RT can be an older RT/DS but only check recent RT/DS to avoid to pick // Depth stencil/RT can be an older RT/DS but only check recent RT/DS to avoid to pick
// some bad data. // some bad data.
Target* dst_match = nullptr; Target* dst_match = nullptr;
for(auto t : m_dst[rev_type]) { for (auto t : m_dst[rev_type])
if (bp == t->m_TEX0.TBP0) { {
if (t->m_age == 0) { if (bp == t->m_TEX0.TBP0)
{
if (t->m_age == 0)
{
dst_match = t; dst_match = t;
break; break;
} else if (t->m_age == 1) { }
else if (t->m_age == 1)
{
dst_match = t; dst_match = t;
} }
} }
} }
if (dst_match) { if (dst_match)
{
GSVector4 sRect(0, 0, 1, 1); GSVector4 sRect(0, 0, 1, 1);
GSVector4 dRect(0, 0, w, h); GSVector4 dRect(0, 0, w, h);
@ -547,10 +594,13 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
int shader; int shader;
bool fmt_16_bits = (psm_s.bpp == 16 && GSLocalMemory::m_psm[dst_match->m_TEX0.PSM].bpp == 16); bool fmt_16_bits = (psm_s.bpp == 16 && GSLocalMemory::m_psm[dst_match->m_TEX0.PSM].bpp == 16);
if (type == DepthStencil) { if (type == DepthStencil)
{
GL_CACHE("TC: Lookup Target(Depth) %dx%d, hit Color (0x%x, %s was %s)", w, h, bp, psm_str(TEX0.PSM), psm_str(dst_match->m_TEX0.PSM)); GL_CACHE("TC: Lookup Target(Depth) %dx%d, hit Color (0x%x, %s was %s)", w, h, bp, psm_str(TEX0.PSM), psm_str(dst_match->m_TEX0.PSM));
shader = (fmt_16_bits) ? ShaderConvert_RGB5A1_TO_FLOAT16 : ShaderConvert_RGBA8_TO_FLOAT32 + psm_s.fmt; shader = (fmt_16_bits) ? ShaderConvert_RGB5A1_TO_FLOAT16 : ShaderConvert_RGBA8_TO_FLOAT32 + psm_s.fmt;
} else { }
else
{
GL_CACHE("TC: Lookup Target(Color) %dx%d, hit Depth (0x%x, %s was %s)", w, h, bp, psm_str(TEX0.PSM), psm_str(dst_match->m_TEX0.PSM)); GL_CACHE("TC: Lookup Target(Color) %dx%d, hit Depth (0x%x, %s was %s)", w, h, bp, psm_str(TEX0.PSM), psm_str(dst_match->m_TEX0.PSM));
shader = (fmt_16_bits) ? ShaderConvert_FLOAT16_TO_RGB5A1 : ShaderConvert_FLOAT32_TO_RGBA8; shader = (fmt_16_bits) ? ShaderConvert_FLOAT16_TO_RGB5A1 : ShaderConvert_FLOAT32_TO_RGBA8;
} }
@ -576,7 +626,8 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
// From a performance point of view, it might cost a little on big upscaling // From a performance point of view, it might cost a little on big upscaling
// but normally few RT are miss so it must remain reasonable. // but normally few RT are miss so it must remain reasonable.
bool supported_fmt = m_can_convert_depth || psm_s.depth == 0; bool supported_fmt = m_can_convert_depth || psm_s.depth == 0;
if (m_preload_frame && TEX0.TBW > 0 && supported_fmt) { if (m_preload_frame && TEX0.TBW > 0 && supported_fmt)
{
GL_INS("Preloading the RT DATA"); GL_INS("Preloading the RT DATA");
// RT doesn't have height but if we use a too big value, we will read outside of the GS memory. // RT doesn't have height but if we use a too big value, we will read outside of the GS memory.
int page0 = TEX0.TBP0 >> 5; int page0 = TEX0.TBP0 >> 5;
@ -587,7 +638,9 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
dst->m_dirty.push_back(GSDirtyRect(GSVector4i(0, 0, TEX0.TBW * 64, max_h), TEX0.PSM)); dst->m_dirty.push_back(GSDirtyRect(GSVector4i(0, 0, TEX0.TBW * 64, max_h), TEX0.PSM));
dst->Update(); dst->Update();
} else { }
else
{
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG
switch (type) { switch (type) {
case RenderTarget: m_renderer->m_dev->ClearRenderTarget(dst->m_texture, 0); break; case RenderTarget: m_renderer->m_dev->ClearRenderTarget(dst->m_texture, 0); break;
@ -620,8 +673,10 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
#endif #endif
// Let's try to find a perfect frame that contains valid data // Let's try to find a perfect frame that contains valid data
for(auto t : m_dst[RenderTarget]) { for (auto t : m_dst[RenderTarget])
if(bp == t->m_TEX0.TBP0 && t->m_end_block >= bp) { {
if (bp == t->m_TEX0.TBP0 && t->m_end_block >= bp)
{
dst = t; dst = t;
GL_CACHE("TC: Lookup Frame %dx%d, perfect hit: %d (0x%x -> 0x%x %s)", w, h, dst->m_texture->GetID(), bp, t->m_end_block, psm_str(TEX0.PSM)); GL_CACHE("TC: Lookup Frame %dx%d, perfect hit: %d (0x%x -> 0x%x %s)", w, h, dst->m_texture->GetID(), bp, t->m_end_block, psm_str(TEX0.PSM));
@ -631,9 +686,12 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
} }
// 2nd try ! Try to find a frame that include the bp // 2nd try ! Try to find a frame that include the bp
if (dst == NULL) { if (dst == NULL)
for(auto t : m_dst[RenderTarget]) { {
if (t->m_TEX0.TBP0 < bp && bp <= t->m_end_block) { for (auto t : m_dst[RenderTarget])
{
if (t->m_TEX0.TBP0 < bp && bp <= t->m_end_block)
{
dst = t; dst = t;
GL_CACHE("TC: Lookup Frame %dx%d, inclusive hit: %d (0x%x, took 0x%x -> 0x%x %s)", w, h, t->m_texture->GetID(), bp, t->m_TEX0.TBP0, t->m_end_block, psm_str(TEX0.PSM)); GL_CACHE("TC: Lookup Frame %dx%d, inclusive hit: %d (0x%x, took 0x%x -> 0x%x %s)", w, h, t->m_texture->GetID(), bp, t->m_TEX0.TBP0, t->m_end_block, psm_str(TEX0.PSM));
@ -644,9 +702,12 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
} }
// 3rd try ! Try to find a frame that doesn't contain valid data (honestly I'm not sure we need to do it) // 3rd try ! Try to find a frame that doesn't contain valid data (honestly I'm not sure we need to do it)
if (dst == NULL) { if (dst == NULL)
for(auto t : m_dst[RenderTarget]) { {
if(bp == t->m_TEX0.TBP0) { for (auto t : m_dst[RenderTarget])
{
if (bp == t->m_TEX0.TBP0)
{
dst = t; dst = t;
GL_CACHE("TC: Lookup Frame %dx%d, empty hit: %d (0x%x -> 0x%x %s)", w, h, dst->m_texture->GetID(), bp, t->m_end_block, psm_str(TEX0.PSM)); GL_CACHE("TC: Lookup Frame %dx%d, empty hit: %d (0x%x -> 0x%x %s)", w, h, dst->m_texture->GetID(), bp, t->m_end_block, psm_str(TEX0.PSM));
@ -690,7 +751,8 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
m_renderer->m_dev->ClearRenderTarget(dst->m_texture, 0); // new frame buffers after reset should be cleared, don't display memory garbage m_renderer->m_dev->ClearRenderTarget(dst->m_texture, 0); // new frame buffers after reset should be cleared, don't display memory garbage
if (m_preload_frame) { if (m_preload_frame)
{
// Load GS data into frame. Game can directly uploads a background or the full image in // Load GS data into frame. Game can directly uploads a background or the full image in
// "CTRC" buffer. It will also avoid various black screen issue in gs dump. // "CTRC" buffer. It will also avoid various black screen issue in gs dump.
// //
@ -737,14 +799,14 @@ void GSTextureCache::InvalidateVideoMemType(int type, uint32 bp)
break; break;
} }
} }
} }
// Goal: invalidate data sent to the GPU when the source (GS memory) is modified // Goal: invalidate data sent to the GPU when the source (GS memory) is modified
// Called each time you want to write to the GS memory // Called each time you want to write to the GS memory
void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, bool target) void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, bool target)
{ {
if(!off) return; // Fixme. Crashes Dual Hearts, maybe others as well. Was fine before r1549. if (!off)
return; // Fixme. Crashes Dual Hearts, maybe others as well. Was fine before r1549.
uint32 bp = off->bp; uint32 bp = off->bp;
uint32 bw = off->bw; uint32 bw = off->bw;
@ -768,7 +830,8 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
} }
uint32 bbp = bp + bw * 0x10; uint32 bbp = bp + bw * 0x10;
if (bw >= 16 && bbp < 16384) { if (bw >= 16 && bbp < 16384)
{
// Detect half of the render target (fix snow engine game) // Detect half of the render target (fix snow engine game)
// Target Page (8KB) have always a width of 64 pixels // Target Page (8KB) have always a width of 64 pixels
// Half of the Target is TBW/2 pages * 8KB / (1 block * 256B) = 0x10 // Half of the Target is TBW/2 pages * 8KB / (1 block * 256B) = 0x10
@ -788,13 +851,15 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
// Haunting ground write frame buffer 0x3000 and expect to write data to 0x3380 // Haunting ground write frame buffer 0x3000 and expect to write data to 0x3380
// Note: the game only does a 0 direct write. If some games expect some real data // Note: the game only does a 0 direct write. If some games expect some real data
// we are screwed. // we are screwed.
if (m_renderer->m_game.title == CRC::HauntingGround) { if (m_renderer->m_game.title == CRC::HauntingGround)
{
uint32 end_block = GSLocalMemory::m_psm[psm].bn(rect.z - 1, rect.w - 1, bp, bw); // Valid only for color formats uint32 end_block = GSLocalMemory::m_psm[psm].bn(rect.z - 1, rect.w - 1, bp, bw); // Valid only for color formats
auto type = RenderTarget; auto type = RenderTarget;
for (auto t : m_dst[type]) for (auto t : m_dst[type])
{ {
if (t->m_TEX0.TBP0 > bp && t->m_end_block <= end_block) { if (t->m_TEX0.TBP0 > bp && t->m_end_block <= end_block)
{
// Haunting ground expect to clean buffer B with a rendering into buffer A. // Haunting ground expect to clean buffer B with a rendering into buffer A.
// Situation is quite messy as it would require to extract the data from the buffer A // Situation is quite messy as it would require to extract the data from the buffer A
// and to move in buffer B. // and to move in buffer B.
@ -880,7 +945,8 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
} }
} }
if(!target) return; if (!target)
return;
for (int type = 0; type < 2; type++) for (int type = 0; type < 2; type++)
{ {
@ -916,7 +982,9 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
delete t; delete t;
continue; continue;
} }
} else if (bp == t->m_TEX0.TBP0) { }
else if (bp == t->m_TEX0.TBP0)
{
// EE writes the ALPHA channel. Mark it as invalid for // EE writes the ALPHA channel. Mark it as invalid for
// the texture cache. Otherwise it will generate a wrong // the texture cache. Otherwise it will generate a wrong
// hit on the texture cache. // hit on the texture cache.
@ -925,7 +993,8 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
} }
// GH: Try to detect texture write that will overlap with a target buffer // GH: Try to detect texture write that will overlap with a target buffer
if(GSUtil::HasSharedBits(psm, t->m_TEX0.PSM)) { if (GSUtil::HasSharedBits(psm, t->m_TEX0.PSM))
{
if (bp < t->m_TEX0.TBP0) if (bp < t->m_TEX0.TBP0)
{ {
uint32 rowsize = bw * 8192; uint32 rowsize = bw * 8192;
@ -955,11 +1024,13 @@ void GSTextureCache::InvalidateVideoMem(GSOffset* off, const GSVector4i& rect, b
// Ben 10 Alien Force : Vilgax Attacks uses a small temporary target for multiple textures (different bw) // Ben 10 Alien Force : Vilgax Attacks uses a small temporary target for multiple textures (different bw)
// It is too complex to handle, and purpose of the code was to handle FMV (large bw). So let's skip small // It is too complex to handle, and purpose of the code was to handle FMV (large bw). So let's skip small
// (128 pixels) target // (128 pixels) target
if (bw > 2 && t->m_TEX0.TBW == bw && t->Inside(bp, bw, psm, rect) && GSUtil::HasCompatibleBits(psm, t->m_TEX0.PSM)) { if (bw > 2 && t->m_TEX0.TBW == bw && t->Inside(bp, bw, psm, rect) && GSUtil::HasCompatibleBits(psm, t->m_TEX0.PSM))
{
uint32 rowsize = bw * 8192u; uint32 rowsize = bw * 8192u;
uint32 offset = (uint32)((bp - t->m_TEX0.TBP0) * 256); uint32 offset = (uint32)((bp - t->m_TEX0.TBP0) * 256);
if(rowsize > 0 && offset % rowsize == 0) { if (rowsize > 0 && offset % rowsize == 0)
{
int y = GSLocalMemory::m_psm[psm].pgs.y * offset / rowsize; int y = GSLocalMemory::m_psm[psm].pgs.y * offset / rowsize;
GL_CACHE("TC: Dirty in the middle of Target(%s) %d (0x%x->0x%x) pos(%d,%d => %d,%d) bw:%u", to_string(type), GL_CACHE("TC: Dirty in the middle of Target(%s) %d (0x%x->0x%x) pos(%d,%d => %d,%d) bw:%u", to_string(type),
@ -987,11 +1058,15 @@ void GSTextureCache::InvalidateLocalMem(GSOffset* off, const GSVector4i& r)
//uint32 bw = off->bw; //uint32 bw = off->bw;
// No depth handling please. // No depth handling please.
if (psm == PSM_PSMZ32 || psm == PSM_PSMZ24 || psm == PSM_PSMZ16 || psm == PSM_PSMZ16S) { if (psm == PSM_PSMZ32 || psm == PSM_PSMZ24 || psm == PSM_PSMZ16 || psm == PSM_PSMZ16S)
{
GL_INS("ERROR: InvalidateLocalMem depth format isn't supported (%d,%d to %d,%d)", r.x, r.y, r.z, r.w); GL_INS("ERROR: InvalidateLocalMem depth format isn't supported (%d,%d to %d,%d)", r.x, r.y, r.z, r.w);
if (m_can_convert_depth) { if (m_can_convert_depth)
for(auto t : m_dst[DepthStencil]) { {
if(GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM)) { for (auto t : m_dst[DepthStencil])
{
if (GSUtil::HasSharedBits(bp, psm, t->m_TEX0.TBP0, t->m_TEX0.PSM))
{
if (GSUtil::HasCompatibleBits(psm, t->m_TEX0.PSM)) if (GSUtil::HasCompatibleBits(psm, t->m_TEX0.PSM))
Read(t, r.rintersect(t->m_valid)); Read(t, r.rintersect(t->m_valid));
} }
@ -1020,16 +1095,21 @@ void GSTextureCache::InvalidateLocalMem(GSOffset* off, const GSVector4i& r)
// note: r.rintersect breaks Wizardry and Chaos Legion // note: r.rintersect breaks Wizardry and Chaos Legion
// Read(t, t->m_valid) works in all tested games but is very slow in GUST titles >< // Read(t, t->m_valid) works in all tested games but is very slow in GUST titles ><
if (GSTextureCache::m_disable_partial_invalidation) { if (GSTextureCache::m_disable_partial_invalidation)
{
Read(t, r.rintersect(t->m_valid)); Read(t, r.rintersect(t->m_valid));
} else { }
else
{
if (r.x == 0 && r.y == 0) // Full screen read? if (r.x == 0 && r.y == 0) // Full screen read?
Read(t, t->m_valid); Read(t, t->m_valid);
else // Block level read? else // Block level read?
Read(t, r.rintersect(t->m_valid)); Read(t, r.rintersect(t->m_valid));
} }
} }
} else { }
else
{
GL_INS("ERROR: InvalidateLocalMem target is a depth format"); GL_INS("ERROR: InvalidateLocalMem target is a depth format");
} }
} }
@ -1115,17 +1195,20 @@ void GSTextureCache::InvalidateVideoMemSubTarget(GSTextureCache::Target* rt)
auto& list = m_dst[RenderTarget]; auto& list = m_dst[RenderTarget];
for(auto i = list.begin(); i != list.end(); ) { for (auto i = list.begin(); i != list.end();)
{
Target* t = *i; Target* t = *i;
if((t->m_TEX0.TBP0 > rt->m_TEX0.TBP0) && (t->m_end_block < rt->m_end_block) && (t->m_TEX0.TBW == rt->m_TEX0.TBW) if ((t->m_TEX0.TBP0 > rt->m_TEX0.TBP0) && (t->m_end_block < rt->m_end_block) && (t->m_TEX0.TBW == rt->m_TEX0.TBW) && (t->m_TEX0.TBP0 < t->m_end_block))
&& (t->m_TEX0.TBP0 < t->m_end_block)) { {
GL_INS("InvalidateVideoMemSubTarget: rt 0x%x -> 0x%x, sub rt 0x%x -> 0x%x", GL_INS("InvalidateVideoMemSubTarget: rt 0x%x -> 0x%x, sub rt 0x%x -> 0x%x",
rt->m_TEX0.TBP0, rt->m_end_block, t->m_TEX0.TBP0, t->m_end_block); rt->m_TEX0.TBP0, rt->m_end_block, t->m_TEX0.TBP0, t->m_end_block);
i = list.erase(i); i = list.erase(i);
delete t; delete t;
} else { }
else
{
++i; ++i;
} }
} }
@ -1140,14 +1223,18 @@ void GSTextureCache::IncAge()
{ {
Source* s = *i; Source* s = *i;
if(s->m_shared_texture) { if (s->m_shared_texture)
{
// Shared textures are temporary only added in the hash set but not in the texture // Shared textures are temporary only added in the hash set but not in the texture
// cache list therefore you can't use RemoveAt // cache list therefore you can't use RemoveAt
i = m_src.m_surfaces.erase(i); i = m_src.m_surfaces.erase(i);
delete s; delete s;
} else { }
else
{
++i; ++i;
if (++s->m_age > maxage) { if (++s->m_age > maxage)
{
m_src.RemoveAt(s); m_src.RemoveAt(s);
} }
} }
@ -1172,7 +1259,8 @@ void GSTextureCache::IncAge()
// probability that game will do it on the current RT. // probability that game will do it on the current RT.
// Variable is cleared here to avoid issue with game that uses a 16 bits // Variable is cleared here to avoid issue with game that uses a 16 bits
// render target // render target
if (t->m_age > 0) { if (t->m_age > 0)
{
// GoW2 uses the effect at the start of the frame // GoW2 uses the effect at the start of the frame
t->m_32_bits_fmt = false; t->m_32_bits_fmt = false;
} }
@ -1185,7 +1273,9 @@ void GSTextureCache::IncAge()
t->m_TEX0.TBP0); t->m_TEX0.TBP0);
delete t; delete t;
} else { }
else
{
++i; ++i;
} }
} }
@ -1226,7 +1316,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
src->m_texture->SetScale(scale); src->m_texture->SetScale(scale);
src->m_end_block = dst->m_end_block; src->m_end_block = dst->m_end_block;
if (psm.pal > 0) { if (psm.pal > 0)
{
// Attach palette for GPU texture conversion // Attach palette for GPU texture conversion
AttachPaletteToSource(src, psm.pal, true); AttachPaletteToSource(src, psm.pal, true);
} }
@ -1265,18 +1356,21 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
int shader = dst->m_type != RenderTarget ? ShaderConvert_FLOAT32_TO_RGBA8 : ShaderConvert_COPY; int shader = dst->m_type != RenderTarget ? ShaderConvert_FLOAT32_TO_RGBA8 : ShaderConvert_COPY;
bool is_8bits = TEX0.PSM == PSM_PSMT8; bool is_8bits = TEX0.PSM == PSM_PSMT8;
if (is_8bits) { if (is_8bits)
{
GL_INS("Reading RT as a packed-indexed 8 bits format"); GL_INS("Reading RT as a packed-indexed 8 bits format");
shader = ShaderConvert_RGBA_TO_8I; shader = ShaderConvert_RGBA_TO_8I;
} }
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG
if (TEX0.PSM == PSM_PSMT4) { if (TEX0.PSM == PSM_PSMT4)
{
GL_INS("ERROR: Reading RT as a packed-indexed 4 bits format is not supported"); GL_INS("ERROR: Reading RT as a packed-indexed 4 bits format is not supported");
} }
#endif #endif
if (GSLocalMemory::m_psm[TEX0.PSM].bpp > 8) { if (GSLocalMemory::m_psm[TEX0.PSM].bpp > 8)
{
src->m_32_bits_fmt = dst->m_32_bits_fmt; src->m_32_bits_fmt = dst->m_32_bits_fmt;
} }
@ -1293,7 +1387,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
int w = (int)(dst->m_texture->GetScale().x * tw); int w = (int)(dst->m_texture->GetScale().x * tw);
int h = (int)(dst->m_texture->GetScale().y * th); int h = (int)(dst->m_texture->GetScale().y * th);
if (is_8bits) { if (is_8bits)
{
// Unscale 8 bits textures, quality won't be nice but format is really awful // Unscale 8 bits textures, quality won't be nice but format is really awful
w = tw; w = tw;
h = th; h = th;
@ -1388,7 +1483,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
// 2/ It doesn't support rescaling of the RT (tw = 1024) // 2/ It doesn't support rescaling of the RT (tw = 1024)
// Maybe it will be more easy to just round the UV value in the Vertex Shader // Maybe it will be more easy to just round the UV value in the Vertex Shader
if (!is_8bits) { if (!is_8bits)
{
// 8 bits handling is special due to unscaling. It is better to not execute this code // 8 bits handling is special due to unscaling. It is better to not execute this code
if (w > dstsize.x) if (w > dstsize.x)
{ {
@ -1416,7 +1512,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
// GH: by default (m_paltex == 0) GSdx converts texture to the 32 bit format // GH: by default (m_paltex == 0) GSdx converts texture to the 32 bit format
// However it is different here. We want to reuse a Render Target as a texture. // However it is different here. We want to reuse a Render Target as a texture.
// Because the texture is already on the GPU, CPU can't convert it. // Because the texture is already on the GPU, CPU can't convert it.
if (psm.pal > 0) { if (psm.pal > 0)
{
AttachPaletteToSource(src, psm.pal, true); AttachPaletteToSource(src, psm.pal, true);
} }
// Disable linear filtering for various GS post-processing effect // Disable linear filtering for various GS post-processing effect
@ -1447,13 +1544,16 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
if ((sRect == dRect).alltrue() && !shader) if ((sRect == dRect).alltrue() && !shader)
{ {
if (half_right) { if (half_right)
{
// You typically hit this code in snow engine game. Dstsize is the size of of Dx/GL RT // You typically hit this code in snow engine game. Dstsize is the size of of Dx/GL RT
// which is arbitrary set to 1280 (biggest RT used by GS). h/w are based on the input texture // which is arbitrary set to 1280 (biggest RT used by GS). h/w are based on the input texture
// so the only reliable way to find the real size of the target is to use the TBW value. // so the only reliable way to find the real size of the target is to use the TBW value.
float real_width = dst->m_TEX0.TBW * 64u * dst->m_texture->GetScale().x; float real_width = dst->m_TEX0.TBW * 64u * dst->m_texture->GetScale().x;
m_renderer->m_dev->CopyRect(sTex, dTex, GSVector4i((int)(real_width / 2.0f), 0, (int)real_width, h)); m_renderer->m_dev->CopyRect(sTex, dTex, GSVector4i((int)(real_width / 2.0f), 0, (int)real_width, h));
} else { }
else
{
m_renderer->m_dev->CopyRect(sTex, dTex, GSVector4i(0, 0, w, h)); // <= likely wrong dstsize.x could be bigger than w m_renderer->m_dev->CopyRect(sTex, dTex, GSVector4i(0, 0, w, h)); // <= likely wrong dstsize.x could be bigger than w
} }
} }
@ -1463,7 +1563,8 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
sRect.z /= sTex->GetWidth(); sRect.z /= sTex->GetWidth();
sRect.w /= sTex->GetHeight(); sRect.w /= sTex->GetHeight();
if (half_right) { if (half_right)
{
sRect.x = sRect.z / 2.0f; sRect.x = sRect.z / 2.0f;
} }
@ -1513,9 +1614,11 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
src->m_texture = m_renderer->m_dev->CreateTexture(tw, th, Get8bitFormat()); src->m_texture = m_renderer->m_dev->CreateTexture(tw, th, Get8bitFormat());
AttachPaletteToSource(src, psm.pal, true); AttachPaletteToSource(src, psm.pal, true);
} }
else { else
{
src->m_texture = m_renderer->m_dev->CreateTexture(tw, th); src->m_texture = m_renderer->m_dev->CreateTexture(tw, th);
if (psm.pal > 0) { if (psm.pal > 0)
{
AttachPaletteToSource(src, psm.pal, false); AttachPaletteToSource(src, psm.pal, false);
} }
} }
@ -1561,19 +1664,23 @@ void GSTextureCache::PrintMemoryUsage()
uint32 tex_rt = 0; uint32 tex_rt = 0;
uint32 rt = 0; uint32 rt = 0;
uint32 dss = 0; uint32 dss = 0;
for(auto s : m_src.m_surfaces) { for (auto s : m_src.m_surfaces)
if(s && !s->m_shared_texture) { {
if (s && !s->m_shared_texture)
{
if (s->m_target) if (s->m_target)
tex_rt += s->m_texture->GetMemUsage(); tex_rt += s->m_texture->GetMemUsage();
else else
tex += s->m_texture->GetMemUsage(); tex += s->m_texture->GetMemUsage();
} }
} }
for(auto t : m_dst[RenderTarget]) { for (auto t : m_dst[RenderTarget])
{
if (t) if (t)
rt += t->m_texture->GetMemUsage(); rt += t->m_texture->GetMemUsage();
} }
for(auto t : m_dst[DepthStencil]) { for (auto t : m_dst[DepthStencil])
{
if (t) if (t)
dss += t->m_texture->GetMemUsage(); dss += t->m_texture->GetMemUsage();
} }
@ -1640,15 +1747,17 @@ GSTextureCache::Source::Source(GSRenderer* r, const GIFRegTEX0& TEX0, const GIFR
m_TEX0 = TEX0; m_TEX0 = TEX0;
m_TEXA = TEXA; m_TEXA = TEXA;
if (dummy_container) { if (dummy_container)
{
// Dummy container only contain a m_texture that is a pointer to another source. // Dummy container only contain a m_texture that is a pointer to another source.
m_write.rect = NULL; m_write.rect = NULL;
m_write.count = 0; m_write.count = 0;
m_repeating = false; m_repeating = false;
}
} else { else
{
memset(m_layer_TEX0, 0, sizeof(m_layer_TEX0)); memset(m_layer_TEX0, 0, sizeof(m_layer_TEX0));
memset(m_valid, 0, sizeof(m_valid)); memset(m_valid, 0, sizeof(m_valid));
@ -1887,7 +1996,8 @@ void GSTextureCache::Source::Flush(uint32 count, int layer)
m_write.count -= count; m_write.count -= count;
} }
bool GSTextureCache::Source::ClutMatch(PaletteKey palette_key) { bool GSTextureCache::Source::ClutMatch(PaletteKey palette_key)
{
return PaletteKeyEqual()(palette_key, m_palette_obj->GetPaletteKey()); return PaletteKeyEqual()(palette_key, m_palette_obj->GetPaletteKey());
} }
@ -1927,15 +2037,19 @@ void GSTextureCache::Target::Update()
GSVector4i r = m_dirty.GetDirtyRectAndClear(m_TEX0, t_size); GSVector4i r = m_dirty.GetDirtyRectAndClear(m_TEX0, t_size);
if (r.rempty()) return; if (r.rempty())
return;
// No handling please // No handling please
if ((m_type == DepthStencil) && !m_depth_supported) { if ((m_type == DepthStencil) && !m_depth_supported)
{
// do the most likely thing a direct write would do, clear it // do the most likely thing a direct write would do, clear it
GL_INS("ERROR: Update DepthStencil dummy"); GL_INS("ERROR: Update DepthStencil dummy");
return; return;
} else if (m_type == DepthStencil && m_renderer->m_game.title == CRC::FFX2) { }
else if (m_type == DepthStencil && m_renderer->m_game.title == CRC::FFX2)
{
GL_INS("ERROR: bad invalidation detected, depth buffer will be cleared"); GL_INS("ERROR: bad invalidation detected, depth buffer will be cleared");
// FFX2 menu. Invalidation of the depth is wrongly done and only the first // FFX2 menu. Invalidation of the depth is wrongly done and only the first
// page is invalidated. Technically a CRC hack will be better but I don't expect // page is invalidated. Technically a CRC hack will be better but I don't expect
@ -2052,7 +2166,8 @@ void GSTextureCache::SourceMap::Add(Source* s, const GIFRegTEX0& TEX0, GSOffset*
void GSTextureCache::SourceMap::RemoveAll() void GSTextureCache::SourceMap::RemoveAll()
{ {
for (auto s : m_surfaces) delete s; for (auto s : m_surfaces)
delete s;
m_surfaces.clear(); m_surfaces.clear();
@ -2118,26 +2233,32 @@ GSTextureCache::Palette::Palette(const GSRenderer* renderer, uint16 pal, bool ne
uint16 palette_size = pal * sizeof(uint32); uint16 palette_size = pal * sizeof(uint32);
m_clut = (uint32*)_aligned_malloc(palette_size, 64); m_clut = (uint32*)_aligned_malloc(palette_size, 64);
memcpy(m_clut, (const uint32*)m_renderer->m_mem.m_clut, palette_size); memcpy(m_clut, (const uint32*)m_renderer->m_mem.m_clut, palette_size);
if (need_gs_texture) { if (need_gs_texture)
{
InitializeTexture(); InitializeTexture();
} }
} }
GSTextureCache::Palette::~Palette() { GSTextureCache::Palette::~Palette()
{
m_renderer->m_dev->Recycle(m_tex_palette); m_renderer->m_dev->Recycle(m_tex_palette);
_aligned_free(m_clut); _aligned_free(m_clut);
} }
GSTexture* GSTextureCache::Palette::GetPaletteGSTexture() { GSTexture* GSTextureCache::Palette::GetPaletteGSTexture()
{
return m_tex_palette; return m_tex_palette;
} }
GSTextureCache::PaletteKey GSTextureCache::Palette::GetPaletteKey() { GSTextureCache::PaletteKey GSTextureCache::Palette::GetPaletteKey()
{
return {m_clut, m_pal}; return {m_clut, m_pal};
} }
void GSTextureCache::Palette::InitializeTexture() { void GSTextureCache::Palette::InitializeTexture()
if (!m_tex_palette) { {
if (!m_tex_palette)
{
// A palette texture is always created with dimensions 256x1 (also in the case that m_pal is 16, thus a 16x1 texture // A palette texture is always created with dimensions 256x1 (also in the case that m_pal is 16, thus a 16x1 texture
// would be enough to store the CLUT data) because the coordinates that the shader uses for // would be enough to store the CLUT data) because the coordinates that the shader uses for
// sampling such texture are always normalized by 255. // sampling such texture are always normalized by 255.
@ -2157,14 +2278,16 @@ void GSTextureCache::Palette::InitializeTexture() {
// it is computed in 16 passes, // it is computed in 16 passes,
// 2) The clut can contain many 0s, so as a way to increase the spread of hashing values for small changes in the input clut the hashing function // 2) The clut can contain many 0s, so as a way to increase the spread of hashing values for small changes in the input clut the hashing function
// is using addition in combination with logical XOR operator; The addition constants are large prime numbers, which may help in achieving what intended. // is using addition in combination with logical XOR operator; The addition constants are large prime numbers, which may help in achieving what intended.
std::size_t GSTextureCache::PaletteKeyHash::operator()(const PaletteKey &key) const { std::size_t GSTextureCache::PaletteKeyHash::operator()(const PaletteKey& key) const
{
uint16 pal = key.pal; uint16 pal = key.pal;
const uint32* clut = key.clut; const uint32* clut = key.clut;
ASSERT((pal & 15) == 0); ASSERT((pal & 15) == 0);
size_t clut_hash = 3831179159; size_t clut_hash = 3831179159;
for (uint16 i = 0; i < pal; i += 16) { for (uint16 i = 0; i < pal; i += 16)
{
clut_hash = (clut_hash + 1488000301) ^ (clut[i ] + 33644011); clut_hash = (clut_hash + 1488000301) ^ (clut[i ] + 33644011);
clut_hash = (clut_hash + 3831179159) ^ (clut[i + 1] + 47627467); clut_hash = (clut_hash + 3831179159) ^ (clut[i + 1] + 47627467);
clut_hash = (clut_hash + 3659574209) ^ (clut[i + 2] + 577038523); clut_hash = (clut_hash + 3659574209) ^ (clut[i + 2] + 577038523);
@ -2190,8 +2313,10 @@ std::size_t GSTextureCache::PaletteKeyHash::operator()(const PaletteKey &key) co
// GSTextureCache::PaletteKeyEqual // GSTextureCache::PaletteKeyEqual
bool GSTextureCache::PaletteKeyEqual::operator()(const PaletteKey &lhs, const PaletteKey &rhs) const { bool GSTextureCache::PaletteKeyEqual::operator()(const PaletteKey& lhs, const PaletteKey& rhs) const
if (lhs.pal != rhs.pal) { {
if (lhs.pal != rhs.pal)
{
return false; return false;
} }
@ -2203,12 +2328,14 @@ bool GSTextureCache::PaletteKeyEqual::operator()(const PaletteKey &lhs, const Pa
GSTextureCache::PaletteMap::PaletteMap(const GSRenderer* renderer) GSTextureCache::PaletteMap::PaletteMap(const GSRenderer* renderer)
: m_renderer(renderer) : m_renderer(renderer)
{ {
for (auto& map : m_maps) { for (auto& map : m_maps)
{
map.reserve(MAX_SIZE); map.reserve(MAX_SIZE);
} }
} }
std::shared_ptr<GSTextureCache::Palette> GSTextureCache::PaletteMap::LookupPalette(uint16 pal, bool need_gs_texture) { std::shared_ptr<GSTextureCache::Palette> GSTextureCache::PaletteMap::LookupPalette(uint16 pal, bool need_gs_texture)
{
ASSERT(pal == 16 || pal == 256); ASSERT(pal == 16 || pal == 256);
// Choose which hash map search into: // Choose which hash map search into:
@ -2223,9 +2350,11 @@ std::shared_ptr<GSTextureCache::Palette> GSTextureCache::PaletteMap::LookupPalet
auto it1 = map.find(palette_key); auto it1 = map.find(palette_key);
if (it1 != map.end()) { if (it1 != map.end())
{
// Clut content match, HIT // Clut content match, HIT
if (need_gs_texture && !it1->second->GetPaletteGSTexture()) { if (need_gs_texture && !it1->second->GetPaletteGSTexture())
{
// Generate GSTexture and upload clut content if needed and not done yet // Generate GSTexture and upload clut content if needed and not done yet
it1->second->InitializeTexture(); it1->second->InitializeTexture();
} }
@ -2234,31 +2363,37 @@ std::shared_ptr<GSTextureCache::Palette> GSTextureCache::PaletteMap::LookupPalet
// No palette with matching clut content, MISS // No palette with matching clut content, MISS
if (map.size() > MAX_SIZE) { if (map.size() > MAX_SIZE)
{
// If the map is too big, try to clean it by disposing and removing unused palettes, before adding the new one // If the map is too big, try to clean it by disposing and removing unused palettes, before adding the new one
GL_INS("WARNING, %u-bit PaletteMap (Size %u): Max size %u exceeded, clearing unused palettes.", pal * sizeof(uint32), map.size(), MAX_SIZE); GL_INS("WARNING, %u-bit PaletteMap (Size %u): Max size %u exceeded, clearing unused palettes.", pal * sizeof(uint32), map.size(), MAX_SIZE);
uint32 current_size = map.size(); uint32 current_size = map.size();
for (auto it = map.begin(); it != map.end(); ) { for (auto it = map.begin(); it != map.end();)
{
// If the palette is unused, there is only one shared pointers holding a reference to the unused Palette object, // If the palette is unused, there is only one shared pointers holding a reference to the unused Palette object,
// and this shared pointer is the one stored in the map itself // and this shared pointer is the one stored in the map itself
if (it->second.use_count() <= 1) { if (it->second.use_count() <= 1)
{
// Palette is unused // Palette is unused
it = map.erase(it); // Erase element from map it = map.erase(it); // Erase element from map
// The palette object should now be gone as the shared pointer to the object in the map is deleted // The palette object should now be gone as the shared pointer to the object in the map is deleted
} }
else { else
{
++it; ++it;
} }
} }
uint32 cleared_palette_count = current_size - (uint32)map.size(); uint32 cleared_palette_count = current_size - (uint32)map.size();
if (cleared_palette_count == 0) { if (cleared_palette_count == 0)
{
GL_INS("ERROR, %u-bit PaletteMap (Size %u): Max size %u exceeded, could not clear any palette, negative performance impact.", pal * sizeof(uint32), map.size(), MAX_SIZE); GL_INS("ERROR, %u-bit PaletteMap (Size %u): Max size %u exceeded, could not clear any palette, negative performance impact.", pal * sizeof(uint32), map.size(), MAX_SIZE);
} }
else { else
{
map.reserve(MAX_SIZE); // Ensure map capacity is not modified by the clearing map.reserve(MAX_SIZE); // Ensure map capacity is not modified by the clearing
GL_INS("INFO, %u-bit PaletteMap (Size %u): Cleared %u palettes.", pal * sizeof(uint32), map.size(), cleared_palette_count); GL_INS("INFO, %u-bit PaletteMap (Size %u): Cleared %u palettes.", pal * sizeof(uint32), map.size(), cleared_palette_count);
} }
@ -2273,10 +2408,11 @@ std::shared_ptr<GSTextureCache::Palette> GSTextureCache::PaletteMap::LookupPalet
return palette; return palette;
} }
void GSTextureCache::PaletteMap::Clear() { void GSTextureCache::PaletteMap::Clear()
for (auto& map : m_maps) { {
for (auto& map : m_maps)
{
map.clear(); // Clear all the nodes of the map, deleting Palette objects managed by shared pointers as they should be unused elsewhere map.clear(); // Clear all the nodes of the map, deleting Palette objects managed by shared pointers as they should be unused elsewhere
map.reserve(MAX_SIZE); // Ensure map capacity is not modified by the clearing map.reserve(MAX_SIZE); // Ensure map capacity is not modified by the clearing
} }
} }

30
plugins/GSdx/Renderers/HW/GSTextureCache.h
View File

@ -28,7 +28,11 @@
class GSTextureCache class GSTextureCache
{ {
public: public:
enum {RenderTarget, DepthStencil}; enum
{
RenderTarget,
DepthStencil
};
class Surface : public GSAlignedClass<32> class Surface : public GSAlignedClass<32>
{ {
@ -54,7 +58,8 @@ public:
bool Overlaps(uint32 bp, uint32 bw, uint32 psm, const GSVector4i& rect); bool Overlaps(uint32 bp, uint32 bw, uint32 psm, const GSVector4i& rect);
}; };
struct PaletteKey { struct PaletteKey
{
const uint32* clut; const uint32* clut;
uint16 pal; uint16 pal;
}; };
@ -86,19 +91,25 @@ public:
void InitializeTexture(); void InitializeTexture();
}; };
struct PaletteKeyHash { struct PaletteKeyHash
{
// Calculate hash // Calculate hash
std::size_t operator()(const PaletteKey& key) const; std::size_t operator()(const PaletteKey& key) const;
}; };
struct PaletteKeyEqual { struct PaletteKeyEqual
{
// Compare pal value and clut contents // Compare pal value and clut contents
bool operator()(const PaletteKey& lhs, const PaletteKey& rhs) const; bool operator()(const PaletteKey& lhs, const PaletteKey& rhs) const;
}; };
class Source : public Surface class Source : public Surface
{ {
struct {GSVector4i* rect; uint32 count;} m_write; struct
{
GSVector4i* rect;
uint32 count;
} m_write;
void Write(const GSVector4i& r, int layer); void Write(const GSVector4i& r, int layer);
void Flush(uint32 count, int layer); void Flush(uint32 count, int layer);
@ -178,7 +189,11 @@ public:
uint32 m_pages[16]; // bitmap of all pages uint32 m_pages[16]; // bitmap of all pages
bool m_used; bool m_used;
SourceMap() : m_used(false) {memset(m_pages, 0, sizeof(m_pages));} SourceMap()
: m_used(false)
{
memset(m_pages, 0, sizeof(m_pages));
}
void Add(Source* s, const GIFRegTEX0& TEX0, GSOffset* off); void Add(Source* s, const GIFRegTEX0& TEX0, GSOffset* off);
void RemoveAll(); void RemoveAll();
@ -249,7 +264,8 @@ public:
bool ShallSearchTextureInsideRt(); bool ShallSearchTextureInsideRt();
const char* to_string(int type) { const char* to_string(int type)
{
return (type == DepthStencil) ? "Depth" : "Color"; return (type == DepthStencil) ? "Depth" : "Color";
} }

7
plugins/GSdx/Renderers/HW/GSVertexHW.h
View File

@ -34,7 +34,12 @@ struct alignas(32) GSVertexHW9
// t.z = union {struct {uint8 r, g, b, a;}; uint32 c0;}; // t.z = union {struct {uint8 r, g, b, a;}; uint32 c0;};
// t.w = union {struct {uint8 ta0, ta1, res, f;}; uint32 c1;} // t.w = union {struct {uint8 ta0, ta1, res, f;}; uint32 c1;}
GSVertexHW9& operator = (GSVertexHW9& v) {t = v.t; p = v.p; return *this;} GSVertexHW9& operator=(GSVertexHW9& v)
{
t = v.t;
p = v.p;
return *this;
}
}; };
#pragma pack(pop) #pragma pack(pop)

1
plugins/GSdx/Renderers/Null/GSDeviceNull.cpp
View File

@ -41,4 +41,3 @@ GSTexture* GSDeviceNull::CreateSurface(int type, int w, int h, int format)
{ {
return new GSTextureNull(type, w, h, format); return new GSTextureNull(type, w, h, format);
} }

1
plugins/GSdx/Renderers/Null/GSDeviceNull.h
View File

@ -39,4 +39,3 @@ public:
bool Create(const std::shared_ptr<GSWnd>& wnd); bool Create(const std::shared_ptr<GSWnd>& wnd);
bool Reset(int w, int h); bool Reset(int w, int h);
}; };

5
plugins/GSdx/Renderers/Null/GSRendererNull.h
View File

@ -28,7 +28,10 @@ class GSRendererNull : public GSRenderer
class GSVertexTraceNull : public GSVertexTrace class GSVertexTraceNull : public GSVertexTrace
{ {
public: public:
GSVertexTraceNull(const GSState* state) : GSVertexTrace(state) {} GSVertexTraceNull(const GSState* state)
: GSVertexTrace(state)
{
}
}; };
protected: protected:

5
plugins/GSdx/Renderers/Null/GSTextureNull.h
View File

@ -25,7 +25,10 @@
class GSTextureNull : public GSTexture class GSTextureNull : public GSTexture
{ {
struct {int type, w, h, format;} m_desc; struct
{
int type, w, h, format;
} m_desc;
public: public:
GSTextureNull(); GSTextureNull();

10
plugins/GSdx/Renderers/OpenGL/GLLoader.h
View File

@ -121,12 +121,14 @@ extern PFNGLTEXTUREPAGECOMMITMENTEXTPROC glTexturePageCommitmentEXT;
#include "PFN_GLLOADER_HPP.h" #include "PFN_GLLOADER_HPP.h"
namespace GLExtension { namespace GLExtension
{
extern bool Has(const std::string& ext); extern bool Has(const std::string& ext);
extern void Set(const std::string& ext, bool v = true); extern void Set(const std::string& ext, bool v = true);
} } // namespace GLExtension
namespace GLLoader { namespace GLLoader
{
void check_gl_requirements(); void check_gl_requirements();
extern bool vendor_id_amd; extern bool vendor_id_amd;
@ -145,4 +147,4 @@ namespace GLLoader {
extern bool found_compatible_GL_ARB_sparse_texture2; extern bool found_compatible_GL_ARB_sparse_texture2;
extern bool found_compatible_sparse_depth; extern bool found_compatible_sparse_depth;
} } // namespace GLLoader

8
plugins/GSdx/Renderers/OpenGL/GLState.cpp
View File

@ -22,7 +22,8 @@
#include "stdafx.h" #include "stdafx.h"
#include "GLState.h" #include "GLState.h"
namespace GLState { namespace GLState
{
GLuint fbo; GLuint fbo;
GSVector2i viewport; GSVector2i viewport;
GSVector4i scissor; GSVector4i scissor;
@ -59,7 +60,8 @@ namespace GLState {
int64 available_vram; int64 available_vram;
void Clear() { void Clear()
{
fbo = 0; fbo = 0;
viewport = GSVector2i(0, 0); viewport = GSVector2i(0, 0);
scissor = GSVector4i(0, 0, 0, 0); scissor = GSVector4i(0, 0, 0, 0);
@ -100,4 +102,4 @@ namespace GLState {
// (256MB are reserved for PBO/IBO/VBO/UBO buffers) // (256MB are reserved for PBO/IBO/VBO/UBO buffers)
available_vram = (4096u - 256u) * 1024u * 1024u; available_vram = (4096u - 256u) * 1024u * 1024u;
} }
} } // namespace GLState

5
plugins/GSdx/Renderers/OpenGL/GLState.h
View File

@ -24,7 +24,8 @@
#include "GSdx.h" #include "GSdx.h"
#include "GSVector.h" #include "GSVector.h"
namespace GLState { namespace GLState
{
extern GLuint fbo; // frame buffer object extern GLuint fbo; // frame buffer object
extern GSVector2i viewport; extern GSVector2i viewport;
extern GSVector4i scissor; extern GSVector4i scissor;
@ -62,4 +63,4 @@ namespace GLState {
extern int64 available_vram; extern int64 available_vram;
extern void Clear(); extern void Clear();
} } // namespace GLState

355
plugins/GSdx/Renderers/OpenGL/GSDeviceOGL.cpp
View File

@ -96,7 +96,8 @@ GSDeviceOGL::GSDeviceOGL()
GSDeviceOGL::~GSDeviceOGL() GSDeviceOGL::~GSDeviceOGL()
{ {
if (m_debug_gl_file) { if (m_debug_gl_file)
{
fclose(m_debug_gl_file); fclose(m_debug_gl_file);
m_debug_gl_file = NULL; m_debug_gl_file = NULL;
} }
@ -143,7 +144,8 @@ GSDeviceOGL::~GSDeviceOGL()
glDeleteSamplers(countof(m_ps_ss), m_ps_ss); glDeleteSamplers(countof(m_ps_ss), m_ps_ss);
for (uint32 key = 0; key < countof(m_om_dss); key++) delete m_om_dss[key]; for (uint32 key = 0; key < countof(m_om_dss); key++)
delete m_om_dss[key];
PboPool::Destroy(); PboPool::Destroy();
@ -154,14 +156,16 @@ GSDeviceOGL::~GSDeviceOGL()
void GSDeviceOGL::GenerateProfilerData() void GSDeviceOGL::GenerateProfilerData()
{ {
if (m_profiler.last_query < 3) { if (m_profiler.last_query < 3)
{
glDeleteQueries(1 << 16, m_profiler.timer_query); glDeleteQueries(1 << 16, m_profiler.timer_query);
return; return;
} }
// Wait latest quey to get valid result // Wait latest quey to get valid result
GLuint available = 0; GLuint available = 0;
while (!available) { while (!available)
{
glGetQueryObjectuiv(m_profiler.timer(), GL_QUERY_RESULT_AVAILABLE, &available); glGetQueryObjectuiv(m_profiler.timer(), GL_QUERY_RESULT_AVAILABLE, &available);
} }
@ -174,7 +178,8 @@ void GSDeviceOGL::GenerateProfilerData()
int first_query = replay > 1 ? m_profiler.last_query / replay : 0; int first_query = replay > 1 ? m_profiler.last_query / replay : 0;
glGetQueryObjectui64v(m_profiler.timer_query[first_query], GL_QUERY_RESULT, &time_start); glGetQueryObjectui64v(m_profiler.timer_query[first_query], GL_QUERY_RESULT, &time_start);
for (uint32 q = first_query + 1; q < m_profiler.last_query; q++) { for (uint32 q = first_query + 1; q < m_profiler.last_query; q++)
{
glGetQueryObjectui64v(m_profiler.timer_query[q], GL_QUERY_RESULT, &time_end); glGetQueryObjectui64v(m_profiler.timer_query[q], GL_QUERY_RESULT, &time_end);
uint64 t = time_end - time_start; uint64 t = time_end - time_start;
times.push_back((double)t * ms); times.push_back((double)t * ms);
@ -193,16 +198,20 @@ void GSDeviceOGL::GenerateProfilerData()
auto minmax_time = std::minmax_element(times.begin(), times.end()); auto minmax_time = std::minmax_element(times.begin(), times.end());
for (auto t : times) mean += t; for (auto t : times)
mean += t;
mean = mean / frames; mean = mean / frames;
for (auto t : times) sd += pow(t-mean, 2); for (auto t : times)
sd += pow(t - mean, 2);
sd = sqrt(sd / frames); sd = sqrt(sd / frames);
uint32 time_repartition[16] = {0}; uint32 time_repartition[16] = {0};
for (auto t : times) { for (auto t : times)
{
uint32 slot = (uint32)(t / 2.0); uint32 slot = (uint32)(t / 2.0);
if (slot >= countof(time_repartition)) { if (slot >= countof(time_repartition))
{
slot = countof(time_repartition) - 1; slot = countof(time_repartition) - 1;
} }
time_repartition[slot]++; time_repartition[slot]++;
@ -215,13 +224,16 @@ void GSDeviceOGL::GenerateProfilerData()
fprintf(stderr, "SD %4.2f ms\n", sd); fprintf(stderr, "SD %4.2f ms\n", sd);
fprintf(stderr, "\n"); fprintf(stderr, "\n");
fprintf(stderr, "Frame Repartition\n"); fprintf(stderr, "Frame Repartition\n");
for (uint32 i = 0; i < countof(time_repartition); i ++) { for (uint32 i = 0; i < countof(time_repartition); i++)
{
fprintf(stderr, "%3u ms => %3u ms\t%4u\n", 2 * i, 2 * (i + 1), time_repartition[i]); fprintf(stderr, "%3u ms => %3u ms\t%4u\n", 2 * i, 2 * (i + 1), time_repartition[i]);
} }
FILE* csv = fopen("GSdx_profile.csv", "w"); FILE* csv = fopen("GSdx_profile.csv", "w");
if (csv) { if (csv)
for (size_t i = 0; i < times.size(); i++) { {
for (size_t i = 0; i < times.size(); i++)
{
fprintf(csv, "%zu,%lf\n", i, times[i]); fprintf(csv, "%zu,%lf\n", i, times[i]);
} }
@ -239,7 +251,8 @@ GSTexture* GSDeviceOGL::CreateSurface(int type, int w, int h, int fmt)
// NOTE: I'm not sure RenderTarget always need to be cleared. It could be costly for big upscale. // NOTE: I'm not sure RenderTarget always need to be cleared. It could be costly for big upscale.
// FIXME: it will be more logical to do it in FetchSurface. This code is only called at first creation // FIXME: it will be more logical to do it in FetchSurface. This code is only called at first creation
// of the texture. However we could reuse a deleted texture. // of the texture. However we could reuse a deleted texture.
if (m_force_texture_clear == 0) { if (m_force_texture_clear == 0)
{
// Clear won't be done if the texture isn't committed. Commit the full texture to ensure // Clear won't be done if the texture isn't committed. Commit the full texture to ensure
// correct behavior of force clear option (debug option) // correct behavior of force clear option (debug option)
t->Commit(); t->Commit();
@ -267,7 +280,8 @@ GSTexture* GSDeviceOGL::FetchSurface(int type, int w, int h, int format)
GSTexture* t = GSDevice::FetchSurface(type, w, h, format); GSTexture* t = GSDevice::FetchSurface(type, w, h, format);
if (m_force_texture_clear) { if (m_force_texture_clear)
{
// Clear won't be done if the texture isn't committed. Commit the full texture to ensure // Clear won't be done if the texture isn't committed. Commit the full texture to ensure
// correct behavior of force clear option (debug option) // correct behavior of force clear option (debug option)
t->Commit(); t->Commit();
@ -302,7 +316,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
// Debug helper // Debug helper
// **************************************************************** // ****************************************************************
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG
if (theApp.GetConfigB("debug_opengl")) { if (theApp.GetConfigB("debug_opengl"))
{
glDebugMessageCallback((GLDEBUGPROC)DebugOutputToFile, NULL); glDebugMessageCallback((GLDEBUGPROC)DebugOutputToFile, NULL);
glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS_ARB); glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS_ARB);
@ -340,7 +355,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0); glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
// Some timers to help profiling // Some timers to help profiling
if (GLLoader::in_replayer) { if (GLLoader::in_replayer)
{
glCreateQueries(GL_TIMESTAMP, 1 << 16, m_profiler.timer_query); glCreateQueries(GL_TIMESTAMP, 1 << 16, m_profiler.timer_query);
} }
} }
@ -371,7 +387,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
{ {
GL_PUSH("GSDeviceOGL::Sampler"); GL_PUSH("GSDeviceOGL::Sampler");
for (uint32 key = 0; key < countof(m_ps_ss); key++) { for (uint32 key = 0; key < countof(m_ps_ss); key++)
{
m_ps_ss[key] = CreateSampler(PSSamplerSelector(key)); m_ps_ss[key] = CreateSampler(PSSamplerSelector(key));
} }
} }
@ -396,7 +413,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
vs = m_shader->Compile("convert.glsl", "vs_main", GL_VERTEX_SHADER, shader.data()); vs = m_shader->Compile("convert.glsl", "vs_main", GL_VERTEX_SHADER, shader.data());
m_convert.vs = vs; m_convert.vs = vs;
for(size_t i = 0; i < countof(m_convert.ps); i++) { for (size_t i = 0; i < countof(m_convert.ps); i++)
{
ps = m_shader->Compile("convert.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data()); ps = m_shader->Compile("convert.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data());
std::string pretty_name = "Convert pipe " + std::to_string(i); std::string pretty_name = "Convert pipe " + std::to_string(i);
m_convert.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps); m_convert.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps);
@ -425,7 +443,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
theApp.LoadResource(IDR_MERGE_GLSL, shader); theApp.LoadResource(IDR_MERGE_GLSL, shader);
for(size_t i = 0; i < countof(m_merge_obj.ps); i++) { for (size_t i = 0; i < countof(m_merge_obj.ps); i++)
{
ps = m_shader->Compile("merge.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data()); ps = m_shader->Compile("merge.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data());
std::string pretty_name = "Merge pipe " + std::to_string(i); std::string pretty_name = "Merge pipe " + std::to_string(i);
m_merge_obj.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps); m_merge_obj.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps);
@ -442,7 +461,8 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
theApp.LoadResource(IDR_INTERLACE_GLSL, shader); theApp.LoadResource(IDR_INTERLACE_GLSL, shader);
for(size_t i = 0; i < countof(m_interlace.ps); i++) { for (size_t i = 0; i < countof(m_interlace.ps); i++)
{
ps = m_shader->Compile("interlace.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data()); ps = m_shader->Compile("interlace.glsl", format("ps_main%d", i), GL_FRAGMENT_SHADER, shader.data());
std::string pretty_name = "Interlace pipe " + std::to_string(i); std::string pretty_name = "Interlace pipe " + std::to_string(i);
m_interlace.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps); m_interlace.ps[i] = m_shader->LinkPipeline(pretty_name, vs, 0, ps);
@ -533,14 +553,19 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
// Get Available Memory // Get Available Memory
// **************************************************************** // ****************************************************************
GLint vram[4] = {0}; GLint vram[4] = {0};
if (GLLoader::vendor_id_amd) { if (GLLoader::vendor_id_amd)
{
// Full vram, remove a small margin for others buffer // Full vram, remove a small margin for others buffer
glGetIntegerv(GL_TEXTURE_FREE_MEMORY_ATI, vram); glGetIntegerv(GL_TEXTURE_FREE_MEMORY_ATI, vram);
} else if (GLExtension::Has("GL_NVX_gpu_memory_info")) { }
else if (GLExtension::Has("GL_NVX_gpu_memory_info"))
{
// GL_GPU_MEMORY_INFO_TOTAL_AVAILABLE_MEMORY_NVX <= give full memory // GL_GPU_MEMORY_INFO_TOTAL_AVAILABLE_MEMORY_NVX <= give full memory
// Available vram // Available vram
glGetIntegerv(GL_GPU_MEMORY_INFO_CURRENT_AVAILABLE_VIDMEM_NVX, vram); glGetIntegerv(GL_GPU_MEMORY_INFO_CURRENT_AVAILABLE_VIDMEM_NVX, vram);
} else { }
else
{
fprintf(stdout, "No extenstion supported to get available memory. Use default value !\n"); fprintf(stdout, "No extenstion supported to get available memory. Use default value !\n");
} }
@ -558,8 +583,7 @@ bool GSDeviceOGL::Create(const std::shared_ptr<GSWnd> &wnd)
GSVector2i tex_font = m_osd.get_texture_font_size(); GSVector2i tex_font = m_osd.get_texture_font_size();
m_font = std::unique_ptr<GSTexture>( m_font = std::unique_ptr<GSTexture>(
new GSTextureOGL(GSTextureOGL::Texture, tex_font.x, tex_font.y, GL_R8, m_fbo_read, false) new GSTextureOGL(GSTextureOGL::Texture, tex_font.x, tex_font.y, GL_R8, m_fbo_read, false));
);
// **************************************************************** // ****************************************************************
// Finish window setup and backbuffer // Finish window setup and backbuffer
@ -608,7 +632,8 @@ void GSDeviceOGL::CreateTextureFX()
// enough but buffer is polluted with noise. Clear will be limited // enough but buffer is polluted with noise. Clear will be limited
// to the mask. // to the mask.
glStencilMask(0xFF); glStencilMask(0xFF);
for (uint32 key = 0; key < countof(m_om_dss); key++) { for (uint32 key = 0; key < countof(m_om_dss); key++)
{
m_om_dss[key] = CreateDepthStencil(OMDepthStencilSelector(key)); m_om_dss[key] = CreateDepthStencil(OMDepthStencilSelector(key));
} }
@ -638,7 +663,8 @@ void GSDeviceOGL::Flip()
{ {
m_wnd->Flip(); m_wnd->Flip();
if (GLLoader::in_replayer) { if (GLLoader::in_replayer)
{
glQueryCounter(m_profiler.timer(), GL_TIMESTAMP); glQueryCounter(m_profiler.timer(), GL_TIMESTAMP);
m_profiler.last_query++; m_profiler.last_query++;
} }
@ -686,7 +712,8 @@ void GSDeviceOGL::DrawIndexedPrimitive(int offset, int count)
void GSDeviceOGL::ClearRenderTarget(GSTexture* t, const GSVector4& c) void GSDeviceOGL::ClearRenderTarget(GSTexture* t, const GSVector4& c)
{ {
if (!t) return; if (!t)
return;
GSTextureOGL* T = static_cast<GSTextureOGL*>(t); GSTextureOGL* T = static_cast<GSTextureOGL*>(t);
if (T->HasBeenCleaned() && !T->IsBackbuffer()) if (T->HasBeenCleaned() && !T->IsBackbuffer())
@ -707,18 +734,20 @@ void GSDeviceOGL::ClearRenderTarget(GSTexture* t, const GSVector4& c)
uint32 old_color_mask = GLState::wrgba; uint32 old_color_mask = GLState::wrgba;
OMSetColorMaskState(); OMSetColorMaskState();
if (T->IsBackbuffer()) { if (T->IsBackbuffer())
{
OMSetFBO(0); OMSetFBO(0);
// glDrawBuffer(GL_BACK); // this is the default when there is no FB // glDrawBuffer(GL_BACK); // this is the default when there is no FB
// 0 will select the first drawbuffer ie GL_BACK // 0 will select the first drawbuffer ie GL_BACK
glClearBufferfv(GL_COLOR, 0, c.v); glClearBufferfv(GL_COLOR, 0, c.v);
} else { }
else
{
OMSetFBO(m_fbo); OMSetFBO(m_fbo);
OMAttachRt(T); OMAttachRt(T);
glClearBufferfv(GL_COLOR, 0, c.v); glClearBufferfv(GL_COLOR, 0, c.v);
} }
OMSetColorMaskState(OMColorMaskSelector(old_color_mask)); OMSetColorMaskState(OMColorMaskSelector(old_color_mask));
@ -730,7 +759,8 @@ void GSDeviceOGL::ClearRenderTarget(GSTexture* t, const GSVector4& c)
void GSDeviceOGL::ClearRenderTarget(GSTexture* t, uint32 c) void GSDeviceOGL::ClearRenderTarget(GSTexture* t, uint32 c)
{ {
if (!t) return; if (!t)
return;
GSVector4 color = GSVector4::rgba32(c) * (1.0f / 255); GSVector4 color = GSVector4::rgba32(c) * (1.0f / 255);
ClearRenderTarget(t, color); ClearRenderTarget(t, color);
@ -738,13 +768,15 @@ void GSDeviceOGL::ClearRenderTarget(GSTexture* t, uint32 c)
void GSDeviceOGL::ClearDepth(GSTexture* t) void GSDeviceOGL::ClearDepth(GSTexture* t)
{ {
if (!t) return; if (!t)
return;
GSTextureOGL* T = static_cast<GSTextureOGL*>(t); GSTextureOGL* T = static_cast<GSTextureOGL*>(t);
GL_PUSH("Clear Depth %d", T->GetID()); GL_PUSH("Clear Depth %d", T->GetID());
if (0 && GLLoader::found_GL_ARB_clear_texture) { if (0 && GLLoader::found_GL_ARB_clear_texture)
{
// I don't know what the driver does but it creates // I don't know what the driver does but it creates
// some slowdowns on Harry Potter PS // some slowdowns on Harry Potter PS
// Maybe it triggers some texture relocations, or maybe // Maybe it triggers some texture relocations, or maybe
@ -754,7 +786,9 @@ void GSDeviceOGL::ClearDepth(GSTexture* t)
// Don't bother with Depth_Stencil insanity // Don't bother with Depth_Stencil insanity
T->Clear(NULL); T->Clear(NULL);
} else { }
else
{
OMSetFBO(m_fbo); OMSetFBO(m_fbo);
// RT must be detached, if RT is too small, depth won't be fully cleared // RT must be detached, if RT is too small, depth won't be fully cleared
// AT tolenico 2 map clip bug // AT tolenico 2 map clip bug
@ -764,9 +798,12 @@ void GSDeviceOGL::ClearDepth(GSTexture* t)
// TODO: check size of scissor before toggling it // TODO: check size of scissor before toggling it
glDisable(GL_SCISSOR_TEST); glDisable(GL_SCISSOR_TEST);
float c = 0.0f; float c = 0.0f;
if (GLState::depth_mask) { if (GLState::depth_mask)
{
glClearBufferfv(GL_DEPTH, 0, &c); glClearBufferfv(GL_DEPTH, 0, &c);
} else { }
else
{
glDepthMask(true); glDepthMask(true);
glClearBufferfv(GL_DEPTH, 0, &c); glClearBufferfv(GL_DEPTH, 0, &c);
glDepthMask(false); glDepthMask(false);
@ -777,7 +814,8 @@ void GSDeviceOGL::ClearDepth(GSTexture* t)
void GSDeviceOGL::ClearStencil(GSTexture* t, uint8 c) void GSDeviceOGL::ClearStencil(GSTexture* t, uint8 c)
{ {
if (!t) return; if (!t)
return;
GSTextureOGL* T = static_cast<GSTextureOGL*>(t); GSTextureOGL* T = static_cast<GSTextureOGL*>(t);
@ -800,15 +838,19 @@ GLuint GSDeviceOGL::CreateSampler(PSSamplerSelector sel)
glCreateSamplers(1, &sampler); glCreateSamplers(1, &sampler);
// Bilinear filtering // Bilinear filtering
if (sel.biln) { if (sel.biln)
{
glSamplerParameteri(sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR); glSamplerParameteri(sampler, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glSamplerParameteri(sampler, GL_TEXTURE_MIN_FILTER, GL_LINEAR); glSamplerParameteri(sampler, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
} else { }
else
{
glSamplerParameteri(sampler, GL_TEXTURE_MAG_FILTER, GL_NEAREST); glSamplerParameteri(sampler, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glSamplerParameteri(sampler, GL_TEXTURE_MIN_FILTER, GL_NEAREST); glSamplerParameteri(sampler, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
} }
switch (static_cast<GS_MIN_FILTER>(sel.triln)) { switch (static_cast<GS_MIN_FILTER>(sel.triln))
{
case GS_MIN_FILTER::Nearest: case GS_MIN_FILTER::Nearest:
// Nop based on biln // Nop based on biln
break; break;
@ -846,7 +888,8 @@ GLuint GSDeviceOGL::CreateSampler(PSSamplerSelector sel)
glSamplerParameteri(sampler, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE); glSamplerParameteri(sampler, GL_TEXTURE_WRAP_R, GL_CLAMP_TO_EDGE);
int anisotropy = theApp.GetConfigI("MaxAnisotropy"); int anisotropy = theApp.GetConfigI("MaxAnisotropy");
if (anisotropy && sel.aniso) { if (anisotropy && sel.aniso)
{
if (GLExtension::Has("GL_ARB_texture_filter_anisotropic")) if (GLExtension::Has("GL_ARB_texture_filter_anisotropic"))
glSamplerParameterf(sampler, GL_TEXTURE_MAX_ANISOTROPY, (float)anisotropy); glSamplerParameterf(sampler, GL_TEXTURE_MAX_ANISOTROPY, (float)anisotropy);
else if (GLExtension::Has("GL_EXT_texture_filter_anisotropic")) else if (GLExtension::Has("GL_EXT_texture_filter_anisotropic"))
@ -911,7 +954,8 @@ void GSDeviceOGL::InitPrimDateTexture(GSTexture* rt, const GSVector4i& area)
void GSDeviceOGL::RecycleDateTexture() void GSDeviceOGL::RecycleDateTexture()
{ {
if (m_date.t) { if (m_date.t)
{
//static_cast<GSTextureOGL*>(m_date.t)->Save(format("/tmp/date_adv_%04ld.csv", GSState::s_n)); //static_cast<GSTextureOGL*>(m_date.t)->Save(format("/tmp/date_adv_%04ld.csv", GSState::s_n));
Recycle(m_date.t); Recycle(m_date.t);
@ -1009,7 +1053,8 @@ void GSDeviceOGL::SelfShaderTestRun(const std::string& dir, const std::string& f
#ifdef __linux__ #ifdef __linux__
// Nouveau actually // Nouveau actually
if (GLLoader::mesa_driver) { if (GLLoader::mesa_driver)
{
if (freopen(out.c_str(), "w", stderr) == NULL) if (freopen(out.c_str(), "w", stderr) == NULL)
fprintf(stderr, "Failed to redirect stderr\n"); fprintf(stderr, "Failed to redirect stderr\n");
} }
@ -1021,7 +1066,8 @@ void GSDeviceOGL::SelfShaderTestRun(const std::string& dir, const std::string& f
#ifdef __linux__ #ifdef __linux__
// Nouveau actually // Nouveau actually
if (GLLoader::mesa_driver) { if (GLLoader::mesa_driver)
{
if (freopen("/dev/tty", "w", stderr) == NULL) if (freopen("/dev/tty", "w", stderr) == NULL)
fprintf(stderr, "Failed to restore stderr\n"); fprintf(stderr, "Failed to restore stderr\n");
} }
@ -1054,15 +1100,18 @@ void GSDeviceOGL::SelfShaderTest()
int nb_shader = 0; int nb_shader = 0;
test = "SW_Blending"; test = "SW_Blending";
for (int colclip = 0; colclip < 2; colclip++) { for (int colclip = 0; colclip < 2; colclip++)
for (int fmt = 0; fmt < 3; fmt++) { {
for (int i = 0; i < 3; i++) { for (int fmt = 0; fmt < 3; fmt++)
{
for (int i = 0; i < 3; i++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 4; sel.tfx = 4;
int ib = (i + 1) % 3; int ib = (i + 1) % 3;
sel.blend_a = i; sel.blend_a = i;
sel.blend_b = ib;; sel.blend_b = ib;
sel.blend_c = i; sel.blend_c = i;
sel.blend_d = i; sel.blend_d = i;
sel.colclip = colclip; sel.colclip = colclip;
@ -1077,7 +1126,8 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "Alpha_Test"; test = "Alpha_Test";
for (int atst = 0; atst < 5; atst++) { for (int atst = 0; atst < 5; atst++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 4; sel.tfx = 4;
@ -1088,7 +1138,8 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "Fbmask__Fog__Shuffle__Read_ba"; test = "Fbmask__Fog__Shuffle__Read_ba";
for (int read_ba = 0; read_ba < 2; read_ba++) { for (int read_ba = 0; read_ba < 2; read_ba++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 4; sel.tfx = 4;
@ -1103,7 +1154,8 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "Date"; test = "Date";
for (int date = 1; date < 7; date++) { for (int date = 1; date < 7; date++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 4; sel.tfx = 4;
@ -1114,7 +1166,8 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "FBA"; test = "FBA";
for (int fmt = 0; fmt < 3; fmt++) { for (int fmt = 0; fmt < 3; fmt++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 4; sel.tfx = 4;
@ -1141,9 +1194,12 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "Tfx__Tcc"; test = "Tfx__Tcc";
for (int channel = 0; channel < 5; channel++) { for (int channel = 0; channel < 5; channel++)
for (int tfx = 0; tfx < 5; tfx++) { {
for (int tcc = 0; tcc < 2; tcc++) { for (int tfx = 0; tfx < 5; tfx++)
{
for (int tcc = 0; tcc < 2; tcc++)
{
PSSelector sel; PSSelector sel;
sel.fst = 1; sel.fst = 1;
@ -1158,14 +1214,21 @@ void GSDeviceOGL::SelfShaderTest()
SelfShaderTestPrint(test, nb_shader); SelfShaderTestPrint(test, nb_shader);
test = "Texture_Sampling"; test = "Texture_Sampling";
for (int depth = 0; depth < 4; depth++) { for (int depth = 0; depth < 4; depth++)
for (int fmt = 0; fmt < 16; fmt++) { {
if ((fmt & 3) == 3) continue; for (int fmt = 0; fmt < 16; fmt++)
{
if ((fmt & 3) == 3)
continue;
for (int ltf = 0; ltf < 2; ltf++) { for (int ltf = 0; ltf < 2; ltf++)
for (int aem = 0; aem < 2; aem++) { {
for (int wms = 1; wms < 4; wms++) { for (int aem = 0; aem < 2; aem++)
for (int wmt = 1; wmt < 4; wmt++) { {
for (int wms = 1; wms < 4; wms++)
{
for (int wmt = 1; wmt < 4; wmt++)
{
PSSelector sel; PSSelector sel;
sel.tfx = 1; sel.tfx = 1;
sel.tcc = 1; sel.tcc = 1;
@ -1352,7 +1415,8 @@ void GSDeviceOGL::StretchRect(GSTexture* sTex, const GSVector4& sRect, GSTexture
// 2/ in case some GSdx code expect thing in dx order. // 2/ in case some GSdx code expect thing in dx order.
// Only flipping the backbuffer is transparent (I hope)... // Only flipping the backbuffer is transparent (I hope)...
GSVector4 flip_sr = sRect; GSVector4 flip_sr = sRect;
if (static_cast<GSTextureOGL*>(dTex)->IsBackbuffer()) { if (static_cast<GSTextureOGL*>(dTex)->IsBackbuffer())
{
flip_sr.y = sRect.w; flip_sr.y = sRect.w;
flip_sr.w = sRect.y; flip_sr.w = sRect.y;
} }
@ -1398,7 +1462,8 @@ void GSDeviceOGL::RenderOsd(GSTexture* dt)
OMSetBlendState((uint8)GSDeviceOGL::m_MERGE_BLEND); OMSetBlendState((uint8)GSDeviceOGL::m_MERGE_BLEND);
OMSetRenderTargets(dt, NULL); OMSetRenderTargets(dt, NULL);
if(m_osd.m_texture_dirty) { if (m_osd.m_texture_dirty)
{
m_osd.upload_texture_atlas(m_font.get()); m_osd.upload_texture_atlas(m_font.get());
} }
@ -1434,14 +1499,16 @@ void GSDeviceOGL::DoMerge(GSTexture* sTex[3], GSVector4* sRect, GSTexture* dTex,
ClearRenderTarget(dTex, c); ClearRenderTarget(dTex, c);
// Upload constant to select YUV algo // Upload constant to select YUV algo
if (feedback_write_2 || feedback_write_1) { if (feedback_write_2 || feedback_write_1)
{
// Write result to feedback loop // Write result to feedback loop
m_misc_cb_cache.EMOD_AC.x = EXTBUF.EMODA; m_misc_cb_cache.EMOD_AC.x = EXTBUF.EMODA;
m_misc_cb_cache.EMOD_AC.y = EXTBUF.EMODC; m_misc_cb_cache.EMOD_AC.y = EXTBUF.EMODC;
m_convert.cb->cache_upload(&m_misc_cb_cache); m_convert.cb->cache_upload(&m_misc_cb_cache);
} }
if (sTex[1] && (PMODE.SLBG == 0 || feedback_write_2_but_blend_bg)) { if (sTex[1] && (PMODE.SLBG == 0 || feedback_write_2_but_blend_bg))
{
// 2nd output is enabled and selected. Copy it to destination so we can blend it with 1st output // 2nd output is enabled and selected. Copy it to destination so we can blend it with 1st output
// Note: value outside of dRect must contains the background color (c) // Note: value outside of dRect must contains the background color (c)
StretchRect(sTex[1], sRect[1], dTex, dRect[1], ShaderConvert_COPY); StretchRect(sTex[1], sRect[1], dTex, dRect[1], ShaderConvert_COPY);
@ -1455,16 +1522,20 @@ void GSDeviceOGL::DoMerge(GSTexture* sTex[3], GSVector4* sRect, GSTexture* dTex,
if (feedback_write_2_but_blend_bg) if (feedback_write_2_but_blend_bg)
ClearRenderTarget(dTex, c); ClearRenderTarget(dTex, c);
if (sTex[0]) { if (sTex[0])
{
if (PMODE.AMOD == 1) // Keep the alpha from the 2nd output if (PMODE.AMOD == 1) // Keep the alpha from the 2nd output
OMSetColorMaskState(OMColorMaskSelector(0x7)); OMSetColorMaskState(OMColorMaskSelector(0x7));
// 1st output is enabled. It must be blended // 1st output is enabled. It must be blended
if (PMODE.MMOD == 1) { if (PMODE.MMOD == 1)
{
// Blend with a constant alpha // Blend with a constant alpha
m_merge_obj.cb->cache_upload(&c.v); m_merge_obj.cb->cache_upload(&c.v);
StretchRect(sTex[0], sRect[0], dTex, dRect[0], m_merge_obj.ps[1], m_MERGE_BLEND, OMColorMaskSelector()); StretchRect(sTex[0], sRect[0], dTex, dRect[0], m_merge_obj.ps[1], m_MERGE_BLEND, OMColorMaskSelector());
} else { }
else
{
// Blend with 2 * input alpha // Blend with 2 * input alpha
StretchRect(sTex[0], sRect[0], dTex, dRect[0], m_merge_obj.ps[0], m_MERGE_BLEND, OMColorMaskSelector()); StretchRect(sTex[0], sRect[0], dTex, dRect[0], m_merge_obj.ps[0], m_MERGE_BLEND, OMColorMaskSelector());
} }
@ -1498,8 +1569,10 @@ void GSDeviceOGL::DoInterlace(GSTexture* sTex, GSTexture* dTex, int shader, bool
void GSDeviceOGL::DoFXAA(GSTexture* sTex, GSTexture* dTex) void GSDeviceOGL::DoFXAA(GSTexture* sTex, GSTexture* dTex)
{ {
// Lazy compile // Lazy compile
if (!m_fxaa.ps) { if (!m_fxaa.ps)
if (!GLLoader::found_GL_ARB_gpu_shader5) { // GL4.0 extension {
if (!GLLoader::found_GL_ARB_gpu_shader5) // GL4.0 extension
{
return; return;
} }
@ -1528,8 +1601,10 @@ void GSDeviceOGL::DoFXAA(GSTexture* sTex, GSTexture* dTex)
void GSDeviceOGL::DoExternalFX(GSTexture* sTex, GSTexture* dTex) void GSDeviceOGL::DoExternalFX(GSTexture* sTex, GSTexture* dTex)
{ {
// Lazy compile // Lazy compile
if (!m_shaderfx.ps) { if (!m_shaderfx.ps)
if (!GLLoader::found_GL_ARB_gpu_shader5) { // GL4.0 extension {
if (!GLLoader::found_GL_ARB_gpu_shader5) // GL4.0 extension
{
return; return;
} }
@ -1545,7 +1620,8 @@ void GSDeviceOGL::DoExternalFX(GSTexture* sTex, GSTexture* dTex)
std::string shader_name(theApp.GetConfigS("shaderfx_glsl")); std::string shader_name(theApp.GetConfigS("shaderfx_glsl"));
std::ifstream fshader(shader_name); std::ifstream fshader(shader_name);
std::stringstream shader; std::stringstream shader;
if (!fshader.good()) { if (!fshader.good())
{
fprintf(stderr, "Error failed to load '%s'. External Shader will be disabled !\n", shader_name.c_str()); fprintf(stderr, "Error failed to load '%s'. External Shader will be disabled !\n", shader_name.c_str());
return; return;
} }
@ -1606,7 +1682,8 @@ void GSDeviceOGL::SetupDATE(GSTexture* rt, GSTexture* ds, const GSVertexPT1* ver
// om // om
OMSetDepthStencilState(m_date.dss); OMSetDepthStencilState(m_date.dss);
if (GLState::blend) { if (GLState::blend)
{
glDisable(GL_BLEND); glDisable(GL_BLEND);
} }
OMSetRenderTargets(NULL, ds, &GLState::scissor); OMSetRenderTargets(NULL, ds, &GLState::scissor);
@ -1624,7 +1701,8 @@ void GSDeviceOGL::SetupDATE(GSTexture* rt, GSTexture* ds, const GSVertexPT1* ver
DrawPrimitive(); DrawPrimitive();
if (GLState::blend) { if (GLState::blend)
{
glEnable(GL_BLEND); glEnable(GL_BLEND);
} }
@ -1655,9 +1733,11 @@ void GSDeviceOGL::PSSetShaderResource(int i, GSTexture* sr)
{ {
ASSERT(i < (int)countof(GLState::tex_unit)); ASSERT(i < (int)countof(GLState::tex_unit));
// Note: Nvidia debgger doesn't support the id 0 (ie the NULL texture) // Note: Nvidia debgger doesn't support the id 0 (ie the NULL texture)
if (sr) { if (sr)
{
GLuint id = static_cast<GSTextureOGL*>(sr)->GetID(); GLuint id = static_cast<GSTextureOGL*>(sr)->GetID();
if (GLState::tex_unit[i] != id) { if (GLState::tex_unit[i] != id)
{
GLState::tex_unit[i] = id; GLState::tex_unit[i] = id;
glBindTextureUnit(i, id); glBindTextureUnit(i, id);
} }
@ -1672,7 +1752,8 @@ void GSDeviceOGL::PSSetShaderResources(GSTexture* sr0, GSTexture* sr1)
void GSDeviceOGL::PSSetSamplerState(GLuint ss) void GSDeviceOGL::PSSetSamplerState(GLuint ss)
{ {
if (GLState::ps_ss != ss) { if (GLState::ps_ss != ss)
{
GLState::ps_ss = ss; GLState::ps_ss = ss;
glBindSampler(0, ss); glBindSampler(0, ss);
} }
@ -1681,14 +1762,18 @@ void GSDeviceOGL::PSSetSamplerState(GLuint ss)
void GSDeviceOGL::OMAttachRt(GSTextureOGL* rt) void GSDeviceOGL::OMAttachRt(GSTextureOGL* rt)
{ {
GLuint id; GLuint id;
if (rt) { if (rt)
{
rt->WasAttached(); rt->WasAttached();
id = rt->GetID(); id = rt->GetID();
} else { }
else
{
id = 0; id = 0;
} }
if (GLState::rt != id) { if (GLState::rt != id)
{
GLState::rt = id; GLState::rt = id;
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, id, 0); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, id, 0);
} }
@ -1697,14 +1782,18 @@ void GSDeviceOGL::OMAttachRt(GSTextureOGL* rt)
void GSDeviceOGL::OMAttachDs(GSTextureOGL* ds) void GSDeviceOGL::OMAttachDs(GSTextureOGL* ds)
{ {
GLuint id; GLuint id;
if (ds) { if (ds)
{
ds->WasAttached(); ds->WasAttached();
id = ds->GetID(); id = ds->GetID();
} else { }
else
{
id = 0; id = 0;
} }
if (GLState::ds != id) { if (GLState::ds != id)
{
GLState::ds = id; GLState::ds = id;
glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, id, 0); glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, id, 0);
} }
@ -1712,7 +1801,8 @@ void GSDeviceOGL::OMAttachDs(GSTextureOGL* ds)
void GSDeviceOGL::OMSetFBO(GLuint fbo) void GSDeviceOGL::OMSetFBO(GLuint fbo)
{ {
if (GLState::fbo != fbo) { if (GLState::fbo != fbo)
{
GLState::fbo = fbo; GLState::fbo = fbo;
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo); glBindFramebuffer(GL_DRAW_FRAMEBUFFER, fbo);
} }
@ -1726,7 +1816,8 @@ void GSDeviceOGL::OMSetDepthStencilState(GSDepthStencilOGL* dss)
void GSDeviceOGL::OMSetColorMaskState(OMColorMaskSelector sel) void GSDeviceOGL::OMSetColorMaskState(OMColorMaskSelector sel)
{ {
if (sel.wrgba != GLState::wrgba) { if (sel.wrgba != GLState::wrgba)
{
GLState::wrgba = sel.wrgba; GLState::wrgba = sel.wrgba;
glColorMaski(0, sel.wr, sel.wg, sel.wb, sel.wa); glColorMaski(0, sel.wr, sel.wg, sel.wb, sel.wa);
@ -1735,37 +1826,45 @@ void GSDeviceOGL::OMSetColorMaskState(OMColorMaskSelector sel)
void GSDeviceOGL::OMSetBlendState(uint8 blend_index, uint8 blend_factor, bool is_blend_constant, bool accumulation_blend) void GSDeviceOGL::OMSetBlendState(uint8 blend_index, uint8 blend_factor, bool is_blend_constant, bool accumulation_blend)
{ {
if (blend_index) { if (blend_index)
if (!GLState::blend) { {
if (!GLState::blend)
{
GLState::blend = true; GLState::blend = true;
glEnable(GL_BLEND); glEnable(GL_BLEND);
} }
if (is_blend_constant && GLState::bf != blend_factor) { if (is_blend_constant && GLState::bf != blend_factor)
{
GLState::bf = blend_factor; GLState::bf = blend_factor;
float bf = (float)blend_factor / 128.0f; float bf = (float)blend_factor / 128.0f;
glBlendColor(bf, bf, bf, bf); glBlendColor(bf, bf, bf, bf);
} }
HWBlend b = GetBlend(blend_index); HWBlend b = GetBlend(blend_index);
if (accumulation_blend) { if (accumulation_blend)
{
b.src = GL_ONE; b.src = GL_ONE;
b.dst = GL_ONE; b.dst = GL_ONE;
} }
if (GLState::eq_RGB != b.op) { if (GLState::eq_RGB != b.op)
{
GLState::eq_RGB = b.op; GLState::eq_RGB = b.op;
glBlendEquationSeparate(b.op, GL_FUNC_ADD); glBlendEquationSeparate(b.op, GL_FUNC_ADD);
} }
if (GLState::f_sRGB != b.src || GLState::f_dRGB != b.dst) { if (GLState::f_sRGB != b.src || GLState::f_dRGB != b.dst)
{
GLState::f_sRGB = b.src; GLState::f_sRGB = b.src;
GLState::f_dRGB = b.dst; GLState::f_dRGB = b.dst;
glBlendFuncSeparate(b.src, b.dst, GL_ONE, GL_ZERO); glBlendFuncSeparate(b.src, b.dst, GL_ONE, GL_ZERO);
} }
}
} else { else
if (GLState::blend) { {
if (GLState::blend)
{
GLState::blend = false; GLState::blend = false;
glDisable(GL_BLEND); glDisable(GL_BLEND);
} }
@ -1777,11 +1876,15 @@ void GSDeviceOGL::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVecto
GSTextureOGL* RT = static_cast<GSTextureOGL*>(rt); GSTextureOGL* RT = static_cast<GSTextureOGL*>(rt);
GSTextureOGL* DS = static_cast<GSTextureOGL*>(ds); GSTextureOGL* DS = static_cast<GSTextureOGL*>(ds);
if (rt == NULL || !RT->IsBackbuffer()) { if (rt == NULL || !RT->IsBackbuffer())
{
OMSetFBO(m_fbo); OMSetFBO(m_fbo);
if (rt) { if (rt)
{
OMAttachRt(RT); OMAttachRt(RT);
} else { }
else
{
OMAttachRt(); OMAttachRt();
} }
@ -1790,8 +1893,9 @@ void GSDeviceOGL::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVecto
OMAttachDs(DS); OMAttachDs(DS);
else else
OMAttachDs(); OMAttachDs();
}
} else { else
{
// Render in the backbuffer // Render in the backbuffer
OMSetFBO(0); OMSetFBO(0);
} }
@ -1818,11 +1922,13 @@ void GSDeviceOGL::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVecto
void GSDeviceOGL::SetupCB(const VSConstantBuffer* vs_cb, const PSConstantBuffer* ps_cb) void GSDeviceOGL::SetupCB(const VSConstantBuffer* vs_cb, const PSConstantBuffer* ps_cb)
{ {
GL_PUSH("UBO"); GL_PUSH("UBO");
if(m_vs_cb_cache.Update(vs_cb)) { if (m_vs_cb_cache.Update(vs_cb))
{
m_vs_cb->upload(vs_cb); m_vs_cb->upload(vs_cb);
} }
if(m_ps_cb_cache.Update(ps_cb)) { if (m_ps_cb_cache.Update(ps_cb))
{
m_ps_cb->upload(ps_cb); m_ps_cb->upload(ps_cb);
} }
} }
@ -1838,10 +1944,13 @@ void GSDeviceOGL::SetupPipeline(const VSSelector& vsel, const GSSelector& gsel,
GLuint ps; GLuint ps;
auto i = m_ps.find(psel); auto i = m_ps.find(psel);
if (i == m_ps.end()) { if (i == m_ps.end())
{
ps = CompilePS(psel); ps = CompilePS(psel);
m_ps[psel] = ps; m_ps[psel] = ps;
} else { }
else
{
ps = i->second; ps = i->second;
} }
@ -1852,9 +1961,12 @@ void GSDeviceOGL::SetupPipeline(const VSSelector& vsel, const GSSelector& gsel,
static PSSelector old_psel; static PSSelector old_psel;
static GLuint old_ps = 0; static GLuint old_ps = 0;
std::string msg(""); std::string msg("");
#define CHECK_STATE(p) if (psel.p != old_psel.p) msg.append(" ").append(#p); #define CHECK_STATE(p) \
if (psel.p != old_psel.p) \
msg.append(" ").append(#p);
if (old_ps != ps) { if (old_ps != ps)
{
CHECK_STATE(tex_fmt); CHECK_STATE(tex_fmt);
CHECK_STATE(dfmt); CHECK_STATE(dfmt);
@ -1922,7 +2034,8 @@ void GSDeviceOGL::DebugOutputToFile(GLenum gl_source, GLenum gl_type, GLuint id,
std::string message(gl_message, gl_length >= 0 ? gl_length : strlen(gl_message)); std::string message(gl_message, gl_length >= 0 ? gl_length : strlen(gl_message));
std::string type, severity, source; std::string type, severity, source;
static int sev_counter = 0; static int sev_counter = 0;
switch(gl_type) { switch (gl_type)
{
case GL_DEBUG_TYPE_ERROR_ARB : type = "Error"; break; case GL_DEBUG_TYPE_ERROR_ARB : type = "Error"; break;
case GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR_ARB : type = "Deprecated bhv"; break; case GL_DEBUG_TYPE_DEPRECATED_BEHAVIOR_ARB : type = "Deprecated bhv"; break;
case GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR_ARB : type = "Undefined bhv"; break; case GL_DEBUG_TYPE_UNDEFINED_BEHAVIOR_ARB : type = "Undefined bhv"; break;
@ -1933,11 +2046,13 @@ void GSDeviceOGL::DebugOutputToFile(GLenum gl_source, GLenum gl_type, GLuint id,
case GL_DEBUG_TYPE_POP_GROUP : return; // Don't print message injected by myself case GL_DEBUG_TYPE_POP_GROUP : return; // Don't print message injected by myself
default : type = "TTT"; break; default : type = "TTT"; break;
} }
switch(gl_severity) { switch (gl_severity)
{
case GL_DEBUG_SEVERITY_HIGH_ARB : severity = "High"; sev_counter++; break; case GL_DEBUG_SEVERITY_HIGH_ARB : severity = "High"; sev_counter++; break;
case GL_DEBUG_SEVERITY_MEDIUM_ARB : severity = "Mid"; break; case GL_DEBUG_SEVERITY_MEDIUM_ARB : severity = "Mid"; break;
case GL_DEBUG_SEVERITY_LOW_ARB : severity = "Low"; break; case GL_DEBUG_SEVERITY_LOW_ARB : severity = "Low"; break;
default : if (id == 0xFEAD) default:
if (id == 0xFEAD)
severity = "Cache"; severity = "Cache";
else if (id == 0xB0B0) else if (id == 0xB0B0)
severity = "REG"; severity = "REG";
@ -1945,7 +2060,8 @@ void GSDeviceOGL::DebugOutputToFile(GLenum gl_source, GLenum gl_type, GLuint id,
severity = "EXTRA"; severity = "EXTRA";
break; break;
} }
switch(gl_source) { switch (gl_source)
{
case GL_DEBUG_SOURCE_API_ARB : source = "API"; break; case GL_DEBUG_SOURCE_API_ARB : source = "API"; break;
case GL_DEBUG_SOURCE_WINDOW_SYSTEM_ARB : source = "WINDOW"; break; case GL_DEBUG_SOURCE_WINDOW_SYSTEM_ARB : source = "WINDOW"; break;
case GL_DEBUG_SOURCE_SHADER_COMPILER_ARB : source = "COMPILER"; break; case GL_DEBUG_SOURCE_SHADER_COMPILER_ARB : source = "COMPILER"; break;
@ -1957,16 +2073,19 @@ void GSDeviceOGL::DebugOutputToFile(GLenum gl_source, GLenum gl_type, GLuint id,
#ifdef _DEBUG #ifdef _DEBUG
// Don't spam noisy information on the terminal // Don't spam noisy information on the terminal
if (gl_severity != GL_DEBUG_SEVERITY_NOTIFICATION) { if (gl_severity != GL_DEBUG_SEVERITY_NOTIFICATION)
{
fprintf(stderr, "T:%s\tID:%d\tS:%s\t=> %s\n", type.c_str(), GSState::s_n, severity.c_str(), message.c_str()); fprintf(stderr, "T:%s\tID:%d\tS:%s\t=> %s\n", type.c_str(), GSState::s_n, severity.c_str(), message.c_str());
} }
#else #else
// Print nouveau shader compiler info // Print nouveau shader compiler info
if (GSState::s_n == 0) { if (GSState::s_n == 0)
{
int t, local, gpr, inst, byte; int t, local, gpr, inst, byte;
int status = sscanf(message.c_str(), "type: %d, local: %d, gpr: %d, inst: %d, bytes: %d", int status = sscanf(message.c_str(), "type: %d, local: %d, gpr: %d, inst: %d, bytes: %d",
&t, &local, &gpr, &inst, &byte); &t, &local, &gpr, &inst, &byte);
if (status == 5) { if (status == 5)
{
m_shader_inst += inst; m_shader_inst += inst;
m_shader_reg += gpr; m_shader_reg += gpr;
fprintf(stderr, "T:%s\t\tS:%s\t=> %s\n", type.c_str(), severity.c_str(), message.c_str()); fprintf(stderr, "T:%s\t\tS:%s\t=> %s\n", type.c_str(), severity.c_str(), message.c_str());
@ -1978,9 +2097,11 @@ void GSDeviceOGL::DebugOutputToFile(GLenum gl_source, GLenum gl_type, GLuint id,
fprintf(m_debug_gl_file, "T:%s\tID:%d\tS:%s\t=> %s\n", type.c_str(), GSState::s_n, severity.c_str(), message.c_str()); fprintf(m_debug_gl_file, "T:%s\tID:%d\tS:%s\t=> %s\n", type.c_str(), GSState::s_n, severity.c_str(), message.c_str());
#ifdef _DEBUG #ifdef _DEBUG
if (sev_counter >= 5) { if (sev_counter >= 5)
{
// Close the file to flush the content on disk before exiting. // Close the file to flush the content on disk before exiting.
if (m_debug_gl_file) { if (m_debug_gl_file)
{
fclose(m_debug_gl_file); fclose(m_debug_gl_file);
m_debug_gl_file = NULL; m_debug_gl_file = NULL;
} }

117
plugins/GSdx/Renderers/OpenGL/GSDeviceOGL.h
View File

@ -34,7 +34,8 @@ extern uint64 g_real_texture_upload_byte;
extern uint64 g_vertex_upload_byte; extern uint64 g_vertex_upload_byte;
#endif #endif
class GSDepthStencilOGL { class GSDepthStencilOGL
{
bool m_depth_enable; bool m_depth_enable;
GLenum m_depth_func; GLenum m_depth_func;
bool m_depth_mask; bool m_depth_mask;
@ -44,8 +45,8 @@ class GSDepthStencilOGL {
GLenum m_stencil_spass_dpass_op; GLenum m_stencil_spass_dpass_op;
public: public:
GSDepthStencilOGL()
GSDepthStencilOGL() : m_depth_enable(false) : m_depth_enable(false)
, m_depth_func(GL_ALWAYS) , m_depth_func(GL_ALWAYS)
, m_depth_mask(0) , m_depth_mask(0)
, m_stencil_enable(false) , m_stencil_enable(false)
@ -57,12 +58,21 @@ public:
void EnableDepth() { m_depth_enable = true; } void EnableDepth() { m_depth_enable = true; }
void EnableStencil() { m_stencil_enable = true; } void EnableStencil() { m_stencil_enable = true; }
void SetDepth(GLenum func, bool mask) { m_depth_func = func; m_depth_mask = mask; } void SetDepth(GLenum func, bool mask)
void SetStencil(GLenum func, GLenum pass) { m_stencil_func = func; m_stencil_spass_dpass_op = pass; } {
m_depth_func = func;
m_depth_mask = mask;
}
void SetStencil(GLenum func, GLenum pass)
{
m_stencil_func = func;
m_stencil_spass_dpass_op = pass;
}
void SetupDepth() void SetupDepth()
{ {
if (GLState::depth != m_depth_enable) { if (GLState::depth != m_depth_enable)
{
GLState::depth = m_depth_enable; GLState::depth = m_depth_enable;
if (m_depth_enable) if (m_depth_enable)
glEnable(GL_DEPTH_TEST); glEnable(GL_DEPTH_TEST);
@ -70,12 +80,15 @@ public:
glDisable(GL_DEPTH_TEST); glDisable(GL_DEPTH_TEST);
} }
if (m_depth_enable) { if (m_depth_enable)
if (GLState::depth_func != m_depth_func) { {
if (GLState::depth_func != m_depth_func)
{
GLState::depth_func = m_depth_func; GLState::depth_func = m_depth_func;
glDepthFunc(m_depth_func); glDepthFunc(m_depth_func);
} }
if (GLState::depth_mask != m_depth_mask) { if (GLState::depth_mask != m_depth_mask)
{
GLState::depth_mask = m_depth_mask; GLState::depth_mask = m_depth_mask;
glDepthMask((GLboolean)m_depth_mask); glDepthMask((GLboolean)m_depth_mask);
} }
@ -84,7 +97,8 @@ public:
void SetupStencil() void SetupStencil()
{ {
if (GLState::stencil != m_stencil_enable) { if (GLState::stencil != m_stencil_enable)
{
GLState::stencil = m_stencil_enable; GLState::stencil = m_stencil_enable;
if (m_stencil_enable) if (m_stencil_enable)
glEnable(GL_STENCIL_TEST); glEnable(GL_STENCIL_TEST);
@ -92,13 +106,16 @@ public:
glDisable(GL_STENCIL_TEST); glDisable(GL_STENCIL_TEST);
} }
if (m_stencil_enable) { if (m_stencil_enable)
{
// Note: here the mask control which bitplane is considered by the operation // Note: here the mask control which bitplane is considered by the operation
if (GLState::stencil_func != m_stencil_func) { if (GLState::stencil_func != m_stencil_func)
{
GLState::stencil_func = m_stencil_func; GLState::stencil_func = m_stencil_func;
glStencilFunc(m_stencil_func, 1, 1); glStencilFunc(m_stencil_func, 1, 1);
} }
if (GLState::stencil_pass != m_stencil_spass_dpass_op) { if (GLState::stencil_pass != m_stencil_spass_dpass_op)
{
GLState::stencil_pass = m_stencil_spass_dpass_op; GLState::stencil_pass = m_stencil_spass_dpass_op;
glStencilOp(GL_KEEP, GL_KEEP, m_stencil_spass_dpass_op); glStencilOp(GL_KEEP, GL_KEEP, m_stencil_spass_dpass_op);
} }
@ -161,8 +178,14 @@ public:
operator uint32() const { return key; } operator uint32() const { return key; }
VSSelector() : key(0) {} VSSelector()
VSSelector(uint32 k) : key(k) {} : key(0)
{
}
VSSelector(uint32 k)
: key(k)
{
}
}; };
struct GSSelector struct GSSelector
@ -183,8 +206,14 @@ public:
operator uint32() const { return key; } operator uint32() const { return key; }
GSSelector() : key(0) {} GSSelector()
GSSelector(uint32 k) : key(k) {} : key(0)
{
}
GSSelector(uint32 k)
: key(k)
{
}
}; };
struct alignas(32) PSConstantBuffer struct alignas(32) PSConstantBuffer
@ -330,7 +359,10 @@ public:
// FIXME is the & useful ? // FIXME is the & useful ?
operator uint64() const { return key; } operator uint64() const { return key; }
PSSelector() : key(0) {} PSSelector()
: key(0)
{
}
}; };
struct PSSamplerSelector struct PSSamplerSelector
@ -353,8 +385,14 @@ public:
operator uint32() { return key; } operator uint32() { return key; }
PSSamplerSelector() : key(0) {} PSSamplerSelector()
PSSamplerSelector(uint32 k) : key(k) {} : key(0)
{
}
PSSamplerSelector(uint32 k)
: key(k)
{
}
}; };
struct OMDepthStencilSelector struct OMDepthStencilSelector
@ -377,8 +415,14 @@ public:
// FIXME is the & useful ? // FIXME is the & useful ?
operator uint32() { return key; } operator uint32() { return key; }
OMDepthStencilSelector() : key(0) {} OMDepthStencilSelector()
OMDepthStencilSelector(uint32 k) : key(k) {} : key(0)
{
}
OMDepthStencilSelector(uint32 k)
: key(k)
{
}
}; };
struct OMColorMaskSelector struct OMColorMaskSelector
@ -406,7 +450,10 @@ public:
// FIXME is the & useful ? // FIXME is the & useful ?
operator uint32() { return key & 0xf; } operator uint32() { return key & 0xf; }
OMColorMaskSelector() : key(0xF) {} OMColorMaskSelector()
: key(0xF)
{
}
OMColorMaskSelector(uint32 c) { wrgba = c; } OMColorMaskSelector(uint32 c) { wrgba = c; }
}; };
@ -441,17 +488,20 @@ private:
GSVertexBufferStateOGL* m_va; // state of the vertex buffer/array GSVertexBufferStateOGL* m_va; // state of the vertex buffer/array
struct { struct
{
GLuint ps[2]; // program object GLuint ps[2]; // program object
GSUniformBufferOGL* cb; // uniform buffer object GSUniformBufferOGL* cb; // uniform buffer object
} m_merge_obj; } m_merge_obj;
struct { struct
{
GLuint ps[4]; // program object GLuint ps[4]; // program object
GSUniformBufferOGL* cb; // uniform buffer object GSUniformBufferOGL* cb; // uniform buffer object
} m_interlace; } m_interlace;
struct { struct
{
GLuint vs; // program object GLuint vs; // program object
GLuint ps[ShaderConvert_Count]; // program object GLuint ps[ShaderConvert_Count]; // program object
GLuint ln; // sampler object GLuint ln; // sampler object
@ -461,26 +511,31 @@ private:
GSUniformBufferOGL* cb; GSUniformBufferOGL* cb;
} m_convert; } m_convert;
struct { struct
{
GLuint ps; GLuint ps;
GSUniformBufferOGL* cb; GSUniformBufferOGL* cb;
} m_fxaa; } m_fxaa;
struct { struct
{
GLuint ps; GLuint ps;
GSUniformBufferOGL* cb; GSUniformBufferOGL* cb;
} m_shaderfx; } m_shaderfx;
struct { struct
{
GSDepthStencilOGL* dss; GSDepthStencilOGL* dss;
GSTexture* t; GSTexture* t;
} m_date; } m_date;
struct { struct
{
GLuint ps; GLuint ps;
} m_shadeboost; } m_shadeboost;
struct { struct
{
uint16 last_query; uint16 last_query;
GLuint timer_query[1 << 16]; GLuint timer_query[1 << 16];

434
plugins/GSdx/Renderers/OpenGL/GSRendererOGL.cpp
View File

@ -51,7 +51,8 @@ void GSRendererOGL::SetupIA(const float& sx, const float& sy)
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev; GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
if (m_userhacks_wildhack && !m_isPackedUV_HackFlag && PRIM->TME && PRIM->FST) { if (m_userhacks_wildhack && !m_isPackedUV_HackFlag && PRIM->TME && PRIM->FST)
{
for (unsigned int i = 0; i < m_vertex.next; i++) for (unsigned int i = 0; i < m_vertex.next; i++)
m_vertex.buff[i].UV &= 0x3FEF3FEF; m_vertex.buff[i].UV &= 0x3FEF3FEF;
} }
@ -62,7 +63,8 @@ void GSRendererOGL::SetupIA(const float& sx, const float& sy)
switch (m_vt.m_primclass) switch (m_vt.m_primclass)
{ {
case GS_POINT_CLASS: case GS_POINT_CLASS:
if (unscale_pt_ln) { if (unscale_pt_ln)
{
m_gs_sel.point = 1; m_gs_sel.point = 1;
vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy); vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy);
} }
@ -71,7 +73,8 @@ void GSRendererOGL::SetupIA(const float& sx, const float& sy)
break; break;
case GS_LINE_CLASS: case GS_LINE_CLASS:
if (unscale_pt_ln) { if (unscale_pt_ln)
{
m_gs_sel.line = 1; m_gs_sel.line = 1;
vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy); vs_cb.PointSize = GSVector2(16.0f * sx, 16.0f * sy);
} }
@ -95,11 +98,14 @@ void GSRendererOGL::SetupIA(const float& sx, const float& sy)
// the extra validation cost of the extra stage. // the extra validation cost of the extra stage.
// //
// Note: keep Geometry Shader in the replayer to ease debug. // Note: keep Geometry Shader in the replayer to ease debug.
if (GLLoader::found_geometry_shader && !m_vt.m_accurate_stq && (m_vertex.next > 32 || GLLoader::in_replayer)) { // <=> 16 sprites (based on Shadow Hearts) if (GLLoader::found_geometry_shader && !m_vt.m_accurate_stq && (m_vertex.next > 32 || GLLoader::in_replayer)) // <=> 16 sprites (based on Shadow Hearts)
{
m_gs_sel.sprite = 1; m_gs_sel.sprite = 1;
t = GL_LINES; t = GL_LINES;
} else { }
else
{
Lines2Sprites(); Lines2Sprites();
t = GL_TRIANGLES; t = GL_TRIANGLES;
@ -121,10 +127,13 @@ void GSRendererOGL::SetupIA(const float& sx, const float& sy)
void GSRendererOGL::EmulateZbuffer() void GSRendererOGL::EmulateZbuffer()
{ {
if (m_context->TEST.ZTE) { if (m_context->TEST.ZTE)
{
m_om_dssel.ztst = m_context->TEST.ZTST; m_om_dssel.ztst = m_context->TEST.ZTST;
m_om_dssel.zwe = !m_context->ZBUF.ZMSK; m_om_dssel.zwe = !m_context->ZBUF.ZMSK;
} else { }
else
{
m_om_dssel.ztst = ZTST_ALWAYS; m_om_dssel.ztst = ZTST_ALWAYS;
} }
@ -137,10 +146,14 @@ void GSRendererOGL::EmulateZbuffer()
//ps_cb.MaxDepth = GSVector4(0.0f, 0.0f, 0.0f, 1.0f); //ps_cb.MaxDepth = GSVector4(0.0f, 0.0f, 0.0f, 1.0f);
m_ps_sel.zclamp = 0; m_ps_sel.zclamp = 0;
if (clamp_z) { if (clamp_z)
if (m_vt.m_primclass == GS_SPRITE_CLASS || m_vt.m_primclass == GS_POINT_CLASS) { {
if (m_vt.m_primclass == GS_SPRITE_CLASS || m_vt.m_primclass == GS_POINT_CLASS)
{
vs_cb.MaxDepth = GSVector2i(max_z); vs_cb.MaxDepth = GSVector2i(max_z);
} else { }
else
{
ps_cb.MaxDepth = GSVector4(0.0f, 0.0f, 0.0f, max_z * ldexpf(1, -32)); ps_cb.MaxDepth = GSVector4(0.0f, 0.0f, 0.0f, max_z * ldexpf(1, -32));
m_ps_sel.zclamp = 1; m_ps_sel.zclamp = 1;
} }
@ -148,7 +161,8 @@ void GSRendererOGL::EmulateZbuffer()
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
// Minor optimization of a corner case (it allow to better emulate some alpha test effects) // Minor optimization of a corner case (it allow to better emulate some alpha test effects)
if (m_om_dssel.ztst == ZTST_GEQUAL && m_vt.m_eq.z && v[0].XYZ.Z == max_z) { if (m_om_dssel.ztst == ZTST_GEQUAL && m_vt.m_eq.z && v[0].XYZ.Z == max_z)
{
GL_DBG("Optimize Z test GEQUAL to ALWAYS (%s)", psm_str(m_context->ZBUF.PSM)); GL_DBG("Optimize Z test GEQUAL to ALWAYS (%s)", psm_str(m_context->ZBUF.PSM));
m_om_dssel.ztst = ZTST_ALWAYS; m_om_dssel.ztst = ZTST_ALWAYS;
} }
@ -159,7 +173,8 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
// Uncomment to disable texture shuffle emulation. // Uncomment to disable texture shuffle emulation.
// m_texture_shuffle = false; // m_texture_shuffle = false;
if (m_texture_shuffle) { if (m_texture_shuffle)
{
m_ps_sel.shuffle = 1; m_ps_sel.shuffle = 1;
m_ps_sel.dfmt = 0; m_ps_sel.dfmt = 0;
@ -185,11 +200,15 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
m_om_csel.wrgba = 0; m_om_csel.wrgba = 0;
// 2 Select the new mask (Please someone put SSE here) // 2 Select the new mask (Please someone put SSE here)
if (rg_mask != 0xFF) { if (rg_mask != 0xFF)
if (write_ba) { {
if (write_ba)
{
GL_INS("Color shuffle %s => B", read_ba ? "B" : "R"); GL_INS("Color shuffle %s => B", read_ba ? "B" : "R");
m_om_csel.wb = 1; m_om_csel.wb = 1;
} else { }
else
{
GL_INS("Color shuffle %s => R", read_ba ? "B" : "R"); GL_INS("Color shuffle %s => R", read_ba ? "B" : "R");
m_om_csel.wr = 1; m_om_csel.wr = 1;
} }
@ -197,11 +216,15 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
m_ps_sel.fbmask = 1; m_ps_sel.fbmask = 1;
} }
if (ba_mask != 0xFF) { if (ba_mask != 0xFF)
if (write_ba) { {
if (write_ba)
{
GL_INS("Color shuffle %s => A", read_ba ? "A" : "G"); GL_INS("Color shuffle %s => A", read_ba ? "A" : "G");
m_om_csel.wa = 1; m_om_csel.wa = 1;
} else { }
else
{
GL_INS("Color shuffle %s => G", read_ba ? "A" : "G"); GL_INS("Color shuffle %s => G", read_ba ? "A" : "G");
m_om_csel.wg = 1; m_om_csel.wg = 1;
} }
@ -209,25 +232,32 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
m_ps_sel.fbmask = 1; m_ps_sel.fbmask = 1;
} }
if (m_ps_sel.fbmask && m_sw_blending) { if (m_ps_sel.fbmask && m_sw_blending)
{
ps_cb.FbMask.r = rg_mask; ps_cb.FbMask.r = rg_mask;
ps_cb.FbMask.g = rg_mask; ps_cb.FbMask.g = rg_mask;
ps_cb.FbMask.b = ba_mask; ps_cb.FbMask.b = ba_mask;
ps_cb.FbMask.a = ba_mask; ps_cb.FbMask.a = ba_mask;
// No blending so hit unsafe path. // No blending so hit unsafe path.
if (!PRIM->ABE) { if (!PRIM->ABE)
{
GL_INS("FBMASK Unsafe SW emulated fb_mask:%x on tex shuffle", fbmask); GL_INS("FBMASK Unsafe SW emulated fb_mask:%x on tex shuffle", fbmask);
m_require_one_barrier = true; m_require_one_barrier = true;
} else { }
else
{
GL_INS("FBMASK SW emulated fb_mask:%x on tex shuffle", fbmask); GL_INS("FBMASK SW emulated fb_mask:%x on tex shuffle", fbmask);
m_require_full_barrier = true; m_require_full_barrier = true;
} }
} else { }
else
{
m_ps_sel.fbmask = 0; m_ps_sel.fbmask = 0;
} }
}
} else { else
{
m_ps_sel.dfmt = GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt; m_ps_sel.dfmt = GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt;
GSVector4i fbmask_v = GSVector4i::load((int)m_context->FRAME.FBMSK); GSVector4i fbmask_v = GSVector4i::load((int)m_context->FRAME.FBMSK);
@ -238,7 +268,8 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
m_ps_sel.fbmask = m_sw_blending && (~ff_fbmask & ~zero_fbmask & 0xF); m_ps_sel.fbmask = m_sw_blending && (~ff_fbmask & ~zero_fbmask & 0xF);
if (m_ps_sel.fbmask) { if (m_ps_sel.fbmask)
{
ps_cb.FbMask = fbmask_v.u8to32(); ps_cb.FbMask = fbmask_v.u8to32();
// Only alpha is special here, I think we can take a very unsafe shortcut // Only alpha is special here, I think we can take a very unsafe shortcut
// Alpha isn't blended on the GS but directly copyied into the RT. // Alpha isn't blended on the GS but directly copyied into the RT.
@ -260,11 +291,14 @@ void GSRendererOGL::EmulateTextureShuffleAndFbmask()
have been invalidated before subsequent Draws are executed. have been invalidated before subsequent Draws are executed.
*/ */
// No blending so hit unsafe path. // No blending so hit unsafe path.
if (!PRIM->ABE || !(~ff_fbmask & ~zero_fbmask & 0x7)) { if (!PRIM->ABE || !(~ff_fbmask & ~zero_fbmask & 0x7))
{
GL_INS("FBMASK Unsafe SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK, GL_INS("FBMASK Unsafe SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK,
(GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32); (GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32);
m_require_one_barrier = true; m_require_one_barrier = true;
} else { }
else
{
// The safe and accurate path (but slow) // The safe and accurate path (but slow)
GL_INS("FBMASK SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK, GL_INS("FBMASK SW emulated fb_mask:%x on %d bits format", m_context->FRAME.FBMSK,
(GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32); (GSLocalMemory::m_psm[m_context->FRAME.PSM].fmt == 2) ? 16 : 32);
@ -282,14 +316,19 @@ void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::
// m_channel_shuffle = false; // m_channel_shuffle = false;
// First let's check we really have a channel shuffle effect // First let's check we really have a channel shuffle effect
if (m_channel_shuffle) { if (m_channel_shuffle)
if (m_game.title == CRC::GT4 || m_game.title == CRC::GT3 || m_game.title == CRC::GTConcept || m_game.title == CRC::TouristTrophy) { {
if (m_game.title == CRC::GT4 || m_game.title == CRC::GT3 || m_game.title == CRC::GTConcept || m_game.title == CRC::TouristTrophy)
{
GL_INS("Gran Turismo RGB Channel"); GL_INS("Gran Turismo RGB Channel");
m_ps_sel.channel = ChannelFetch_RGB; m_ps_sel.channel = ChannelFetch_RGB;
m_context->TEX0.TFX = TFX_DECAL; m_context->TEX0.TFX = TFX_DECAL;
*rt = tex->m_from_target; *rt = tex->m_from_target;
} else if (m_game.title == CRC::Tekken5) { }
if (m_context->FRAME.FBW == 1) { else if (m_game.title == CRC::Tekken5)
{
if (m_context->FRAME.FBW == 1)
{
// Used in stages: Secret Garden, Acid Rain, Moonlit Wilderness // Used in stages: Secret Garden, Acid Rain, Moonlit Wilderness
GL_INS("Tekken5 RGB Channel"); GL_INS("Tekken5 RGB Channel");
m_ps_sel.channel = ChannelFetch_RGB; m_ps_sel.channel = ChannelFetch_RGB;
@ -298,23 +337,32 @@ void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::
// Minus current draw call // Minus current draw call
m_skip = 12 * (3 + 3 + 1) - 1; m_skip = 12 * (3 + 3 + 1) - 1;
*rt = tex->m_from_target; *rt = tex->m_from_target;
} else { }
else
{
// Could skip model drawing if wrongly detected // Could skip model drawing if wrongly detected
m_channel_shuffle = false; m_channel_shuffle = false;
} }
} else if ((tex->m_texture->GetType() == GSTexture::DepthStencil) && !(tex->m_32_bits_fmt)) { }
else if ((tex->m_texture->GetType() == GSTexture::DepthStencil) && !(tex->m_32_bits_fmt))
{
// So far 2 games hit this code path. Urban Chaos and Tales of Abyss // So far 2 games hit this code path. Urban Chaos and Tales of Abyss
// UC: will copy depth to green channel // UC: will copy depth to green channel
// ToA: will copy depth to alpha channel // ToA: will copy depth to alpha channel
if ((m_context->FRAME.FBMSK & 0xFF0000) == 0xFF0000) { if ((m_context->FRAME.FBMSK & 0xFF0000) == 0xFF0000)
{
// Green channel is masked // Green channel is masked
GL_INS("Tales Of Abyss Crazyness (MSB 16b depth to Alpha)"); GL_INS("Tales Of Abyss Crazyness (MSB 16b depth to Alpha)");
m_ps_sel.tales_of_abyss_hle = 1; m_ps_sel.tales_of_abyss_hle = 1;
} else { }
else
{
GL_INS("Urban Chaos Crazyness (Green extraction)"); GL_INS("Urban Chaos Crazyness (Green extraction)");
m_ps_sel.urban_chaos_hle = 1; m_ps_sel.urban_chaos_hle = 1;
} }
} else if (m_index.tail <= 64 && m_context->CLAMP.WMT == 3) { }
else if (m_index.tail <= 64 && m_context->CLAMP.WMT == 3)
{
// Blood will tell. I think it is channel effect too but again // Blood will tell. I think it is channel effect too but again
// implemented in a different way. I don't want to add more CRC stuff. So // implemented in a different way. I don't want to add more CRC stuff. So
// let's disable channel when the signature is different // let's disable channel when the signature is different
@ -323,23 +371,29 @@ void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::
// handled above. // handled above.
GL_INS("Maybe not a channel!"); GL_INS("Maybe not a channel!");
m_channel_shuffle = false; m_channel_shuffle = false;
} else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MAXU & 0x8) == 8)) { }
else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MAXU & 0x8) == 8))
{
// Read either blue or Alpha. Let's go for Blue ;) // Read either blue or Alpha. Let's go for Blue ;)
// MGS3/Kill Zone // MGS3/Kill Zone
GL_INS("Blue channel"); GL_INS("Blue channel");
m_ps_sel.channel = ChannelFetch_BLUE; m_ps_sel.channel = ChannelFetch_BLUE;
} else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MINU & 0x8) == 0)) { }
else if (m_context->CLAMP.WMS == 3 && ((m_context->CLAMP.MINU & 0x8) == 0))
{
// Read either Red or Green. Let's check the V coordinate. 0-1 is likely top so // Read either Red or Green. Let's check the V coordinate. 0-1 is likely top so
// red. 2-3 is likely bottom so green (actually depends on texture base pointer offset) // red. 2-3 is likely bottom so green (actually depends on texture base pointer offset)
bool green = PRIM->FST && (m_vertex.buff[0].V & 32); bool green = PRIM->FST && (m_vertex.buff[0].V & 32);
if (green && (m_context->FRAME.FBMSK & 0x00FFFFFF) == 0x00FFFFFF) { if (green && (m_context->FRAME.FBMSK & 0x00FFFFFF) == 0x00FFFFFF)
{
// Typically used in Terminator 3 // Typically used in Terminator 3
int blue_mask = m_context->FRAME.FBMSK >> 24; int blue_mask = m_context->FRAME.FBMSK >> 24;
int green_mask = ~blue_mask & 0xFF; int green_mask = ~blue_mask & 0xFF;
int blue_shift = -1; int blue_shift = -1;
// Note: potentially we could also check the value of the clut // Note: potentially we could also check the value of the clut
switch (m_context->FRAME.FBMSK >> 24) { switch (m_context->FRAME.FBMSK >> 24)
{
case 0xFF: ASSERT(0); break; case 0xFF: ASSERT(0); break;
case 0xFE: blue_shift = 1; break; case 0xFE: blue_shift = 1; break;
case 0xFC: blue_shift = 2; break; case 0xFC: blue_shift = 2; break;
@ -354,31 +408,40 @@ void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::
int green_shift = 8 - blue_shift; int green_shift = 8 - blue_shift;
dev->SetupCBMisc(GSVector4i(blue_mask, blue_shift, green_mask, green_shift)); dev->SetupCBMisc(GSVector4i(blue_mask, blue_shift, green_mask, green_shift));
if (blue_shift >= 0) { if (blue_shift >= 0)
{
GL_INS("Green/Blue channel (%d, %d)", blue_shift, green_shift); GL_INS("Green/Blue channel (%d, %d)", blue_shift, green_shift);
m_ps_sel.channel = ChannelFetch_GXBY; m_ps_sel.channel = ChannelFetch_GXBY;
m_context->FRAME.FBMSK = 0x00FFFFFF; m_context->FRAME.FBMSK = 0x00FFFFFF;
} else { }
else
{
GL_INS("Green channel (wrong mask) (fbmask %x)", m_context->FRAME.FBMSK >> 24); GL_INS("Green channel (wrong mask) (fbmask %x)", m_context->FRAME.FBMSK >> 24);
m_ps_sel.channel = ChannelFetch_GREEN; m_ps_sel.channel = ChannelFetch_GREEN;
} }
}
} else if (green) { else if (green)
{
GL_INS("Green channel"); GL_INS("Green channel");
m_ps_sel.channel = ChannelFetch_GREEN; m_ps_sel.channel = ChannelFetch_GREEN;
} else { }
else
{
// Pop // Pop
GL_INS("Red channel"); GL_INS("Red channel");
m_ps_sel.channel = ChannelFetch_RED; m_ps_sel.channel = ChannelFetch_RED;
} }
} else { }
else
{
GL_INS("Channel not supported"); GL_INS("Channel not supported");
m_channel_shuffle = false; m_channel_shuffle = false;
} }
} }
// Effect is really a channel shuffle effect so let's cheat a little // Effect is really a channel shuffle effect so let's cheat a little
if (m_channel_shuffle) { if (m_channel_shuffle)
{
dev->PSSetShaderResource(4, tex->m_from_target); dev->PSSetShaderResource(4, tex->m_from_target);
m_require_one_barrier = true; m_require_one_barrier = true;
@ -395,8 +458,9 @@ void GSRendererOGL::EmulateChannelShuffle(GSTexture** rt, const GSTextureCache::
m_vertex.head = m_vertex.tail = m_vertex.next = 2; m_vertex.head = m_vertex.tail = m_vertex.next = 2;
m_index.tail = 2; m_index.tail = 2;
}
} else { else
{
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG
dev->PSSetShaderResource(4, NULL); dev->PSSetShaderResource(4, NULL);
#endif #endif
@ -410,12 +474,14 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
bool sw_blending = false; bool sw_blending = false;
// No blending so early exit // No blending so early exit
if (!(PRIM->ABE || m_env.PABE.PABE || (PRIM->AA1 && m_vt.m_primclass == GS_LINE_CLASS))) { if (!(PRIM->ABE || m_env.PABE.PABE || (PRIM->AA1 && m_vt.m_primclass == GS_LINE_CLASS)))
{
dev->OMSetBlendState(); dev->OMSetBlendState();
return; return;
} }
if (m_env.PABE.PABE) { if (m_env.PABE.PABE)
{
// Breath of Fire Dragon Quarter, Strawberry Shortcake, Super Robot Wars, Cartoon Network Racing. // Breath of Fire Dragon Quarter, Strawberry Shortcake, Super Robot Wars, Cartoon Network Racing.
GL_INS("PABE mode ENABLED"); GL_INS("PABE mode ENABLED");
m_ps_sel.pabe = 1; m_ps_sel.pabe = 1;
@ -438,12 +504,14 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
// Warning no break on purpose // Warning no break on purpose
// Note: the [[fallthrough]] attribute tell compilers not to complain about not having breaks. // Note: the [[fallthrough]] attribute tell compilers not to complain about not having breaks.
switch (m_sw_blending) { switch (m_sw_blending)
{
case ACC_BLEND_ULTRA: case ACC_BLEND_ULTRA:
sw_blending |= true; sw_blending |= true;
[[fallthrough]]; [[fallthrough]];
case ACC_BLEND_FULL: case ACC_BLEND_FULL:
if (!m_vt.m_alpha.valid && (ALPHA.C == 0)) GetAlphaMinMax(); if (!m_vt.m_alpha.valid && (ALPHA.C == 0))
GetAlphaMinMax();
sw_blending |= (ALPHA.A != ALPHA.B) && ((ALPHA.C == 0 && m_vt.m_alpha.max > 128) || (ALPHA.C == 2 && ALPHA.FIX > 128u)); sw_blending |= (ALPHA.A != ALPHA.B) && ((ALPHA.C == 0 && m_vt.m_alpha.max > 128) || (ALPHA.C == 2 && ALPHA.FIX > 128u));
[[fallthrough]]; [[fallthrough]];
case ACC_BLEND_HIGH: case ACC_BLEND_HIGH:
@ -463,29 +531,37 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
} }
// Color clip // Color clip
if (m_env.COLCLAMP.CLAMP == 0) { if (m_env.COLCLAMP.CLAMP == 0)
{
// Safe FBMASK, avoid hitting accumulation mode on 16bit, // Safe FBMASK, avoid hitting accumulation mode on 16bit,
// fixes shadows in Superman shadows of Apokolips. // fixes shadows in Superman shadows of Apokolips.
const bool sw_fbmask_colclip = !m_require_one_barrier && m_ps_sel.fbmask; const bool sw_fbmask_colclip = !m_require_one_barrier && m_ps_sel.fbmask;
const bool free_colclip = m_prim_overlap == PRIM_OVERLAP_NO || blend_non_recursive || sw_fbmask_colclip; const bool free_colclip = m_prim_overlap == PRIM_OVERLAP_NO || blend_non_recursive || sw_fbmask_colclip;
GL_DBG("COLCLIP Info (Blending: %d/%d/%d/%d, SW FBMASK: %d, OVERLAP: %d)", GL_DBG("COLCLIP Info (Blending: %d/%d/%d/%d, SW FBMASK: %d, OVERLAP: %d)",
ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D, sw_fbmask_colclip, m_prim_overlap); ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D, sw_fbmask_colclip, m_prim_overlap);
if (free_colclip) { if (free_colclip)
{
// The fastest algo that requires a single pass // The fastest algo that requires a single pass
GL_INS("COLCLIP Free mode ENABLED"); GL_INS("COLCLIP Free mode ENABLED");
m_ps_sel.colclip = 1; m_ps_sel.colclip = 1;
sw_blending = true; sw_blending = true;
accumulation_blend = false; // disable the HDR algo accumulation_blend = false; // disable the HDR algo
} else if (accumulation_blend) { }
else if (accumulation_blend)
{
// A fast algo that requires 2 passes // A fast algo that requires 2 passes
GL_INS("COLCLIP Fast HDR mode ENABLED"); GL_INS("COLCLIP Fast HDR mode ENABLED");
m_ps_sel.hdr = 1; m_ps_sel.hdr = 1;
sw_blending = true; // Enable sw blending for the HDR algo sw_blending = true; // Enable sw blending for the HDR algo
} else if (sw_blending) { }
else if (sw_blending)
{
// A slow algo that could requires several passes (barely used) // A slow algo that could requires several passes (barely used)
GL_INS("COLCLIP SW mode ENABLED"); GL_INS("COLCLIP SW mode ENABLED");
m_ps_sel.colclip = 1; m_ps_sel.colclip = 1;
} else { }
else
{
GL_INS("COLCLIP HDR mode ENABLED"); GL_INS("COLCLIP HDR mode ENABLED");
m_ps_sel.hdr = 1; m_ps_sel.hdr = 1;
} }
@ -497,7 +573,8 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
// Switch DATE_GL42 with DATE_GL45 in such cases to ensure accuracy. // Switch DATE_GL42 with DATE_GL45 in such cases to ensure accuracy.
// No mix of COLCLIP + sw blend + DATE_GL42, neither sw fbmask + DATE_GL42. // No mix of COLCLIP + sw blend + DATE_GL42, neither sw fbmask + DATE_GL42.
// Note: Do the swap after colclip to avoid adding extra conditions. // Note: Do the swap after colclip to avoid adding extra conditions.
if (sw_blending && DATE_GL42) { if (sw_blending && DATE_GL42)
{
GL_PERF("DATE: Swap DATE_GL42 with DATE_GL45"); GL_PERF("DATE: Swap DATE_GL42 with DATE_GL45");
m_require_full_barrier = true; m_require_full_barrier = true;
DATE_GL42 = false; DATE_GL42 = false;
@ -509,16 +586,19 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
GL_INS("BLEND_INFO: %d/%d/%d/%d. Clamp:%d. Prim:%d number %d (drawlist %d) (sw %d)", GL_INS("BLEND_INFO: %d/%d/%d/%d. Clamp:%d. Prim:%d number %d (drawlist %d) (sw %d)",
ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D, m_env.COLCLAMP.CLAMP, m_vt.m_primclass, m_vertex.next, m_drawlist.size(), sw_blending); ALPHA.A, ALPHA.B, ALPHA.C, ALPHA.D, m_env.COLCLAMP.CLAMP, m_vt.m_primclass, m_vertex.next, m_drawlist.size(), sw_blending);
#endif #endif
if (sw_blending) { if (sw_blending)
{
m_ps_sel.blend_a = ALPHA.A; m_ps_sel.blend_a = ALPHA.A;
m_ps_sel.blend_b = ALPHA.B; m_ps_sel.blend_b = ALPHA.B;
m_ps_sel.blend_c = ALPHA.C; m_ps_sel.blend_c = ALPHA.C;
m_ps_sel.blend_d = ALPHA.D; m_ps_sel.blend_d = ALPHA.D;
if (accumulation_blend) { if (accumulation_blend)
{
// Keep HW blending to do the addition/subtraction // Keep HW blending to do the addition/subtraction
dev->OMSetBlendState(blend_index, 0, false, true); dev->OMSetBlendState(blend_index, 0, false, true);
if (ALPHA.A == 2) { if (ALPHA.A == 2)
{
// The blend unit does a reverse subtraction so it means // The blend unit does a reverse subtraction so it means
// the shader must output a positive value. // the shader must output a positive value.
// Replace 0 - Cs by Cs - 0 // Replace 0 - Cs by Cs - 0
@ -529,8 +609,9 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
m_ps_sel.blend_d = 2; m_ps_sel.blend_d = 2;
// Note accumulation_blend doesn't require a barrier // Note accumulation_blend doesn't require a barrier
}
} else { else
{
// Disable HW blending // Disable HW blending
dev->OMSetBlendState(); dev->OMSetBlendState();
@ -538,16 +619,22 @@ void GSRendererOGL::EmulateBlending(bool& DATE_GL42, bool& DATE_GL45)
} }
// Require the fix alpha vlaue // Require the fix alpha vlaue
if (ALPHA.C == 2) { if (ALPHA.C == 2)
{
ps_cb.TA_Af.a = (float)ALPHA.FIX / 128.0f; ps_cb.TA_Af.a = (float)ALPHA.FIX / 128.0f;
} }
} else { }
else
{
m_ps_sel.clr1 = !!(blend_flag & BLEND_C_CLR); m_ps_sel.clr1 = !!(blend_flag & BLEND_C_CLR);
if (m_ps_sel.dfmt == 1 && ALPHA.C == 1) { if (m_ps_sel.dfmt == 1 && ALPHA.C == 1)
{
// 24 bits doesn't have an alpha channel so use 1.0f fix factor as equivalent // 24 bits doesn't have an alpha channel so use 1.0f fix factor as equivalent
const uint8 hacked_blend_index = blend_index + 3; // +3 <=> +1 on C const uint8 hacked_blend_index = blend_index + 3; // +3 <=> +1 on C
dev->OMSetBlendState(hacked_blend_index, 128, true); dev->OMSetBlendState(hacked_blend_index, 128, true);
} else { }
else
{
dev->OMSetBlendState(blend_index, ALPHA.FIX, (ALPHA.C == 2)); dev->OMSetBlendState(blend_index, ALPHA.FIX, (ALPHA.C == 2));
} }
} }
@ -581,7 +668,8 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
break; break;
case TriFiltering::PS2: case TriFiltering::PS2:
if (need_mipmap && m_mipmap != 2) { if (need_mipmap && m_mipmap != 2)
{
trilinear = m_context->TEX1.MMIN; trilinear = m_context->TEX1.MMIN;
trilinear_auto = true; trilinear_auto = true;
} }
@ -603,14 +691,16 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
// Performance note: // Performance note:
// 1/ Don't set 0 as it is the default value // 1/ Don't set 0 as it is the default value
// 2/ Only keep aem when it is useful (avoid useless shader permutation) // 2/ Only keep aem when it is useful (avoid useless shader permutation)
if (m_ps_sel.shuffle) { if (m_ps_sel.shuffle)
{
// Force a 32 bits access (normally shuffle is done on 16 bits) // Force a 32 bits access (normally shuffle is done on 16 bits)
// m_ps_sel.tex_fmt = 0; // removed as an optimization // m_ps_sel.tex_fmt = 0; // removed as an optimization
m_ps_sel.aem = m_env.TEXA.AEM; m_ps_sel.aem = m_env.TEXA.AEM;
ASSERT(tex->m_target); ASSERT(tex->m_target);
// Require a float conversion if the texure is a depth otherwise uses Integral scaling // Require a float conversion if the texure is a depth otherwise uses Integral scaling
if (psm.depth) { if (psm.depth)
{
m_ps_sel.depth_fmt = (tex->m_texture->GetType() != GSTexture::DepthStencil) ? 3 : 1; m_ps_sel.depth_fmt = (tex->m_texture->GetType() != GSTexture::DepthStencil) ? 3 : 1;
m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate
} }
@ -626,8 +716,9 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
bilinear &= m_vt.IsLinear(); bilinear &= m_vt.IsLinear();
vs_cb.TextureOffset = RealignTargetTextureCoordinate(tex); vs_cb.TextureOffset = RealignTargetTextureCoordinate(tex);
}
} else if (tex->m_target) { else if (tex->m_target)
{
// Use an old target. AEM and index aren't resolved it must be done // Use an old target. AEM and index aren't resolved it must be done
// on the GPU // on the GPU
@ -636,7 +727,8 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
m_ps_sel.aem = m_env.TEXA.AEM; m_ps_sel.aem = m_env.TEXA.AEM;
// Don't upload AEM if format is 32 bits // Don't upload AEM if format is 32 bits
if (cpsm.fmt) { if (cpsm.fmt)
{
GSVector4 ta(m_env.TEXA & GSVector4i::x000000ff()); GSVector4 ta(m_env.TEXA & GSVector4i::x000000ff());
ta /= 255.0f; ta /= 255.0f;
// FIXME rely on compiler for the optimization // FIXME rely on compiler for the optimization
@ -645,7 +737,8 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
} }
// Select the index format // Select the index format
if (tex->m_palette) { if (tex->m_palette)
{
// FIXME Potentially improve fmt field in GSLocalMemory // FIXME Potentially improve fmt field in GSLocalMemory
if (m_context->TEX0.PSM == PSM_PSMT4HL) if (m_context->TEX0.PSM == PSM_PSMT4HL)
m_ps_sel.tex_fmt |= 1 << 2; m_ps_sel.tex_fmt |= 1 << 2;
@ -661,14 +754,17 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
} }
// Depth format // Depth format
if (tex->m_texture->GetType() == GSTexture::DepthStencil) { if (tex->m_texture->GetType() == GSTexture::DepthStencil)
{
// Require a float conversion if the texure is a depth format // Require a float conversion if the texure is a depth format
m_ps_sel.depth_fmt = (psm.bpp == 16) ? 2 : 1; m_ps_sel.depth_fmt = (psm.bpp == 16) ? 2 : 1;
m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate
// Don't force interpolation on depth format // Don't force interpolation on depth format
bilinear &= m_vt.IsLinear(); bilinear &= m_vt.IsLinear();
} else if (psm.depth) { }
else if (psm.depth)
{
// Use Integral scaling // Use Integral scaling
m_ps_sel.depth_fmt = 3; m_ps_sel.depth_fmt = 3;
m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate m_vs_sel.int_fst = !PRIM->FST; // select float/int coordinate
@ -678,25 +774,30 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
} }
vs_cb.TextureOffset = RealignTargetTextureCoordinate(tex); vs_cb.TextureOffset = RealignTargetTextureCoordinate(tex);
}
} else if (tex->m_palette) { else if (tex->m_palette)
{
// Use a standard 8 bits texture. AEM is already done on the CLUT // Use a standard 8 bits texture. AEM is already done on the CLUT
// Therefore you only need to set the index // Therefore you only need to set the index
// m_ps_sel.aem = 0; // removed as an optimization // m_ps_sel.aem = 0; // removed as an optimization
// Note 4 bits indexes are converted to 8 bits // Note 4 bits indexes are converted to 8 bits
m_ps_sel.tex_fmt = 3 << 2; m_ps_sel.tex_fmt = 3 << 2;
}
} else { else
{
// Standard texture. Both index and AEM expansion were already done by the CPU. // Standard texture. Both index and AEM expansion were already done by the CPU.
// m_ps_sel.tex_fmt = 0; // removed as an optimization // m_ps_sel.tex_fmt = 0; // removed as an optimization
// m_ps_sel.aem = 0; // removed as an optimization // m_ps_sel.aem = 0; // removed as an optimization
} }
if (m_context->TEX0.TFX == TFX_MODULATE && m_vt.m_eq.rgba == 0xFFFF && m_vt.m_min.c.eq(GSVector4i(128))) { if (m_context->TEX0.TFX == TFX_MODULATE && m_vt.m_eq.rgba == 0xFFFF && m_vt.m_min.c.eq(GSVector4i(128)))
{
// Micro optimization that reduces GPU load (removes 5 instructions on the FS program) // Micro optimization that reduces GPU load (removes 5 instructions on the FS program)
m_ps_sel.tfx = TFX_DECAL; m_ps_sel.tfx = TFX_DECAL;
} else { }
else
{
m_ps_sel.tfx = m_context->TEX0.TFX; m_ps_sel.tfx = m_context->TEX0.TFX;
} }
@ -717,16 +818,21 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
ps_cb.WH = WH; ps_cb.WH = WH;
ps_cb.HalfTexel = GSVector4(-0.5f, 0.5f).xxyy() / WH.zwzw(); ps_cb.HalfTexel = GSVector4(-0.5f, 0.5f).xxyy() / WH.zwzw();
if (complex_wms_wmt) { if (complex_wms_wmt)
{
ps_cb.MskFix = GSVector4i(m_context->CLAMP.MINU, m_context->CLAMP.MINV, m_context->CLAMP.MAXU, m_context->CLAMP.MAXV); ps_cb.MskFix = GSVector4i(m_context->CLAMP.MINU, m_context->CLAMP.MINV, m_context->CLAMP.MAXU, m_context->CLAMP.MAXV);
ps_cb.MinMax = GSVector4(ps_cb.MskFix) / WH.xyxy(); ps_cb.MinMax = GSVector4(ps_cb.MskFix) / WH.xyxy();
} else if (trilinear_manual) { }
else if (trilinear_manual)
{
// Reuse MinMax for mipmap parameter to avoid an extension of the UBO // Reuse MinMax for mipmap parameter to avoid an extension of the UBO
ps_cb.MinMax.x = (float)m_context->TEX1.K / 16.0f; ps_cb.MinMax.x = (float)m_context->TEX1.K / 16.0f;
ps_cb.MinMax.y = float(1 << m_context->TEX1.L); ps_cb.MinMax.y = float(1 << m_context->TEX1.L);
ps_cb.MinMax.z = float(m_lod.x); // Offset because first layer is m_lod, dunno if we can do better ps_cb.MinMax.z = float(m_lod.x); // Offset because first layer is m_lod, dunno if we can do better
ps_cb.MinMax.w = float(m_lod.y); ps_cb.MinMax.w = float(m_lod.y);
} else if (trilinear_auto) { }
else if (trilinear_auto)
{
tex->m_texture->GenerateMipmap(); tex->m_texture->GenerateMipmap();
} }
@ -735,7 +841,8 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
ps_cb.TC_OH_TS = GSVector4(1 / 16.0f, 1 / 16.0f, m_userhacks_tcoffset_x, m_userhacks_tcoffset_y) / WH.xyxy(); ps_cb.TC_OH_TS = GSVector4(1 / 16.0f, 1 / 16.0f, m_userhacks_tcoffset_x, m_userhacks_tcoffset_y) / WH.xyxy();
// Must be done after all coordinates math // Must be done after all coordinates math
if (m_context->HasFixedTEX0() && !PRIM->FST) { if (m_context->HasFixedTEX0() && !PRIM->FST)
{
m_ps_sel.invalid_tex0 = 1; m_ps_sel.invalid_tex0 = 1;
// Use invalid size to denormalize ST coordinate // Use invalid size to denormalize ST coordinate
ps_cb.WH.x = (float)(1 << m_context->stack.TEX0.TW); ps_cb.WH.x = (float)(1 << m_context->stack.TEX0.TW);
@ -748,20 +855,26 @@ void GSRendererOGL::EmulateTextureSampler(const GSTextureCache::Source* tex)
// Only enable clamping in CLAMP mode. REGION_CLAMP will be done manually in the shader // Only enable clamping in CLAMP mode. REGION_CLAMP will be done manually in the shader
m_ps_ssel.tau = (wms != CLAMP_CLAMP); m_ps_ssel.tau = (wms != CLAMP_CLAMP);
m_ps_ssel.tav = (wmt != CLAMP_CLAMP); m_ps_ssel.tav = (wmt != CLAMP_CLAMP);
if (shader_emulated_sampler) { if (shader_emulated_sampler)
{
m_ps_ssel.biln = 0; m_ps_ssel.biln = 0;
m_ps_ssel.aniso = 0; m_ps_ssel.aniso = 0;
m_ps_ssel.triln = 0; m_ps_ssel.triln = 0;
} else { }
else
{
m_ps_ssel.biln = bilinear; m_ps_ssel.biln = bilinear;
// Aniso filtering doesn't work with textureLod so use texture (automatic_lod) instead. // Aniso filtering doesn't work with textureLod so use texture (automatic_lod) instead.
// Enable aniso only for triangles. Sprites are flat so aniso is likely useless (it would save perf for others primitives). // Enable aniso only for triangles. Sprites are flat so aniso is likely useless (it would save perf for others primitives).
const bool anisotropic = m_vt.m_primclass == GS_TRIANGLE_CLASS && !trilinear_manual; const bool anisotropic = m_vt.m_primclass == GS_TRIANGLE_CLASS && !trilinear_manual;
m_ps_ssel.aniso = anisotropic; m_ps_ssel.aniso = anisotropic;
m_ps_ssel.triln = trilinear; m_ps_ssel.triln = trilinear;
if (trilinear_manual) { if (trilinear_manual)
{
m_ps_sel.manual_lod = 1; m_ps_sel.manual_lod = 1;
} else if (trilinear_auto || anisotropic) { }
else if (trilinear_auto || anisotropic)
{
m_ps_sel.automatic_lod = 1; m_ps_sel.automatic_lod = 1;
} }
} }
@ -788,7 +901,8 @@ GSRendererOGL::PRIM_OVERLAP GSRendererOGL::PrimitiveOverlap()
m_drawlist.clear(); m_drawlist.clear();
size_t i = 0; size_t i = 0;
while (i < count) { while (i < count)
{
// In order to speed up comparison a bounding-box is accumulated. It removes a // In order to speed up comparison a bounding-box is accumulated. It removes a
// loop so code is much faster (check game virtua fighter). Besides it allow to check // loop so code is much faster (check game virtua fighter). Besides it allow to check
// properly the Y order. // properly the Y order.
@ -801,7 +915,8 @@ GSRendererOGL::PRIM_OVERLAP GSRendererOGL::PrimitiveOverlap()
all = all.xyxy().blend(all.zwzw(), all > all.zwxy()); all = all.xyxy().blend(all.zwzw(), all > all.zwxy());
size_t j = i + 2; size_t j = i + 2;
while (j < count) { while (j < count)
{
GSVector4i sprite = GSVector4i(v[j].m[1]).upl16(GSVector4i(v[j + 1].m[1])).upl16().xzyw(); GSVector4i sprite = GSVector4i(v[j].m[1]).upl16(GSVector4i(v[j + 1].m[1])).upl16().xzyw();
sprite = sprite.xyxy().blend(sprite.zwzw(), sprite > sprite.zwxy()); sprite = sprite.xyxy().blend(sprite.zwzw(), sprite > sprite.zwxy());
@ -812,9 +927,12 @@ GSRendererOGL::PRIM_OVERLAP GSRendererOGL::PrimitiveOverlap()
ASSERT(all.x <= all.z); ASSERT(all.x <= all.z);
ASSERT(all.y <= all.w); ASSERT(all.y <= all.w);
if (all.rintersect(sprite).rempty()) { if (all.rintersect(sprite).rempty())
{
all = all.runion_ordered(sprite); all = all.runion_ordered(sprite);
} else { }
else
{
overlap = PRIM_OVERLAP_YES; overlap = PRIM_OVERLAP_YES;
break; break;
} }
@ -873,18 +991,25 @@ void GSRendererOGL::SendDraw()
{ {
GSDeviceOGL* dev = (GSDeviceOGL*)m_dev; GSDeviceOGL* dev = (GSDeviceOGL*)m_dev;
if (!m_require_full_barrier && m_require_one_barrier) { if (!m_require_full_barrier && m_require_one_barrier)
{
// Need only a single barrier // Need only a single barrier
glTextureBarrier(); glTextureBarrier();
dev->DrawIndexedPrimitive(); dev->DrawIndexedPrimitive();
} else if (!m_require_full_barrier) { }
else if (!m_require_full_barrier)
{
// Don't need any barrier // Don't need any barrier
dev->DrawIndexedPrimitive(); dev->DrawIndexedPrimitive();
} else if (m_prim_overlap == PRIM_OVERLAP_NO) { }
else if (m_prim_overlap == PRIM_OVERLAP_NO)
{
// Need full barrier but a single barrier will be enough // Need full barrier but a single barrier will be enough
glTextureBarrier(); glTextureBarrier();
dev->DrawIndexedPrimitive(); dev->DrawIndexedPrimitive();
} else if (m_vt.m_primclass == GS_SPRITE_CLASS) { }
else if (m_vt.m_primclass == GS_SPRITE_CLASS)
{
size_t nb_vertex = (m_gs_sel.sprite == 1) ? 2 : 6; size_t nb_vertex = (m_gs_sel.sprite == 1) ? 2 : 6;
GL_PUSH("Split the draw (SPRITE)"); GL_PUSH("Split the draw (SPRITE)");
@ -903,12 +1028,15 @@ void GSRendererOGL::SendDraw()
m_index.tail / nb_vertex, m_drawlist.size(), message.c_str()); m_index.tail / nb_vertex, m_drawlist.size(), message.c_str());
#endif #endif
for (size_t count, p = 0, n = 0; n < m_drawlist.size(); p += count, ++n) { for (size_t count, p = 0, n = 0; n < m_drawlist.size(); p += count, ++n)
{
count = m_drawlist[n] * nb_vertex; count = m_drawlist[n] * nb_vertex;
glTextureBarrier(); glTextureBarrier();
dev->DrawIndexedPrimitive(p, count); dev->DrawIndexedPrimitive(p, count);
} }
} else { }
else
{
// FIXME: Investigate: a dynamic check to pack as many primitives as possibles // FIXME: Investigate: a dynamic check to pack as many primitives as possibles
// I'm nearly sure GSdx already have this kind of code (maybe we can adapt GSDirtyRect) // I'm nearly sure GSdx already have this kind of code (maybe we can adapt GSDirtyRect)
size_t nb_vertex = GSUtil::GetClassVertexCount(m_vt.m_primclass); size_t nb_vertex = GSUtil::GetClassVertexCount(m_vt.m_primclass);
@ -917,7 +1045,8 @@ void GSRendererOGL::SendDraw()
GL_PERF("Split single draw in %d draw", m_index.tail / nb_vertex); GL_PERF("Split single draw in %d draw", m_index.tail / nb_vertex);
for (size_t p = 0; p < m_index.tail; p += nb_vertex) { for (size_t p = 0; p < m_index.tail; p += nb_vertex)
{
glTextureBarrier(); glTextureBarrier();
dev->DrawIndexedPrimitive(p, nb_vertex); dev->DrawIndexedPrimitive(p, nb_vertex);
} }
@ -948,8 +1077,7 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
tex && tex->m_texture ? tex->m_texture->GetID() : -1, tex && tex->m_texture ? tex->m_texture->GetID() : -1,
area_in.x, area_in.y, area_in.z, area_in.w, area_in.x, area_in.y, area_in.z, area_in.w,
rt ? rt->GetID() : -1, ds ? ds->GetID() : -1, rt ? rt->GetID() : -1, ds ? ds->GetID() : -1,
area_out.x, area_out.y, area_out.z, area_out.w area_out.x, area_out.y, area_out.z, area_out.w);
);
#endif #endif
GSTexture* hdr_rt = NULL; GSTexture* hdr_rt = NULL;
@ -985,15 +1113,19 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
m_prim_overlap = PrimitiveOverlap(); m_prim_overlap = PrimitiveOverlap();
// Detect framebuffer read that will need special handling // Detect framebuffer read that will need special handling
if ((m_context->FRAME.Block() == m_context->TEX0.TBP0) && PRIM->TME && m_sw_blending) { if ((m_context->FRAME.Block() == m_context->TEX0.TBP0) && PRIM->TME && m_sw_blending)
if ((m_context->FRAME.FBMSK == 0x00FFFFFF) && (m_vt.m_primclass == GS_TRIANGLE_CLASS)) { {
if ((m_context->FRAME.FBMSK == 0x00FFFFFF) && (m_vt.m_primclass == GS_TRIANGLE_CLASS))
{
// This pattern is used by several games to emulate a stencil (shadow) // This pattern is used by several games to emulate a stencil (shadow)
// Ratchet & Clank, Jak do alpha integer multiplication (tfx) which is mostly equivalent to +1/-1 // Ratchet & Clank, Jak do alpha integer multiplication (tfx) which is mostly equivalent to +1/-1
// Tri-Ace (Star Ocean 3/RadiataStories/VP2) uses a palette to handle the +1/-1 // Tri-Ace (Star Ocean 3/RadiataStories/VP2) uses a palette to handle the +1/-1
GL_DBG("Source and Target are the same! Let's sample the framebuffer"); GL_DBG("Source and Target are the same! Let's sample the framebuffer");
m_ps_sel.tex_is_fb = 1; m_ps_sel.tex_is_fb = 1;
m_require_full_barrier = true; m_require_full_barrier = true;
} else if (m_prim_overlap != PRIM_OVERLAP_NO) { }
else if (m_prim_overlap != PRIM_OVERLAP_NO)
{
// Note: It is fine if the texture fits in a single GS page. First access will cache // Note: It is fine if the texture fits in a single GS page. First access will cache
// the page in the GS texture buffer. // the page in the GS texture buffer.
GL_INS("ERROR: Source and Target are the same!"); GL_INS("ERROR: Source and Target are the same!");
@ -1069,20 +1201,25 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// Blend // Blend
if (!IsOpaque() && rt) { if (!IsOpaque() && rt)
{
EmulateBlending(DATE_GL42, DATE_GL45); EmulateBlending(DATE_GL42, DATE_GL45);
} else { }
else
{
dev->OMSetBlendState(); // No blending please dev->OMSetBlendState(); // No blending please
} }
if (m_ps_sel.dfmt == 1) { if (m_ps_sel.dfmt == 1)
{
// Disable writing of the alpha channel // Disable writing of the alpha channel
m_om_csel.wa = 0; m_om_csel.wa = 0;
} }
// DATE setup, no DATE_GL45 please // DATE setup, no DATE_GL45 please
if (DATE && !DATE_GL45) { if (DATE && !DATE_GL45)
{
GSVector4i dRect = ComputeBoundingBox(rtscale, rtsize); GSVector4i dRect = ComputeBoundingBox(rtscale, rtsize);
// Reduce the quantity of clean function // Reduce the quantity of clean function
@ -1091,11 +1228,16 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// Must be done here to avoid any GL state pertubation (clear function...) // Must be done here to avoid any GL state pertubation (clear function...)
// Create an r32ui image that will containt primitive ID // Create an r32ui image that will containt primitive ID
if (DATE_GL42) { if (DATE_GL42)
{
dev->InitPrimDateTexture(rt, dRect); dev->InitPrimDateTexture(rt, dRect);
} else if (DATE_one) { }
else if (DATE_one)
{
dev->ClearStencil(ds, 1); dev->ClearStencil(ds, 1);
} else { }
else
{
GSVector4 src = GSVector4(dRect) / GSVector4(rtsize.x, rtsize.y).xyxy(); GSVector4 src = GSVector4(dRect) / GSVector4(rtsize.x, rtsize.y).xyxy();
GSVector4 dst = src * 2.0f - 1.0f; GSVector4 dst = src * 2.0f - 1.0f;
@ -1147,14 +1289,19 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// GS_SPRITE_CLASS are already flat (either by CPU or the GS) // GS_SPRITE_CLASS are already flat (either by CPU or the GS)
m_ps_sel.iip = (m_vt.m_primclass == GS_SPRITE_CLASS) ? 1 : PRIM->IIP; m_ps_sel.iip = (m_vt.m_primclass == GS_SPRITE_CLASS) ? 1 : PRIM->IIP;
if (DATE_GL45) { if (DATE_GL45)
{
m_ps_sel.date = 5 + m_context->TEST.DATM; m_ps_sel.date = 5 + m_context->TEST.DATM;
} else if (DATE_one) { }
else if (DATE_one)
{
m_require_one_barrier = true; m_require_one_barrier = true;
m_ps_sel.date = 5 + m_context->TEST.DATM; m_ps_sel.date = 5 + m_context->TEST.DATM;
m_om_dssel.date = 1; m_om_dssel.date = 1;
m_om_dssel.date_one = 1; m_om_dssel.date_one = 1;
} else if (DATE) { }
else if (DATE)
{
if (DATE_GL42) if (DATE_GL42)
m_ps_sel.date = 1 + m_context->TEST.DATM; m_ps_sel.date = 1 + m_context->TEST.DATM;
else else
@ -1191,7 +1338,8 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
bool ate_RGBA_then_Z = false; bool ate_RGBA_then_Z = false;
bool ate_RGB_then_ZA = false; bool ate_RGB_then_ZA = false;
uint8 ps_atst = 0; uint8 ps_atst = 0;
if (ate_first_pass & ate_second_pass) { if (ate_first_pass & ate_second_pass)
{
GL_DBG("Complex Alpha Test"); GL_DBG("Complex Alpha Test");
const bool commutative_depth = (m_om_dssel.ztst == ZTST_GEQUAL && m_vt.m_eq.z) || (m_om_dssel.ztst == ZTST_ALWAYS); const bool commutative_depth = (m_om_dssel.ztst == ZTST_GEQUAL && m_vt.m_eq.z) || (m_om_dssel.ztst == ZTST_ALWAYS);
const bool commutative_alpha = (m_context->ALPHA.C != 1); // when either Alpha Src or a constant const bool commutative_alpha = (m_context->ALPHA.C != 1); // when either Alpha Src or a constant
@ -1200,38 +1348,48 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
ate_RGB_then_ZA = (m_context->TEST.AFAIL == AFAIL_RGB_ONLY) & commutative_depth & commutative_alpha; ate_RGB_then_ZA = (m_context->TEST.AFAIL == AFAIL_RGB_ONLY) & commutative_depth & commutative_alpha;
} }
if (ate_RGBA_then_Z) { if (ate_RGBA_then_Z)
{
GL_DBG("Alternate ATE handling: ate_RGBA_then_Z"); GL_DBG("Alternate ATE handling: ate_RGBA_then_Z");
// Render all color but don't update depth // Render all color but don't update depth
// ATE is disabled here // ATE is disabled here
m_om_dssel.zwe = false; m_om_dssel.zwe = false;
} else if (ate_RGB_then_ZA) { }
else if (ate_RGB_then_ZA)
{
GL_DBG("Alternate ATE handling: ate_RGB_then_ZA"); GL_DBG("Alternate ATE handling: ate_RGB_then_ZA");
// Render RGB color but don't update depth/alpha // Render RGB color but don't update depth/alpha
// ATE is disabled here // ATE is disabled here
m_om_dssel.zwe = false; m_om_dssel.zwe = false;
m_om_csel.wa = false; m_om_csel.wa = false;
} else { }
else
{
EmulateAtst(ps_cb.FogColor_AREF, ps_atst, false); EmulateAtst(ps_cb.FogColor_AREF, ps_atst, false);
m_ps_sel.atst = ps_atst; m_ps_sel.atst = ps_atst;
} }
if (tex) { if (tex)
{
EmulateTextureSampler(tex); EmulateTextureSampler(tex);
} else { }
else
{
m_ps_sel.tfx = 4; m_ps_sel.tfx = 4;
} }
// Always bind the RT. This way special effect can use it. // Always bind the RT. This way special effect can use it.
dev->PSSetShaderResource(3, rt); dev->PSSetShaderResource(3, rt);
if (m_game.title == CRC::ICO) { if (m_game.title == CRC::ICO)
{
GSVertex* v = &m_vertex.buff[0]; GSVertex* v = &m_vertex.buff[0];
const GSVideoMode mode = GetVideoMode(); const GSVideoMode mode = GetVideoMode();
if (tex && m_vt.m_primclass == GS_SPRITE_CLASS && m_vertex.next == 2 && PRIM->ABE && // Blend texture if (tex && m_vt.m_primclass == GS_SPRITE_CLASS && m_vertex.next == 2 && PRIM->ABE && // Blend texture
((v[1].U == 8200 && v[1].V == 7176 && mode == GSVideoMode::NTSC) || // at display resolution 512x448 ((v[1].U == 8200 && v[1].V == 7176 && mode == GSVideoMode::NTSC) || // at display resolution 512x448
(v[1].U == 8200 && v[1].V == 8200 && mode == GSVideoMode::PAL)) && // at display resolution 512x512 (v[1].U == 8200 && v[1].V == 8200 && mode == GSVideoMode::PAL)) && // at display resolution 512x512
tex->m_TEX0.PSM == PSM_PSMT8H) { // i.e. read the alpha channel of a 32 bits texture tex->m_TEX0.PSM == PSM_PSMT8H) // i.e. read the alpha channel of a 32 bits texture
{
// Note potentially we can limit to TBP0:0x2800 // Note potentially we can limit to TBP0:0x2800
// Depth buffer was moved so GSdx will invalide it which means a // Depth buffer was moved so GSdx will invalide it which means a
@ -1248,7 +1406,8 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
dev->PSSetShaderResource(4, ds); dev->PSSetShaderResource(4, ds);
// We need the palette to convert the depth to the correct alpha value. // We need the palette to convert the depth to the correct alpha value.
if (!tex->m_palette) { if (!tex->m_palette)
{
uint16 pal = GSLocalMemory::m_psm[tex->m_TEX0.PSM].pal; uint16 pal = GSLocalMemory::m_psm[tex->m_TEX0.PSM].pal;
m_tc->AttachPaletteToSource(tex, pal, true); m_tc->AttachPaletteToSource(tex, pal, true);
dev->PSSetShaderResource(1, tex->m_palette); dev->PSSetShaderResource(1, tex->m_palette);
@ -1277,7 +1436,8 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
if (ds) if (ds)
ds->CommitRegion(GSVector2i(commitRect.z, commitRect.w)); ds->CommitRegion(GSVector2i(commitRect.z, commitRect.w));
if (DATE_GL42) { if (DATE_GL42)
{
GL_PUSH("Date GL42"); GL_PUSH("Date GL42");
// It could be good idea to use stencil in the same time. // It could be good idea to use stencil in the same time.
// Early stencil test will reduce the number of atomic-load operation // Early stencil test will reduce the number of atomic-load operation
@ -1306,13 +1466,16 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
dev->Barrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT); dev->Barrier(GL_SHADER_IMAGE_ACCESS_BARRIER_BIT);
} }
if (m_ps_sel.hdr) { if (m_ps_sel.hdr)
{
hdr_rt = dev->CreateTexture(rtsize.x, rtsize.y, GL_RGBA32F); hdr_rt = dev->CreateTexture(rtsize.x, rtsize.y, GL_RGBA32F);
dev->CopyRectConv(rt, hdr_rt, ComputeBoundingBox(rtscale, rtsize), false); dev->CopyRectConv(rt, hdr_rt, ComputeBoundingBox(rtscale, rtsize), false);
dev->OMSetRenderTargets(hdr_rt, ds, &scissor); dev->OMSetRenderTargets(hdr_rt, ds, &scissor);
} else { }
else
{
dev->OMSetRenderTargets(rt, ds, &scissor); dev->OMSetRenderTargets(rt, ds, &scissor);
} }
@ -1362,10 +1525,13 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// Depth test should be disabled when depth writes are masked and similarly, Alpha test must be disabled // Depth test should be disabled when depth writes are masked and similarly, Alpha test must be disabled
// when writes to all of the alpha bits in the Framebuffer are masked. // when writes to all of the alpha bits in the Framebuffer are masked.
if (ate_RGBA_then_Z) { if (ate_RGBA_then_Z)
{
z = !m_context->ZBUF.ZMSK; z = !m_context->ZBUF.ZMSK;
r = g = b = a = false; r = g = b = a = false;
} else if (ate_RGB_then_ZA) { }
else if (ate_RGB_then_ZA)
{
z = !m_context->ZBUF.ZMSK; z = !m_context->ZBUF.ZMSK;
a = (m_context->FRAME.FBMSK & 0xFF000000) != 0xFF000000; a = (m_context->FRAME.FBMSK & 0xFF000000) != 0xFF000000;
r = g = b = false; r = g = b = false;
@ -1386,7 +1552,8 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
} }
} }
if (DATE_GL42) { if (DATE_GL42)
{
dev->RecycleDateTexture(); dev->RecycleDateTexture();
} }
@ -1394,7 +1561,8 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
// Warning: EndScene must be called before StretchRect otherwise // Warning: EndScene must be called before StretchRect otherwise
// vertices will be overwritten. Trust me you don't want to do that. // vertices will be overwritten. Trust me you don't want to do that.
if (hdr_rt) { if (hdr_rt)
{
GSVector4 dRect(ComputeBoundingBox(rtscale, rtsize)); GSVector4 dRect(ComputeBoundingBox(rtscale, rtsize));
GSVector4 sRect = dRect / GSVector4(rtsize.x, rtsize.y).xyxy(); GSVector4 sRect = dRect / GSVector4(rtsize.x, rtsize.y).xyxy();
dev->StretchRect(hdr_rt, sRect, rt, dRect, ShaderConvert_MOD_256, false); dev->StretchRect(hdr_rt, sRect, rt, dRect, ShaderConvert_MOD_256, false);

8
plugins/GSdx/Renderers/OpenGL/GSRendererOGL.h
View File

@ -27,13 +27,15 @@
class GSRendererOGL final : public GSRendererHW class GSRendererOGL final : public GSRendererHW
{ {
enum PRIM_OVERLAP { enum PRIM_OVERLAP
{
PRIM_OVERLAP_UNKNOW, PRIM_OVERLAP_UNKNOW,
PRIM_OVERLAP_YES, PRIM_OVERLAP_YES,
PRIM_OVERLAP_NO PRIM_OVERLAP_NO
}; };
enum ACC_BLEND { enum ACC_BLEND
{
ACC_BLEND_NONE = 0, ACC_BLEND_NONE = 0,
ACC_BLEND_BASIC = 1, ACC_BLEND_BASIC = 1,
ACC_BLEND_MEDIUM = 2, ACC_BLEND_MEDIUM = 2,
@ -73,7 +75,7 @@ class GSRendererOGL final : public GSRendererHW
public: public:
GSRendererOGL(); GSRendererOGL();
virtual ~GSRendererOGL() {}; virtual ~GSRendererOGL() {}
bool CreateDevice(GSDevice* dev); bool CreateDevice(GSDevice* dev);

98
plugins/GSdx/Renderers/OpenGL/GSShaderOGL.cpp
View File

@ -29,9 +29,9 @@
#include "GSdxResources.h" #include "GSdxResources.h"
#endif #endif
GSShaderOGL::GSShaderOGL(bool debug) : GSShaderOGL::GSShaderOGL(bool debug)
m_pipeline(0), : m_pipeline(0)
m_debug_shader(debug) , m_debug_shader(debug)
{ {
theApp.LoadResource(IDR_COMMON_GLSL, m_common_header); theApp.LoadResource(IDR_COMMON_GLSL, m_common_header);
@ -45,8 +45,10 @@ GSShaderOGL::~GSShaderOGL()
printf("Delete %zu Shaders, %zu Programs, %zu Pipelines\n", printf("Delete %zu Shaders, %zu Programs, %zu Pipelines\n",
m_shad_to_delete.size(), m_prog_to_delete.size(), m_pipe_to_delete.size()); m_shad_to_delete.size(), m_prog_to_delete.size(), m_pipe_to_delete.size());
for (auto s : m_shad_to_delete) glDeleteShader(s); for (auto s : m_shad_to_delete)
for (auto p : m_prog_to_delete) glDeleteProgram(p); glDeleteShader(s);
for (auto p : m_prog_to_delete)
glDeleteProgram(p);
glDeleteProgramPipelines(m_pipe_to_delete.size(), &m_pipe_to_delete[0]); glDeleteProgramPipelines(m_pipe_to_delete.size(), &m_pipe_to_delete[0]);
} }
@ -91,7 +93,8 @@ void GSShaderOGL::BindProgram(GLuint vs, GLuint gs, GLuint ps)
{ {
GLuint p = LinkProgram(vs, gs, ps); GLuint p = LinkProgram(vs, gs, ps);
if (GLState::program != p) { if (GLState::program != p)
{
GLState::program = p; GLState::program = p;
glUseProgram(p); glUseProgram(p);
} }
@ -99,7 +102,8 @@ void GSShaderOGL::BindProgram(GLuint vs, GLuint gs, GLuint ps)
void GSShaderOGL::BindProgram(GLuint p) void GSShaderOGL::BindProgram(GLuint p)
{ {
if (GLState::program != p) { if (GLState::program != p)
{
GLState::program = p; GLState::program = p;
glUseProgram(p); glUseProgram(p);
} }
@ -109,12 +113,14 @@ void GSShaderOGL::BindPipeline(GLuint vs, GLuint gs, GLuint ps)
{ {
BindPipeline(m_pipeline); BindPipeline(m_pipeline);
if (GLState::vs != vs) { if (GLState::vs != vs)
{
GLState::vs = vs; GLState::vs = vs;
glUseProgramStages(m_pipeline, GL_VERTEX_SHADER_BIT, vs); glUseProgramStages(m_pipeline, GL_VERTEX_SHADER_BIT, vs);
} }
if (GLState::gs != gs) { if (GLState::gs != gs)
{
GLState::gs = gs; GLState::gs = gs;
glUseProgramStages(m_pipeline, GL_GEOMETRY_SHADER_BIT, gs); glUseProgramStages(m_pipeline, GL_GEOMETRY_SHADER_BIT, gs);
} }
@ -133,12 +139,14 @@ void GSShaderOGL::BindPipeline(GLuint vs, GLuint gs, GLuint ps)
void GSShaderOGL::BindPipeline(GLuint pipe) void GSShaderOGL::BindPipeline(GLuint pipe)
{ {
if (GLState::pipeline != pipe) { if (GLState::pipeline != pipe)
{
GLState::pipeline = pipe; GLState::pipeline = pipe;
glBindProgramPipeline(pipe); glBindProgramPipeline(pipe);
} }
if (GLState::program) { if (GLState::program)
{
GLState::program = 0; GLState::program = 0;
glUseProgram(0); glUseProgram(0);
} }
@ -146,15 +154,18 @@ void GSShaderOGL::BindPipeline(GLuint pipe)
bool GSShaderOGL::ValidateShader(GLuint s) bool GSShaderOGL::ValidateShader(GLuint s)
{ {
if (!m_debug_shader) return true; if (!m_debug_shader)
return true;
GLint status = 0; GLint status = 0;
glGetShaderiv(s, GL_COMPILE_STATUS, &status); glGetShaderiv(s, GL_COMPILE_STATUS, &status);
if (status) return true; if (status)
return true;
GLint log_length = 0; GLint log_length = 0;
glGetShaderiv(s, GL_INFO_LOG_LENGTH, &log_length); glGetShaderiv(s, GL_INFO_LOG_LENGTH, &log_length);
if (log_length > 0) { if (log_length > 0)
{
char* log = new char[log_length]; char* log = new char[log_length];
glGetShaderInfoLog(s, log_length, NULL, log); glGetShaderInfoLog(s, log_length, NULL, log);
fprintf(stderr, "%s", log); fprintf(stderr, "%s", log);
@ -167,15 +178,18 @@ bool GSShaderOGL::ValidateShader(GLuint s)
bool GSShaderOGL::ValidateProgram(GLuint p) bool GSShaderOGL::ValidateProgram(GLuint p)
{ {
if (!m_debug_shader) return true; if (!m_debug_shader)
return true;
GLint status = 0; GLint status = 0;
glGetProgramiv(p, GL_LINK_STATUS, &status); glGetProgramiv(p, GL_LINK_STATUS, &status);
if (status) return true; if (status)
return true;
GLint log_length = 0; GLint log_length = 0;
glGetProgramiv(p, GL_INFO_LOG_LENGTH, &log_length); glGetProgramiv(p, GL_INFO_LOG_LENGTH, &log_length);
if (log_length > 0) { if (log_length > 0)
{
char* log = new char[log_length]; char* log = new char[log_length];
glGetProgramInfoLog(p, log_length, NULL, log); glGetProgramInfoLog(p, log_length, NULL, log);
fprintf(stderr, "%s", log); fprintf(stderr, "%s", log);
@ -188,18 +202,21 @@ bool GSShaderOGL::ValidateProgram(GLuint p)
bool GSShaderOGL::ValidatePipeline(GLuint p) bool GSShaderOGL::ValidatePipeline(GLuint p)
{ {
if (!m_debug_shader) return true; if (!m_debug_shader)
return true;
// FIXME: might be mandatory to validate the pipeline // FIXME: might be mandatory to validate the pipeline
glValidateProgramPipeline(p); glValidateProgramPipeline(p);
GLint status = 0; GLint status = 0;
glGetProgramPipelineiv(p, GL_VALIDATE_STATUS, &status); glGetProgramPipelineiv(p, GL_VALIDATE_STATUS, &status);
if (status) return true; if (status)
return true;
GLint log_length = 0; GLint log_length = 0;
glGetProgramPipelineiv(p, GL_INFO_LOG_LENGTH, &log_length); glGetProgramPipelineiv(p, GL_INFO_LOG_LENGTH, &log_length);
if (log_length > 0) { if (log_length > 0)
{
char* log = new char[log_length]; char* log = new char[log_length];
glGetProgramPipelineInfoLog(p, log_length, NULL, log); glGetProgramPipelineInfoLog(p, log_length, NULL, log);
fprintf(stderr, "%s", log); fprintf(stderr, "%s", log);
@ -218,10 +235,13 @@ std::string GSShaderOGL::GenGlslHeader(const std::string& entry, GLenum type, co
header += "#extension GL_ARB_shading_language_420pack: require\n"; header += "#extension GL_ARB_shading_language_420pack: require\n";
// Need GL version 410 // Need GL version 410
header += "#extension GL_ARB_separate_shader_objects: require\n"; header += "#extension GL_ARB_separate_shader_objects: require\n";
if (GLLoader::found_GL_ARB_shader_image_load_store) { if (GLLoader::found_GL_ARB_shader_image_load_store)
{
// Need GL version 420 // Need GL version 420
header += "#extension GL_ARB_shader_image_load_store: require\n"; header += "#extension GL_ARB_shader_image_load_store: require\n";
} else { }
else
{
header += "#define DISABLE_GL42_image\n"; header += "#define DISABLE_GL42_image\n";
} }
if (GLLoader::vendor_id_amd || GLLoader::vendor_id_intel) if (GLLoader::vendor_id_amd || GLLoader::vendor_id_intel)
@ -234,7 +254,8 @@ std::string GSShaderOGL::GenGlslHeader(const std::string& entry, GLenum type, co
header += "#define pGL_ES 0\n"; header += "#define pGL_ES 0\n";
// Allow to puts several shader in 1 files // Allow to puts several shader in 1 files
switch (type) { switch (type)
{
case GL_VERTEX_SHADER: case GL_VERTEX_SHADER:
header += "#define VERTEX_SHADER 1\n"; header += "#define VERTEX_SHADER 1\n";
break; break;
@ -244,7 +265,8 @@ std::string GSShaderOGL::GenGlslHeader(const std::string& entry, GLenum type, co
case GL_FRAGMENT_SHADER: case GL_FRAGMENT_SHADER:
header += "#define FRAGMENT_SHADER 1\n"; header += "#define FRAGMENT_SHADER 1\n";
break; break;
default: ASSERT(0); default:
ASSERT(0);
} }
// Select the entry point ie the main function // Select the entry point ie the main function
@ -276,7 +298,8 @@ GLuint GSShaderOGL::Compile(const std::string& glsl_file, const std::string& ent
bool status = ValidateProgram(program); bool status = ValidateProgram(program);
if (!status) { if (!status)
{
// print extra info // print extra info
fprintf(stderr, "%s (entry %s, prog %d) :", glsl_file.c_str(), entry.c_str(), program); fprintf(stderr, "%s (entry %s, prog %d) :", glsl_file.c_str(), entry.c_str(), program);
fprintf(stderr, "\n%s", macro_sel.c_str()); fprintf(stderr, "\n%s", macro_sel.c_str());
@ -312,7 +335,8 @@ GLuint GSShaderOGL::CompileShader(const std::string& glsl_file, const std::strin
bool status = ValidateShader(shader); bool status = ValidateShader(shader);
if (!status) { if (!status)
{
// print extra info // print extra info
fprintf(stderr, "%s (entry %s, prog %d) :", glsl_file.c_str(), entry.c_str(), shader); fprintf(stderr, "%s (entry %s, prog %d) :", glsl_file.c_str(), entry.c_str(), shader);
fprintf(stderr, "\n%s", macro_sel.c_str()); fprintf(stderr, "\n%s", macro_sel.c_str());
@ -331,7 +355,8 @@ GLuint GSShaderOGL::CompileShader(const std::string& glsl_file, const std::strin
// GLSL improvement (unfortunately). // GLSL improvement (unfortunately).
int GSShaderOGL::DumpAsm(const std::string& file, GLuint p) int GSShaderOGL::DumpAsm(const std::string& file, GLuint p)
{ {
if (!GLLoader::vendor_id_nvidia) return 0; if (!GLLoader::vendor_id_nvidia)
return 0;
GLint binaryLength; GLint binaryLength;
glGetProgramiv(p, GL_PROGRAM_BINARY_LENGTH, &binaryLength); glGetProgramiv(p, GL_PROGRAM_BINARY_LENGTH, &binaryLength);
@ -345,18 +370,24 @@ int GSShaderOGL::DumpAsm(const std::string& file, GLuint p)
// Search the magic number "!!" // Search the magic number "!!"
int asm_ = 0; int asm_ = 0;
while (asm_ < binaryLength && (binary[asm_] != '!' || binary[asm_+1] != '!')) { while (asm_ < binaryLength && (binary[asm_] != '!' || binary[asm_ + 1] != '!'))
{
asm_ += 1; asm_ += 1;
} }
int instructions = -1; int instructions = -1;
if (asm_ < binaryLength) { if (asm_ < binaryLength)
{
// Now print asm as text // Now print asm as text
char* asm_txt = strtok(&binary[asm_], "\n"); char* asm_txt = strtok(&binary[asm_], "\n");
while (asm_txt != NULL && (strncmp(asm_txt, "END", 3) || !strncmp(asm_txt, "ENDIF", 5))) { while (asm_txt != NULL && (strncmp(asm_txt, "END", 3) || !strncmp(asm_txt, "ENDIF", 5)))
if (!strncmp(asm_txt, "OUT", 3) || !strncmp(asm_txt, "TEMP", 4) || !strncmp(asm_txt, "LONG", 4)) { {
if (!strncmp(asm_txt, "OUT", 3) || !strncmp(asm_txt, "TEMP", 4) || !strncmp(asm_txt, "LONG", 4))
{
instructions = 0; instructions = 0;
} else if (instructions >= 0) { }
else if (instructions >= 0)
{
if (instructions == 0) if (instructions == 0)
fprintf(outfile, "\n"); fprintf(outfile, "\n");
instructions++; instructions++;
@ -369,7 +400,8 @@ int GSShaderOGL::DumpAsm(const std::string& file, GLuint p)
} }
fclose(outfile); fclose(outfile);
if (instructions < 0) { if (instructions < 0)
{
// RAW dump in case of error // RAW dump in case of error
fprintf(stderr, "Error: failed to find the number of instructions!\n"); fprintf(stderr, "Error: failed to find the number of instructions!\n");
outfile = fopen(file.c_str(), "wb"); outfile = fopen(file.c_str(), "wb");

3
plugins/GSdx/Renderers/OpenGL/GSShaderOGL.h
View File

@ -21,7 +21,8 @@
#pragma once #pragma once
class GSShaderOGL { class GSShaderOGL
{
GLuint m_pipeline; GLuint m_pipeline;
std::unordered_map<uint32, GLuint> m_program; std::unordered_map<uint32, GLuint> m_program;
const bool m_debug_shader; const bool m_debug_shader;

6
plugins/GSdx/Renderers/OpenGL/GSTextureCacheOGL.cpp
View File

@ -123,13 +123,15 @@ void GSTextureCacheOGL::Read(Source* t, const GSVector4i& r)
// FIXME Create a get function to avoid the useless copy // FIXME Create a get function to avoid the useless copy
// Note: With openGL 4.5 you can use glGetTextureSubImage // Note: With openGL 4.5 you can use glGetTextureSubImage
if (GSTexture* offscreen = m_renderer->m_dev->CreateOffscreen(r.width(), r.height())) { if (GSTexture* offscreen = m_renderer->m_dev->CreateOffscreen(r.width(), r.height()))
{
m_renderer->m_dev->CopyRect(t->m_texture, offscreen, r); m_renderer->m_dev->CopyRect(t->m_texture, offscreen, r);
GSTexture::GSMap m; GSTexture::GSMap m;
GSVector4i r_offscreen(0, 0, r.width(), r.height()); GSVector4i r_offscreen(0, 0, r.width(), r.height());
if (offscreen->Map(m, &r_offscreen)) { if (offscreen->Map(m, &r_offscreen))
{
GSOffset* off = m_renderer->m_mem.GetOffset(TEX0.TBP0, TEX0.TBW, TEX0.PSM); GSOffset* off = m_renderer->m_mem.GetOffset(TEX0.TBP0, TEX0.TBW, TEX0.PSM);
m_renderer->m_mem.WritePixel32(m.bits, m.pitch, off, r); m_renderer->m_mem.WritePixel32(m.bits, m.pitch, off, r);

133
plugins/GSdx/Renderers/OpenGL/GSTextureOGL.cpp
View File

@ -30,7 +30,8 @@ extern uint64 g_real_texture_upload_byte;
#endif #endif
// FIXME OGL4: investigate, only 1 unpack buffer always bound // FIXME OGL4: investigate, only 1 unpack buffer always bound
namespace PboPool { namespace PboPool
{
const uint32 m_pbo_size = 64 * 1024 * 1024; const uint32 m_pbo_size = 64 * 1024 * 1024;
const uint32 m_seg_size = 16 * 1024 * 1024; const uint32 m_seg_size = 16 * 1024 * 1024;
@ -49,7 +50,8 @@ namespace PboPool {
const GLbitfield map_flags = common_flags | GL_MAP_FLUSH_EXPLICIT_BIT; const GLbitfield map_flags = common_flags | GL_MAP_FLUSH_EXPLICIT_BIT;
const GLbitfield create_flags = common_flags | GL_CLIENT_STORAGE_BIT; const GLbitfield create_flags = common_flags | GL_CLIENT_STORAGE_BIT;
void Init() { void Init()
{
glGenBuffers(1, &m_buffer); glGenBuffers(1, &m_buffer);
BindPbo(); BindPbo();
@ -60,19 +62,22 @@ namespace PboPool {
m_map = (char*)glMapBufferRange(GL_PIXEL_UNPACK_BUFFER, 0, m_pbo_size, map_flags); m_map = (char*)glMapBufferRange(GL_PIXEL_UNPACK_BUFFER, 0, m_pbo_size, map_flags);
m_offset = 0; m_offset = 0;
for (size_t i = 0; i < countof(m_fence); i++) { for (size_t i = 0; i < countof(m_fence); i++)
{
m_fence[i] = 0; m_fence[i] = 0;
} }
UnbindPbo(); UnbindPbo();
} }
char* Map(uint32 size) { char* Map(uint32 size)
{
char* map; char* map;
// Note: keep offset aligned for SSE/AVX // Note: keep offset aligned for SSE/AVX
m_size = (size + 63) & ~0x3F; m_size = (size + 63) & ~0x3F;
if (m_size > m_pbo_size) { if (m_size > m_pbo_size)
{
fprintf(stderr, "BUG: PBO too small %u but need %u\n", m_pbo_size, m_size); fprintf(stderr, "BUG: PBO too small %u but need %u\n", m_pbo_size, m_size);
} }
@ -87,41 +92,50 @@ namespace PboPool {
return map; return map;
} }
void Unmap() { void Unmap()
{
glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, m_offset, m_size); glFlushMappedBufferRange(GL_PIXEL_UNPACK_BUFFER, m_offset, m_size);
} }
uptr Offset() { uptr Offset()
{
return m_offset; return m_offset;
} }
void Destroy() { void Destroy()
{
m_map = NULL; m_map = NULL;
m_offset = 0; m_offset = 0;
for (size_t i = 0; i < countof(m_fence); i++) { for (size_t i = 0; i < countof(m_fence); i++)
{
glDeleteSync(m_fence[i]); glDeleteSync(m_fence[i]);
} }
glDeleteBuffers(1, &m_buffer); glDeleteBuffers(1, &m_buffer);
} }
void BindPbo() { void BindPbo()
{
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, m_buffer); glBindBuffer(GL_PIXEL_UNPACK_BUFFER, m_buffer);
} }
void Sync() { void Sync()
{
uint32 segment_current = m_offset / m_seg_size; uint32 segment_current = m_offset / m_seg_size;
uint32 segment_next = (m_offset + m_size) / m_seg_size; uint32 segment_next = (m_offset + m_size) / m_seg_size;
if (segment_current != segment_next) { if (segment_current != segment_next)
if (segment_next >= countof(m_fence)) { {
if (segment_next >= countof(m_fence))
{
segment_next = 0; segment_next = 0;
} }
// Align current transfer on the start of the segment // Align current transfer on the start of the segment
m_offset = m_seg_size * segment_next; m_offset = m_seg_size * segment_next;
if (m_size > m_seg_size) { if (m_size > m_seg_size)
{
fprintf(stderr, "BUG: PBO Map size %u is bigger than a single segment %u. Crossing more than one fence is not supported yet, texture data may be corrupted.\n", m_size, m_seg_size); fprintf(stderr, "BUG: PBO Map size %u is bigger than a single segment %u. Crossing more than one fence is not supported yet, texture data may be corrupted.\n", m_size, m_seg_size);
// TODO Synchronize all crossed fences // TODO Synchronize all crossed fences
} }
@ -130,10 +144,12 @@ namespace PboPool {
m_fence[segment_current] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); m_fence[segment_current] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
// Check next segment is free // Check next segment is free
if (m_fence[segment_next]) { if (m_fence[segment_next])
{
GLenum status = glClientWaitSync(m_fence[segment_next], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED); GLenum status = glClientWaitSync(m_fence[segment_next], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
// Potentially it doesn't work on AMD driver which might always return GL_CONDITION_SATISFIED // Potentially it doesn't work on AMD driver which might always return GL_CONDITION_SATISFIED
if (status != GL_ALREADY_SIGNALED) { if (status != GL_ALREADY_SIGNALED)
{
GL_PERF("GL_PIXEL_UNPACK_BUFFER: Sync Sync (%x)! Buffer too small ?", status); GL_PERF("GL_PIXEL_UNPACK_BUFFER: Sync Sync (%x)! Buffer too small ?", status);
} }
@ -143,14 +159,16 @@ namespace PboPool {
} }
} }
void UnbindPbo() { void UnbindPbo()
{
glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0); glBindBuffer(GL_PIXEL_UNPACK_BUFFER, 0);
} }
void EndTransfer() { void EndTransfer()
{
m_offset += m_size; m_offset += m_size;
} }
} } // namespace PboPool
GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read, bool mipmap) GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read, bool mipmap)
: m_clean(false), m_generate_mipmap(true), m_local_buffer(nullptr), m_r_x(0), m_r_y(0), m_r_w(0), m_r_h(0), m_layer(0) : m_clean(false), m_generate_mipmap(true), m_local_buffer(nullptr), m_r_x(0), m_r_y(0), m_r_w(0), m_r_h(0), m_layer(0)
@ -166,7 +184,8 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read,
m_max_layer = 1; m_max_layer = 1;
// Bunch of constant parameter // Bunch of constant parameter
switch (m_format) { switch (m_format)
{
// 1 Channel integer // 1 Channel integer
case GL_R32UI: case GL_R32UI:
case GL_R32I: case GL_R32I:
@ -240,7 +259,8 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read,
ASSERT(0); ASSERT(0);
} }
switch (m_type) { switch (m_type)
{
case GSTexture::Backbuffer: case GSTexture::Backbuffer:
return; // backbuffer isn't a real texture return; // backbuffer isn't a real texture
case GSTexture::Offscreen: case GSTexture::Offscreen:
@ -259,7 +279,8 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read,
break; break;
} }
switch (m_format) { switch (m_format)
{
case GL_R16UI: case GL_R16UI:
case GL_R8: case GL_R8:
m_sparse &= GLLoader::found_compatible_GL_ARB_sparse_texture2; m_sparse &= GLLoader::found_compatible_GL_ARB_sparse_texture2;
@ -294,21 +315,26 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read,
// Create a gl object (texture isn't allocated here) // Create a gl object (texture isn't allocated here)
glCreateTextures(GL_TEXTURE_2D, 1, &m_texture_id); glCreateTextures(GL_TEXTURE_2D, 1, &m_texture_id);
if (m_format == GL_R8) { if (m_format == GL_R8)
{
// Emulate DX behavior, beside it avoid special code in shader to differentiate // Emulate DX behavior, beside it avoid special code in shader to differentiate
// palette texture from a GL_RGBA target or a GL_R texture. // palette texture from a GL_RGBA target or a GL_R texture.
glTextureParameteri(m_texture_id, GL_TEXTURE_SWIZZLE_A, GL_RED); glTextureParameteri(m_texture_id, GL_TEXTURE_SWIZZLE_A, GL_RED);
} }
if (m_sparse) { if (m_sparse)
{
GSVector2i old_size = m_size; GSVector2i old_size = m_size;
m_size = RoundUpPage(m_size); m_size = RoundUpPage(m_size);
if (m_size != old_size) { if (m_size != old_size)
{
fprintf(stderr, "Sparse texture size (%dx%d) isn't a multiple of gpu page size (%dx%d)\n", fprintf(stderr, "Sparse texture size (%dx%d) isn't a multiple of gpu page size (%dx%d)\n",
old_size.x, old_size.y, m_gpu_page_size.x, m_gpu_page_size.y); old_size.x, old_size.y, m_gpu_page_size.x, m_gpu_page_size.y);
} }
glTextureParameteri(m_texture_id, GL_TEXTURE_SPARSE_ARB, true); glTextureParameteri(m_texture_id, GL_TEXTURE_SPARSE_ARB, true);
} else { }
else
{
m_committed_size = m_size; m_committed_size = m_size;
} }
@ -316,7 +342,8 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, int format, GLuint fbo_read,
static int every_512 = 0; static int every_512 = 0;
GLState::available_vram -= m_mem_usage; GLState::available_vram -= m_mem_usage;
if ((GLState::available_vram < 0) && (every_512 % 512 == 0)) { if ((GLState::available_vram < 0) && (every_512 % 512 == 0))
{
fprintf(stderr, "Available VRAM is very low (%lld), a crash is expected! Enable conservative buffer allocation or reduce upscaling!\n", GLState::available_vram); fprintf(stderr, "Available VRAM is very low (%lld), a crash is expected! Enable conservative buffer allocation or reduce upscaling!\n", GLState::available_vram);
every_512++; every_512++;
// Pull emergency break // Pull emergency break
@ -334,7 +361,8 @@ GSTextureOGL::~GSTextureOGL()
GLState::rt = 0; GLState::rt = 0;
if (m_texture_id == GLState::ds) if (m_texture_id == GLState::ds)
GLState::ds = 0; GLState::ds = 0;
for (size_t i = 0; i < countof(GLState::tex_unit); i++) { for (size_t i = 0; i < countof(GLState::tex_unit); i++)
{
if (m_texture_id == GLState::tex_unit[i]) if (m_texture_id == GLState::tex_unit[i])
GLState::tex_unit[i] = 0; GLState::tex_unit[i] = 0;
} }
@ -406,7 +434,8 @@ bool GSTextureOGL::Update(const GSVector4i& r, const void* data, int pitch, int
// PERF: slow path of the texture upload. Dunno if we could do better maybe check if TC can keep row_byte == pitch // PERF: slow path of the texture upload. Dunno if we could do better maybe check if TC can keep row_byte == pitch
// Note: row_byte != pitch // Note: row_byte != pitch
for (int h = 0; h < r.height(); h++) { for (int h = 0; h < r.height(); h++)
{
memcpy(map, src, row_byte); memcpy(map, src, row_byte);
map += row_byte; map += row_byte;
src += pitch; src += pitch;
@ -440,7 +469,8 @@ bool GSTextureOGL::Map(GSMap& m, const GSVector4i* _r, int layer)
uint32 row_byte = r.width() << m_int_shift; uint32 row_byte = r.width() << m_int_shift;
m.pitch = row_byte; m.pitch = row_byte;
if (m_type == GSTexture::Offscreen) { if (m_type == GSTexture::Offscreen)
{
// The fastest way will be to use a PBO to read the data asynchronously. Unfortunately GSdx // The fastest way will be to use a PBO to read the data asynchronously. Unfortunately GSdx
// architecture is waiting the data right now. // architecture is waiting the data right now.
@ -467,7 +497,9 @@ bool GSTextureOGL::Map(GSMap& m, const GSVector4i* _r, int layer)
m.bits = m_local_buffer; m.bits = m_local_buffer;
return true; return true;
} else if (m_type == GSTexture::Texture || m_type == GSTexture::RenderTarget) { }
else if (m_type == GSTexture::Texture || m_type == GSTexture::RenderTarget)
{
GL_PUSH_("Upload Texture %d", m_texture_id); // POP is in Unmap GL_PUSH_("Upload Texture %d", m_texture_id); // POP is in Unmap
m_clean = false; m_clean = false;
@ -495,7 +527,8 @@ bool GSTextureOGL::Map(GSMap& m, const GSVector4i* _r, int layer)
void GSTextureOGL::Unmap() void GSTextureOGL::Unmap()
{ {
if (m_type == GSTexture::Texture || m_type == GSTexture::RenderTarget) { if (m_type == GSTexture::Texture || m_type == GSTexture::RenderTarget)
{
PboPool::Unmap(); PboPool::Unmap();
@ -514,7 +547,8 @@ void GSTextureOGL::Unmap()
void GSTextureOGL::GenerateMipmap() void GSTextureOGL::GenerateMipmap()
{ {
if (m_generate_mipmap && m_max_layer > 1) { if (m_generate_mipmap && m_max_layer > 1)
{
glGenerateTextureMipmap(m_texture_id); glGenerateTextureMipmap(m_texture_id);
m_generate_mipmap = false; m_generate_mipmap = false;
} }
@ -524,12 +558,16 @@ void GSTextureOGL::CommitPages(const GSVector2i& region, bool commit)
{ {
GLState::available_vram += m_mem_usage; GLState::available_vram += m_mem_usage;
if (commit) { if (commit)
if (m_committed_size.x == 0) { {
if (m_committed_size.x == 0)
{
// Nothing allocated so far // Nothing allocated so far
GL_INS("CommitPages initial %dx%d of %u", region.x, region.y, m_texture_id); GL_INS("CommitPages initial %dx%d of %u", region.x, region.y, m_texture_id);
glTexturePageCommitmentEXT(m_texture_id, GL_TEX_LEVEL_0, 0, 0, 0, region.x, region.y, 1, commit); glTexturePageCommitmentEXT(m_texture_id, GL_TEX_LEVEL_0, 0, 0, 0, region.x, region.y, 1, commit);
} else { }
else
{
GL_INS("CommitPages extend %dx%d to %dx%d of %u", m_committed_size.x, m_committed_size.y, region.x, region.y, m_texture_id); GL_INS("CommitPages extend %dx%d to %dx%d of %u", m_committed_size.x, m_committed_size.y, region.x, region.y, m_texture_id);
int w = region.x - m_committed_size.x; int w = region.x - m_committed_size.x;
int h = region.y - m_committed_size.y; int h = region.y - m_committed_size.y;
@ -539,8 +577,9 @@ void GSTextureOGL::CommitPages(const GSVector2i& region, bool commit)
glTexturePageCommitmentEXT(m_texture_id, GL_TEX_LEVEL_0, 0, m_committed_size.y, 0, region.x, h, 1, commit); glTexturePageCommitmentEXT(m_texture_id, GL_TEX_LEVEL_0, 0, m_committed_size.y, 0, region.x, h, 1, commit);
} }
m_committed_size = region; m_committed_size = region;
}
} else { else
{
// Release everything // Release everything
GL_INS("CommitPages release of %u", m_texture_id); GL_INS("CommitPages release of %u", m_texture_id);
@ -565,9 +604,12 @@ bool GSTextureOGL::Save(const std::string& fn)
GSPng::Format fmt = GSPng::RGB_PNG; GSPng::Format fmt = GSPng::RGB_PNG;
#endif #endif
if (IsBackbuffer()) { if (IsBackbuffer())
{
glReadPixels(0, 0, m_committed_size.x, m_committed_size.y, GL_RGBA, GL_UNSIGNED_BYTE, image.get()); glReadPixels(0, 0, m_committed_size.x, m_committed_size.y, GL_RGBA, GL_UNSIGNED_BYTE, image.get());
} else if(IsDss()) { }
else if (IsDss())
{
glBindFramebuffer(GL_READ_FRAMEBUFFER, m_fbo_read); glBindFramebuffer(GL_READ_FRAMEBUFFER, m_fbo_read);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, m_texture_id, 0); glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D, m_texture_id, 0);
@ -576,18 +618,23 @@ bool GSTextureOGL::Save(const std::string& fn)
glBindFramebuffer(GL_READ_FRAMEBUFFER, 0); glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
fmt = GSPng::RGB_A_PNG; fmt = GSPng::RGB_A_PNG;
} else if(m_format == GL_R32I) { }
else if (m_format == GL_R32I)
{
// Note: 4.5 function used for accurate DATE // Note: 4.5 function used for accurate DATE
// barely used outside of dev and not sparse anyway // barely used outside of dev and not sparse anyway
glGetTextureImage(m_texture_id, 0, GL_RED_INTEGER, GL_INT, buf_size, image.get()); glGetTextureImage(m_texture_id, 0, GL_RED_INTEGER, GL_INT, buf_size, image.get());
fmt = GSPng::R32I_PNG; fmt = GSPng::R32I_PNG;
} else { }
else
{
glBindFramebuffer(GL_READ_FRAMEBUFFER, m_fbo_read); glBindFramebuffer(GL_READ_FRAMEBUFFER, m_fbo_read);
glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, m_texture_id, 0); glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, m_texture_id, 0);
if (m_format == GL_RGBA8) { if (m_format == GL_RGBA8)
{
glReadPixels(0, 0, m_committed_size.x, m_committed_size.y, GL_RGBA, GL_UNSIGNED_BYTE, image.get()); glReadPixels(0, 0, m_committed_size.x, m_committed_size.y, GL_RGBA, GL_UNSIGNED_BYTE, image.get());
} }
else if (m_format == GL_R16UI) else if (m_format == GL_R16UI)

5
plugins/GSdx/Renderers/OpenGL/GSTextureOGL.h
View File

@ -23,7 +23,8 @@
#include "Renderers/Common/GSTexture.h" #include "Renderers/Common/GSTexture.h"
namespace PboPool { namespace PboPool
{
inline void BindPbo(); inline void BindPbo();
inline void UnbindPbo(); inline void UnbindPbo();
inline void Sync(); inline void Sync();
@ -35,7 +36,7 @@ namespace PboPool {
void Init(); void Init();
void Destroy(); void Destroy();
} } // namespace PboPool
class GSTextureOGL final : public GSTexture class GSTextureOGL final : public GSTexture
{ {

22
plugins/GSdx/Renderers/OpenGL/GSUniformBufferOGL.h
View File

@ -28,7 +28,8 @@ extern uint64 g_uniform_upload_byte;
#endif #endif
class GSUniformBufferOGL { class GSUniformBufferOGL
{
GLuint m_buffer; // data object GLuint m_buffer; // data object
GLuint m_index; // GLSL slot GLuint m_index; // GLSL slot
uint32 m_size; // size of the data uint32 m_size; // size of the data
@ -49,7 +50,8 @@ public:
void bind() void bind()
{ {
if (GLState::ubo != m_buffer) { if (GLState::ubo != m_buffer)
{
GLState::ubo = m_buffer; GLState::ubo = m_buffer;
glBindBuffer(GL_UNIFORM_BUFFER, m_buffer); glBindBuffer(GL_UNIFORM_BUFFER, m_buffer);
} }
@ -82,7 +84,8 @@ public:
void cache_upload(const void* src) void cache_upload(const void* src)
{ {
if (memcmp(m_cache, src, m_size) != 0) { if (memcmp(m_cache, src, m_size) != 0)
{
memcpy(m_cache, src, m_size); memcpy(m_cache, src, m_size);
upload(src); upload(src);
} }
@ -97,7 +100,8 @@ public:
#define UBO_BUFFER_SIZE (4 * 1024 * 1024) #define UBO_BUFFER_SIZE (4 * 1024 * 1024)
class GSUniformBufferStorageOGL { class GSUniformBufferStorageOGL
{
GLuint m_buffer; // data object GLuint m_buffer; // data object
GLuint m_index; // GLSL slot GLuint m_index; // GLSL slot
uint32 m_size; // size of the data uint32 m_size; // size of the data
@ -105,8 +109,8 @@ class GSUniformBufferStorageOGL {
uint32 m_offset; uint32 m_offset;
public: public:
GSUniformBufferStorageOGL(GLuint index, uint32 size) : m_index(index) GSUniformBufferStorageOGL(GLuint index, uint32 size)
, m_size(size), m_offset(0) : m_index(index) , m_size(size) , m_offset(0)
{ {
glGenBuffers(1, &m_buffer); glGenBuffers(1, &m_buffer);
bind(); bind();
@ -116,7 +120,8 @@ public:
void bind() void bind()
{ {
if (GLState::ubo != m_buffer) { if (GLState::ubo != m_buffer)
{
GLState::ubo = m_buffer; GLState::ubo = m_buffer;
glBindBuffer(GL_UNIFORM_BUFFER, m_buffer); glBindBuffer(GL_UNIFORM_BUFFER, m_buffer);
} }
@ -159,7 +164,8 @@ public:
m_offset = 0; m_offset = 0;
} }
~GSUniformBufferStorageOGL() { ~GSUniformBufferStorageOGL()
{
glDeleteBuffers(1, &m_buffer); glDeleteBuffers(1, &m_buffer);
} }
}; };

99
plugins/GSdx/Renderers/OpenGL/GSVertexArrayOGL.h
View File

@ -27,7 +27,8 @@
extern uint64 g_vertex_upload_byte; extern uint64 g_vertex_upload_byte;
#endif #endif
struct GSInputLayoutOGL { struct GSInputLayoutOGL
{
GLint location; GLint location;
GLint size; GLint size;
GLenum type; GLenum type;
@ -37,7 +38,8 @@ struct GSInputLayoutOGL {
}; };
template <int STRIDE> template <int STRIDE>
class GSBufferOGL { class GSBufferOGL
{
size_t m_start; size_t m_start;
size_t m_count; size_t m_count;
size_t m_limit; size_t m_limit;
@ -60,7 +62,8 @@ class GSBufferOGL {
m_limit = 1u << (1u + (size_t)std::log2(count - 1u)); m_limit = 1u << (1u + (size_t)std::log2(count - 1u));
m_quarter_shift = (size_t)std::log2(m_limit * STRIDE) - 2; m_quarter_shift = (size_t)std::log2(m_limit * STRIDE) - 2;
for (size_t i = 0; i < 5; i++) { for (size_t i = 0; i < 5; i++)
{
m_fence[i] = 0; m_fence[i] = 0;
} }
@ -80,14 +83,17 @@ class GSBufferOGL {
glBufferStorage(m_target, STRIDE * m_limit, NULL, create_flags); glBufferStorage(m_target, STRIDE * m_limit, NULL, create_flags);
m_buffer_ptr = (uint8*)glMapBufferRange(m_target, 0, STRIDE * m_limit, map_flags); m_buffer_ptr = (uint8*)glMapBufferRange(m_target, 0, STRIDE * m_limit, map_flags);
if (!m_buffer_ptr) { if (!m_buffer_ptr)
{
fprintf(stderr, "Failed to map buffer\n"); fprintf(stderr, "Failed to map buffer\n");
throw GSDXError(); throw GSDXError();
} }
} }
~GSBufferOGL() { ~GSBufferOGL()
for (size_t i = 0; i < 5; i++) { {
for (size_t i = 0; i < 5; i++)
{
glDeleteSync(m_fence[i]); glDeleteSync(m_fence[i]);
} }
glDeleteBuffers(1, &m_buffer_name); glDeleteBuffers(1, &m_buffer_name);
@ -108,14 +114,16 @@ class GSBufferOGL {
size_t offset = m_start * STRIDE; size_t offset = m_start * STRIDE;
size_t length = m_count * STRIDE; size_t length = m_count * STRIDE;
if (m_count > (m_limit - m_start) ) { if (m_count > (m_limit - m_start))
{
size_t current_chunk = offset >> m_quarter_shift; size_t current_chunk = offset >> m_quarter_shift;
#ifdef ENABLE_OGL_DEBUG_FENCE #ifdef ENABLE_OGL_DEBUG_FENCE
fprintf(stderr, "%x: Wrap buffer\n", m_target); fprintf(stderr, "%x: Wrap buffer\n", m_target);
fprintf(stderr, "%x: Insert a fence in chunk %zu\n", m_target, current_chunk); fprintf(stderr, "%x: Insert a fence in chunk %zu\n", m_target, current_chunk);
#endif #endif
ASSERT(current_chunk > 0 && current_chunk < 5); ASSERT(current_chunk > 0 && current_chunk < 5);
if (m_fence[current_chunk] == 0) { if (m_fence[current_chunk] == 0)
{
m_fence[current_chunk] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); m_fence[current_chunk] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
} }
@ -124,10 +132,12 @@ class GSBufferOGL {
offset = 0; offset = 0;
// Only check first chunk // Only check first chunk
if (m_fence[0]) { if (m_fence[0])
{
#ifdef ENABLE_OGL_DEBUG_FENCE #ifdef ENABLE_OGL_DEBUG_FENCE
GLenum status = glClientWaitSync(m_fence[0], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED); GLenum status = glClientWaitSync(m_fence[0], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
if (status != GL_ALREADY_SIGNALED) { if (status != GL_ALREADY_SIGNALED)
{
fprintf(stderr, "%x: Sync Sync! Buffer too small\n", m_target); fprintf(stderr, "%x: Sync Sync! Buffer too small\n", m_target);
} }
#else #else
@ -141,13 +151,15 @@ class GSBufferOGL {
// Protect buffer with fences // Protect buffer with fences
size_t current_chunk = offset >> m_quarter_shift; size_t current_chunk = offset >> m_quarter_shift;
size_t next_chunk = (offset + length) >> m_quarter_shift; size_t next_chunk = (offset + length) >> m_quarter_shift;
for (size_t c = current_chunk + 1; c <= next_chunk; c++) { for (size_t c = current_chunk + 1; c <= next_chunk; c++)
{
#ifdef ENABLE_OGL_DEBUG_FENCE #ifdef ENABLE_OGL_DEBUG_FENCE
fprintf(stderr, "%x: Insert a fence in chunk %d\n", m_target, c - 1); fprintf(stderr, "%x: Insert a fence in chunk %d\n", m_target, c - 1);
#endif #endif
ASSERT(c > 0 && c < 5); ASSERT(c > 0 && c < 5);
m_fence[c - 1] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0); m_fence[c - 1] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
if (m_fence[c]) { if (m_fence[c])
{
#ifdef ENABLE_OGL_DEBUG_FENCE #ifdef ENABLE_OGL_DEBUG_FENCE
GLenum status = glClientWaitSync(m_fence[c], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED); GLenum status = glClientWaitSync(m_fence[c], GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
#else #else
@ -157,7 +169,8 @@ class GSBufferOGL {
m_fence[c] = 0; m_fence[c] = 0;
#ifdef ENABLE_OGL_DEBUG_FENCE #ifdef ENABLE_OGL_DEBUG_FENCE
if (status != GL_ALREADY_SIGNALED) { if (status != GL_ALREADY_SIGNALED)
{
fprintf(stderr, "%x: Sync Sync! Buffer too small\n", m_target); fprintf(stderr, "%x: Sync Sync! Buffer too small\n", m_target);
} }
#endif #endif
@ -211,10 +224,10 @@ class GSBufferOGL {
} }
size_t GetStart() { return m_start; } size_t GetStart() { return m_start; }
}; };
class GSVertexBufferStateOGL { class GSVertexBufferStateOGL
{
std::unique_ptr<GSBufferOGL<sizeof(GSVertexPT1)>> m_vb; std::unique_ptr<GSBufferOGL<sizeof(GSVertexPT1)>> m_vb;
std::unique_ptr<GSBufferOGL<sizeof(uint32)>> m_ib; std::unique_ptr<GSBufferOGL<sizeof(uint32)>> m_ib;
@ -226,7 +239,8 @@ class GSVertexBufferStateOGL {
GSVertexBufferStateOGL(const GSVertexBufferStateOGL&) = delete; GSVertexBufferStateOGL(const GSVertexBufferStateOGL&) = delete;
public: public:
GSVertexBufferStateOGL(const std::vector<GSInputLayoutOGL>& layout) : m_topology(0), m_layout(layout) GSVertexBufferStateOGL(const std::vector<GSInputLayoutOGL>& layout)
: m_topology(0), m_layout(layout)
{ {
glGenVertexArrays(1, &m_va); glGenVertexArrays(1, &m_va);
glBindVertexArray(m_va); glBindVertexArray(m_va);
@ -250,15 +264,20 @@ public:
void set_internal_format() void set_internal_format()
{ {
for (const auto &l : m_layout) { for (const auto& l : m_layout)
{
// Note this function need both a vertex array object and a GL_ARRAY_BUFFER buffer // Note this function need both a vertex array object and a GL_ARRAY_BUFFER buffer
glEnableVertexAttribArray(l.location); glEnableVertexAttribArray(l.location);
switch (l.type) { switch (l.type)
{
case GL_UNSIGNED_SHORT: case GL_UNSIGNED_SHORT:
case GL_UNSIGNED_INT: case GL_UNSIGNED_INT:
if (l.normalize) { if (l.normalize)
{
glVertexAttribPointer(l.location, l.size, l.type, l.normalize, l.stride, l.offset); glVertexAttribPointer(l.location, l.size, l.type, l.normalize, l.stride, l.offset);
} else { }
else
{
// Rule: when shader use integral (not normalized) you must use glVertexAttribIPointer (note the extra I) // Rule: when shader use integral (not normalized) you must use glVertexAttribIPointer (note the extra I)
glVertexAttribIPointer(l.location, l.size, l.type, l.stride, l.offset); glVertexAttribIPointer(l.location, l.size, l.type, l.stride, l.offset);
} }
@ -286,13 +305,18 @@ public:
void SetTopology(GLenum topology) { m_topology = topology; } void SetTopology(GLenum topology) { m_topology = topology; }
void* MapVB(size_t count) { void* MapVB(size_t count)
{
void* ptr; void* ptr;
while (true) { while (true)
try { {
try
{
ptr = m_vb->map(count); ptr = m_vb->map(count);
break; break;
} catch (GSDXErrorGlVertexArrayTooSmall) { }
catch (GSDXErrorGlVertexArrayTooSmall)
{
GL_INS("GL vertex buffer is too small"); GL_INS("GL vertex buffer is too small");
m_vb.reset(new GSBufferOGL<sizeof(GSVertexPT1)>(GL_ARRAY_BUFFER, count)); m_vb.reset(new GSBufferOGL<sizeof(GSVertexPT1)>(GL_ARRAY_BUFFER, count));
@ -304,12 +328,17 @@ public:
return ptr; return ptr;
} }
void UnmapVB() { m_vb->unmap(); } void UnmapVB() { m_vb->unmap(); }
void UploadVB(const void* vertices, size_t count) { void UploadVB(const void* vertices, size_t count)
while (true) { {
try { while (true)
{
try
{
m_vb->upload(vertices, count); m_vb->upload(vertices, count);
break; break;
} catch (GSDXErrorGlVertexArrayTooSmall) { }
catch (GSDXErrorGlVertexArrayTooSmall)
{
GL_INS("GL vertex buffer is too small"); GL_INS("GL vertex buffer is too small");
m_vb.reset(new GSBufferOGL<sizeof(GSVertexPT1)>(GL_ARRAY_BUFFER, count)); m_vb.reset(new GSBufferOGL<sizeof(GSVertexPT1)>(GL_ARRAY_BUFFER, count));
@ -319,12 +348,17 @@ public:
} }
} }
void UploadIB(const void* index, size_t count) { void UploadIB(const void* index, size_t count)
while (true) { {
try { while (true)
{
try
{
m_ib->upload(index, count); m_ib->upload(index, count);
break; break;
} catch (GSDXErrorGlVertexArrayTooSmall) { }
catch (GSDXErrorGlVertexArrayTooSmall)
{
GL_INS("GL index buffer is too small"); GL_INS("GL index buffer is too small");
m_ib.reset(new GSBufferOGL<sizeof(uint32)>(GL_ELEMENT_ARRAY_BUFFER, count)); m_ib.reset(new GSBufferOGL<sizeof(uint32)>(GL_ELEMENT_ARRAY_BUFFER, count));
@ -336,5 +370,4 @@ public:
{ {
glDeleteVertexArrays(1, &m_va); glDeleteVertexArrays(1, &m_va);
} }
}; };

96
plugins/GSdx/Renderers/SW/GSDrawScanline.cpp
View File

@ -272,7 +272,8 @@ void GSDrawScanline::SetupPrim(const GSVertexSW* vertex, const uint32* index, co
c = c.upl16(c.zwxy()); c = c.upl16(c.zwxy());
if(sel.tfx == TFX_NONE) c = c.srl16(7); if (sel.tfx == TFX_NONE)
c = c.srl16(7);
m_local.c.rb = c.xxxx(); m_local.c.rb = c.xxxx();
m_local.c.ga = c.zzzz(); m_local.c.ga = c.zzzz();
@ -418,7 +419,8 @@ void GSDrawScanline::SetupPrim(const GSVertexSW* vertex, const uint32* index, co
c = c.upl16(c.zwxy()); c = c.upl16(c.zwxy());
if(sel.tfx == TFX_NONE) c = c.srl16(7); if (sel.tfx == TFX_NONE)
c = c.srl16(7);
m_local.c.rb = c.xxxx(); m_local.c.rb = c.xxxx();
m_local.c.ga = c.zzzz(); m_local.c.ga = c.zzzz();
@ -596,7 +598,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
case ZTST_GREATER: test |= zso <= zdo; break; case ZTST_GREATER: test |= zso <= zdo; break;
} }
if(test.alltrue()) continue; if (test.alltrue())
continue;
} }
} }
@ -904,7 +907,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
ga2 = c00.srl16(8); ga2 = c00.srl16(8);
} }
if(sel.lcm) lodf = m_global.lod.f; if (sel.lcm)
lodf = m_global.lod.f;
lodf = lodf.srl16(1); lodf = lodf.srl16(1);
@ -1042,16 +1046,19 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
{ {
case TFX_MODULATE: case TFX_MODULATE:
ga = ga.modulate16<1>(gaf).clamp8(); ga = ga.modulate16<1>(gaf).clamp8();
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_DECAL: case TFX_DECAL:
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_HIGHLIGHT: case TFX_HIGHLIGHT:
ga = ga.mix16(!sel.tcc ? gaf.srl16(7) : ga.addus8(gaf.srl16(7))); ga = ga.mix16(!sel.tcc ? gaf.srl16(7) : ga.addus8(gaf.srl16(7)));
break; break;
case TFX_HIGHLIGHT2: case TFX_HIGHLIGHT2:
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_NONE: case TFX_NONE:
ga = sel.iip ? gaf.srl16(7) : gaf; ga = sel.iip ? gaf.srl16(7) : gaf;
@ -1089,7 +1096,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
// TestAlpha // TestAlpha
if(!TestAlpha(test, fm, zm, ga)) continue; if (!TestAlpha(test, fm, zm, ga))
continue;
// ColorTFX // ColorTFX
@ -1180,7 +1188,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
} }
} }
if(test.alltrue()) continue; if (test.alltrue())
continue;
} }
// WriteMask // WriteMask
@ -1479,12 +1488,13 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
} }
} }
} }
} } while (0);
while(0);
if(sel.edge) break; if (sel.edge)
break;
if(steps <= 0) break; if (steps <= 0)
break;
// Step // Step
@ -1718,7 +1728,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
case ZTST_GREATER: test |= zso <= zdo; break; case ZTST_GREATER: test |= zso <= zdo; break;
} }
if(test.alltrue()) continue; if (test.alltrue())
continue;
} }
} }
@ -2039,7 +2050,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
ga2 = c00.srl16(8); ga2 = c00.srl16(8);
} }
if(sel.lcm) lodf = m_global.lod.f; if (sel.lcm)
lodf = m_global.lod.f;
lodf = lodf.srl16(1); lodf = lodf.srl16(1);
@ -2174,16 +2186,19 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
{ {
case TFX_MODULATE: case TFX_MODULATE:
ga = ga.modulate16<1>(gaf).clamp8(); ga = ga.modulate16<1>(gaf).clamp8();
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_DECAL: case TFX_DECAL:
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_HIGHLIGHT: case TFX_HIGHLIGHT:
ga = ga.mix16(!sel.tcc ? gaf.srl16(7) : ga.addus8(gaf.srl16(7))); ga = ga.mix16(!sel.tcc ? gaf.srl16(7) : ga.addus8(gaf.srl16(7)));
break; break;
case TFX_HIGHLIGHT2: case TFX_HIGHLIGHT2:
if(!sel.tcc) ga = ga.mix16(gaf.srl16(7)); if (!sel.tcc)
ga = ga.mix16(gaf.srl16(7));
break; break;
case TFX_NONE: case TFX_NONE:
ga = sel.iip ? gaf.srl16(7) : gaf; ga = sel.iip ? gaf.srl16(7) : gaf;
@ -2221,7 +2236,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
// TestAlpha // TestAlpha
if(!TestAlpha(test, fm, zm, ga)) continue; if (!TestAlpha(test, fm, zm, ga))
continue;
// ColorTFX // ColorTFX
@ -2307,7 +2323,8 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
} }
} }
if(test.alltrue()) continue; if (test.alltrue())
continue;
} }
// WriteMask // WriteMask
@ -2589,12 +2606,13 @@ void GSDrawScanline::DrawScanline(int pixels, int left, int top, const GSVertexS
} }
} }
} }
} } while (0);
while(0);
if(sel.edge) break; if (sel.edge)
break;
if(steps <= 0) break; if (steps <= 0)
break;
// Step // Step
@ -2683,15 +2701,18 @@ bool GSDrawScanline::TestAlpha(T& test, T& fm, T& zm, const T& ga)
switch (sel.afail) switch (sel.afail)
{ {
case AFAIL_FB_ONLY: case AFAIL_FB_ONLY:
if(!sel.zwrite) return true; if (!sel.zwrite)
return true;
break; break;
case AFAIL_ZB_ONLY: case AFAIL_ZB_ONLY:
if(!sel.fwrite) return true; if (!sel.fwrite)
return true;
break; break;
case AFAIL_RGB_ONLY: case AFAIL_RGB_ONLY:
if(!sel.zwrite && sel.fpsm == 1) return true; if (!sel.zwrite && sel.fpsm == 1)
return true;
break; break;
} }
@ -2732,7 +2753,8 @@ bool GSDrawScanline::TestAlpha(T& test, T& fm, T& zm, const T& ga)
{ {
case AFAIL_KEEP: case AFAIL_KEEP:
test |= t; test |= t;
if(test.alltrue()) return false; if (test.alltrue())
return false;
break; break;
case AFAIL_FB_ONLY: case AFAIL_FB_ONLY:
@ -2757,7 +2779,8 @@ bool GSDrawScanline::TestAlpha(T& test, T& fm, T& zm, const T& ga)
static const int s_offsets[] = {0, 2, 8, 10, 16, 18, 24, 26}; // columnTable16[0] static const int s_offsets[] = {0, 2, 8, 10, 16, 18, 24, 26}; // columnTable16[0]
template<class T> void GSDrawScanline::WritePixel(const T& src, int addr, int i, uint32 psm) template <class T>
void GSDrawScanline::WritePixel(const T& src, int addr, int i, uint32 psm)
{ {
uint8* dst = (uint8*)m_global.vm + addr * 2 + s_offsets[i] * 2; uint8* dst = (uint8*)m_global.vm + addr * 2 + s_offsets[i] * 2;
@ -2871,7 +2894,8 @@ void GSDrawScanline::DrawRect(const GSVector4i& r, const GSVertexSW& v)
template <class T, bool masked> template <class T, bool masked>
void GSDrawScanline::DrawRectT(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, uint32 c, uint32 m) void GSDrawScanline::DrawRectT(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, uint32 c, uint32 m)
{ {
if(m == 0xffffffff) return; if (m == 0xffffffff)
return;
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
@ -2896,7 +2920,8 @@ void GSDrawScanline::DrawRectT(const int* RESTRICT row, const int* RESTRICT col,
color = color.andnot(mask); color = color.andnot(mask);
c = c & (~m); c = c & (~m);
if(masked) ASSERT(mask.u32[0] != 0); if (masked)
ASSERT(mask.u32[0] != 0);
GSVector4i br = r.ralign<Align_Inside>(GSVector2i(8 * 4 / sizeof(T), 8)); GSVector4i br = r.ralign<Align_Inside>(GSVector2i(8 * 4 / sizeof(T), 8));
@ -2922,7 +2947,8 @@ void GSDrawScanline::DrawRectT(const int* RESTRICT row, const int* RESTRICT col,
template <class T, bool masked> template <class T, bool masked>
void GSDrawScanline::FillRect(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, uint32 c, uint32 m) void GSDrawScanline::FillRect(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, uint32 c, uint32 m)
{ {
if(r.x >= r.z) return; if (r.x >= r.z)
return;
T* vm = (T*)m_global.vm; T* vm = (T*)m_global.vm;
@ -2942,7 +2968,8 @@ void GSDrawScanline::FillRect(const int* RESTRICT row, const int* RESTRICT col,
template <class T, bool masked> template <class T, bool masked>
void GSDrawScanline::FillBlock(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, const GSVector8i& c, const GSVector8i& m) void GSDrawScanline::FillBlock(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, const GSVector8i& c, const GSVector8i& m)
{ {
if(r.x >= r.z) return; if (r.x >= r.z)
return;
T* vm = (T*)m_global.vm; T* vm = (T*)m_global.vm;
@ -2971,7 +2998,8 @@ void GSDrawScanline::FillBlock(const int* RESTRICT row, const int* RESTRICT col,
template <class T, bool masked> template <class T, bool masked>
void GSDrawScanline::FillBlock(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, const GSVector4i& c, const GSVector4i& m) void GSDrawScanline::FillBlock(const int* RESTRICT row, const int* RESTRICT col, const GSVector4i& r, const GSVector4i& c, const GSVector4i& m)
{ {
if(r.x >= r.z) return; if (r.x >= r.z)
return;
T* vm = (T*)m_global.vm; T* vm = (T*)m_global.vm;

5
plugins/GSdx/Renderers/SW/GSDrawScanline.h
View File

@ -86,5 +86,8 @@ public:
#endif #endif
void PrintStats() {m_ds_map.PrintStats();} void PrintStats()
{
m_ds_map.PrintStats();
}
}; };

33
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.cpp
View File

@ -43,9 +43,12 @@ GSDrawScanlineCodeGenerator::GSDrawScanlineCodeGenerator(void* param, uint64 key
if (m_sel.breakpoint) if (m_sel.breakpoint)
db(0xCC); db(0xCC);
try { try
{
Generate(); Generate();
} catch (std::exception& e) { }
catch (std::exception& e)
{
fprintf(stderr, "ERR:GSDrawScanlineCodeGenerator %s\n", e.what()); fprintf(stderr, "ERR:GSDrawScanlineCodeGenerator %s\n", e.what());
} }
} }
@ -63,7 +66,6 @@ void GSDrawScanlineCodeGenerator::modulate16(const Xmm& a, const Operand& f, uin
vpsllw(a, shift + 1); vpsllw(a, shift + 1);
vpmulhw(a, f); vpmulhw(a, f);
} }
} }
else else
{ {
@ -133,7 +135,8 @@ void GSDrawScanlineCodeGenerator::clamp16(const Xmm& a, const Xmm& temp)
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
// Greg: why ? // Greg: why ?
if(m_cpu.has(util::Cpu::tAVX2)) { if (m_cpu.has(util::Cpu::tAVX2))
{
ASSERT(a.isYMM()); ASSERT(a.isYMM());
vpermq(Ymm(a.getIdx()), Ymm(a.getIdx()), _MM_SHUFFLE(3, 1, 2, 0)); // this sucks vpermq(Ymm(a.getIdx()), Ymm(a.getIdx()), _MM_SHUFFLE(3, 1, 2, 0)); // this sucks
} }
@ -227,13 +230,18 @@ void GSDrawScanlineCodeGenerator::split16_2x8(const Xmm& l, const Xmm& h, const
if (m_cpu.has(util::Cpu::tAVX)) if (m_cpu.has(util::Cpu::tAVX))
{ {
if (src == h) { if (src == h)
{
vpsllw(l, src, 8); vpsllw(l, src, 8);
vpsrlw(h, 8); vpsrlw(h, 8);
} else if (src == l) { }
else if (src == l)
{
vpsrlw(h, src, 8); vpsrlw(h, src, 8);
vpsllw(l, 8); vpsllw(l, 8);
} else { }
else
{
vpsllw(l, src, 8); vpsllw(l, src, 8);
vpsrlw(h, src, 8); vpsrlw(h, src, 8);
} }
@ -241,11 +249,16 @@ void GSDrawScanlineCodeGenerator::split16_2x8(const Xmm& l, const Xmm& h, const
} }
else else
{ {
if (src == h) { if (src == h)
{
movdqa(l, src); movdqa(l, src);
} else if (src == l) { }
else if (src == l)
{
movdqa(h, src); movdqa(h, src);
} else { }
else
{
movdqa(l, src); movdqa(l, src);
movdqa(h, src); movdqa(h, src);
} }

116
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.x64.avx.cpp
View File

@ -1668,18 +1668,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(_rb, _dst_rb); break; break;
case 2: vpxor(_rb, _rb); break; case 1:
vmovdqa(_rb, _dst_rb);
break;
case 2:
vpxor(_rb, _rb);
break;
} }
// rb = rb.sub16(c[abb * 2 + 0]); // rb = rb.sub16(c[abb * 2 + 0]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(_rb, xmm5); break; case 0:
case 1: vpsubw(_rb, _dst_rb); break; vpsubw(_rb, xmm5);
case 2: break; break;
case 1:
vpsubw(_rb, _dst_rb);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -1708,9 +1718,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(_rb, xmm5); break; case 0:
case 1: vpaddw(_rb, _dst_rb); break; vpaddw(_rb, xmm5);
case 2: break; break;
case 1:
vpaddw(_rb, _dst_rb);
break;
case 2:
break;
} }
} }
else else
@ -1719,9 +1734,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(_rb, _dst_rb); break; break;
case 2: vpxor(_rb, _rb); break; case 1:
vmovdqa(_rb, _dst_rb);
break;
case 2:
vpxor(_rb, _rb);
break;
} }
} }
@ -1752,18 +1772,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(_ga, _dst_ga); break; break;
case 2: vpxor(_ga, _ga); break; case 1:
vmovdqa(_ga, _dst_ga);
break;
case 2:
vpxor(_ga, _ga);
break;
} }
// ga = ga.sub16(c[abeb * 2 + 1]); // ga = ga.sub16(c[abeb * 2 + 1]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(_ga, xmm5); break; case 0:
case 1: vpsubw(_ga, _dst_ga); break; vpsubw(_ga, xmm5);
case 2: break; break;
case 1:
vpsubw(_ga, _dst_ga);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -1777,9 +1807,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(_ga, xmm5); break; case 0:
case 1: vpaddw(_ga, _dst_ga); break; vpaddw(_ga, xmm5);
case 2: break; break;
case 1:
vpaddw(_ga, _dst_ga);
break;
case 2:
break;
} }
} }
else else
@ -1788,9 +1823,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(_ga, _dst_ga); break; break;
case 2: vpxor(_ga, _ga); break; case 1:
vmovdqa(_ga, _dst_ga);
break;
case 2:
vpxor(_ga, _ga);
break;
} }
} }
@ -1843,7 +1883,6 @@ void GSDrawScanlineCodeGenerator::WriteFrame_AVX()
vpaddw(xmm2, ptr[rax + sizeof(GSVector4i) * 0]); vpaddw(xmm2, ptr[rax + sizeof(GSVector4i) * 0]);
vpaddw(xmm3, ptr[rax + sizeof(GSVector4i) * 1]); vpaddw(xmm3, ptr[rax + sizeof(GSVector4i) * 1]);
} }
if (m_sel.colclamp == 0) if (m_sel.colclamp == 0)
@ -1997,12 +2036,16 @@ void GSDrawScanlineCodeGenerator::WritePixel_AVX(const Xmm& src, const Reg64& ad
switch (psm) switch (psm)
{ {
case 0: case 0:
if(i == 0) vmovd(dst, src); if (i == 0)
else vpextrd(dst, src, i); vmovd(dst, src);
else
vpextrd(dst, src, i);
break; break;
case 1: case 1:
if(i == 0) vmovd(eax, src); if (i == 0)
else vpextrd(eax, src, i); vmovd(eax, src);
else
vpextrd(eax, src, i);
xor(eax, dst); xor(eax, dst);
and(eax, 0xffffff); and(eax, 0xffffff);
xor(dst, eax); xor(dst, eax);
@ -2033,14 +2076,19 @@ void GSDrawScanlineCodeGenerator::ReadTexel_AVX(const Xmm& dst, const Xmm& addr,
const Address& src = m_sel.tlu ? ptr[_m_local__gd__clut + rax * 4] : ptr[_m_local__gd__tex + rax * 4]; const Address& src = m_sel.tlu ? ptr[_m_local__gd__clut + rax * 4] : ptr[_m_local__gd__tex + rax * 4];
// Extract address offset // Extract address offset
if(i == 0) vmovd(eax, addr); if (i == 0)
else vpextrd(eax, addr, i); vmovd(eax, addr);
else
vpextrd(eax, addr, i);
// If clut, load the value as a byte index // If clut, load the value as a byte index
if(m_sel.tlu) movzx(eax, byte[_m_local__gd__tex + rax]); if (m_sel.tlu)
movzx(eax, byte[_m_local__gd__tex + rax]);
if(i == 0) vmovd(dst, src); if (i == 0)
else vpinsrd(dst, src, i); vmovd(dst, src);
else
vpinsrd(dst, src, i);
} }
// Gather example (AVX2). Not faster on Haswell but potentially better on recent CPU // Gather example (AVX2). Not faster on Haswell but potentially better on recent CPU

121
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.x64.avx2.cpp
View File

@ -2559,18 +2559,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm5, ymm0); break; break;
case 2: vpxor(ymm5, ymm5); break; case 1:
vmovdqa(ymm5, ymm0);
break;
case 2:
vpxor(ymm5, ymm5);
break;
} }
// rb = rb.sub16(c[abb * 2 + 0]); // rb = rb.sub16(c[abb * 2 + 0]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(ymm5, ymm4); break; case 0:
case 1: vpsubw(ymm5, ymm0); break; vpsubw(ymm5, ymm4);
case 2: break; break;
case 1:
vpsubw(ymm5, ymm0);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2599,9 +2609,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(ymm5, ymm4); break; case 0:
case 1: vpaddw(ymm5, ymm0); break; vpaddw(ymm5, ymm4);
case 2: break; break;
case 1:
vpaddw(ymm5, ymm0);
break;
case 2:
break;
} }
} }
else else
@ -2610,9 +2625,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm5, ymm0); break; break;
case 2: vpxor(ymm5, ymm5); break; case 1:
vmovdqa(ymm5, ymm0);
break;
case 2:
vpxor(ymm5, ymm5);
break;
} }
} }
@ -2643,18 +2663,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm6, ymm1); break; break;
case 2: vpxor(ymm6, ymm6); break; case 1:
vmovdqa(ymm6, ymm1);
break;
case 2:
vpxor(ymm6, ymm6);
break;
} }
// ga = ga.sub16(c[abeb * 2 + 1]); // ga = ga.sub16(c[abeb * 2 + 1]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(ymm6, ymm4); break; case 0:
case 1: vpsubw(ymm6, ymm1); break; vpsubw(ymm6, ymm4);
case 2: break; break;
case 1:
vpsubw(ymm6, ymm1);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2668,9 +2698,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(ymm6, ymm4); break; case 0:
case 1: vpaddw(ymm6, ymm1); break; vpaddw(ymm6, ymm4);
case 2: break; break;
case 1:
vpaddw(ymm6, ymm1);
break;
case 2:
break;
} }
} }
else else
@ -2679,9 +2714,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm6, ymm1); break; break;
case 2: vpxor(ymm6, ymm6); break; case 1:
vmovdqa(ymm6, ymm1);
break;
case 2:
vpxor(ymm6, ymm6);
break;
} }
} }
@ -2918,19 +2958,25 @@ void GSDrawScanlineCodeGenerator::WritePixel(const Xmm& src, const RegLong& addr
switch (psm) switch (psm)
{ {
case 0: case 0:
if(j == 0) vmovd(dst, src); if (j == 0)
else vpextrd(dst, src, j); vmovd(dst, src);
else
vpextrd(dst, src, j);
break; break;
case 1: case 1:
if(j == 0) vmovd(eax, src); if (j == 0)
else vpextrd(eax, src, j); vmovd(eax, src);
else
vpextrd(eax, src, j);
xor(eax, dst); xor(eax, dst);
and(eax, 0xffffff); and(eax, 0xffffff);
xor(dst, eax); xor(dst, eax);
break; break;
case 2: case 2:
if(j == 0) vmovd(eax, src); if (j == 0)
else vpextrw(eax, src, j * 2); vmovd(eax, src);
else
vpextrw(eax, src, j * 2);
mov(dst, ax); mov(dst, ax);
break; break;
} }
@ -3042,13 +3088,18 @@ void GSDrawScanlineCodeGenerator::ReadTexel(const Ymm& dst, const Ymm& addr, uin
const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4]; const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4];
if(i == 0) vmovd(eax, Xmm(addr.getIdx())); if (i == 0)
else vpextrd(eax, Xmm(addr.getIdx()), i); vmovd(eax, Xmm(addr.getIdx()));
else
vpextrd(eax, Xmm(addr.getIdx()), i);
if(m_sel.tlu) movzx(eax, byte[ebx + eax]); if (m_sel.tlu)
movzx(eax, byte[ebx + eax]);
if(i == 0) vmovd(Xmm(dst.getIdx()), src); if (i == 0)
else vpinsrd(Xmm(dst.getIdx()), src, i); vmovd(Xmm(dst.getIdx()), src);
else
vpinsrd(Xmm(dst.getIdx()), src, i);
} }

121
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.x86.avx.cpp
View File

@ -2472,18 +2472,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(xmm5, xmm0); break; break;
case 2: vpxor(xmm5, xmm5); break; case 1:
vmovdqa(xmm5, xmm0);
break;
case 2:
vpxor(xmm5, xmm5);
break;
} }
// rb = rb.sub16(c[abb * 2 + 0]); // rb = rb.sub16(c[abb * 2 + 0]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(xmm5, xmm4); break; case 0:
case 1: vpsubw(xmm5, xmm0); break; vpsubw(xmm5, xmm4);
case 2: break; break;
case 1:
vpsubw(xmm5, xmm0);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2512,9 +2522,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(xmm5, xmm4); break; case 0:
case 1: vpaddw(xmm5, xmm0); break; vpaddw(xmm5, xmm4);
case 2: break; break;
case 1:
vpaddw(xmm5, xmm0);
break;
case 2:
break;
} }
} }
else else
@ -2523,9 +2538,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(xmm5, xmm0); break; break;
case 2: vpxor(xmm5, xmm5); break; case 1:
vmovdqa(xmm5, xmm0);
break;
case 2:
vpxor(xmm5, xmm5);
break;
} }
} }
@ -2556,18 +2576,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(xmm6, xmm1); break; break;
case 2: vpxor(xmm6, xmm6); break; case 1:
vmovdqa(xmm6, xmm1);
break;
case 2:
vpxor(xmm6, xmm6);
break;
} }
// ga = ga.sub16(c[abeb * 2 + 1]); // ga = ga.sub16(c[abeb * 2 + 1]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(xmm6, xmm4); break; case 0:
case 1: vpsubw(xmm6, xmm1); break; vpsubw(xmm6, xmm4);
case 2: break; break;
case 1:
vpsubw(xmm6, xmm1);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2581,9 +2611,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(xmm6, xmm4); break; case 0:
case 1: vpaddw(xmm6, xmm1); break; vpaddw(xmm6, xmm4);
case 2: break; break;
case 1:
vpaddw(xmm6, xmm1);
break;
case 2:
break;
} }
} }
else else
@ -2592,9 +2627,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_AVX()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(xmm6, xmm1); break; break;
case 2: vpxor(xmm6, xmm6); break; case 1:
vmovdqa(xmm6, xmm1);
break;
case 2:
vpxor(xmm6, xmm6);
break;
} }
} }
@ -2785,19 +2825,25 @@ void GSDrawScanlineCodeGenerator::WritePixel_AVX(const Xmm& src, const Reg32& ad
switch (psm) switch (psm)
{ {
case 0: case 0:
if(i == 0) vmovd(dst, src); if (i == 0)
else vpextrd(dst, src, i); vmovd(dst, src);
else
vpextrd(dst, src, i);
break; break;
case 1: case 1:
if(i == 0) vmovd(eax, src); if (i == 0)
else vpextrd(eax, src, i); vmovd(eax, src);
else
vpextrd(eax, src, i);
xor(eax, dst); xor(eax, dst);
and(eax, 0xffffff); and(eax, 0xffffff);
xor(dst, eax); xor(dst, eax);
break; break;
case 2: case 2:
if(i == 0) vmovd(eax, src); if (i == 0)
else vpextrw(eax, src, i * 2); vmovd(eax, src);
else
vpextrw(eax, src, i * 2);
mov(dst, ax); mov(dst, ax);
break; break;
} }
@ -2878,13 +2924,18 @@ void GSDrawScanlineCodeGenerator::ReadTexel_AVX(const Xmm& dst, const Xmm& addr,
const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4]; const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4];
if(i == 0) vmovd(eax, addr); if (i == 0)
else vpextrd(eax, addr, i); vmovd(eax, addr);
else
vpextrd(eax, addr, i);
if(m_sel.tlu) movzx(eax, byte[ebx + eax]); if (m_sel.tlu)
movzx(eax, byte[ebx + eax]);
if(i == 0) vmovd(dst, src); if (i == 0)
else vpinsrd(dst, src, i); vmovd(dst, src);
else
vpinsrd(dst, src, i);
} }
#endif #endif

121
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.x86.avx2.cpp
View File

@ -2477,18 +2477,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm5, ymm0); break; break;
case 2: vpxor(ymm5, ymm5); break; case 1:
vmovdqa(ymm5, ymm0);
break;
case 2:
vpxor(ymm5, ymm5);
break;
} }
// rb = rb.sub16(c[abb * 2 + 0]); // rb = rb.sub16(c[abb * 2 + 0]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(ymm5, ymm4); break; case 0:
case 1: vpsubw(ymm5, ymm0); break; vpsubw(ymm5, ymm4);
case 2: break; break;
case 1:
vpsubw(ymm5, ymm0);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2517,9 +2527,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(ymm5, ymm4); break; case 0:
case 1: vpaddw(ymm5, ymm0); break; vpaddw(ymm5, ymm4);
case 2: break; break;
case 1:
vpaddw(ymm5, ymm0);
break;
case 2:
break;
} }
} }
else else
@ -2528,9 +2543,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm5, ymm0); break; break;
case 2: vpxor(ymm5, ymm5); break; case 1:
vmovdqa(ymm5, ymm0);
break;
case 2:
vpxor(ymm5, ymm5);
break;
} }
} }
@ -2561,18 +2581,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm6, ymm1); break; break;
case 2: vpxor(ymm6, ymm6); break; case 1:
vmovdqa(ymm6, ymm1);
break;
case 2:
vpxor(ymm6, ymm6);
break;
} }
// ga = ga.sub16(c[abeb * 2 + 1]); // ga = ga.sub16(c[abeb * 2 + 1]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: vpsubw(ymm6, ymm4); break; case 0:
case 1: vpsubw(ymm6, ymm1); break; vpsubw(ymm6, ymm4);
case 2: break; break;
case 1:
vpsubw(ymm6, ymm1);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2586,9 +2616,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: vpaddw(ymm6, ymm4); break; case 0:
case 1: vpaddw(ymm6, ymm1); break; vpaddw(ymm6, ymm4);
case 2: break; break;
case 1:
vpaddw(ymm6, ymm1);
break;
case 2:
break;
} }
} }
else else
@ -2597,9 +2632,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: vmovdqa(ymm6, ymm1); break; break;
case 2: vpxor(ymm6, ymm6); break; case 1:
vmovdqa(ymm6, ymm1);
break;
case 2:
vpxor(ymm6, ymm6);
break;
} }
} }
@ -2836,19 +2876,25 @@ void GSDrawScanlineCodeGenerator::WritePixel(const Xmm& src, const Reg32& addr,
switch (psm) switch (psm)
{ {
case 0: case 0:
if(j == 0) vmovd(dst, src); if (j == 0)
else vpextrd(dst, src, j); vmovd(dst, src);
else
vpextrd(dst, src, j);
break; break;
case 1: case 1:
if(j == 0) vmovd(eax, src); if (j == 0)
else vpextrd(eax, src, j); vmovd(eax, src);
else
vpextrd(eax, src, j);
xor(eax, dst); xor(eax, dst);
and(eax, 0xffffff); and(eax, 0xffffff);
xor(dst, eax); xor(dst, eax);
break; break;
case 2: case 2:
if(j == 0) vmovd(eax, src); if (j == 0)
else vpextrw(eax, src, j * 2); vmovd(eax, src);
else
vpextrw(eax, src, j * 2);
mov(dst, ax); mov(dst, ax);
break; break;
} }
@ -2960,13 +3006,18 @@ void GSDrawScanlineCodeGenerator::ReadTexel(const Ymm& dst, const Ymm& addr, uin
const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4]; const Address& src = m_sel.tlu ? ptr[edx + eax * 4] : ptr[ebx + eax * 4];
if(i == 0) vmovd(eax, Xmm(addr.getIdx())); if (i == 0)
else vpextrd(eax, Xmm(addr.getIdx()), i); vmovd(eax, Xmm(addr.getIdx()));
else
vpextrd(eax, Xmm(addr.getIdx()), i);
if(m_sel.tlu) movzx(eax, byte[ebx + eax]); if (m_sel.tlu)
movzx(eax, byte[ebx + eax]);
if(i == 0) vmovd(Xmm(dst.getIdx()), src); if (i == 0)
else vpinsrd(Xmm(dst.getIdx()), src, i); vmovd(Xmm(dst.getIdx()), src);
else
vpinsrd(Xmm(dst.getIdx()), src, i);
} }

121
plugins/GSdx/Renderers/SW/GSDrawScanlineCodeGenerator.x86.cpp
View File

@ -2481,18 +2481,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: movdqa(xmm5, xmm0); break; break;
case 2: pxor(xmm5, xmm5); break; case 1:
movdqa(xmm5, xmm0);
break;
case 2:
pxor(xmm5, xmm5);
break;
} }
// rb = rb.sub16(c[abb * 2 + 0]); // rb = rb.sub16(c[abb * 2 + 0]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: psubw(xmm5, xmm4); break; case 0:
case 1: psubw(xmm5, xmm0); break; psubw(xmm5, xmm4);
case 2: break; break;
case 1:
psubw(xmm5, xmm0);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2521,9 +2531,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: paddw(xmm5, xmm4); break; case 0:
case 1: paddw(xmm5, xmm0); break; paddw(xmm5, xmm4);
case 2: break; break;
case 1:
paddw(xmm5, xmm0);
break;
case 2:
break;
} }
} }
else else
@ -2532,9 +2547,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: movdqa(xmm5, xmm0); break; break;
case 2: pxor(xmm5, xmm5); break; case 1:
movdqa(xmm5, xmm0);
break;
case 2:
pxor(xmm5, xmm5);
break;
} }
} }
@ -2566,18 +2586,28 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.aba) switch (m_sel.aba)
{ {
case 0: break; case 0:
case 1: movdqa(xmm6, xmm1); break; break;
case 2: pxor(xmm6, xmm6); break; case 1:
movdqa(xmm6, xmm1);
break;
case 2:
pxor(xmm6, xmm6);
break;
} }
// ga = ga.sub16(c[abeb * 2 + 1]); // ga = ga.sub16(c[abeb * 2 + 1]);
switch (m_sel.abb) switch (m_sel.abb)
{ {
case 0: psubw(xmm6, xmm4); break; case 0:
case 1: psubw(xmm6, xmm1); break; psubw(xmm6, xmm4);
case 2: break; break;
case 1:
psubw(xmm6, xmm1);
break;
case 2:
break;
} }
if (!(m_sel.fpsm == 1 && m_sel.abc == 1)) if (!(m_sel.fpsm == 1 && m_sel.abc == 1))
@ -2591,9 +2621,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: paddw(xmm6, xmm4); break; case 0:
case 1: paddw(xmm6, xmm1); break; paddw(xmm6, xmm4);
case 2: break; break;
case 1:
paddw(xmm6, xmm1);
break;
case 2:
break;
} }
} }
else else
@ -2602,9 +2637,14 @@ void GSDrawScanlineCodeGenerator::AlphaBlend_SSE()
switch (m_sel.abd) switch (m_sel.abd)
{ {
case 0: break; case 0:
case 1: movdqa(xmm6, xmm1); break; break;
case 2: pxor(xmm6, xmm6); break; case 1:
movdqa(xmm6, xmm1);
break;
case 2:
pxor(xmm6, xmm6);
break;
} }
} }
@ -2798,14 +2838,18 @@ void GSDrawScanlineCodeGenerator::WritePixel_SSE(const Xmm& src, const Reg32& ad
switch (psm) switch (psm)
{ {
case 0: case 0:
if(i == 0) movd(dst, src); if (i == 0)
else { movd(dst, src);
else
{
pextrd(dst, src, i); pextrd(dst, src, i);
} }
break; break;
case 1: case 1:
if(i == 0) movd(eax, src); if (i == 0)
else { movd(eax, src);
else
{
pextrd(eax, src, i); pextrd(eax, src, i);
} }
xor(eax, dst); xor(eax, dst);
@ -2813,8 +2857,10 @@ void GSDrawScanlineCodeGenerator::WritePixel_SSE(const Xmm& src, const Reg32& ad
xor(dst, eax); xor(dst, eax);
break; break;
case 2: case 2:
if(i == 0) movd(eax, src); if (i == 0)
else pextrw(eax, src, i * 2); movd(eax, src);
else
pextrw(eax, src, i * 2);
mov(dst, ax); mov(dst, ax);
break; break;
} }
@ -2895,13 +2941,18 @@ void GSDrawScanlineCodeGenerator::ReadTexel_SSE(const Xmm& dst, const Xmm& addr,
ASSERT(i == 0 || m_cpu.has(util::Cpu::tSSE41)); ASSERT(i == 0 || m_cpu.has(util::Cpu::tSSE41));
if(i == 0) movd(eax, addr); if (i == 0)
else pextrd(eax, addr, i); movd(eax, addr);
else
pextrd(eax, addr, i);
if(m_sel.tlu) movzx(eax, byte[ebx + eax]); if (m_sel.tlu)
movzx(eax, byte[ebx + eax]);
if(i == 0) movd(dst, src); if (i == 0)
else pinsrd(dst, src, i); movd(dst, src);
else
pinsrd(dst, src, i);
} }
#endif #endif

105
plugins/GSdx/Renderers/SW/GSRasterizer.cpp
View File

@ -26,7 +26,8 @@
int GSRasterizerData::s_counter = 0; int GSRasterizerData::s_counter = 0;
static int compute_best_thread_height(int threads) { static int compute_best_thread_height(int threads)
{
// - for more threads screen segments should be smaller to better distribute the pixels // - for more threads screen segments should be smaller to better distribute the pixels
// - but not too small to keep the threading overhead low // - but not too small to keep the threading overhead low
// - ideal value between 3 and 5, or log2(64 / number of threads) // - ideal value between 3 and 5, or log2(64 / number of threads)
@ -70,7 +71,8 @@ GSRasterizer::~GSRasterizer()
{ {
_aligned_free(m_scanline); _aligned_free(m_scanline);
if(m_edge.buff != NULL) vmfree(m_edge.buff, sizeof(GSVertexSW) * 2048); if (m_edge.buff != NULL)
vmfree(m_edge.buff, sizeof(GSVertexSW) * 2048);
delete m_ds; delete m_ds;
} }
@ -106,7 +108,8 @@ int GSRasterizer::FindMyNextScanline(int top) const
if (m_scanline[i] == 0) if (m_scanline[i] == 0)
{ {
while(m_scanline[++i] == 0); while (m_scanline[++i] == 0)
;
top = i << m_thread_height; top = i << m_thread_height;
} }
@ -135,7 +138,8 @@ void GSRasterizer::Draw(GSRasterizerData* data)
{ {
GSPerfMonAutoTimer pmat(m_perfmon, GSPerfMon::WorkerDraw0 + m_id); GSPerfMonAutoTimer pmat(m_perfmon, GSPerfMon::WorkerDraw0 + m_id);
if(data->vertex != NULL && data->vertex_count == 0 || data->index != NULL && data->index_count == 0) return; if (data->vertex != NULL && data->vertex_count == 0 || data->index != NULL && data->index_count == 0)
return;
m_pixels.actual = 0; m_pixels.actual = 0;
m_pixels.total = 0; m_pixels.total = 0;
@ -177,13 +181,19 @@ void GSRasterizer::Draw(GSRasterizerData* data)
if (index != NULL) if (index != NULL)
{ {
do {DrawLine(vertex, index); index += 2;} do
while(index < index_end); {
DrawLine(vertex, index);
index += 2;
} while (index < index_end);
} }
else else
{ {
do {DrawLine(vertex, tmp_index); vertex += 2;} do
while(vertex < vertex_end); {
DrawLine(vertex, tmp_index);
vertex += 2;
} while (vertex < vertex_end);
} }
break; break;
@ -192,13 +202,19 @@ void GSRasterizer::Draw(GSRasterizerData* data)
if (index != NULL) if (index != NULL)
{ {
do {DrawTriangle(vertex, index); index += 3;} do
while(index < index_end); {
DrawTriangle(vertex, index);
index += 3;
} while (index < index_end);
} }
else else
{ {
do {DrawTriangle(vertex, tmp_index); vertex += 3;} do
while(vertex < vertex_end); {
DrawTriangle(vertex, tmp_index);
vertex += 3;
} while (vertex < vertex_end);
} }
break; break;
@ -207,13 +223,19 @@ void GSRasterizer::Draw(GSRasterizerData* data)
if (index != NULL) if (index != NULL)
{ {
do {DrawSprite(vertex, index); index += 2;} do
while(index < index_end); {
DrawSprite(vertex, index);
index += 2;
} while (index < index_end);
} }
else else
{ {
do {DrawSprite(vertex, tmp_index); vertex += 2;} do
while(vertex < vertex_end); {
DrawSprite(vertex, tmp_index);
vertex += 2;
} while (vertex < vertex_end);
} }
break; break;
@ -370,7 +392,8 @@ void GSRasterizer::DrawLine(const GSVertexSW* vertex, const uint32* index)
} }
} }
if(--steps == 0) break; if (--steps == 0)
break;
edge += dedge; edge += dedge;
} }
@ -428,7 +451,8 @@ void GSRasterizer::DrawTriangle(const GSVertexSW* vertex, const uint32* index)
// if(i == 1) => y0 == y1 < y2 // if(i == 1) => y0 == y1 < y2
// if(i == 4) => y0 < y1 == y2 // if(i == 4) => y0 < y1 == y2
if(m1 == 7) return; // y0 == y1 == y2 if (m1 == 7) // y0 == y1 == y2
return;
GSVector4 tbf = y0011.xzxz(y1221).ceil(); GSVector4 tbf = y0011.xzxz(y1221).ceil();
GSVector4 tbmax = tbf.max(m_fscissor_y); GSVector4 tbmax = tbf.max(m_fscissor_y);
@ -447,7 +471,8 @@ void GSRasterizer::DrawTriangle(const GSVertexSW* vertex, const uint32* index)
int m2 = cross.upl(cross == GSVector4::zero()).mask(); int m2 = cross.upl(cross == GSVector4::zero()).mask();
if(m2 & 2) return; if (m2 & 2)
return;
m2 &= 1; m2 &= 1;
@ -615,7 +640,8 @@ void GSRasterizer::DrawTriangle(const GSVertexSW* vertex, const uint32* index)
// if(i == 1) => y0 == y1 < y2 // if(i == 1) => y0 == y1 < y2
// if(i == 4) => y0 < y1 == y2 // if(i == 4) => y0 < y1 == y2
if(m1 == 7) return; // y0 == y1 == y2 if (m1 == 7)
return; // y0 == y1 == y2
GSVector4 tbf = y0011.xzxz(y1221).ceil(); GSVector4 tbf = y0011.xzxz(y1221).ceil();
GSVector4 tbmax = tbf.max(m_fscissor_y); GSVector4 tbmax = tbf.max(m_fscissor_y);
@ -634,7 +660,8 @@ void GSRasterizer::DrawTriangle(const GSVertexSW* vertex, const uint32* index)
int m2 = cross.upl(cross == GSVector4::zero()).mask(); int m2 = cross.upl(cross == GSVector4::zero()).mask();
if(m2 & 2) return; if (m2 & 2)
return;
m2 &= 1; m2 &= 1;
@ -795,7 +822,8 @@ void GSRasterizer::DrawSprite(const GSVertexSW* vertex, const uint32* index)
r = r.rintersect(m_scissor); r = r.rintersect(m_scissor);
if(r.rempty()) return; if (r.rempty())
return;
GSVertexSW scan = v[0]; GSVertexSW scan = v[0];
@ -860,7 +888,8 @@ void GSRasterizer::DrawSprite(const GSVertexSW* vertex, const uint32* index)
DrawScanline(r.width(), r.left, r.top, scan); DrawScanline(r.width(), r.left, r.top, scan);
} }
if(++r.top >= r.bottom) break; if (++r.top >= r.bottom)
break;
scan.t += dedge.t; scan.t += dedge.t;
} }
@ -897,7 +926,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
top = tb.extract32<0>(); // max(t, st) top = tb.extract32<0>(); // max(t, st)
bottom = tb.extract32<3>(); // min(b, sb) bottom = tb.extract32<3>(); // min(b, sb)
if(top >= bottom) return; if (top >= bottom)
return;
edge = v0; edge = v0;
dedge = dv / dv.p.yyyy(); dedge = dv / dv.p.yyyy();
@ -909,7 +939,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
top = tb.extract32<1>(); // max(b, st) top = tb.extract32<1>(); // max(b, st)
bottom = tb.extract32<2>(); // min(t, sb) bottom = tb.extract32<2>(); // min(t, sb)
if(top >= bottom) return; if (top >= bottom)
return;
edge = v1; edge = v1;
dedge = dv / dv.p.yyyy(); dedge = dv / dv.p.yyyy();
@ -938,7 +969,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
e++; e++;
} }
if(++top >= bottom) break; if (++top >= bottom)
break;
edge += dedge; edge += dedge;
x += dx; x += dx;
@ -960,7 +992,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
e++; e++;
} }
if(++top >= bottom) break; if (++top >= bottom)
break;
edge += dedge; edge += dedge;
x += dx; x += dx;
@ -983,7 +1016,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
left = lr.extract32<0>(); // max(l, sl) left = lr.extract32<0>(); // max(l, sl)
right = lr.extract32<3>(); // min(r, sr) right = lr.extract32<3>(); // min(r, sr)
if(left >= right) return; if (left >= right)
return;
edge = v0; edge = v0;
dedge = dv / dv.p.xxxx(); dedge = dv / dv.p.xxxx();
@ -995,7 +1029,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
left = lr.extract32<1>(); // max(r, sl) left = lr.extract32<1>(); // max(r, sl)
right = lr.extract32<2>(); // min(l, sr) right = lr.extract32<2>(); // min(l, sr)
if(left >= right) return; if (left >= right)
return;
edge = v1; edge = v1;
dedge = dv / dv.p.xxxx(); dedge = dv / dv.p.xxxx();
@ -1024,7 +1059,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
e++; e++;
} }
if(++left >= right) break; if (++left >= right)
break;
edge += dedge; edge += dedge;
y += dy; y += dy;
@ -1046,7 +1082,8 @@ void GSRasterizer::DrawEdge(const GSVertexSW& v0, const GSVertexSW& v1, const GS
e++; e++;
} }
if(++left >= right) break; if (++left >= right)
break;
edge += dedge; edge += dedge;
y += dy; y += dy;
@ -1088,8 +1125,7 @@ void GSRasterizer::Flush(const GSVertexSW* vertex, const uint32* index, const GS
int top = e->_pad.i32[2]; int top = e->_pad.i32[2];
DrawScanline(pixels, left, top, *e++); DrawScanline(pixels, left, top, *e++);
} } while (e < ee);
while(e < ee);
} }
else else
{ {
@ -1100,8 +1136,7 @@ void GSRasterizer::Flush(const GSVertexSW* vertex, const uint32* index, const GS
int top = e->_pad.i32[2]; int top = e->_pad.i32[2];
DrawEdge(pixels, left, top, *e++); DrawEdge(pixels, left, top, *e++);
} } while (e < ee);
while(e < ee);
} }
m_edge.count = 0; m_edge.count = 0;

14
plugins/GSdx/Renderers/SW/GSRasterizer.h
View File

@ -64,7 +64,8 @@ public:
virtual ~GSRasterizerData() virtual ~GSRasterizerData()
{ {
if(buff != NULL) _aligned_free(buff); if (buff != NULL)
_aligned_free(buff);
} }
}; };
@ -82,7 +83,13 @@ protected:
DrawRectPtr m_dr; DrawRectPtr m_dr;
public: public:
IDrawScanline() : m_sp(NULL), m_ds(NULL), m_de(NULL), m_dr(NULL) {} IDrawScanline()
: m_sp(NULL)
, m_ds(NULL)
, m_de(NULL)
, m_dr(NULL)
{
}
virtual ~IDrawScanline() {} virtual ~IDrawScanline() {}
virtual void BeginDraw(const GSRasterizerData* data) = 0; virtual void BeginDraw(const GSRasterizerData* data) = 0;
@ -195,7 +202,8 @@ protected:
public: public:
virtual ~GSRasterizerList(); virtual ~GSRasterizerList();
template<class DS> static IRasterizer* Create(int threads, GSPerfMon* perfmon) template <class DS>
static IRasterizer* Create(int threads, GSPerfMon* perfmon)
{ {
threads = std::max<int>(threads, 0); threads = std::max<int>(threads, 0);

121
plugins/GSdx/Renderers/SW/GSRendererSW.cpp
View File

@ -44,10 +44,12 @@ GSRendererSW::GSRendererSW(int threads)
m_output = (uint8*)_aligned_malloc(1024 * 1024 * sizeof(uint32), 32); m_output = (uint8*)_aligned_malloc(1024 * 1024 * sizeof(uint32), 32);
for (uint32 i = 0; i < countof(m_fzb_pages); i++) { for (uint32 i = 0; i < countof(m_fzb_pages); i++)
{
m_fzb_pages[i] = 0; m_fzb_pages[i] = 0;
} }
for (uint32 i = 0; i < countof(m_tex_pages); i++) { for (uint32 i = 0; i < countof(m_tex_pages); i++)
{
m_tex_pages[i] = 0; m_tex_pages[i] = 0;
} }
@ -70,7 +72,8 @@ GSRendererSW::GSRendererSW(int threads)
// Reset handler with the auto flush hack enabled on the SW renderer. // Reset handler with the auto flush hack enabled on the SW renderer.
// Some games run better without the hack so rely on ini/gui option. // Some games run better without the hack so rely on ini/gui option.
if (!GLLoader::in_replayer && theApp.GetConfigB("autoflush_sw")) { if (!GLLoader::in_replayer && theApp.GetConfigB("autoflush_sw"))
{
m_userhacks_auto_flush = true; m_userhacks_auto_flush = true;
ResetHandlers(); ResetHandlers();
} }
@ -193,7 +196,8 @@ GSTexture* GSRendererSW::GetFeedbackOutput()
int dummy; int dummy;
// It is enough to emulate Xenosaga cutscene. (or any game that will do a basic loopback) // It is enough to emulate Xenosaga cutscene. (or any game that will do a basic loopback)
for (int i = 0; i < 2; i++) { for (int i = 0; i < 2; i++)
{
if (m_regs->EXTBUF.EXBP == m_regs->DISP[i].DISPFB.Block()) if (m_regs->EXTBUF.EXBP == m_regs->DISP[i].DISPFB.Block())
return GetOutput(i, dummy); return GetOutput(i, dummy);
} }
@ -402,8 +406,7 @@ void GSRendererSW::Draw()
(float)(v->XYZ.X - context->XYOFFSET.OFX) / 16, (float)(v->XYZ.X - context->XYOFFSET.OFX) / 16,
(float)(v->XYZ.Y - context->XYOFFSET.OFY) / 16, (float)(v->XYZ.Y - context->XYOFFSET.OFY) / 16,
PRIM->FST ? (float)(v->U) / 16 : v->ST.S / (PRIM->PRIM == GS_SPRITE ? vn->RGBAQ.Q : v->RGBAQ.Q), PRIM->FST ? (float)(v->U) / 16 : v->ST.S / (PRIM->PRIM == GS_SPRITE ? vn->RGBAQ.Q : v->RGBAQ.Q),
PRIM->FST ? (float)(v->V) / 16 : v->ST.T / (PRIM->PRIM == GS_SPRITE ? vn->RGBAQ.Q : v->RGBAQ.Q) PRIM->FST ? (float)(v->V) / 16 : v->ST.T / (PRIM->PRIM == GS_SPRITE ? vn->RGBAQ.Q : v->RGBAQ.Q));
);
} }
} }
} }
@ -467,7 +470,8 @@ void GSRendererSW::Draw()
if (s_savet && s_n >= s_saven && PRIM->TME) if (s_savet && s_n >= s_saven && PRIM->TME)
{ {
if (texture_shuffle) { if (texture_shuffle)
{
// Dump the RT in 32 bits format. It helps to debug texture shuffle effect // Dump the RT in 32 bits format. It helps to debug texture shuffle effect
s = format("%05d_f%lld_itexraw_%05x_32bits.bmp", s_n, frame, (int)m_context->TEX0.TBP0); s = format("%05d_f%lld_itexraw_%05x_32bits.bmp", s_n, frame, (int)m_context->TEX0.TBP0);
m_mem.SaveBMP(m_dump_root + s, m_context->TEX0.TBP0, m_context->TEX0.TBW, 0, 1 << m_context->TEX0.TW, 1 << m_context->TEX0.TH); m_mem.SaveBMP(m_dump_root + s, m_context->TEX0.TBP0, m_context->TEX0.TBW, 0, 1 << m_context->TEX0.TW, 1 << m_context->TEX0.TH);
@ -480,7 +484,8 @@ void GSRendererSW::Draw()
if (s_save && s_n >= s_saven) if (s_save && s_n >= s_saven)
{ {
if (texture_shuffle) { if (texture_shuffle)
{
// Dump the RT in 32 bits format. It helps to debug texture shuffle effect // Dump the RT in 32 bits format. It helps to debug texture shuffle effect
s = format("%05d_f%lld_rt0_%05x_32bits.bmp", s_n, frame, m_context->FRAME.Block()); s = format("%05d_f%lld_rt0_%05x_32bits.bmp", s_n, frame, m_context->FRAME.Block());
m_mem.SaveBMP(m_dump_root + s, m_context->FRAME.Block(), m_context->FRAME.FBW, 0, GetFrameRect().width(), 512); m_mem.SaveBMP(m_dump_root + s, m_context->FRAME.Block(), m_context->FRAME.FBW, 0, GetFrameRect().width(), 512);
@ -503,7 +508,8 @@ void GSRendererSW::Draw()
if (s_save && s_n >= s_saven) if (s_save && s_n >= s_saven)
{ {
if (texture_shuffle) { if (texture_shuffle)
{
// Dump the RT in 32 bits format. It helps to debug texture shuffle effect // Dump the RT in 32 bits format. It helps to debug texture shuffle effect
s = format("%05d_f%lld_rt1_%05x_32bits.bmp", s_n, frame, m_context->FRAME.Block()); s = format("%05d_f%lld_rt1_%05x_32bits.bmp", s_n, frame, m_context->FRAME.Block());
m_mem.SaveBMP(m_dump_root + s, m_context->FRAME.Block(), m_context->FRAME.FBW, 0, GetFrameRect().width(), 512); m_mem.SaveBMP(m_dump_root + s, m_context->FRAME.Block(), m_context->FRAME.FBW, 0, GetFrameRect().width(), 512);
@ -624,14 +630,22 @@ void GSRendererSW::Sync(int reason)
int pixels = m_rl->GetPixels(); int pixels = m_rl->GetPixels();
if(LOG) {fprintf(s_fp, "sync n=%d r=%d t=%llu p=%d %c\n", s_n, reason, t, pixels, t > 10000000 ? '*' : ' '); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "sync n=%d r=%d t=%llu p=%d %c\n", s_n, reason, t, pixels, t > 10000000 ? '*' : ' ');
fflush(s_fp);
}
m_perfmon.Put(GSPerfMon::Fillrate, pixels); m_perfmon.Put(GSPerfMon::Fillrate, pixels);
} }
void GSRendererSW::InvalidateVideoMem(const GIFRegBITBLTBUF& BITBLTBUF, const GSVector4i& r) void GSRendererSW::InvalidateVideoMem(const GIFRegBITBLTBUF& BITBLTBUF, const GSVector4i& r)
{ {
if(LOG) {fprintf(s_fp, "w %05x %u %u, %d %d %d %d\n", BITBLTBUF.DBP, BITBLTBUF.DBW, BITBLTBUF.DPSM, r.x, r.y, r.z, r.w); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "w %05x %u %u, %d %d %d %d\n", BITBLTBUF.DBP, BITBLTBUF.DBW, BITBLTBUF.DPSM, r.x, r.y, r.z, r.w);
fflush(s_fp);
}
GSOffset* off = m_mem.GetOffset(BITBLTBUF.DBP, BITBLTBUF.DBW, BITBLTBUF.DPSM); GSOffset* off = m_mem.GetOffset(BITBLTBUF.DBP, BITBLTBUF.DBW, BITBLTBUF.DPSM);
@ -657,7 +671,11 @@ void GSRendererSW::InvalidateVideoMem(const GIFRegBITBLTBUF& BITBLTBUF, const GS
void GSRendererSW::InvalidateLocalMem(const GIFRegBITBLTBUF& BITBLTBUF, const GSVector4i& r, bool clut) void GSRendererSW::InvalidateLocalMem(const GIFRegBITBLTBUF& BITBLTBUF, const GSVector4i& r, bool clut)
{ {
if(LOG) {fprintf(s_fp, "%s %05x %u %u, %d %d %d %d\n", clut ? "rp" : "r", BITBLTBUF.SBP, BITBLTBUF.SBW, BITBLTBUF.SPSM, r.x, r.y, r.z, r.w); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "%s %05x %u %u, %d %d %d %d\n", clut ? "rp" : "r", BITBLTBUF.SBP, BITBLTBUF.SBW, BITBLTBUF.SPSM, r.x, r.y, r.z, r.w);
fflush(s_fp);
}
if (!m_rl->IsSynced()) if (!m_rl->IsSynced())
{ {
@ -679,8 +697,10 @@ void GSRendererSW::InvalidateLocalMem(const GIFRegBITBLTBUF& BITBLTBUF, const GS
void GSRendererSW::UsePages(const uint32* pages, const int type) void GSRendererSW::UsePages(const uint32* pages, const int type)
{ {
for(const uint32* p = pages; *p != GSOffset::EOP; p++) { for (const uint32* p = pages; *p != GSOffset::EOP; p++)
switch (type) { {
switch (type)
{
case 0: case 0:
ASSERT((m_fzb_pages[*p] & 0xFFFF) < USHRT_MAX); ASSERT((m_fzb_pages[*p] & 0xFFFF) < USHRT_MAX);
m_fzb_pages[*p] += 1; m_fzb_pages[*p] += 1;
@ -693,15 +713,18 @@ void GSRendererSW::UsePages(const uint32* pages, const int type)
ASSERT(m_tex_pages[*p] < USHRT_MAX); ASSERT(m_tex_pages[*p] < USHRT_MAX);
m_tex_pages[*p] += 1; m_tex_pages[*p] += 1;
break; break;
default:break; default:
break;
} }
} }
} }
void GSRendererSW::ReleasePages(const uint32* pages, const int type) void GSRendererSW::ReleasePages(const uint32* pages, const int type)
{ {
for(const uint32* p = pages; *p != GSOffset::EOP; p++) { for (const uint32* p = pages; *p != GSOffset::EOP; p++)
switch (type) { {
switch (type)
{
case 0: case 0:
ASSERT((m_fzb_pages[*p] & 0xFFFF) > 0); ASSERT((m_fzb_pages[*p] & 0xFFFF) > 0);
m_fzb_pages[*p] -= 1; m_fzb_pages[*p] -= 1;
@ -714,7 +737,8 @@ void GSRendererSW::ReleasePages(const uint32* pages, const int type)
ASSERT(m_tex_pages[*p] > 0); ASSERT(m_tex_pages[*p] > 0);
m_tex_pages[*p] -= 1; m_tex_pages[*p] -= 1;
break; break;
default:break; default:
break;
} }
} }
} }
@ -772,7 +796,11 @@ bool GSRendererSW::CheckTargetPages(const uint32* fb_pages, const uint32* zb_pag
{ {
if (used) if (used)
{ {
if(LOG) {fprintf(s_fp, "syncpoint 0\n"); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "syncpoint 0\n");
fflush(s_fp);
}
res = true; res = true;
} }
@ -833,7 +861,11 @@ bool GSRendererSW::CheckTargetPages(const uint32* fb_pages, const uint32* zb_pag
{ {
if (used) if (used)
{ {
if(LOG) {fprintf(s_fp, "syncpoint 1\n"); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "syncpoint 1\n");
fflush(s_fp);
}
res = true; res = true;
} }
@ -851,7 +883,11 @@ bool GSRendererSW::CheckTargetPages(const uint32* fb_pages, const uint32* zb_pag
{ {
if (m_fzb_pages[*p] & 0xffff0000) if (m_fzb_pages[*p] & 0xffff0000)
{ {
if(LOG) {fprintf(s_fp, "syncpoint 2\n"); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "syncpoint 2\n");
fflush(s_fp);
}
res = true; res = true;
@ -866,7 +902,11 @@ bool GSRendererSW::CheckTargetPages(const uint32* fb_pages, const uint32* zb_pag
{ {
if (m_fzb_pages[*p] & 0x0000ffff) if (m_fzb_pages[*p] & 0x0000ffff)
{ {
if(LOG) {fprintf(s_fp, "syncpoint 3\n"); fflush(s_fp);} if (LOG)
{
fprintf(s_fp, "syncpoint 3\n");
fflush(s_fp);
}
res = true; res = true;
@ -987,7 +1027,8 @@ bool GSRendererSW::GetScanlineGlobalData(SharedData* data)
zwrite || ztest ? m_context->ZBUF.Block() : 0xfffff, m_context->ZBUF.PSM, zwrite || ztest ? m_context->ZBUF.Block() : 0xfffff, m_context->ZBUF.PSM,
PRIM->TME ? m_context->TEX0.TBP0 : 0xfffff, m_context->TEX0.PSM, (int)m_context->TEX0.TW, (int)m_context->TEX0.TH); PRIM->TME ? m_context->TEX0.TBP0 : 0xfffff, m_context->TEX0.PSM, (int)m_context->TEX0.TW, (int)m_context->TEX0.TH);
*/ */
if(!fwrite && !zwrite) return false; if (!fwrite && !zwrite)
return false;
gd.sel.fwrite = fwrite; gd.sel.fwrite = fwrite;
gd.sel.ftest = ftest; gd.sel.ftest = ftest;
@ -1037,7 +1078,11 @@ bool GSRendererSW::GetScanlineGlobalData(SharedData* data)
GSTextureCacheSW::Texture* t = m_tc->Lookup(TEX0, env.TEXA); GSTextureCacheSW::Texture* t = m_tc->Lookup(TEX0, env.TEXA);
if(t == NULL) {ASSERT(0); return false;} if (t == NULL)
{
ASSERT(0);
return false;
}
data->SetSource(t, r, 0); data->SetSource(t, r, 0);
@ -1131,7 +1176,11 @@ bool GSRendererSW::GetScanlineGlobalData(SharedData* data)
GSTextureCacheSW::Texture* t = m_tc->Lookup(MIP_TEX0, env.TEXA, gd.sel.tw + 3); GSTextureCacheSW::Texture* t = m_tc->Lookup(MIP_TEX0, env.TEXA, gd.sel.tw + 3);
if(t == NULL) {ASSERT(0); return false;} if (t == NULL)
{
ASSERT(0);
return false;
}
GSVector4i r; GSVector4i r;
@ -1399,23 +1448,28 @@ GSRendererSW::SharedData::~SharedData()
{ {
ReleasePages(); ReleasePages();
if(global.clut) _aligned_free(global.clut); if (global.clut)
if(global.dimx) _aligned_free(global.dimx); _aligned_free(global.clut);
if (global.dimx)
_aligned_free(global.dimx);
if(LOG) {fprintf(s_fp, "[%d] done t=%lld p=%d | %d %d %d | %08x_%08x\n", if (LOG)
{
fprintf(s_fp, "[%d] done t=%lld p=%d | %d %d %d | %08x_%08x\n",
counter, counter,
__rdtsc() - start, pixels, __rdtsc() - start, pixels,
primclass, vertex_count, index_count, primclass, vertex_count, index_count,
global.sel.hi, global.sel.lo global.sel.hi, global.sel.lo);
); fflush(s_fp);
fflush(s_fp);} }
} }
//static TransactionScope::Lock s_lock; //static TransactionScope::Lock s_lock;
void GSRendererSW::SharedData::UsePages(const uint32* fb_pages, int fpsm, const uint32* zb_pages, int zpsm) void GSRendererSW::SharedData::UsePages(const uint32* fb_pages, int fpsm, const uint32* zb_pages, int zpsm)
{ {
if(m_using_pages) return; if (m_using_pages)
return;
{ {
//TransactionScope scope(s_lock); //TransactionScope scope(s_lock);
@ -1446,7 +1500,8 @@ void GSRendererSW::SharedData::UsePages(const uint32* fb_pages, int fpsm, const
void GSRendererSW::SharedData::ReleasePages() void GSRendererSW::SharedData::ReleasePages()
{ {
if(!m_using_pages) return; if (!m_using_pages)
return;
{ {
//TransactionScope scope(s_lock); //TransactionScope scope(s_lock);

7
plugins/GSdx/Renderers/SW/GSRendererSW.h
View File

@ -47,7 +47,12 @@ class GSRendererSW : public GSRenderer
int m_zpsm; int m_zpsm;
bool m_using_pages; bool m_using_pages;
TextureLevel m_tex[7 + 1]; // NULL terminated TextureLevel m_tex[7 + 1]; // NULL terminated
enum {SyncNone, SyncSource, SyncTarget} m_syncpoint; enum
{
SyncNone,
SyncSource,
SyncTarget
} m_syncpoint;
public: public:
SharedData(GSRendererSW* parent); SharedData(GSRendererSW* parent);

21
plugins/GSdx/Renderers/SW/GSScanlineEnvironment.h
View File

@ -94,21 +94,17 @@ union GSScanlineSelector
uint64 key; uint64 key;
GSScanlineSelector() = default; GSScanlineSelector() = default;
GSScanlineSelector(uint64 k) : key(k) {} GSScanlineSelector(uint64 k)
: key(k)
{
}
operator uint32() const { return lo; } operator uint32() const { return lo; }
operator uint64() const { return key; } operator uint64() const { return key; }
bool IsSolidRect() const bool IsSolidRect() const
{ {
return prim == GS_SPRITE_CLASS return prim == GS_SPRITE_CLASS && iip == 0 && tfx == TFX_NONE && abe == 0 && ztst <= 1 && atst <= 1 && date == 0 && fge == 0;
&& iip == 0
&& tfx == TFX_NONE
&& abe == 0
&& ztst <= 1
&& atst <= 1
&& date == 0
&& fge == 0;
} }
void Print() const void Print() const
@ -319,14 +315,15 @@ struct GSScanlineConstantData : public GSAlignedClass<32>
1.0f 1.0f
}; };
for (size_t n = 0; n < countof(log2_coef); ++n) { for (size_t n = 0; n < countof(log2_coef); ++n)
for (size_t i = 0; i < 4; ++i) { {
for (size_t i = 0; i < 4; ++i)
{
m_log2_coef_128b[n][i] = log2_coef[n]; m_log2_coef_128b[n][i] = log2_coef[n];
m_log2_coef_256b[n][i] = log2_coef[n]; m_log2_coef_256b[n][i] = log2_coef[n];
m_log2_coef_256b[n][i + 4] = log2_coef[n]; m_log2_coef_256b[n][i + 4] = log2_coef[n];
} }
} }
} }
}; };

7
plugins/GSdx/Renderers/SW/GSSetupPrimCodeGenerator.cpp
View File

@ -36,7 +36,8 @@ GSSetupPrimCodeGenerator::GSSetupPrimCodeGenerator(void* param, uint64 key, void
m_en.t = m_sel.fb && m_sel.tfx != TFX_NONE ? 1 : 0; m_en.t = m_sel.fb && m_sel.tfx != TFX_NONE ? 1 : 0;
m_en.c = m_sel.fb && !(m_sel.tfx == TFX_DECAL && m_sel.tcc) ? 1 : 0; m_en.c = m_sel.fb && !(m_sel.tfx == TFX_DECAL && m_sel.tcc) ? 1 : 0;
try { try
{
#if _M_SSE >= 0x501 #if _M_SSE >= 0x501
Generate_AVX2(); Generate_AVX2();
#else #else
@ -45,7 +46,9 @@ GSSetupPrimCodeGenerator::GSSetupPrimCodeGenerator(void* param, uint64 key, void
else else
Generate_SSE(); Generate_SSE();
#endif #endif
} catch (std::exception& e) { }
catch (std::exception& e)
{
fprintf(stderr, "ERR:GSSetupPrimCodeGenerator %s\n", e.what()); fprintf(stderr, "ERR:GSSetupPrimCodeGenerator %s\n", e.what());
} }
} }

5
plugins/GSdx/Renderers/SW/GSSetupPrimCodeGenerator.h
View File

@ -33,7 +33,10 @@ class GSSetupPrimCodeGenerator : public GSCodeGenerator
GSScanlineLocalData& m_local; GSScanlineLocalData& m_local;
bool m_rip; bool m_rip;
struct {uint32 z:1, f:1, t:1, c:1;} m_en; struct
{
uint32 z : 1, f : 1, t : 1, c : 1;
} m_en;
#if _M_SSE < 0x501 #if _M_SSE < 0x501
void Generate_SSE(); void Generate_SSE();

42
plugins/GSdx/Renderers/SW/GSSetupPrimCodeGenerator.x86.avx2.cpp
View File

@ -103,8 +103,10 @@ void GSSetupPrimCodeGenerator::Depth_AVX2()
{ {
// m_local.d[i].z = dz * shift[1 + i]; // m_local.d[i].z = dz * shift[1 + i];
if(i < 4) vmulps(ymm0, ymm2, Ymm(4 + i)); if (i < 4)
else vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm0, ymm2, Ymm(4 + i));
else
vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]);
vmovaps(ptr[&m_local.d[i].z], ymm0); vmovaps(ptr[&m_local.d[i].z], ymm0);
} }
@ -112,8 +114,10 @@ void GSSetupPrimCodeGenerator::Depth_AVX2()
{ {
// m_local.d[i].f = GSVector8i(df * m_shift[i]).xxzzlh(); // m_local.d[i].f = GSVector8i(df * m_shift[i]).xxzzlh();
if(i < 4) vmulps(ymm0, ymm1, Ymm(4 + i)); if (i < 4)
else vmulps(ymm0, ymm1, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm0, ymm1, Ymm(4 + i));
else
vmulps(ymm0, ymm1, ptr[g_const->m_shift_256b[i + 1]]);
vcvttps2dq(ymm0, ymm0); vcvttps2dq(ymm0, ymm0);
vpshuflw(ymm0, ymm0, _MM_SHUFFLE(2, 2, 0, 0)); vpshuflw(ymm0, ymm0, _MM_SHUFFLE(2, 2, 0, 0));
vpshufhw(ymm0, ymm0, _MM_SHUFFLE(2, 2, 0, 0)); vpshufhw(ymm0, ymm0, _MM_SHUFFLE(2, 2, 0, 0));
@ -189,8 +193,10 @@ void GSSetupPrimCodeGenerator::Texture_AVX2()
{ {
// GSVector8 v = dstq * shift[1 + i]; // GSVector8 v = dstq * shift[1 + i];
if(i < 4) vmulps(ymm2, ymm1, Ymm(4 + i)); if (i < 4)
else vmulps(ymm2, ymm1, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm2, ymm1, Ymm(4 + i));
else
vmulps(ymm2, ymm1, ptr[g_const->m_shift_256b[i + 1]]);
if (m_sel.fst) if (m_sel.fst)
{ {
@ -252,15 +258,19 @@ void GSSetupPrimCodeGenerator::Color_AVX2()
{ {
// GSVector8i r = GSVector8i(dr * shift[1 + i]).ps32(); // GSVector8i r = GSVector8i(dr * shift[1 + i]).ps32();
if(i < 4) vmulps(ymm0, ymm2, Ymm(4 + i)); if (i < 4)
else vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm0, ymm2, Ymm(4 + i));
else
vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]);
vcvttps2dq(ymm0, ymm0); vcvttps2dq(ymm0, ymm0);
vpackssdw(ymm0, ymm0); vpackssdw(ymm0, ymm0);
// GSVector4i b = GSVector8i(db * shift[1 + i]).ps32(); // GSVector4i b = GSVector8i(db * shift[1 + i]).ps32();
if(i < 4) vmulps(ymm1, ymm3, Ymm(4 + i)); if (i < 4)
else vmulps(ymm1, ymm3, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm1, ymm3, Ymm(4 + i));
else
vmulps(ymm1, ymm3, ptr[g_const->m_shift_256b[i + 1]]);
vcvttps2dq(ymm1, ymm1); vcvttps2dq(ymm1, ymm1);
vpackssdw(ymm1, ymm1); vpackssdw(ymm1, ymm1);
@ -284,15 +294,19 @@ void GSSetupPrimCodeGenerator::Color_AVX2()
{ {
// GSVector8i g = GSVector8i(dg * shift[1 + i]).ps32(); // GSVector8i g = GSVector8i(dg * shift[1 + i]).ps32();
if(i < 4) vmulps(ymm0, ymm2, Ymm(4 + i)); if (i < 4)
else vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm0, ymm2, Ymm(4 + i));
else
vmulps(ymm0, ymm2, ptr[g_const->m_shift_256b[i + 1]]);
vcvttps2dq(ymm0, ymm0); vcvttps2dq(ymm0, ymm0);
vpackssdw(ymm0, ymm0); vpackssdw(ymm0, ymm0);
// GSVector8i a = GSVector8i(da * shift[1 + i]).ps32(); // GSVector8i a = GSVector8i(da * shift[1 + i]).ps32();
if(i < 4) vmulps(ymm1, ymm3, Ymm(4 + i)); if (i < 4)
else vmulps(ymm1, ymm3, ptr[g_const->m_shift_256b[i + 1]]); vmulps(ymm1, ymm3, Ymm(4 + i));
else
vmulps(ymm1, ymm3, ptr[g_const->m_shift_256b[i + 1]]);
vcvttps2dq(ymm1, ymm1); vcvttps2dq(ymm1, ymm1);
vpackssdw(ymm1, ymm1); vpackssdw(ymm1, ymm1);

3
plugins/GSdx/Renderers/SW/GSTextureCacheSW.cpp
View File

@ -109,7 +109,8 @@ void GSTextureCacheSW::InvalidatePages(const uint32* pages, uint32 psm)
void GSTextureCacheSW::RemoveAll() void GSTextureCacheSW::RemoveAll()
{ {
for(auto i : m_textures) delete i; for (auto i : m_textures)
delete i;
m_textures.clear(); m_textures.clear();

1
plugins/GSdx/Renderers/SW/GSVertexSW.h
View File

@ -263,4 +263,3 @@ struct alignas(32) GSVertexSW2
}; };
#endif #endif

14
plugins/GSdx/Window/GSCaptureDlg.cpp
View File

@ -24,14 +24,14 @@
#include "GSCaptureDlg.h" #include "GSCaptureDlg.h"
#define BeginEnumSysDev(clsid, pMoniker) \ #define BeginEnumSysDev(clsid, pMoniker) \
{CComPtr<ICreateDevEnum> pDevEnum4$##clsid; \ { \
CComPtr<ICreateDevEnum> pDevEnum4$##clsid; \
pDevEnum4$##clsid.CoCreateInstance(CLSID_SystemDeviceEnum); \ pDevEnum4$##clsid.CoCreateInstance(CLSID_SystemDeviceEnum); \
CComPtr<IEnumMoniker> pClassEnum4$##clsid; \ CComPtr<IEnumMoniker> pClassEnum4$##clsid; \
if(SUCCEEDED(pDevEnum4$##clsid->CreateClassEnumerator(clsid, &pClassEnum4$##clsid, 0)) \ if (SUCCEEDED(pDevEnum4$##clsid->CreateClassEnumerator(clsid, &pClassEnum4$##clsid, 0)) && pClassEnum4$##clsid) \
&& pClassEnum4$##clsid) \
{ \ { \
for (CComPtr<IMoniker> pMoniker; pClassEnum4$##clsid->Next(1, &pMoniker, 0) == S_OK; pMoniker = NULL) \ for (CComPtr<IMoniker> pMoniker; pClassEnum4$##clsid->Next(1, &pMoniker, 0) == S_OK; pMoniker = NULL) \
{ \ {
#define EndEnumSysDev }}} #define EndEnumSysDev }}}
@ -55,7 +55,8 @@ int GSCaptureDlg::GetSelCodec(Codec& c)
if (ComboBoxGetSelData(IDC_CODECS, data)) if (ComboBoxGetSelData(IDC_CODECS, data))
{ {
if(data == 0) return 2; if (data == 0)
return 2;
c = *(Codec*)data; c = *(Codec*)data;
@ -63,7 +64,8 @@ int GSCaptureDlg::GetSelCodec(Codec& c)
{ {
c.moniker->BindToObject(NULL, NULL, __uuidof(IBaseFilter), (void**)&c.filter); c.moniker->BindToObject(NULL, NULL, __uuidof(IBaseFilter), (void**)&c.filter);
if(!c.filter) return 0; if (!c.filter)
return 0;
} }
return 1; return 1;

10
plugins/GSdx/Window/GSDialog.cpp
View File

@ -91,11 +91,13 @@ INT_PTR CALLBACK GSDialog::DialogProc(HWND hWnd, UINT message, WPARAM wParam, LP
UINT GSDialog::GetTooltipStructSize() UINT GSDialog::GetTooltipStructSize()
{ {
DLLGETVERSIONPROC dllGetVersion = (DLLGETVERSIONPROC)GetProcAddress(GetModuleHandle(L"ComCtl32.dll"), "DllGetVersion"); DLLGETVERSIONPROC dllGetVersion = (DLLGETVERSIONPROC)GetProcAddress(GetModuleHandle(L"ComCtl32.dll"), "DllGetVersion");
if (dllGetVersion) { if (dllGetVersion)
{
DLLVERSIONINFO2 dllversion = {0}; DLLVERSIONINFO2 dllversion = {0};
dllversion.info1.cbSize = sizeof(DLLVERSIONINFO2); dllversion.info1.cbSize = sizeof(DLLVERSIONINFO2);
if (dllGetVersion((DLLVERSIONINFO*)&dllversion) == S_OK) { if (dllGetVersion((DLLVERSIONINFO*)&dllversion) == S_OK)
{
// Minor, then major version. // Minor, then major version.
DWORD version = MAKELONG(dllversion.info1.dwMinorVersion, dllversion.info1.dwMajorVersion); DWORD version = MAKELONG(dllversion.info1.dwMinorVersion, dllversion.info1.dwMajorVersion);
DWORD tooltip_v3 = MAKELONG(0, 6); DWORD tooltip_v3 = MAKELONG(0, 6);
@ -269,7 +271,8 @@ void GSDialog::ComboBoxFixDroppedWidth(UINT id)
{ {
size.cx += 10; size.cx += 10;
if(size.cx > width) width = size.cx; if (size.cx > width)
width = size.cx;
} }
delete[] buff; delete[] buff;
@ -307,7 +310,6 @@ void GSDialog::OpenFileDialog(UINT id, const wchar_t *title)
SendMessage(GetDlgItem(m_hWnd, id), WM_SETTEXT, 0, (LPARAM)filename); SendMessage(GetDlgItem(m_hWnd, id), WM_SETTEXT, 0, (LPARAM)filename);
SetCurrentDirectory(current_directory); SetCurrentDirectory(current_directory);
} }
void GSDialog::AddTooltip(UINT id) void GSDialog::AddTooltip(UINT id)

11
plugins/GSdx/Window/GSSetting.h
View File

@ -30,10 +30,10 @@ struct GSSetting
std::string note; std::string note;
template <typename T> template <typename T>
explicit GSSetting(T value, const char* name, const char* note) : explicit GSSetting(T value, const char* name, const char* note)
value(static_cast<int32_t>(value)), : value(static_cast<int32_t>(value))
name(name), , name(name)
note(note) , note(note)
{ {
} }
}; };
@ -45,7 +45,8 @@ const char* dialog_message(int ID, bool* updateText = NULL);
#endif #endif
#ifndef _WIN32 #ifndef _WIN32
enum { enum
{
IDC_FILTER, IDC_FILTER,
IDC_HALF_SCREEN_TS, IDC_HALF_SCREEN_TS,
IDC_TRI_FILTER, IDC_TRI_FILTER,

71
plugins/GSdx/Window/GSSettingsDlg.cpp
View File

@ -377,9 +377,10 @@ void GSSettingsDlg::UpdateControls()
// Shader Configuration Dialog // Shader Configuration Dialog
GSShaderDlg::GSShaderDlg() : GSShaderDlg::GSShaderDlg()
GSDialog(IDD_SHADER) : GSDialog(IDD_SHADER)
{} {
}
void GSShaderDlg::OnInit() void GSShaderDlg::OnInit()
{ {
@ -486,7 +487,8 @@ bool GSShaderDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
wprintf(text, "%d", m_contrast); wprintf(text, "%d", m_contrast);
SetDlgItemText(m_hWnd, IDC_CONTRAST_VALUE, text); SetDlgItemText(m_hWnd, IDC_CONTRAST_VALUE, text);
} }
} break; }
break;
case WM_COMMAND: case WM_COMMAND:
{ {
@ -531,7 +533,8 @@ bool GSShaderDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
theApp.SetConfig("shaderfx_conf", output.c_str()); theApp.SetConfig("shaderfx_conf", output.c_str());
EndDialog(m_hWnd, id); EndDialog(m_hWnd, id);
} break; }
break;
case IDC_SHADEBOOST: case IDC_SHADEBOOST:
UpdateControls(); UpdateControls();
case IDC_SHADER_FX: case IDC_SHADER_FX:
@ -551,14 +554,18 @@ bool GSShaderDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
case IDCANCEL: case IDCANCEL:
{ {
EndDialog(m_hWnd, IDCANCEL); EndDialog(m_hWnd, IDCANCEL);
} break;
} }
break;
}
}
break;
} break; case WM_CLOSE:
EndDialog(m_hWnd, IDCANCEL);
break;
case WM_CLOSE:EndDialog(m_hWnd, IDCANCEL); break; default:
return false;
default: return false;
} }
@ -798,19 +805,24 @@ bool GSHacksDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
theApp.SetConfig("UserHacks_TCOffsetY", (int)SendMessage(GetDlgItem(m_hWnd, IDC_TCOFFSETY), UDM_GETPOS, 0, 0)); theApp.SetConfig("UserHacks_TCOffsetY", (int)SendMessage(GetDlgItem(m_hWnd, IDC_TCOFFSETY), UDM_GETPOS, 0, 0));
EndDialog(m_hWnd, id); EndDialog(m_hWnd, id);
} break; }
break;
case IDCANCEL: case IDCANCEL:
{ {
EndDialog(m_hWnd, IDCANCEL); EndDialog(m_hWnd, IDCANCEL);
} break;
} }
break;
}
}
break;
} break; case WM_CLOSE:
EndDialog(m_hWnd, IDCANCEL);
break;
case WM_CLOSE:EndDialog(m_hWnd, IDCANCEL); break; default:
return false;
default: return false;
} }
return true; return true;
@ -818,9 +830,10 @@ bool GSHacksDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
// OSD Configuration Dialog // OSD Configuration Dialog
GSOSDDlg::GSOSDDlg() : GSOSDDlg::GSOSDDlg()
GSDialog(IDD_OSD) : GSDialog(IDD_OSD)
{} {
}
void GSOSDDlg::OnInit() void GSOSDDlg::OnInit()
{ {
@ -951,7 +964,8 @@ bool GSOSDDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
wprintf(text, "%d", m_color.b); wprintf(text, "%d", m_color.b);
SetDlgItemText(m_hWnd, IDC_OSD_COLOR_BLUE_AMOUNT, text); SetDlgItemText(m_hWnd, IDC_OSD_COLOR_BLUE_AMOUNT, text);
} }
} break; }
break;
case WM_COMMAND: case WM_COMMAND:
{ {
@ -974,7 +988,8 @@ bool GSOSDDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
theApp.SetConfig("osd_monitor_enabled", (int)IsDlgButtonChecked(m_hWnd, IDC_OSD_MONITOR)); theApp.SetConfig("osd_monitor_enabled", (int)IsDlgButtonChecked(m_hWnd, IDC_OSD_MONITOR));
EndDialog(m_hWnd, id); EndDialog(m_hWnd, id);
} break; }
break;
case IDC_OSD_LOG: case IDC_OSD_LOG:
if (HIWORD(wParam) == BN_CLICKED) if (HIWORD(wParam) == BN_CLICKED)
UpdateControls(); UpdateControls();
@ -992,14 +1007,18 @@ bool GSOSDDlg::OnMessage(UINT message, WPARAM wParam, LPARAM lParam)
case IDCANCEL: case IDCANCEL:
{ {
EndDialog(m_hWnd, IDCANCEL); EndDialog(m_hWnd, IDCANCEL);
} break;
} }
break;
}
}
break;
} break; case WM_CLOSE:
EndDialog(m_hWnd, IDCANCEL);
break;
case WM_CLOSE:EndDialog(m_hWnd, IDCANCEL); break; default:
return false;
default: return false;
} }

7
plugins/GSdx/Window/GSSettingsDlg.h
View File

@ -83,7 +83,12 @@ class GSSettingsDlg : public GSDialog
std::string name; std::string name;
std::string id; std::string id;
D3D_FEATURE_LEVEL level; D3D_FEATURE_LEVEL level;
Adapter(const std::string &n, const std::string &i, const D3D_FEATURE_LEVEL &l) : name(n), id(i), level(l) {} Adapter(const std::string& n, const std::string& i, const D3D_FEATURE_LEVEL& l)
: name(n)
, id(i)
, level(l)
{
}
}; };
std::vector<GSSetting> m_renderers; std::vector<GSSetting> m_renderers;

25
plugins/GSdx/Window/GSWnd.h
View File

@ -31,8 +31,11 @@ protected:
bool m_managed; // set true when we're attached to a 3rdparty window that's amanged by the emulator bool m_managed; // set true when we're attached to a 3rdparty window that's amanged by the emulator
public: public:
GSWnd() : m_managed(false) {}; GSWnd()
virtual ~GSWnd() {}; : m_managed(false)
{
}
virtual ~GSWnd() {}
virtual bool Create(const std::string& title, int w, int h) = 0; virtual bool Create(const std::string& title, int w, int h) = 0;
virtual bool Attach(void* handle, bool managed = true) = 0; virtual bool Attach(void* handle, bool managed = true) = 0;
@ -44,16 +47,15 @@ public:
virtual GSVector4i GetClientRect() = 0; virtual GSVector4i GetClientRect() = 0;
virtual bool SetWindowText(const char* title) = 0; virtual bool SetWindowText(const char* title) = 0;
virtual void AttachContext() {}; virtual void AttachContext() {}
virtual void DetachContext() {}; virtual void DetachContext() {}
virtual void Show() = 0; virtual void Show() = 0;
virtual void Hide() = 0; virtual void Hide() = 0;
virtual void HideFrame() = 0; virtual void HideFrame() = 0;
virtual void Flip() {}; virtual void Flip() {}
virtual void SetVSync(int vsync) {}; virtual void SetVSync(int vsync) {}
}; };
class GSWndGL : public GSWnd class GSWndGL : public GSWnd
@ -73,8 +75,13 @@ protected:
virtual bool HasLateVsyncSupport() = 0; virtual bool HasLateVsyncSupport() = 0;
public: public:
GSWndGL() : m_ctx_attached(false), m_vsync_change_requested(false), m_vsync(0) {}; GSWndGL()
virtual ~GSWndGL() {}; : m_ctx_attached(false)
, m_vsync_change_requested(false)
, m_vsync(0)
{
}
virtual ~GSWndGL() {}
virtual bool Create(const std::string& title, int w, int h) = 0; virtual bool Create(const std::string& title, int w, int h) = 0;
virtual bool Attach(void* handle, bool managed = true) = 0; virtual bool Attach(void* handle, bool managed = true) = 0;

12
plugins/GSdx/Window/GSWndDX.cpp
View File

@ -179,7 +179,8 @@ GSVector4i GSWndDX::GetClientRect()
bool GSWndDX::SetWindowText(const char* title) bool GSWndDX::SetWindowText(const char* title)
{ {
if(!m_managed) return false; if (!m_managed)
return false;
const size_t tmp_size = strlen(title) + 1; const size_t tmp_size = strlen(title) + 1;
std::wstring tmp(tmp_size, L'#'); std::wstring tmp(tmp_size, L'#');
@ -191,7 +192,8 @@ bool GSWndDX::SetWindowText(const char* title)
void GSWndDX::Show() void GSWndDX::Show()
{ {
if(!m_managed) return; if (!m_managed)
return;
SetForegroundWindow(m_hWnd); SetForegroundWindow(m_hWnd);
ShowWindow(m_hWnd, SW_SHOWNORMAL); ShowWindow(m_hWnd, SW_SHOWNORMAL);
@ -200,14 +202,16 @@ void GSWndDX::Show()
void GSWndDX::Hide() void GSWndDX::Hide()
{ {
if(!m_managed) return; if (!m_managed)
return;
ShowWindow(m_hWnd, SW_HIDE); ShowWindow(m_hWnd, SW_HIDE);
} }
void GSWndDX::HideFrame() void GSWndDX::HideFrame()
{ {
if(!m_managed) return; if (!m_managed)
return;
SetWindowLong(m_hWnd, GWL_STYLE, GetWindowLong(m_hWnd, GWL_STYLE) & ~(WS_CAPTION | WS_THICKFRAME)); SetWindowLong(m_hWnd, GWL_STYLE, GetWindowLong(m_hWnd, GWL_STYLE) & ~(WS_CAPTION | WS_THICKFRAME));
SetWindowPos(m_hWnd, NULL, 0, 0, 0, 0, SWP_NOSIZE | SWP_NOMOVE | SWP_NOZORDER | SWP_NOACTIVATE); SetWindowPos(m_hWnd, NULL, 0, 0, 0, 0, SWP_NOSIZE | SWP_NOMOVE | SWP_NOZORDER | SWP_NOACTIVATE);

51
plugins/GSdx/Window/GSWndEGL.cpp
View File

@ -29,27 +29,31 @@ int GSWndEGL::SelectPlatform()
{ {
// Check the supported extension // Check the supported extension
const char* client_extensions = eglQueryString(EGL_NO_DISPLAY, EGL_EXTENSIONS); const char* client_extensions = eglQueryString(EGL_NO_DISPLAY, EGL_EXTENSIONS);
if (!client_extensions) { if (!client_extensions)
{
fprintf(stderr, "EGL: Client extension not supported\n"); fprintf(stderr, "EGL: Client extension not supported\n");
return 0; return 0;
} }
fprintf(stdout, "EGL: Supported extensions: %s\n", client_extensions); fprintf(stdout, "EGL: Supported extensions: %s\n", client_extensions);
// Check platform extensions are supported (Note: there are core in 1.5) // Check platform extensions are supported (Note: there are core in 1.5)
if (!strstr(client_extensions, "EGL_EXT_platform_base")) { if (!strstr(client_extensions, "EGL_EXT_platform_base"))
{
fprintf(stderr, "EGL: Dynamic platform selection isn't supported\n"); fprintf(stderr, "EGL: Dynamic platform selection isn't supported\n");
return 0; return 0;
} }
// Finally we can select the platform // Finally we can select the platform
#if GS_EGL_X11 #if GS_EGL_X11
if (strstr(client_extensions, "EGL_EXT_platform_x11")) { if (strstr(client_extensions, "EGL_EXT_platform_x11"))
{
fprintf(stdout, "EGL: select X11 platform\n"); fprintf(stdout, "EGL: select X11 platform\n");
return EGL_PLATFORM_X11_KHR; return EGL_PLATFORM_X11_KHR;
} }
#endif #endif
#if GS_EGL_WL #if GS_EGL_WL
if (strstr(client_extensions, "EGL_EXT_platform_wayland")) { if (strstr(client_extensions, "EGL_EXT_platform_wayland"))
{
fprintf(stdout, "EGL: select Wayland platform\n"); fprintf(stdout, "EGL: select Wayland platform\n");
return EGL_PLATFORM_WAYLAND_KHR; return EGL_PLATFORM_WAYLAND_KHR;
} }
@ -111,7 +115,8 @@ void GSWndEGL::CreateContext(int major, int minor)
m_eglContext = eglCreateContext(m_eglDisplay, eglConfig, EGL_NO_CONTEXT, contextAttribs); m_eglContext = eglCreateContext(m_eglDisplay, eglConfig, EGL_NO_CONTEXT, contextAttribs);
EGLint status = eglGetError(); EGLint status = eglGetError();
if (status == EGL_BAD_ATTRIBUTE || status == EGL_BAD_MATCH) { if (status == EGL_BAD_ATTRIBUTE || status == EGL_BAD_MATCH)
{
// Radeon/Gallium don't support advance attribute. Fallback to random value // Radeon/Gallium don't support advance attribute. Fallback to random value
// Note: Intel gives an EGL_BAD_MATCH. I don't know why but let's by stubborn and retry. // Note: Intel gives an EGL_BAD_MATCH. I don't know why but let's by stubborn and retry.
fprintf(stderr, "EGL: warning your driver doesn't support advance openGL context attributes\n"); fprintf(stderr, "EGL: warning your driver doesn't support advance openGL context attributes\n");
@ -133,7 +138,8 @@ void GSWndEGL::CreateContext(int major, int minor)
void GSWndEGL::AttachContext() void GSWndEGL::AttachContext()
{ {
if (!IsContextAttached()) { if (!IsContextAttached())
{
// The setting of the API is local to a thread. This function // The setting of the API is local to a thread. This function
// can be called from 2 threads. // can be called from 2 threads.
BindAPI(); BindAPI();
@ -146,7 +152,8 @@ void GSWndEGL::AttachContext()
void GSWndEGL::DetachContext() void GSWndEGL::DetachContext()
{ {
if (IsContextAttached()) { if (IsContextAttached())
{
//fprintf(stderr, "Detach the context\n"); //fprintf(stderr, "Detach the context\n");
eglMakeCurrent(m_eglDisplay, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT); eglMakeCurrent(m_eglDisplay, EGL_NO_SURFACE, EGL_NO_SURFACE, EGL_NO_CONTEXT);
m_ctx_attached = false; m_ctx_attached = false;
@ -161,7 +168,8 @@ void GSWndEGL::BindAPI()
{ {
eglBindAPI(EGL_OPENGL_API); eglBindAPI(EGL_OPENGL_API);
EGLenum api = eglQueryAPI(); EGLenum api = eglQueryAPI();
if (api != EGL_OPENGL_API) { if (api != EGL_OPENGL_API)
{
fprintf(stderr, "EGL: Failed to bind the OpenGL API got 0x%x instead\n", api); fprintf(stderr, "EGL: Failed to bind the OpenGL API got 0x%x instead\n", api);
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
@ -198,7 +206,8 @@ void GSWndEGL::Detach()
bool GSWndEGL::Create(const std::string& title, int w, int h) bool GSWndEGL::Create(const std::string& title, int w, int h)
{ {
if(w <= 0 || h <= 0) { if (w <= 0 || h <= 0)
{
w = theApp.GetConfigI("ModeWidth"); w = theApp.GetConfigI("ModeWidth");
h = theApp.GetConfigI("ModeHeight"); h = theApp.GetConfigI("ModeHeight");
} }
@ -217,7 +226,8 @@ bool GSWndEGL::Create(const std::string& title, int w, int h)
void* GSWndEGL::GetProcAddress(const char* name, bool opt) void* GSWndEGL::GetProcAddress(const char* name, bool opt)
{ {
void* ptr = (void*)eglGetProcAddress(name); void* ptr = (void*)eglGetProcAddress(name);
if (ptr == nullptr) { if (ptr == nullptr)
{
if (theApp.GetConfigB("debug_opengl")) if (theApp.GetConfigB("debug_opengl"))
fprintf(stderr, "Failed to find %s\n", name); fprintf(stderr, "Failed to find %s\n", name);
@ -267,12 +277,14 @@ void GSWndEGL::OpenEGLDisplay()
// Create an EGL display from the native display // Create an EGL display from the native display
m_eglDisplay = eglGetPlatformDisplay(m_platform, native_display, nullptr); m_eglDisplay = eglGetPlatformDisplay(m_platform, native_display, nullptr);
if (m_eglDisplay == EGL_NO_DISPLAY) { if (m_eglDisplay == EGL_NO_DISPLAY)
{
fprintf(stderr, "EGL: Failed to open a display! (0x%x)\n", eglGetError()); fprintf(stderr, "EGL: Failed to open a display! (0x%x)\n", eglGetError());
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
if (!eglInitialize(m_eglDisplay, nullptr, nullptr)) { if (!eglInitialize(m_eglDisplay, nullptr, nullptr))
{
fprintf(stderr, "EGL: Failed to initialize the display! (0x%x)\n", eglGetError()); fprintf(stderr, "EGL: Failed to initialize the display! (0x%x)\n", eglGetError());
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
@ -305,7 +317,8 @@ void *GSWndEGL_X11::CreateNativeWindow(int w, int h)
XMapWindow (m_NativeDisplay, m_NativeWindow); XMapWindow (m_NativeDisplay, m_NativeWindow);
#endif #endif
if (m_NativeDisplay == nullptr) { if (m_NativeDisplay == nullptr)
{
fprintf(stderr, "EGL X11: display wasn't created before the window\n"); fprintf(stderr, "EGL X11: display wasn't created before the window\n");
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
@ -318,7 +331,8 @@ void *GSWndEGL_X11::CreateNativeWindow(int w, int h)
m_NativeWindow = xcb_generate_id(c); m_NativeWindow = xcb_generate_id(c);
if (m_NativeWindow == 0) { if (m_NativeWindow == 0)
{
fprintf(stderr, "EGL X11: failed to create the native window\n"); fprintf(stderr, "EGL X11: failed to create the native window\n");
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} }
@ -341,7 +355,8 @@ void *GSWndEGL_X11::AttachNativeWindow(void *handle)
void GSWndEGL_X11::DestroyNativeResources() void GSWndEGL_X11::DestroyNativeResources()
{ {
if (m_NativeDisplay) { if (m_NativeDisplay)
{
XCloseDisplay(m_NativeDisplay); XCloseDisplay(m_NativeDisplay);
m_NativeDisplay = nullptr; m_NativeDisplay = nullptr;
} }
@ -349,7 +364,8 @@ void GSWndEGL_X11::DestroyNativeResources()
bool GSWndEGL_X11::SetWindowText(const char* title) bool GSWndEGL_X11::SetWindowText(const char* title)
{ {
if (!m_managed) return true; if (!m_managed)
return true;
xcb_connection_t* c = XGetXCBConnection(m_NativeDisplay); xcb_connection_t* c = XGetXCBConnection(m_NativeDisplay);
@ -393,7 +409,8 @@ void *GSWndEGL_WL::AttachNativeWindow(void *handle)
void GSWndEGL_WL::DestroyNativeResources() void GSWndEGL_WL::DestroyNativeResources()
{ {
if (m_NativeDisplay) { if (m_NativeDisplay)
{
wl_display_disconnect(m_NativeDisplay); wl_display_disconnect(m_NativeDisplay);
m_NativeDisplay = nullptr; m_NativeDisplay = nullptr;
} }

6
plugins/GSdx/Window/GSWndEGL.h
View File

@ -71,9 +71,9 @@ public:
void Flip() final; void Flip() final;
// Deprecated API // Deprecated API
void Show() final {}; void Show() final {}
void Hide() final {}; void Hide() final {}
void HideFrame() final {}; // DX9 API void HideFrame() final {} // DX9 API
virtual void* GetDisplay() = 0; // GSopen1 API virtual void* GetDisplay() = 0; // GSopen1 API
virtual void* GetHandle() = 0; // DX API virtual void* GetHandle() = 0; // DX API

46
plugins/GSdx/Window/GSWndWGL.cpp
View File

@ -30,10 +30,13 @@ static void win_error(const wchar_t* msg, bool fatal = true)
if (errorID) if (errorID)
fprintf(stderr, "WIN API ERROR:%ld\t", errorID); fprintf(stderr, "WIN API ERROR:%ld\t", errorID);
if (fatal) { if (fatal)
{
MessageBox(NULL, msg, L"ERROR", MB_OK | MB_ICONEXCLAMATION); MessageBox(NULL, msg, L"ERROR", MB_OK | MB_ICONEXCLAMATION);
throw GSDXRecoverableError(); throw GSDXRecoverableError();
} else { }
else
{
fprintf(stderr, "ERROR:%ls\n", msg); fprintf(stderr, "ERROR:%ls\n", msg);
} }
} }
@ -105,7 +108,8 @@ void GSWndWGL::CreateContext(int major, int minor)
win_error(L"Failed to init wglCreateContextAttribsARB function pointer"); win_error(L"Failed to init wglCreateContextAttribsARB function pointer");
HGLRC context30 = wglCreateContextAttribsARB(m_NativeDisplay, NULL, context_attribs); HGLRC context30 = wglCreateContextAttribsARB(m_NativeDisplay, NULL, context_attribs);
if (!context30) { if (!context30)
{
win_error(L"Failed to create a 3.x context with standard flags", false); win_error(L"Failed to create a 3.x context with standard flags", false);
// retry with more compatible option for (Mesa on Windows, OpenGL on WINE) // retry with more compatible option for (Mesa on Windows, OpenGL on WINE)
context_attribs[2 * 2 + 1] = 0; context_attribs[2 * 2 + 1] = 0;
@ -125,7 +129,8 @@ void GSWndWGL::CreateContext(int major, int minor)
void GSWndWGL::AttachContext() void GSWndWGL::AttachContext()
{ {
if (!IsContextAttached()) { if (!IsContextAttached())
{
wglMakeCurrent(m_NativeDisplay, m_context); wglMakeCurrent(m_NativeDisplay, m_context);
m_ctx_attached = true; m_ctx_attached = true;
} }
@ -133,7 +138,8 @@ void GSWndWGL::AttachContext()
void GSWndWGL::DetachContext() void GSWndWGL::DetachContext()
{ {
if (IsContextAttached()) { if (IsContextAttached())
{
wglMakeCurrent(NULL, NULL); wglMakeCurrent(NULL, NULL);
m_ctx_attached = false; m_ctx_attached = false;
} }
@ -145,10 +151,13 @@ void GSWndWGL::PopulateWndGlFunction()
// To ease the process, extension management is itself an extension. Clever isn't it! // To ease the process, extension management is itself an extension. Clever isn't it!
PFNWGLGETEXTENSIONSSTRINGARBPROC wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC)wglGetProcAddress("wglGetExtensionsStringARB"); PFNWGLGETEXTENSIONSSTRINGARBPROC wglGetExtensionsStringARB = (PFNWGLGETEXTENSIONSSTRINGARBPROC)wglGetProcAddress("wglGetExtensionsStringARB");
if (wglGetExtensionsStringARB) { if (wglGetExtensionsStringARB)
{
const char* ext = wglGetExtensionsStringARB(m_NativeDisplay); const char* ext = wglGetExtensionsStringARB(m_NativeDisplay);
m_has_late_vsync = m_swapinterval && ext && strstr(ext, "WGL_EXT_swap_control_tear"); m_has_late_vsync = m_swapinterval && ext && strstr(ext, "WGL_EXT_swap_control_tear");
} else { }
else
{
m_has_late_vsync = false; m_has_late_vsync = false;
} }
} }
@ -173,7 +182,8 @@ void GSWndWGL::Detach()
// The window still need to be closed // The window still need to be closed
DetachContext(); DetachContext();
if (m_context) wglDeleteContext(m_context); if (m_context)
wglDeleteContext(m_context);
m_context = NULL; m_context = NULL;
CloseWGLDisplay(); CloseWGLDisplay();
@ -184,14 +194,12 @@ void GSWndWGL::Detach()
DestroyWindow(m_NativeWindow); DestroyWindow(m_NativeWindow);
m_NativeWindow = NULL; m_NativeWindow = NULL;
} }
} }
void GSWndWGL::OpenWGLDisplay() void GSWndWGL::OpenWGLDisplay()
{ {
GLuint PixelFormat; // Holds The Results After Searching For A Match GLuint PixelFormat; // Holds The Results After Searching For A Match
PIXELFORMATDESCRIPTOR pfd = // pfd Tells Windows How We Want Things To Be PIXELFORMATDESCRIPTOR pfd = // pfd Tells Windows How We Want Things To Be
{ {
sizeof(PIXELFORMATDESCRIPTOR), // Size Of This Pixel Format Descriptor sizeof(PIXELFORMATDESCRIPTOR), // Size Of This Pixel Format Descriptor
1, // Version Number 1, // Version Number
@ -239,7 +247,8 @@ void GSWndWGL::CloseWGLDisplay()
// modifications // modifications
bool GSWndWGL::Create(const std::string& title, int w, int h) bool GSWndWGL::Create(const std::string& title, int w, int h)
{ {
if(m_NativeWindow) return false; if (m_NativeWindow)
return false;
m_managed = true; m_managed = true;
@ -293,7 +302,8 @@ bool GSWndWGL::Create(const std::string& title, int w, int h)
std::wstring tmp = std::wstring(title.begin(), title.end()); std::wstring tmp = std::wstring(title.begin(), title.end());
m_NativeWindow = CreateWindow(wc.lpszClassName, tmp.c_str(), style, r.left, r.top, r.width(), r.height(), NULL, NULL, wc.hInstance, (LPVOID)this); m_NativeWindow = CreateWindow(wc.lpszClassName, tmp.c_str(), style, r.left, r.top, r.width(), r.height(), NULL, NULL, wc.hInstance, (LPVOID)this);
if (m_NativeWindow == NULL) return false; if (m_NativeWindow == NULL)
return false;
OpenWGLDisplay(); OpenWGLDisplay();
@ -327,7 +337,8 @@ void* GSWndWGL::GetProcAddress(const char* name, bool opt)
ptr = (void *)GetProcAddress(module, name); ptr = (void *)GetProcAddress(module, name);
} }
#endif #endif
if (ptr == NULL) { if (ptr == NULL)
{
if (theApp.GetConfigB("debug_opengl")) if (theApp.GetConfigB("debug_opengl"))
fprintf(stderr, "Failed to find %s\n", name); fprintf(stderr, "Failed to find %s\n", name);
@ -344,7 +355,8 @@ void GSWndWGL::SetSwapInterval()
// m_swapinterval uses an integer as parameter // m_swapinterval uses an integer as parameter
// 0 -> disable vsync // 0 -> disable vsync
// n -> wait n frame // n -> wait n frame
if (m_swapinterval) m_swapinterval(m_vsync); if (m_swapinterval)
m_swapinterval(m_vsync);
} }
void GSWndWGL::Flip() void GSWndWGL::Flip()
@ -357,7 +369,8 @@ void GSWndWGL::Flip()
void GSWndWGL::Show() void GSWndWGL::Show()
{ {
if (!m_managed) return; if (!m_managed)
return;
// Used by GSReplay // Used by GSReplay
SetForegroundWindow(m_NativeWindow); SetForegroundWindow(m_NativeWindow);
@ -378,7 +391,8 @@ void GSWndWGL::HideFrame()
bool GSWndWGL::SetWindowText(const char* title) bool GSWndWGL::SetWindowText(const char* title)
{ {
if (!m_managed) return false; if (!m_managed)
return false;
const size_t tmp_size = strlen(title) + 1; const size_t tmp_size = strlen(title) + 1;
std::wstring tmp(tmp_size, L'#'); std::wstring tmp(tmp_size, L'#');

2
plugins/GSdx/Window/GSWndWGL.h
View File

@ -45,7 +45,7 @@ class GSWndWGL : public GSWndGL
public: public:
GSWndWGL(); GSWndWGL();
virtual ~GSWndWGL() {}; virtual ~GSWndWGL() {}
bool Create(const std::string& title, int w, int h); bool Create(const std::string& title, int w, int h);
bool Attach(void* handle, bool managed = true); bool Attach(void* handle, bool managed = true);

Some files were not shown because too many files have changed in this diff Show More