libretro/pcsx2
1
0
Fork 0
mirror of https://github.com/libretro/pcsx2.git synced 2024-12-28 20:54:46 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

gsdx glsl: extend hw shader to sample depth texture

Browse Source 
Will use integral coordinate to avoid any rescaling.

Bilinear interpolation isn't supported. I don't think it is allowed to
filter a depth texture anyway.
This commit is contained in:
Gregory Hainaut 2016-04-23 12:06:10 +02:00
parent 583de1bf0b
commit fda511a949
4 changed files with 197 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
plugins/GSdx/GSDeviceOGL.cpp
View File

@ -777,6 +777,7 @@ GLuint GSDeviceOGL::CompilePS(PSSelector sel)
+ format("#define PS_WMT %d\n", sel.wmt)
+ format("#define PS_TEX_FMT %d\n", sel.tex_fmt)
+ format("#define PS_DFMT %d\n", sel.dfmt)
+ format("#define PS_DEPTH_FMT %d\n", sel.depth_fmt)
+ format("#define PS_AEM %d\n", sel.aem)
+ format("#define PS_TFX %d\n", sel.tfx)
+ format("#define PS_TCC %d\n", sel.tcc)

3
plugins/GSdx/GSDeviceOGL.h
View File

@ -247,6 +247,7 @@ class GSDeviceOGL final : public GSDevice
// Format
uint32 tex_fmt:4;
uint32 dfmt:2;
uint32 depth_fmt:2;
// Alpha extension/Correction
uint32 aem:1;
uint32 fba:1;
@ -270,7 +271,7 @@ class GSDeviceOGL final : public GSDevice
uint32 write_rg:1;
uint32 fbmask:1;
uint32 _free1:2;
//uint32 _free1:0;
// *** Word 2
// Blend and Colclip

99
plugins/GSdx/res/glsl/tfx_fs.glsl
View File

@ -175,6 +175,94 @@ mat4 sample_4p(vec4 u)
return c;
}
//////////////////////////////////////////////////////////////////////
// Depth sampling
//////////////////////////////////////////////////////////////////////
vec4 fetch_c(ivec2 uv)
{
return texelFetch(TextureSampler, ivec2(uv), 0);
}
ivec2 clamp_wrap_uv_depth(ivec2 uv)
{
ivec2 uv_out = uv;
// Keep the full precision
// It allow to multiply the ScalingFactor before the 1/16 coeff
ivec4 mask = ivec4(MskFix) << 4;
#if PS_WMS == PS_WMT
#if PS_WMS == 2
uv_out = clamp(uv, mask.xy, mask.zw);
#elif PS_WMS == 3
uv_out = (uv & mask.xy) | mask.zw;
#endif
#else // PS_WMS != PS_WMT
#if PS_WMS == 2
uv_out.x = clamp(uv, mask.x, mask.z);
#elif PS_WMS == 3
uv_out.x = (uv.x & mask.x) | mask.z;
#endif
#if PS_WMT == 2
uv_out.y = clamp(uv, mask.y, mask.w);
#elif PS_WMT == 3
uv_out.y = (uv.y & mask.y) | mask.w;
#endif
#endif
return uv_out;
}
vec4 sample_depth(vec2 st)
{
vec2 uv_f = vec2(clamp_wrap_uv_depth(ivec2(st))) * vec2(ScalingFactor.xy) * vec2(1.0f/16.0f);
ivec2 uv = ivec2(uv_f);
vec4 t;
#if PS_DEPTH_FMT == 1
// Based on ps_main11 of convert
// 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);
vec4 res = fract(vec4(fetch_c(uv).r) * bitSh);
t = (res - res.xxyz * bitMsk) * 256.0f;
#elif PS_DEPTH_FMT == 2
// Based on ps_main12 of convert
// 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(fetch_c(uv).r) * bitSh) & bitMsk;
t = vec4(color) * vec4(8.0f, 8.0f, 8.0f, 128.0f);
#elif PS_DEPTH_FMT == 3
// Convert a RGBA/RGB5A1 color texture into a RGBA/RGB5A1 color texture
t = fetch_c(uv) * 255.0f;
#endif
// warning t ranges from 0 to 255
#if (PS_AEM_FMT == FMT_24)
t.a = ( (PS_AEM == 0) || any(bvec3(t.rgb)) ) ? 255.0f * TA.x : 0.0f;
#elif (PS_AEM_FMT == FMT_16)
t.a = t.a >= 128.0f ? 255.0f * TA.y : ( (PS_AEM == 0) || any(bvec3(t.rgb)) ) ? 255.0f * TA.x : 0.0f;
#endif
return t;
}
//////////////////////////////////////////////////////////////////////
vec4 sample_color(vec2 st)
{
#if (PS_TCOFFSETHACK == 1)
@ -328,10 +416,17 @@ vec4 ps_color()
{
//FIXME: maybe we can set gl_Position.w = q in VS
#if (PS_FST == 0)
vec4 T = sample_color(PSin.t_float.xy / vec2(PSin.t_float.w));
vec2 st = PSin.t_float.xy / vec2(PSin.t_float.w);
#else
// Note xy are normalized coordinate
vec4 T = sample_color(PSin.t_int.xy);
vec2 st = PSin.t_int.xy;
#endif
#if (PS_DEPTH_FMT > 0)
// Integral coordinate
vec4 T = sample_depth(PSin.t_int.zw);
#else
vec4 T = sample_color(st);
#endif
#if PS_IIP == 1

99
plugins/GSdx/res/glsl_source.h
View File

@ -1019,6 +1019,94 @@ static const char* const tfx_fs_all_glsl =
" return c;\n"
"}\n"
"\n"
"//////////////////////////////////////////////////////////////////////\n"
"// Depth sampling\n"
"//////////////////////////////////////////////////////////////////////\n"
"vec4 fetch_c(ivec2 uv)\n"
"{\n"
" return texelFetch(TextureSampler, ivec2(uv), 0);\n"
"}\n"
"\n"
"ivec2 clamp_wrap_uv_depth(ivec2 uv)\n"
"{\n"
" ivec2 uv_out = uv;\n"
"\n"
" // Keep the full precision\n"
" // It allow to multiply the ScalingFactor before the 1/16 coeff\n"
" ivec4 mask = ivec4(MskFix) << 4;\n"
"\n"
"#if PS_WMS == PS_WMT\n"
"\n"
"#if PS_WMS == 2\n"
" uv_out = clamp(uv, mask.xy, mask.zw);\n"
"#elif PS_WMS == 3\n"
" uv_out = (uv & mask.xy) | mask.zw;\n"
"#endif\n"
"\n"
"#else // PS_WMS != PS_WMT\n"
"\n"
"#if PS_WMS == 2\n"
" uv_out.x = clamp(uv, mask.x, mask.z);\n"
"#elif PS_WMS == 3\n"
" uv_out.x = (uv.x & mask.x) | mask.z;\n"
"#endif\n"
"\n"
"#if PS_WMT == 2\n"
" uv_out.y = clamp(uv, mask.y, mask.w);\n"
"#elif PS_WMT == 3\n"
" uv_out.y = (uv.y & mask.y) | mask.w;\n"
"#endif\n"
"\n"
"#endif\n"
"\n"
" return uv_out;\n"
"}\n"
"\n"
"vec4 sample_depth(vec2 st)\n"
"{\n"
" vec2 uv_f = vec2(clamp_wrap_uv_depth(ivec2(st))) * vec2(ScalingFactor.xy) * vec2(1.0f/16.0f);\n"
" ivec2 uv = ivec2(uv_f);\n"
"\n"
" vec4 t;\n"
"#if PS_DEPTH_FMT == 1\n"
" // Based on ps_main11 of convert\n"
"\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"
" vec4 res = fract(vec4(fetch_c(uv).r) * bitSh);\n"
"\n"
" t = (res - res.xxyz * bitMsk) * 256.0f;\n"
"\n"
"#elif PS_DEPTH_FMT == 2\n"
" // Based on ps_main12 of convert\n"
"\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(fetch_c(uv).r) * bitSh) & bitMsk;\n"
"\n"
" t = vec4(color) * vec4(8.0f, 8.0f, 8.0f, 128.0f);\n"
"\n"
"#elif PS_DEPTH_FMT == 3\n"
" // Convert a RGBA/RGB5A1 color texture into a RGBA/RGB5A1 color texture\n"
" t = fetch_c(uv) * 255.0f;\n"
"\n"
"#endif\n"
"\n"
" // warning t ranges from 0 to 255\n"
"#if (PS_AEM_FMT == FMT_24)\n"
" t.a = ( (PS_AEM == 0) || any(bvec3(t.rgb)) ) ? 255.0f * TA.x : 0.0f;\n"
"#elif (PS_AEM_FMT == FMT_16)\n"
" t.a = t.a >= 128.0f ? 255.0f * TA.y : ( (PS_AEM == 0) || any(bvec3(t.rgb)) ) ? 255.0f * TA.x : 0.0f;\n"
"#endif\n"
"\n"
"\n"
" return t;\n"
"}\n"
"//////////////////////////////////////////////////////////////////////\n"
"\n"
"vec4 sample_color(vec2 st)\n"
"{\n"
"#if (PS_TCOFFSETHACK == 1)\n"
@ -1172,10 +1260,17 @@ static const char* const tfx_fs_all_glsl =
"{\n"
" //FIXME: maybe we can set gl_Position.w = q in VS\n"
"#if (PS_FST == 0)\n"
" vec4 T = sample_color(PSin.t_float.xy / vec2(PSin.t_float.w));\n"
" vec2 st = PSin.t_float.xy / vec2(PSin.t_float.w);\n"
"#else\n"
" // Note xy are normalized coordinate\n"
" vec4 T = sample_color(PSin.t_int.xy);\n"
" vec2 st = PSin.t_int.xy;\n"
"#endif\n"
"\n"
"#if (PS_DEPTH_FMT > 0)\n"
" // Integral coordinate\n"
" vec4 T = sample_depth(PSin.t_int.zw);\n"
"#else\n"
" vec4 T = sample_color(st);\n"
"#endif\n"
"\n"
"#if PS_IIP == 1\n"