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;
}
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
// instead to emulate it with shader

2
plugins/GSdx/GSDeviceOGL.h
View File

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

5
plugins/GSdx/GSTextureCache.cpp
View File

@ -291,8 +291,7 @@ GSTextureCache::Target* GSTextureCache::LookupTarget(const GIFRegTEX0& TEX0, int
if (type == DepthStencil) {
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;
ASSERT(shader <= 14);
int shader = 13 + GSLocalMemory::m_psm[TEX0.PSM].fmt;
m_renderer->m_dev->StretchRect(t->m_texture, sRect, dst->m_texture, dRect, shader, false);
} else {
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) {
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

98
plugins/GSdx/res/glsl/convert.glsl
View File

@ -88,17 +88,6 @@ vec4 ps_crt(uint i)
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
void ps_main0()
{
@ -143,7 +132,7 @@ void ps_main1()
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));
#endif
@ -155,20 +144,17 @@ void ps_main1()
void ps_main10()
{
// Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture
vec4 c = sample_c();
const float exp2_32 = exp2(32.0f);
SV_Target1 = uint(exp2_32 * c.r);
SV_Target1 = uint(exp2(32.0f) * sample_c().r);
}
#endif
#ifdef ps_main11
void ps_main11()
{
const float exp2_32 = exp2(32.0f);
const vec4 bitSh = vec4(256.0*256.0*256.0, 256.0*256.0, 256.0, 1.0);
// Convert a GL_FLOAT32 depth texture into a RGBA color texture
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);
// Convert a GL_FLOAT32 depth texture into a RGBA texture
vec4 res = fract(vec4(sample_c().r) * bitSh);
SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f;
@ -176,44 +162,65 @@ void ps_main11()
#endif
#ifdef ps_main12
//out float gl_FragDepth;
void ps_main12()
{
// Convert a RRGBA texture into a float depth texture
// FIXME: I'm afraid of the accuracy
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);
gl_FragDepth = dot(sample_c(), bitSh);
// Convert a GL_FLOAT32 (only 16 lsb) depth into a RGB5A1 color texture
const vec4 bitSh = vec4(exp2(32.0f), exp2(27.0f), exp2(22.0f), exp2(17.0f));
const uvec4 bitMsk = uvec4(0x1F, 0x1F, 0x1F, 0x1);
uvec4 color = uvec4(vec4(sample_c().r) * bitSh) & bitMsk;
SV_Target0 = vec4(color) / vec4(32.0f, 32.0f, 32.0f, 1.0f);
}
#endif
#ifdef ps_main13
//out float gl_FragDepth;
void ps_main13()
{
// Convert a RRGBA texture into a float depth texture
// FIXME: I'm afraid of the accuracy
const vec4 bitSh = vec4(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f), exp(-8.0f)) * vec4(255.0);
gl_FragDepth = dot(sample_c(), bitSh);
}
#endif
#ifdef ps_main14
void ps_main14()
{
// 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
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);
const vec3 bitSh = vec3(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f)) * vec3(255.0);
gl_FragDepth = dot(sample_c().rgb, bitSh);
}
#endif
#ifdef ps_main14
//out float gl_FragDepth;
void ps_main14()
{
// 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(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0)) * vec2(255.0/256.0);
gl_FragDepth = dot(sample_c().rg, bitSh);
}
#endif
#ifdef 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
@ -222,7 +229,7 @@ void ps_main15()
// boost on MGS3
//
// Hypothesis wrong in Prince of Persia ... Seriously WTF !
//#define ONLY_BLUE;
//#define ONLY_BLUE;
// Convert a RGBA texture into a 8 bits packed texture
// Input column: 8x2 RGBA pixels
@ -236,7 +243,7 @@ void ps_main15()
float c;
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 txN = tb.x | (int(gl_FragCoord.x) & 7);
@ -306,6 +313,17 @@ void ps_main7()
#endif
#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
{
highp uvec4 p = uvec4(PSin_p);

102
plugins/GSdx/res/glsl_source.h
View File

@ -113,17 +113,6 @@ static const char* convert_glsl =
" 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"
"#ifdef ps_main0\n"
"void ps_main0()\n"
"{\n"
@ -168,7 +157,7 @@ static const char* convert_glsl =
"\n"
" highp uvec4 i = uvec4(c * vec4(uint(0x001f), uint(0x03e0), uint(0x7c00), uint(0x8000)));\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"
"#endif\n"
"\n"
@ -180,20 +169,17 @@ static const char* convert_glsl =
"void ps_main10()\n"
"{\n"
" // Convert a GL_FLOAT32 depth texture into a 32 bits UINT texture\n"
" vec4 c = sample_c();\n"
" const float exp2_32 = exp2(32.0f);\n"
" SV_Target1 = uint(exp2_32 * c.r);\n"
" SV_Target1 = uint(exp2(32.0f) * sample_c().r);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main11\n"
"void ps_main11()\n"
"{\n"
" const float exp2_32 = exp2(32.0f);\n"
" const vec4 bitSh = vec4(256.0*256.0*256.0, 256.0*256.0, 256.0, 1.0);\n"
" // Convert a GL_FLOAT32 depth texture into a RGBA color texture\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"
"\n"
" // Convert a GL_FLOAT32 depth texture into a RGBA texture\n"
" vec4 res = fract(vec4(sample_c().r) * bitSh);\n"
"\n"
" SV_Target0 = (res - res.xxyz * bitMsk) * 256.0f/255.0f;\n"
@ -201,45 +187,66 @@ static const char* convert_glsl =
"#endif\n"
"\n"
"#ifdef ps_main12\n"
"//out float gl_FragDepth;\n"
"void ps_main12()\n"
"{\n"
" // Convert a RRGBA texture into a float depth texture\n"
" // FIXME: I'm afraid of the accuracy\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"
" gl_FragDepth = dot(sample_c(), bitSh);\n"
" // Convert a GL_FLOAT32 (only 16 lsb) depth into a RGB5A1 color texture\n"
" const vec4 bitSh = vec4(exp2(32.0f), exp2(27.0f), exp2(22.0f), exp2(17.0f));\n"
" const uvec4 bitMsk = uvec4(0x1F, 0x1F, 0x1F, 0x1);\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"
"#endif\n"
"\n"
"#ifdef ps_main13\n"
"//out float gl_FragDepth;\n"
"void ps_main13()\n"
"{\n"
" // Convert a RRGBA texture into a float depth texture\n"
" // FIXME: I'm afraid of the accuracy\n"
" const vec4 bitSh = vec4(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f), exp(-8.0f)) * vec4(255.0);\n"
" gl_FragDepth = dot(sample_c(), bitSh);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main14\n"
"void ps_main14()\n"
"{\n"
" // Same as above but without the alpha channel (24 bits Z)\n"
"\n"
" // Convert a RRGBA texture into a float depth texture\n"
" // FIXME: I'm afraid of the accuracy\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"
" const vec3 bitSh = vec3(exp2(-32.0f), exp2(-24.0f), exp2(-16.0f)) * vec3(255.0);\n"
" gl_FragDepth = dot(sample_c().rgb, bitSh);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main14\n"
"//out float gl_FragDepth;\n"
"void ps_main14()\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(1.0/(256.0*256.0*256.0), 1.0/(256.0*256.0)) * vec2(255.0/256.0);\n"
" gl_FragDepth = dot(sample_c().rg, bitSh);\n"
"}\n"
"#endif\n"
"\n"
"#ifdef ps_main15\n"
"void ps_main15()\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"
" // Potential speed optimization. There is a high probability that\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"
" //\n"
" // Hypothesis wrong in Prince of Persia ... Seriously WTF !\n"
"//#define ONLY_BLUE;\n"
" //#define ONLY_BLUE;\n"
"\n"
" // Convert a RGBA texture into a 8 bits packed texture\n"
" // Input column: 8x2 RGBA pixels\n"
@ -261,7 +268,7 @@ static const char* convert_glsl =
" float c;\n"
"\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"
" int ty = tb.y | (int(gl_FragCoord.y) & 1);\n"
" int txN = tb.x | (int(gl_FragCoord.x) & 7);\n"
@ -331,6 +338,17 @@ static const char* convert_glsl =
"#endif\n"
"\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"
"{\n"
" highp uvec4 p = uvec4(PSin_p);\n"
@ -369,11 +387,11 @@ static const char* convert_glsl =
"\n"
" const float PI = 3.14159265359f;\n"
"\n"
" vec2 texdim = vec2(textureSize(TextureSampler, 0)); \n"
" vec2 texdim = vec2(textureSize(TextureSampler, 0));\n"
"\n"
" vec4 c;\n"
" if (dFdy(PSin_t.y) * PSin_t.y > 0.5f) {\n"
" c = sample_c(); \n"
" c = sample_c();\n"
" } else {\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"