gsdx ogl: correct most of Z-depth issue

Best setting if you driver support GL_NV_depth_buffer_float => GL_NV_Depth = 1 & logz = 0
Otherwise => GL_NV_Depth = 0 & logz = 1

Explanation of the bug:
Dx z position ranges from 0.0f to 1.0f (FS ranges 0.0f to 1.0f)
GL z Position ranges from -1.0f to 1.0f (FS ranges 0.0f to 1.0f)

Why it sucks:
GS small depth value will be "mapped" to -1.0f. In others all small values will be 1.0f! Terrible lost
of accuraccy.

The GL_NV_depth_buffer_float extension allow to set the near plane as -1.0f.
So
"GL z Position ranges from -1.0f to 1.0f (FS ranges 0.0f to 1.0f)"
will become
"GL z Position ranges from -1.0f to 1.0f (FS ranges -1.0f to 1.0f)"
and therefore
"z posision [0.0f;1.0f] will map to FS [0.0f;1.0f]" as DX

Yes we just get back all precision lost previously :)
However you need hardware (intel?) and driver support (free driver?/gles?) :(





git-svn-id: http://pcsx2.googlecode.com/svn/trunk@5860 96395faa-99c1-11dd-bbfe-3dabce05a288

plugins/GSdx/GLLoader.cpp

@@ -118,6 +118,8 @@ PFNGLMAKETEXTUREHANDLERESIDENTARBPROC  gl_MakeTextureHandleResidentARB = NULL;
 PFNGLMAKETEXTUREHANDLENONRESIDENTARBPROC gl_MakeTextureHandleNonResidentARB = NULL;
 PFNGLUNIFORMHANDLEUI64VARBPROC         gl_UniformHandleui64vARB        = NULL;
 PFNGLPROGRAMUNIFORMHANDLEUI64VARBPROC  gl_ProgramUniformHandleui64vARB = NULL;
+
+PFNGLDEPTHRANGEDNVPROC                 gl_DepthRangedNV                = NULL;
 #endif
 
 namespace GLLoader {

plugins/GSdx/GLLoader.h

@@ -183,6 +183,8 @@ extern PFNGLMAKETEXTUREHANDLENONRESIDENTARBPROC gl_MakeTextureHandleNonResidentA
 extern PFNGLUNIFORMHANDLEUI64VARBPROC           gl_UniformHandleui64vARB;
 extern PFNGLPROGRAMUNIFORMHANDLEUI64VARBPROC    gl_ProgramUniformHandleui64vARB;
 
+extern PFNGLDEPTHRANGEDNVPROC                   gl_DepthRangedNV;
+
 #else
 #define gl_ActiveTexture glActiveTexture
 #define gl_BlendColor glBlendColor

plugins/GSdx/GSDeviceOGL.cpp

@@ -167,7 +167,7 @@ bool GSDeviceOGL::Create(GSWnd* wnd)
 	// ****************************************************************
 	// Various object
 	// ****************************************************************
-	m_shader = new GSShaderOGL(!!theApp.GetConfig("debug_ogl_shader", 1));
+	m_shader = new GSShaderOGL(!!theApp.GetConfig("debug_ogl_shader", 1), !!theApp.GetConfig("GL_NV_Depth", 0));
 
 	gl_GenFramebuffers(1, &m_fbo);
 	gl_GenFramebuffers(1, &m_fbo_read);

plugins/GSdx/GSShaderOGL.cpp

@@ -23,8 +23,9 @@
 #include "GSShaderOGL.h"
 #include "GLState.h"
 
-GSShaderOGL::GSShaderOGL(bool debug) :
+GSShaderOGL::GSShaderOGL(bool debug, bool nv_depth) :
 	m_debug_shader(debug),
+	m_nv_depth(nv_depth),
 	m_vs_sub_count(0),
 	m_ps_sub_count(0)
 {
@@ -397,6 +398,10 @@ std::string GSShaderOGL::GenGlslHeader(const std::string& entry, GLenum type, co
 		header += "#define ENABLE_BINDLESS_TEX\n";
 	}
 
+	if (m_nv_depth) {
+		// Specific nvidia extension that seem to help for z fighting
+		header += "#define NV_DEPTH\n";
+	}
 
 #else
 	header = "#version 300 es\n";

plugins/GSdx/GSShaderOGL.h

@@ -25,6 +25,7 @@ class GSShaderOGL {
 	GLuint m_pipeline;
 	hash_map<uint64, GLuint > m_single_prog;
 	const bool m_debug_shader;
+	const bool m_nv_depth;
 	GLuint m_vs_sub_count;
 	GLuint m_ps_sub_count;
 
@@ -46,7 +47,7 @@ class GSShaderOGL {
 	GLuint LinkNewProgram();
 
 	public:
-	GSShaderOGL(bool debug);
+	GSShaderOGL(bool debug, bool nv_depth);
 	~GSShaderOGL();
 
 	void GS(GLuint s);

plugins/GSdx/GSTextureFXOGL.cpp

@@ -59,6 +59,10 @@ void GSDeviceOGL::CreateTextureFX()
 
 	// Help to debug FS in apitrace
 	m_apitrace = CompilePS(PSSelector());
+
+	if (!!theApp.GetConfig("GL_NV_Depth", 0)) {
+		gl_DepthRangedNV(-1.0f, 1.0f);
+	}
 }
 
 GSDepthStencilOGL* GSDeviceOGL::CreateDepthStencil(OMDepthStencilSelector dssel)

plugins/GSdx/GSVertexArrayOGL.h

@@ -234,8 +234,12 @@ public:
 			switch (layout[i].type) {
 				case GL_UNSIGNED_SHORT:
 				case GL_UNSIGNED_INT:
-					// Rule: when shader use integral (not normalized) you must use gl_VertexAttribIPointer (note the extra I)
+					if (layout[i].normalize) {
-					gl_VertexAttribIPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].stride, layout[i].offset);
+						gl_VertexAttribPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].normalize,  layout[i].stride, layout[i].offset);
+					} else {
+						// Rule: when shader use integral (not normalized) you must use gl_VertexAttribIPointer (note the extra I)
+						gl_VertexAttribIPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].stride, layout[i].offset);
+					}
 					break;
 				default:
 					gl_VertexAttribPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].normalize,  layout[i].stride, layout[i].offset);

plugins/GSdx/GSWnd.cpp

@@ -120,6 +120,7 @@ void GSWndGL::PopulateGlFunction()
 	*(void**)&(gl_UniformHandleui64vARB) = GetProcAddress("glUniformHandleui64vARB", true);
 	*(void**)&(gl_ProgramUniformHandleui64vARB) = GetProcAddress("glProgramUniformHandleui64vARB", true);
 
+	*(void**)&(gl_DepthRangedNV) = GetProcAddress("glDepthRangedNV", true);
 
 #endif
 }

plugins/GSdx/res/glsl_source.h

@@ -665,6 +665,7 @@ static const char* tfx_glsl =
 	"};\n"
 	"\n"
 	"const float exp_min32 = exp2(-32.0f);\n"
+	"const float exp_min31 = exp2(-31.0f);\n"
 	"\n"
 	"#ifdef SUBROUTINE_GL40\n"
 	"// Function pointer type\n"
@@ -722,7 +723,7 @@ static const char* tfx_glsl =
 	"\n"
 	"void vs_main()\n"
 	"{\n"
-	"    uint z;\n"
+	"    highp uint z;\n"
 	"    if(VS_BPPZ == 1) // 24\n"
 	"        z = i_z & uint(0xffffff);\n"
 	"    else if(VS_BPPZ == 2) // 16\n"
@@ -734,16 +735,26 @@ static const char* tfx_glsl =
 	"    // example: ceil(afterseveralvertextransformations(y = 133)) => 134 => line 133 stays empty\n"
 	"    // input granularity is 1/16 pixel, anything smaller than that won't step drawing up/left by one pixel\n"
 	"    // example: 133.0625 (133 + 1/16) should start from line 134, ceil(133.0625 - 0.05) still above 133\n"
+	"    vec4 p;\n"
 	"\n"
-	"    vec3 p = vec3(i_p, z) - vec3(0.05f, 0.05f, 0.0f);\n"
+	"    p.xy = vec2(i_p) - vec2(0.05f, 0.05f);\n"
-	"    p = p * vec3(VertexScale, exp_min32) - vec3(VertexOffset, 0.0f);\n"
+	"    p.xy = p.xy * VertexScale - VertexOffset;\n"
-	"\n"
+	"    p.w = 1.0f;\n"
-	"    if(VS_LOGZ == 1)\n"
+	"#ifdef NV_DEPTH\n"
-	"    {\n"
+	"    if(VS_LOGZ == 1) {\n"
-	"        p.z = log2(1.0f + float(z)) / 32.0f;\n"
+	"        p.z = log2(float(1u+z)) / 32.0f;\n"
+	"    } else {\n"
+	"        p.z = float(z) * exp_min32;\n"
 	"    }\n"
+	"#else\n"
+	"    if(VS_LOGZ == 1) {\n"
+	"        p.z = log2(float(1u+z)) / 31.0f - 1.0f;\n"
+	"    } else {\n"
+	"        p.z = float(z) * exp_min31 - 1.0f;\n"
+	"    }\n"
+	"#endif\n"
 	"\n"
-	"    gl_Position = vec4(p, 1.0f); // NOTE I don't know if it is possible to merge POSITION_OUT and gl_Position\n"
+	"    gl_Position = p;\n"
 	"\n"
 	"    texture_coord();\n"
 	"\n"

plugins/GSdx/res/tfx.glsl

@@ -106,6 +106,7 @@ layout(std140, binding = 20) uniform cb20
 };
 
 const float exp_min32 = exp2(-32.0f);
+const float exp_min31 = exp2(-31.0f);
 
 #ifdef SUBROUTINE_GL40
 // Function pointer type
@@ -163,7 +164,7 @@ void texture_coord()
 
 void vs_main()
 {
-    uint z;
+    highp uint z;
     if(VS_BPPZ == 1) // 24
         z = i_z & uint(0xffffff);
     else if(VS_BPPZ == 2) // 16
@@ -175,16 +176,26 @@ void vs_main()
     // example: ceil(afterseveralvertextransformations(y = 133)) => 134 => line 133 stays empty
     // input granularity is 1/16 pixel, anything smaller than that won't step drawing up/left by one pixel
     // example: 133.0625 (133 + 1/16) should start from line 134, ceil(133.0625 - 0.05) still above 133
+    vec4 p;
 
-    vec3 p = vec3(i_p, z) - vec3(0.05f, 0.05f, 0.0f);
+    p.xy = vec2(i_p) - vec2(0.05f, 0.05f);
-    p = p * vec3(VertexScale, exp_min32) - vec3(VertexOffset, 0.0f);
+    p.xy = p.xy * VertexScale - VertexOffset;
-
+    p.w = 1.0f;
-    if(VS_LOGZ == 1)
+#ifdef NV_DEPTH
-    {
+    if(VS_LOGZ == 1) {
-        p.z = log2(1.0f + float(z)) / 32.0f;
+        p.z = log2(float(1u+z)) / 32.0f;
+    } else {
+        p.z = float(z) * exp_min32;
     }
+#else
+    if(VS_LOGZ == 1) {
+        p.z = log2(float(1u+z)) / 31.0f - 1.0f;
+    } else {
+        p.z = float(z) * exp_min31 - 1.0f;
+    }
+#endif
 
-    gl_Position = vec4(p, 1.0f); // NOTE I don't know if it is possible to merge POSITION_OUT and gl_Position
+    gl_Position = p;
 
     texture_coord();