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

gsdx-ogl: move depth conversion shader 

Add 2 new shaders:
* ps_main12: cast a 16 bit depth to a RGB5A1 color
* ps_main16: cast a a RGB5A1 color to a 16 bit depth

Shader might be used in future commit as it seems Silent Hill uses this
kind of format.

Fix tab/indentation too
This commit is contained in:
Gregory Hainaut 2015-09-08 11:19:07 +02:00
parent 4eed4ca3a1
commit ca9b5ce11d
5 changed files with 181 additions and 145 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
plugins/GSdx/GSDeviceOGL.cpp
View File

@ -950,7 +950,8 @@ void GSDeviceOGL::StretchRect(GSTexture* sTex, const GSVector4& sRect, GSTexture
return; return;
} }
bool draw_in_depth = (ps == m_convert.ps[12] || ps == m_convert.ps[13] || ps == m_convert.ps[14]); bool draw_in_depth = (ps == m_convert.ps[13] || ps == m_convert.ps[14] ||
ps == m_convert.ps[15] || ps == m_convert.ps[16]);
// Performance optimization. It might be faster to use a framebuffer blit for standard case // Performance optimization. It might be faster to use a framebuffer blit for standard case
// instead to emulate it with shader // instead to emulate it with shader

2
plugins/GSdx/GSDeviceOGL.h
View File

@ -411,7 +411,7 @@ class GSDeviceOGL : public GSDevice
struct { struct {
GLuint vs; // program object GLuint vs; // program object
GLuint ps[16]; // program object GLuint ps[18]; // program object
GLuint ln; // sampler object GLuint ln; // sampler object
GLuint pt; // sampler object GLuint pt; // sampler object
GSDepthStencilOGL* dss; GSDepthStencilOGL* dss;

5
plugins/GSdx/GSTextureCache.cpp
View File

@ -291,8 +291,7 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
if (type == DepthStencil) { if (type == DepthStencil) {
GL_CACHE("TC: Lookup Target(Depth) %dx%d, hit Color (0x%x, F:0x%x)", w, h, bp, TEX0.PSM); GL_CACHE("TC: Lookup Target(Depth) %dx%d, hit Color (0x%x, F:0x%x)", w, h, bp, TEX0.PSM);
int shader = 12 + GSLocalMemory::m_psm[TEX0.PSM].fmt; int shader = 13 + GSLocalMemory::m_psm[TEX0.PSM].fmt;
ASSERT(shader <= 14);
m_renderer->m_dev->StretchRect(t->m_texture, sRect, dst->m_texture, dRect, shader, false); m_renderer->m_dev->StretchRect(t->m_texture, sRect, dst->m_texture, dRect, shader, false);
} else { } else {
GL_CACHE("TC: Lookup Target(Color) %dx%d, hit Depth (0x%x, F:0x%x)", w, h, bp, TEX0.PSM); GL_CACHE("TC: Lookup Target(Color) %dx%d, hit Depth (0x%x, F:0x%x)", w, h, bp, TEX0.PSM);
@ -870,7 +869,7 @@ GSTextureCache::Source* GSTextureCache::CreateSource(const GIFRegTEX0& TEX0, con
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 = 15; // ask a conversion to 8 bits format shader = 17; // ask a conversion to 8 bits format
} }
#ifdef ENABLE_OGL_DEBUG #ifdef ENABLE_OGL_DEBUG

158
plugins/GSdx/res/glsl/convert.glsl
View File

@ -85,18 +85,7 @@ vec4 ps_crt(uint i)
vec4(0, 0, 1, 0), vec4(0, 0, 1, 0),
vec4(1, 1, 1, 0) vec4(1, 1, 1, 0)
); );
return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f); return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);
}
vec4 ps_scanlines(uint i)
{
vec4 mask[2] =
{
vec4(1, 1, 1, 0),
vec4(0, 0, 0, 0)
};
return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);
} }
#ifdef ps_main0 #ifdef ps_main0
@ -109,41 +98,41 @@ void ps_main0()
#ifdef ps_main1 #ifdef ps_main1
void ps_main1() void ps_main1()
{ {
// Input Color is RGBA8 // Input Color is RGBA8
// We want to output a pixel on the PSMCT16* format // We want to output a pixel on the PSMCT16* format
// A1-BGR5 // A1-BGR5
#if 0 #if 0
// Note: dot is a good idea from pseudo. However we must be careful about float accuraccy. // Note: dot is a good idea from pseudo. However we must be careful about float accuraccy.
// Here a global idea example: // Here a global idea example:
// //
// SV_Target1 = dot(round(sample_c() * vec4(31.f, 31.f, 31.f, 1.f)), vec4(1.f, 32.f, 1024.f, 32768.f)); // SV_Target1 = dot(round(sample_c() * vec4(31.f, 31.f, 31.f, 1.f)), vec4(1.f, 32.f, 1024.f, 32768.f));
// //
// For me this code is more accurate but it will require some tests // For me this code is more accurate but it will require some tests
vec4 c = sample_c() * 255.0f + 0.5f; // Denormalize value to avoid float precision issue vec4 c = sample_c() * 255.0f + 0.5f; // Denormalize value to avoid float precision issue
// shift Red: -3 // shift Red: -3
// shift Green: -3 + 5 // shift Green: -3 + 5
// shift Blue: -3 + 10 // shift Blue: -3 + 10
// shift Alpha: -7 + 15 // shift Alpha: -7 + 15
highp uvec4 i = uvec4(c * vec4(1/8.0f, 4.0f, 128.0f, 256.0f)); // Shift value highp uvec4 i = uvec4(c * vec4(1/8.0f, 4.0f, 128.0f, 256.0f)); // Shift value
// bit field operation requires GL4 HW. Could be nice to merge it with step/mix below // bit field operation requires GL4 HW. Could be nice to merge it with step/mix below
SV_Target1 = (i.r & uint(0x001f)) | (i.g & uint(0x03e0)) | (i.b & uint(0x7c00)) | (i.a & uint(0x8000)); SV_Target1 = (i.r & uint(0x001f)) | (i.g & uint(0x03e0)) | (i.b & uint(0x7c00)) | (i.a & uint(0x8000));
#else #else
// Old code which is likely wrong. // Old code which is likely wrong.
vec4 c = sample_c(); vec4 c = sample_c();
c.a *= 256.0f / 127.0f; // hm, 0.5 won't give us 1.0 if we just multiply with 2 c.a *= 256.0f / 127.0f; // hm, 0.5 won't give us 1.0 if we just multiply with 2
highp uvec4 i = uvec4(c * vec4(uint(0x001f), uint(0x03e0), uint(0x7c00), uint(0x8000))); highp uvec4 i = uvec4(c * vec4(uint(0x001f), uint(0x03e0), uint(0x7c00), uint(0x8000)));
// bit field operation requires GL4 HW. Could be nice to merge it with step/mix below // bit field operation requires GL4 HW.
SV_Target1 = (i.x & uint(0x001f)) | (i.y & uint(0x03e0)) | (i.z & uint(0x7c00)) | (i.w & uint(0x8000)); SV_Target1 = (i.x & uint(0x001f)) | (i.y & uint(0x03e0)) | (i.z & uint(0x7c00)) | (i.w & uint(0x8000));
#endif #endif
@ -154,66 +143,84 @@ void ps_main1()
#ifdef ps_main10 #ifdef ps_main10
void ps_main10() void ps_main10()
{ {
// Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture // Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture
vec4 c = sample_c(); SV_Target1 = uint(exp2(32.0f) * sample_c().r);
const float exp2_32 = exp2(32.0f);
SV_Target1 = uint(exp2_32 * c.r);
} }
#endif #endif
#ifdef ps_main11 #ifdef ps_main11
void ps_main11() void ps_main11()
{ {
const float exp2_32 = exp2(32.0f); // Convert a GL_FLOAT32 depth texture into a RGBA color texture
const vec4 bitSh = vec4(256.0*256.0*256.0, 256.0*256.0, 256.0, 1.0); const vec4 bitSh = vec4(exp2(24.0f), exp2(16.0f), exp2(8.0f), exp2(0.0f));
const vec4 bitMsk = vec4(0.0, 1.0/256.0, 1.0/256.0, 1.0/256.0); const vec4 bitMsk = vec4(0.0, 1.0/256.0, 1.0/256.0, 1.0/256.0);
// Convert a GL_FLOAT32 depth texture into a RGBA texture vec4 res = fract(vec4(sample_c().r) * bitSh);
vec4 res = fract(vec4(sample_c().r) * bitSh);
SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f; SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f;
} }
#endif #endif
#ifdef ps_main12 #ifdef ps_main12
//out float gl_FragDepth;
void ps_main12() void ps_main12()
{ {
// Convert a RRGBA texture into a float depth texture // Convert a GL_FLOAT32 (only 16 lsb) depth into a RGB5A1 color texture
// FIXME: I'm afraid of the accuracy const vec4 bitSh = vec4(exp2(32.0f), exp2(27.0f), exp2(22.0f), exp2(17.0f));
const vec4 bitSh = vec4(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0), 1.0/256.0, 1.0) * vec4(255.0/256.0); const uvec4 bitMsk = uvec4(0x1F, 0x1F, 0x1F, 0x1);
gl_FragDepth = dot(sample_c(), bitSh); uvec4 color = uvec4(vec4(sample_c().r) * bitSh) & bitMsk;
SV_Target0 = vec4(color) / vec4(32.0f, 32.0f, 32.0f, 1.0f);
} }
#endif #endif
#ifdef ps_main13 #ifdef ps_main13
//out float gl_FragDepth;
void ps_main13() void ps_main13()
{ {
// Same as above but without the alpha channel (24 bits Z) // Convert a RRGBA texture into a float depth texture
// FIXME: I'm afraid of the accuracy
// Convert a RRGBA texture into a float depth texture const vec4 bitSh = vec4(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f), exp(-8.0f)) * vec4(255.0);
// FIXME: I'm afraid of the accuracy gl_FragDepth = dot(sample_c(), bitSh);
const vec3 bitSh = vec3(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0), 1.0/256.0) * vec3(255.0/256.0);
gl_FragDepth = dot(sample_c().rgb, bitSh);
} }
#endif #endif
#ifdef ps_main14 #ifdef ps_main14
//out float gl_FragDepth;
void ps_main14() void ps_main14()
{ {
// Same as above but without the A/B channels (16 bits Z) // Same as above but without the alpha channel (24 bits Z)
// Convert a RRGBA texture into a float depth texture // Convert a RRGBA texture into a float depth texture
// FIXME: I'm afraid of the accuracy // FIXME: I'm afraid of the accuracy
const vec2 bitSh = vec2(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0)) * vec2(255.0/256.0); const vec3 bitSh = vec3(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f)) * vec3(255.0);
gl_FragDepth = dot(sample_c().rg, bitSh); gl_FragDepth = dot(sample_c().rgb, bitSh);
} }
#endif #endif
#ifdef ps_main15 #ifdef ps_main15
void ps_main15() void ps_main15()
{
// Same as above but without the A/B channels (16 bits Z)
// Convert a RRGBA texture into a float depth texture
// FIXME: I'm afraid of the accuracy
const vec2 bitSh = vec2(exp2(-32.0f), exp2(-24.0f)) * vec2(255.0);
gl_FragDepth = dot(sample_c().rg, bitSh);
}
#endif
#ifdef ps_main16
void ps_main16()
{
// Convert a RGB5A1 (saved as RGBA8) color to a 16 bit Z
// FIXME: I'm afraid of the accuracy
const vec4 bitSh = vec4(exp2(-32.0f), exp2(-27.0f), exp2(-22.0f), exp(-17.0f));
// Trunc color to drop useless lsb
vec4 color = trunc(sample_c() * vec4(255.0f) / vec4(8.0f, 8.0f, 8.0f, 128.0f));
gl_FragDepth = dot(vec4(color), bitSh);
}
#endif
#ifdef ps_main17
void ps_main17()
{ {
// Potential speed optimization. There is a high probability that // Potential speed optimization. There is a high probability that
@ -222,7 +229,7 @@ void ps_main15()
// boost on MGS3 // boost on MGS3
// //
// Hypothesis wrong in Prince of Persia ... Seriously WTF ! // Hypothesis wrong in Prince of Persia ... Seriously WTF !
//#define ONLY_BLUE; //#define ONLY_BLUE;
// Convert a RGBA texture into a 8 bits packed texture // Convert a RGBA texture into a 8 bits packed texture
// Input column: 8x2 RGBA pixels // Input column: 8x2 RGBA pixels
@ -236,7 +243,7 @@ void ps_main15()
float c; float c;
uvec2 sel = uvec2(gl_FragCoord.xy) % uvec2(16u, 16u); uvec2 sel = uvec2(gl_FragCoord.xy) % uvec2(16u, 16u);
ivec2 tb = ((ivec2(gl_FragCoord.xy) & ~ivec2(15, 3)) >> 1u); ivec2 tb = ((ivec2(gl_FragCoord.xy) & ~ivec2(15, 3)) >> 1);
int ty = tb.y | (int(gl_FragCoord.y) & 1); int ty = tb.y | (int(gl_FragCoord.y) & 1);
int txN = tb.x | (int(gl_FragCoord.x) & 7); int txN = tb.x | (int(gl_FragCoord.x) & 7);
@ -299,18 +306,29 @@ void ps_main7()
{ {
vec4 c = sample_c(); vec4 c = sample_c();
c.a = dot(c.rgb, vec3(0.299, 0.587, 0.114)); c.a = dot(c.rgb, vec3(0.299, 0.587, 0.114));
SV_Target0 = c; SV_Target0 = c;
} }
#endif #endif
#ifdef ps_main5 #ifdef ps_main5
vec4 ps_scanlines(uint i)
{
vec4 mask[2] =
{
vec4(1, 1, 1, 0),
vec4(0, 0, 0, 0)
};
return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);
}
void ps_main5() // scanlines void ps_main5() // scanlines
{ {
highp uvec4 p = uvec4(PSin_p); highp uvec4 p = uvec4(PSin_p);
vec4 c = ps_scanlines(p.y % 2u); vec4 c = ps_scanlines(p.y % 2u);
SV_Target0 = c; SV_Target0 = c;
} }
@ -319,9 +337,9 @@ void ps_main5() // scanlines
#ifdef ps_main6 #ifdef ps_main6
void ps_main6() // diagonal void ps_main6() // diagonal
{ {
highp uvec4 p = uvec4(PSin_p); highp uvec4 p = uvec4(PSin_p);
vec4 c = ps_crt((p.x + (p.y % 3u)) % 3u); vec4 c = ps_crt((p.x + (p.y % 3u)) % 3u);
SV_Target0 = c; SV_Target0 = c;
} }
@ -330,9 +348,9 @@ void ps_main6() // diagonal
#ifdef ps_main8 #ifdef ps_main8
void ps_main8() // triangular void ps_main8() // triangular
{ {
highp uvec4 p = uvec4(PSin_p); highp uvec4 p = uvec4(PSin_p);
vec4 c = ps_crt(((p.x + ((p.y >> 1u) & 1u) * 3u) >> 1u) % 3u); vec4 c = ps_crt(((p.x + ((p.y >> 1u) & 1u) * 3u) >> 1u) % 3u);
SV_Target0 = c; SV_Target0 = c;
} }
@ -344,14 +362,14 @@ void ps_main9()
const float PI = 3.14159265359f; const float PI = 3.14159265359f;
vec2 texdim = vec2(textureSize(TextureSampler, 0)); vec2 texdim = vec2(textureSize(TextureSampler, 0));
vec4 c; vec4 c;
if (dFdy(PSin_t.y) * PSin_t.y > 0.5f) { if (dFdy(PSin_t.y) * PSin_t.y > 0.5f) {
c = sample_c(); c = sample_c();
} else { } else {
float factor = (0.9f - 0.4f * cos(2.0f * PI * PSin_t.y * texdim.y)); float factor = (0.9f - 0.4f * cos(2.0f * PI * PSin_t.y * texdim.y));
c = factor * texture(TextureSampler, vec2(PSin_t.x, (floor(PSin_t.y * texdim.y) + 0.5f) / texdim.y)); c = factor * texture(TextureSampler, vec2(PSin_t.x, (floor(PSin_t.y * texdim.y) + 0.5f) / texdim.y));
} }
SV_Target0 = c; SV_Target0 = c;

158
plugins/GSdx/res/glsl_source.h
View File

@ -110,18 +110,7 @@ static const char* convert_glsl =
" vec4(0, 0, 1, 0),\n" " vec4(0, 0, 1, 0),\n"
" vec4(1, 1, 1, 0)\n" " vec4(1, 1, 1, 0)\n"
" );\n" " );\n"
" return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);\n" " return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);\n"
"}\n"
"\n"
"vec4 ps_scanlines(uint i)\n"
"{\n"
" vec4 mask[2] =\n"
" {\n"
" vec4(1, 1, 1, 0),\n"
" vec4(0, 0, 0, 0)\n"
" };\n"
"\n"
" return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);\n"
"}\n" "}\n"
"\n" "\n"
"#ifdef ps_main0\n" "#ifdef ps_main0\n"
@ -134,41 +123,41 @@ static const char* convert_glsl =
"#ifdef ps_main1\n" "#ifdef ps_main1\n"
"void ps_main1()\n" "void ps_main1()\n"
"{\n" "{\n"
" // Input Color is RGBA8\n" " // Input Color is RGBA8\n"
"\n" "\n"
" // We want to output a pixel on the PSMCT16* format\n" " // We want to output a pixel on the PSMCT16* format\n"
" // A1-BGR5\n" " // A1-BGR5\n"
"\n" "\n"
"#if 0\n" "#if 0\n"
" // Note: dot is a good idea from pseudo. However we must be careful about float accuraccy.\n" " // Note: dot is a good idea from pseudo. However we must be careful about float accuraccy.\n"
" // Here a global idea example:\n" " // Here a global idea example:\n"
" //\n" " //\n"
" // SV_Target1 = dot(round(sample_c() * vec4(31.f, 31.f, 31.f, 1.f)), vec4(1.f, 32.f, 1024.f, 32768.f));\n" " // SV_Target1 = dot(round(sample_c() * vec4(31.f, 31.f, 31.f, 1.f)), vec4(1.f, 32.f, 1024.f, 32768.f));\n"
" //\n" " //\n"
"\n" "\n"
" // For me this code is more accurate but it will require some tests\n" " // For me this code is more accurate but it will require some tests\n"
"\n" "\n"
" vec4 c = sample_c() * 255.0f + 0.5f; // Denormalize value to avoid float precision issue\n" " vec4 c = sample_c() * 255.0f + 0.5f; // Denormalize value to avoid float precision issue\n"
"\n" "\n"
" // shift Red: -3\n" " // shift Red: -3\n"
" // shift Green: -3 + 5\n" " // shift Green: -3 + 5\n"
" // shift Blue: -3 + 10\n" " // shift Blue: -3 + 10\n"
" // shift Alpha: -7 + 15\n" " // shift Alpha: -7 + 15\n"
" highp uvec4 i = uvec4(c * vec4(1/8.0f, 4.0f, 128.0f, 256.0f)); // Shift value\n" " highp uvec4 i = uvec4(c * vec4(1/8.0f, 4.0f, 128.0f, 256.0f)); // Shift value\n"
"\n" "\n"
" // bit field operation requires GL4 HW. Could be nice to merge it with step/mix below\n" " // bit field operation requires GL4 HW. Could be nice to merge it with step/mix below\n"
" SV_Target1 = (i.r & uint(0x001f)) | (i.g & uint(0x03e0)) | (i.b & uint(0x7c00)) | (i.a & uint(0x8000));\n" " SV_Target1 = (i.r & uint(0x001f)) | (i.g & uint(0x03e0)) | (i.b & uint(0x7c00)) | (i.a & uint(0x8000));\n"
"\n" "\n"
"#else\n" "#else\n"
" // Old code which is likely wrong.\n" " // Old code which is likely wrong.\n"
"\n" "\n"
" vec4 c = sample_c();\n" " vec4 c = sample_c();\n"
"\n" "\n"
" c.a *= 256.0f / 127.0f; // hm, 0.5 won't give us 1.0 if we just multiply with 2\n" " c.a *= 256.0f / 127.0f; // hm, 0.5 won't give us 1.0 if we just multiply with 2\n"
"\n" "\n"
" highp uvec4 i = uvec4(c * vec4(uint(0x001f), uint(0x03e0), uint(0x7c00), uint(0x8000)));\n" " highp uvec4 i = uvec4(c * vec4(uint(0x001f), uint(0x03e0), uint(0x7c00), uint(0x8000)));\n"
"\n" "\n"
" // bit field operation requires GL4 HW. Could be nice to merge it with step/mix below\n" " // bit field operation requires GL4 HW.\n"
" SV_Target1 = (i.x & uint(0x001f)) | (i.y & uint(0x03e0)) | (i.z & uint(0x7c00)) | (i.w & uint(0x8000));\n" " SV_Target1 = (i.x & uint(0x001f)) | (i.y & uint(0x03e0)) | (i.z & uint(0x7c00)) | (i.w & uint(0x8000));\n"
"#endif\n" "#endif\n"
"\n" "\n"
@ -179,67 +168,85 @@ static const char* convert_glsl =
"#ifdef ps_main10\n" "#ifdef ps_main10\n"
"void ps_main10()\n" "void ps_main10()\n"
"{\n" "{\n"
" // Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture\n" " // Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture\n"
" vec4 c = sample_c();\n" " SV_Target1 = uint(exp2(32.0f) * sample_c().r);\n"
" const float exp2_32 = exp2(32.0f);\n"
" SV_Target1 = uint(exp2_32 * c.r);\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main11\n" "#ifdef ps_main11\n"
"void ps_main11()\n" "void ps_main11()\n"
"{\n" "{\n"
" const float exp2_32 = exp2(32.0f);\n" " // Convert a GL_FLOAT32 depth texture into a RGBA color texture\n"
" const vec4 bitSh = vec4(256.0*256.0*256.0, 256.0*256.0, 256.0, 1.0);\n" " const vec4 bitSh = vec4(exp2(24.0f), exp2(16.0f), exp2(8.0f), exp2(0.0f));\n"
" const vec4 bitMsk = vec4(0.0, 1.0/256.0, 1.0/256.0, 1.0/256.0);\n" " const vec4 bitMsk = vec4(0.0, 1.0/256.0, 1.0/256.0, 1.0/256.0);\n"
"\n" "\n"
" // Convert a GL_FLOAT32 depth texture into a RGBA texture\n" " vec4 res = fract(vec4(sample_c().r) * bitSh);\n"
" vec4 res = fract(vec4(sample_c().r) * bitSh);\n"
"\n" "\n"
" SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f;\n" " SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f;\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main12\n" "#ifdef ps_main12\n"
"//out float gl_FragDepth;\n"
"void ps_main12()\n" "void ps_main12()\n"
"{\n" "{\n"
" // Convert a RRGBA texture into a float depth texture\n" " // Convert a GL_FLOAT32 (only 16 lsb) depth into a RGB5A1 color texture\n"
" // FIXME: I'm afraid of the accuracy\n" " const vec4 bitSh = vec4(exp2(32.0f), exp2(27.0f), exp2(22.0f), exp2(17.0f));\n"
" const vec4 bitSh = vec4(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0), 1.0/256.0, 1.0) * vec4(255.0/256.0);\n" " const uvec4 bitMsk = uvec4(0x1F, 0x1F, 0x1F, 0x1);\n"
" gl_FragDepth = dot(sample_c(), bitSh);\n" " uvec4 color = uvec4(vec4(sample_c().r) * bitSh) & bitMsk;\n"
"\n"
" SV_Target0 = vec4(color) / vec4(32.0f, 32.0f, 32.0f, 1.0f);\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main13\n" "#ifdef ps_main13\n"
"//out float gl_FragDepth;\n"
"void ps_main13()\n" "void ps_main13()\n"
"{\n" "{\n"
" // Same as above but without the alpha channel (24 bits Z)\n" " // Convert a RRGBA texture into a float depth texture\n"
"\n" " // FIXME: I'm afraid of the accuracy\n"
" // Convert a RRGBA texture into a float depth texture\n" " const vec4 bitSh = vec4(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f), exp(-8.0f)) * vec4(255.0);\n"
" // FIXME: I'm afraid of the accuracy\n" " gl_FragDepth = dot(sample_c(), bitSh);\n"
" const vec3 bitSh = vec3(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0), 1.0/256.0) * vec3(255.0/256.0);\n"
" gl_FragDepth = dot(sample_c().rgb, bitSh);\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main14\n" "#ifdef ps_main14\n"
"//out float gl_FragDepth;\n"
"void ps_main14()\n" "void ps_main14()\n"
"{\n" "{\n"
" // Same as above but without the A/B channels (16 bits Z)\n" " // Same as above but without the alpha channel (24 bits Z)\n"
"\n" "\n"
" // Convert a RRGBA texture into a float depth texture\n" " // Convert a RRGBA texture into a float depth texture\n"
" // FIXME: I'm afraid of the accuracy\n" " // FIXME: I'm afraid of the accuracy\n"
" const vec2 bitSh = vec2(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0)) * vec2(255.0/256.0);\n" " const vec3 bitSh = vec3(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f)) * vec3(255.0);\n"
" gl_FragDepth = dot(sample_c().rg, bitSh);\n" " gl_FragDepth = dot(sample_c().rgb, bitSh);\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main15\n" "#ifdef ps_main15\n"
"void ps_main15()\n" "void ps_main15()\n"
"{\n" "{\n"
" // Same as above but without the A/B channels (16 bits Z)\n"
"\n"
" // Convert a RRGBA texture into a float depth texture\n"
" // FIXME: I'm afraid of the accuracy\n"
" const vec2 bitSh = vec2(exp2(-32.0f), exp2(-24.0f)) * vec2(255.0);\n"
" gl_FragDepth = dot(sample_c().rg, bitSh);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main16\n"
"void ps_main16()\n"
"{\n"
" // Convert a RGB5A1 (saved as RGBA8) color to a 16 bit Z\n"
" // FIXME: I'm afraid of the accuracy\n"
" const vec4 bitSh = vec4(exp2(-32.0f), exp2(-27.0f), exp2(-22.0f), exp(-17.0f));\n"
" // Trunc color to drop useless lsb\n"
" vec4 color = trunc(sample_c() * vec4(255.0f) / vec4(8.0f, 8.0f, 8.0f, 128.0f));\n"
" gl_FragDepth = dot(vec4(color), bitSh);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main17\n"
"void ps_main17()\n"
"{\n"
"\n" "\n"
" // Potential speed optimization. There is a high probability that\n" " // Potential speed optimization. There is a high probability that\n"
" // game only want to extract a single channel (blue). It will allow\n" " // game only want to extract a single channel (blue). It will allow\n"
@ -247,7 +254,7 @@ static const char* convert_glsl =
" // boost on MGS3\n" " // boost on MGS3\n"
" //\n" " //\n"
" // Hypothesis wrong in Prince of Persia ... Seriously WTF !\n" " // Hypothesis wrong in Prince of Persia ... Seriously WTF !\n"
"//#define ONLY_BLUE;\n" " //#define ONLY_BLUE;\n"
"\n" "\n"
" // Convert a RGBA texture into a 8 bits packed texture\n" " // Convert a RGBA texture into a 8 bits packed texture\n"
" // Input column: 8x2 RGBA pixels\n" " // Input column: 8x2 RGBA pixels\n"
@ -261,7 +268,7 @@ static const char* convert_glsl =
" float c;\n" " float c;\n"
"\n" "\n"
" uvec2 sel = uvec2(gl_FragCoord.xy) % uvec2(16u, 16u);\n" " uvec2 sel = uvec2(gl_FragCoord.xy) % uvec2(16u, 16u);\n"
" ivec2 tb = ((ivec2(gl_FragCoord.xy) & ~ivec2(15, 3)) >> 1u);\n" " ivec2 tb = ((ivec2(gl_FragCoord.xy) & ~ivec2(15, 3)) >> 1);\n"
"\n" "\n"
" int ty = tb.y | (int(gl_FragCoord.y) & 1);\n" " int ty = tb.y | (int(gl_FragCoord.y) & 1);\n"
" int txN = tb.x | (int(gl_FragCoord.x) & 7);\n" " int txN = tb.x | (int(gl_FragCoord.x) & 7);\n"
@ -324,18 +331,29 @@ static const char* convert_glsl =
"{\n" "{\n"
" vec4 c = sample_c();\n" " vec4 c = sample_c();\n"
"\n" "\n"
" c.a = dot(c.rgb, vec3(0.299, 0.587, 0.114));\n" " c.a = dot(c.rgb, vec3(0.299, 0.587, 0.114));\n"
"\n" "\n"
" SV_Target0 = c;\n" " SV_Target0 = c;\n"
"}\n" "}\n"
"#endif\n" "#endif\n"
"\n" "\n"
"#ifdef ps_main5\n" "#ifdef ps_main5\n"
"vec4 ps_scanlines(uint i)\n"
"{\n"
" vec4 mask[2] =\n"
" {\n"
" vec4(1, 1, 1, 0),\n"
" vec4(0, 0, 0, 0)\n"
" };\n"
"\n"
" return sample_c() * clamp((mask[i] + 0.5f), 0.0f, 1.0f);\n"
"}\n"
"\n"
"void ps_main5() // scanlines\n" "void ps_main5() // scanlines\n"
"{\n" "{\n"
" highp uvec4 p = uvec4(PSin_p);\n" " highp uvec4 p = uvec4(PSin_p);\n"
"\n" "\n"
" vec4 c = ps_scanlines(p.y % 2u);\n" " vec4 c = ps_scanlines(p.y % 2u);\n"
"\n" "\n"
" SV_Target0 = c;\n" " SV_Target0 = c;\n"
"}\n" "}\n"
@ -344,9 +362,9 @@ static const char* convert_glsl =
"#ifdef ps_main6\n" "#ifdef ps_main6\n"
"void ps_main6() // diagonal\n" "void ps_main6() // diagonal\n"
"{\n" "{\n"
" highp uvec4 p = uvec4(PSin_p);\n" " highp uvec4 p = uvec4(PSin_p);\n"
"\n" "\n"
" vec4 c = ps_crt((p.x + (p.y % 3u)) % 3u);\n" " vec4 c = ps_crt((p.x + (p.y % 3u)) % 3u);\n"
"\n" "\n"
" SV_Target0 = c;\n" " SV_Target0 = c;\n"
"}\n" "}\n"
@ -355,9 +373,9 @@ static const char* convert_glsl =
"#ifdef ps_main8\n" "#ifdef ps_main8\n"
"void ps_main8() // triangular\n" "void ps_main8() // triangular\n"
"{\n" "{\n"
" highp uvec4 p = uvec4(PSin_p);\n" " highp uvec4 p = uvec4(PSin_p);\n"
"\n" "\n"
" vec4 c = ps_crt(((p.x + ((p.y >> 1u) & 1u) * 3u) >> 1u) % 3u);\n" " vec4 c = ps_crt(((p.x + ((p.y >> 1u) & 1u) * 3u) >> 1u) % 3u);\n"
"\n" "\n"
" SV_Target0 = c;\n" " SV_Target0 = c;\n"
"}\n" "}\n"
@ -369,14 +387,14 @@ static const char* convert_glsl =
"\n" "\n"
" const float PI = 3.14159265359f;\n" " const float PI = 3.14159265359f;\n"
"\n" "\n"
" vec2 texdim = vec2(textureSize(TextureSampler, 0)); \n" " vec2 texdim = vec2(textureSize(TextureSampler, 0));\n"
"\n" "\n"
" vec4 c;\n" " vec4 c;\n"
" if (dFdy(PSin_t.y) * PSin_t.y > 0.5f) {\n" " if (dFdy(PSin_t.y) * PSin_t.y > 0.5f) {\n"
" c = sample_c(); \n" " c = sample_c();\n"
" } else {\n" " } else {\n"
" float factor = (0.9f - 0.4f * cos(2.0f * PI * PSin_t.y * texdim.y));\n" " float factor = (0.9f - 0.4f * cos(2.0f * PI * PSin_t.y * texdim.y));\n"
" c = factor * texture(TextureSampler, vec2(PSin_t.x, (floor(PSin_t.y * texdim.y) + 0.5f) / texdim.y));\n" " c = factor * texture(TextureSampler, vec2(PSin_t.x, (floor(PSin_t.y * texdim.y) + 0.5f) / texdim.y));\n"
" }\n" " }\n"
"\n" "\n"
" SV_Target0 = c;\n" " SV_Target0 = c;\n"