zzogl: GLSL is working again for AMD gpu. Nvidia test is welcome

* properly release shader in release mode
* set stream format every time an array buffer is bound


git-svn-id: http://pcsx2.googlecode.com/svn/trunk@5219 96395faa-99c1-11dd-bbfe-3dabce05a288

plugins/zzogl-pg/opengl/GSVertexArrayOGL.h

@@ -165,10 +165,14 @@ class GSVertexBufferStateOGL {
 	GLuint m_va;
 	GLenum m_topology;
 
+	// DEBUG
+	vector<GSInputLayoutOGL> layout_store;
+
 public:
 	GSVertexBufferStateOGL(size_t stride, GSInputLayoutOGL* layout, uint32 layout_nbr)
 	{
 		glGenVertexArrays(1, &m_va);
+		layout_store.clear();
 
 		m_vb = new GSBufferOGL(GL_ARRAY_BUFFER, stride);
 		m_ib = new GSBufferOGL(GL_ELEMENT_ARRAY_BUFFER, sizeof(uint32));
@@ -188,23 +192,39 @@ public:
 		m_vb->bind();
 	}
 
+	void set_internal_format()
+	{
+		for (vector<GSInputLayoutOGL>::iterator it = layout_store.begin(); it != layout_store.end(); it++)
+			set_internal_format(*it);
+	}
+
+	void set_internal_format(GSInputLayoutOGL& layout) {
+		// Note this function need both a vertex array object and a GL_ARRAY_BUFFER buffer
+		glEnableVertexAttribArray(layout.index);
+		switch (layout.type) {
+			case GL_UNSIGNED_SHORT:
+			case GL_UNSIGNED_INT:
+			case GL_SHORT:
+			case GL_INT:
+				// Rule: when shader use integral (not normalized) you must use glVertexAttribIPointer (note the extra I)
+				if (layout.normalize == false)
+					glVertexAttribIPointer(layout.index, layout.size, layout.type, layout.stride, layout.offset);
+				else
+					glVertexAttribPointer(layout.index, layout.size, layout.type, layout.normalize,  layout.stride, layout.offset);
+				break;
+			default:
+				glVertexAttribPointer(layout.index, layout.size, layout.type, layout.normalize,  layout.stride, layout.offset);
+				break;
+		}
+	}
+
 	void set_internal_format(GSInputLayoutOGL* layout, uint32 layout_nbr)
 	{
 		for (uint i = 0; i < layout_nbr; i++) {
-			// Note this function need both a vertex array object and a GL_ARRAY_BUFFER buffer
+			// DEBUG
-			glEnableVertexAttribArray(layout[i].index);
+			layout_store.push_back(layout[i]);
-			switch (layout[i].type) {
+
-				case GL_UNSIGNED_SHORT:
+			set_internal_format(layout[i]);
-				case GL_UNSIGNED_INT:
-				case GL_SHORT:
-				case GL_INT:
-					// Rule: when shader use integral (not normalized) you must use glVertexAttribIPointer (note the extra I)
-					glVertexAttribIPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].stride, layout[i].offset);
-					break;
-				default:
-					glVertexAttribPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].normalize,  layout[i].stride, layout[i].offset);
-					break;
-			}
 		}
 	}
 

plugins/zzogl-pg/opengl/ZZGl.h

@@ -84,6 +84,9 @@ static __forceinline void GL_STENCILFUNC_SET()
 	glStencilFunc(s_stencilfunc, s_stencilref, s_stencilmask); 
 }
 
+#ifdef GLSL4_API
+#include "ZZoglShaders.h"
+#endif
 // sets the data stream
 static __forceinline void SET_STREAM()
 {
@@ -92,6 +95,8 @@ static __forceinline void SET_STREAM()
 	glSecondaryColorPointerEXT(4, GL_UNSIGNED_BYTE, sizeof(VertexGPU), (void*)12);
 	glTexCoordPointer(3, GL_FLOAT, sizeof(VertexGPU), (void*)16);
 	glVertexPointer(4, GL_SHORT, sizeof(VertexGPU), (void*)0);
+#else
+	vertex_array->set_internal_format();
 #endif
 }
 

plugins/zzogl-pg/opengl/ZZoglCreate.cpp

@@ -40,7 +40,6 @@ typedef void (APIENTRYP _PFNSWAPINTERVAL)(int);
 
 map<string, GLbyte> mapGLExtensions;
 
-extern bool LoadEffects();
 extern bool ZZshLoadExtraEffects();
 extern FRAGMENTSHADER* ZZshLoadShadeEffect(int type, int texfilter, int fog, int testaem, int exactcolor, const clampInfo& clamp, int context, bool* pbFailed);
 
@@ -575,10 +574,9 @@ bool ZZCreate(int _width, int _height)
 	GL_REPORT_ERROR();
 
 #ifdef GLSL4_API
-	// FIXME maybe GL_UNSIGNED_SHORT could be better than GL_SHORT
 	GSInputLayoutOGL vert_format[] =
 	{
-		{0 , 4 , GL_SHORT          , GL_FALSE , sizeof(VertexGPU) , (const GLvoid*)(0) }  , // vertex
+		{0 , 2 , GL_SHORT , GL_FALSE , sizeof(VertexGPU) , (const GLvoid*)(0) }  , // vertex
 		{1 , 4 , GL_UNSIGNED_BYTE  , GL_TRUE  , sizeof(VertexGPU) , (const GLvoid*)(8) }  , // color
 		{2 , 4 , GL_UNSIGNED_BYTE  , GL_TRUE , sizeof(VertexGPU) , (const GLvoid*)(12) } , // z value. FIXME WTF 4 unsigned byte, why not a full integer
 		{3 , 3 , GL_FLOAT          , GL_FALSE , sizeof(VertexGPU) , (const GLvoid*)(16) } , // tex coord
@@ -596,7 +594,7 @@ bool ZZCreate(int _width, int _height)
             glBindBuffer(GL_ARRAY_BUFFER, g_vboBuffers[i]);
             glBufferData(GL_ARRAY_BUFFER, 0x100*sizeof(VertexGPU), NULL, GL_STREAM_DRAW);
 #ifdef GLSL4_API
-			vertex_array->set_internal_format(vert_format, 4);
+			vertex_array->set_internal_format();
 #endif
         }
         vb_buffer_allocated = true; // mark the buffer allocated
@@ -680,7 +678,7 @@ bool ZZCreate(int _width, int _height)
 	glGenBuffers(1, &vboRect);
 	glBindBuffer(GL_ARRAY_BUFFER, vboRect);
 #ifdef GLSL4_API
-	vertex_array->set_internal_format(vert_format, 4);
+	vertex_array->set_internal_format();
 #endif
 
 	vector<VertexGPU> verts(4);

plugins/zzogl-pg/opengl/ZZoglShaders.h

@@ -137,8 +137,7 @@ struct GlobalUniform {
 		float linear[2*4];
 	};
 	void SettleFloat(uint indice, const float* v) {
-		assert(indice >= 0);
+		assert(indice + 3 < 2*4);
-		assert(indice + 3 < sizeof(linear));
 		linear[indice+0] = v[0];
 		linear[indice+1] = v[1];
 		linear[indice+2] = v[2];
@@ -163,8 +162,7 @@ struct ConstantUniform {
 		float linear[8*4];
 	};
 	void SettleFloat(uint indice, const float* v) {
-		assert(indice >= 0);
+		assert(indice + 3 < 8*4);
-		assert(indice + 3 < sizeof(linear));
 		linear[indice+0] = v[0];
 		linear[indice+1] = v[1];
 		linear[indice+2] = v[2];
@@ -191,8 +189,7 @@ struct FragmentUniform {
 		float linear[13*4];
 	};
 	void SettleFloat(uint indice, const float* v) {
-		assert(indice >= 0);
+		assert(indice + 3 < 13*4);
-		assert(indice + 3 < sizeof(linear));
 		linear[indice+0] = v[0];
 		linear[indice+1] = v[1];
 		linear[indice+2] = v[2];
@@ -209,8 +206,7 @@ struct VertexUniform {
 		float linear[3*4];
 	};
 	void SettleFloat(uint indice, const float* v) {
-		assert(indice >= 0);
+		assert(indice + 3 <  3*4);
-		assert(indice + 3 < sizeof(linear) / 4);
 		linear[indice+0] = v[0];
 		linear[indice+1] = v[1];
 		linear[indice+2] = v[2];
@@ -340,8 +336,7 @@ struct SamplerParam {
 
 	SamplerParam() : unit(-1), texid(0), target(0) {}
 
-	void set_unit(int new_unit = 0) {
+	void set_unit(int new_unit) {
-		assert(new_unit >= 0);
 		assert(new_unit < 10);
 		unit = new_unit;
 		target = g_texture_target[new_unit];
@@ -359,6 +354,10 @@ struct SamplerParam {
 	void set_texture(GLuint new_texid) {
 		texid = new_texid;
 	}
+
+	void release_texture() {
+		texid = 0;
+	}
 };
 
 struct FRAGMENTSHADER
@@ -393,7 +392,7 @@ struct FRAGMENTSHADER
 		//sBitwiseANDY         = 4;
 		//sInterlace           = 5;
 		//sCLUT                = 6;
-		samplers[sMemory+0].set_unit(0);
+		samplers[sMemory].set_unit(0);
 		samplers[sMemory+1].set_unit(0); // Dual context. Use same unit
 		samplers[sFinal].set_unit(1);
 		samplers[sBitwiseANDX].set_unit(6);
@@ -437,17 +436,26 @@ struct FRAGMENTSHADER
 	}
 
 	void enable_texture() {
-		samplers[0+context].enable_texture(); // sMemory is dual context
+		samplers[sMemory+context].enable_texture(); // sMemory is dual context
 		for (int i = 2; i < 7; i++)
 			samplers[i].enable_texture();
 	}
 
 	void set_texture(ZZshParameter param, GLuint texid) {
-		if (param == 0) // sMemory is dual context
+		if (param == sMemory) // sMemory is dual context
-			samplers[0+context].set_texture(texid);
+			samplers[sMemory+context].set_texture(texid);
 		else
 			samplers[param].set_texture(texid);
 	}
+
+	void release_prog() {
+		if(program) {
+			glDeleteProgram(program);
+			program = 0;
+		}
+		for (uint i = 0; i < 7 ; i++)
+			samplers[i].release_texture();
+	}
 };
 #endif
 

plugins/zzogl-pg/opengl/ZZoglShadersGLSL4.cpp

@@ -124,8 +124,8 @@ GSVertexBufferStateOGL *vertex_array;
 COMMONSHADER g_cs;
 static GLuint s_pipeline = 0;
 
-FRAGMENTSHADER ppsDebug;
+//FRAGMENTSHADER ppsDebug;
-FRAGMENTSHADER ppsDebug2;
+//FRAGMENTSHADER ppsDebug2;
 
 //------------------ Code
 
@@ -392,7 +392,7 @@ inline bool LoadShaderFromFile(ZZshProgram& program, const char* DefineString, c
 		return false;
 	}
 
-	ZZLog::Error_Log("Used shader for %s... Ok",name);
+	ZZLog::Error_Log("Compile shader for %s... Ok",name);
 	return true;
 }
 
@@ -422,25 +422,22 @@ void ZZshSetupShader() {
 	PutParametersAndRun(vs, ps);
 }
 
-static void ZZshSetShader(VERTEXSHADER* vs, FRAGMENTSHADER* ps) {
-	if (vs == NULL || ps == NULL) return;
-
-	FRAGMENTSHADER* debug = ps;
-
-	glUseProgramStages(s_pipeline, GL_VERTEX_SHADER_BIT, vs->program);
-	glUseProgramStages(s_pipeline, GL_FRAGMENT_SHADER_BIT, ps->program);
-	//PutParametersAndRun(vs, ps);
-	GL_REPORT_ERRORD();
-}
-
 void ZZshSetVertexShader(ZZshShaderLink prog) {
 	g_vsprog = prog;
-	ZZshSetShader((VERTEXSHADER*)(g_vsprog.link), (FRAGMENTSHADER*)(g_psprog.link)) ;
+
+	VERTEXSHADER* vs = (VERTEXSHADER*)g_vsprog.link;
+	if (!vs) return;
+
+	glUseProgramStages(s_pipeline, GL_VERTEX_SHADER_BIT, vs->program);
 }
 
 void ZZshSetPixelShader(ZZshShaderLink prog) {
 	g_psprog = prog;
-	ZZshSetShader((VERTEXSHADER*)(g_vsprog.link), (FRAGMENTSHADER*)(g_psprog.link)) ;
+
+	FRAGMENTSHADER* ps = (FRAGMENTSHADER*)g_psprog.link;
+	if (!ps) return;
+
+	glUseProgramStages(s_pipeline, GL_FRAGMENT_SHADER_BIT, ps->program);
 }
 
 //------------------------------------------------------------------------------------------------------------------
@@ -482,14 +479,13 @@ void PutParametersInProgram(VERTEXSHADER* vs, FRAGMENTSHADER* ps) {
 	fragment_buffer->bind();
 	fragment_buffer->upload((void*)&ps->uniform_buffer[ps->context]);
 
-	g_cs.enable_texture();
-
 #ifdef ENABLE_MARKER
 	char* debug = new char[100];
 	sprintf(debug, "FS(%d): enable texture", ps->program);
 	if (GLEW_GREMEDY_string_marker) glStringMarkerGREMEDY(0, debug);
 #endif
 
+	g_cs.enable_texture();
 	ps->enable_texture();
 }
 
@@ -547,13 +543,8 @@ static __forceinline bool LOAD_PS(char* DefineString, const char* name, FRAGMENT
 inline bool LoadEffects()
 {
 	// clear the textures
-	for(u32 i = 0; i < ArraySize(ppsTexture); ++i) {
+	for(u32 i = 0; i < ArraySize(ppsTexture); ++i)
-		SAFE_RELEASE_PROG(ppsTexture[i].prog);
+		ppsTexture[i].release_prog();
-	}
-
-#ifndef _DEBUG
-	memset(ppsTexture, 0, sizeof(ppsTexture));
-#endif
 
 	return true;
 }
@@ -623,8 +614,8 @@ bool ZZshLoadExtraEffects() {
 	if (!LOAD_PS(DefineString, "Convert32to16PS", ppsConvert32to16, cgfProf, 0, "")) bLoadSuccess = false;
 
 	// DEBUG
-	if (!LOAD_PS(DefineString, "ZeroDebugPS", ppsDebug, cgfProf, 0, "")) bLoadSuccess = false;
+	// if (!LOAD_PS(DefineString, "ZeroDebugPS", ppsDebug, cgfProf, 0, "")) bLoadSuccess = false;
-	if (!LOAD_PS(DefineString, "ZeroDebug2PS", ppsDebug2, cgfProf, 0, "")) bLoadSuccess = false;
+	// if (!LOAD_PS(DefineString, "ZeroDebug2PS", ppsDebug2, cgfProf, 0, "")) bLoadSuccess = false;
 
 	GL_REPORT_ERRORD();
 	return true;
@@ -656,7 +647,7 @@ static ZZshProgram LoadShaderFromType(int type, int texfilter, int texwrap, int
 	if (!LoadShaderFromFile(program, DefineString, name, GL_FRAGMENT_SHADER))
 		return LoadShaderFromFile(program, DefineString, name, GL_FRAGMENT_SHADER, true);
 
-	ZZLog::Debug_Log("Used shader for type:%d filter:%d wrap:%d for:%d depth:%d aem:%d color:%d decompression:%d ctx:%d... Ok \n", type, texfilter, texwrap, fog, writedepth, testaem, exactcolor, ps, context);
+	ZZLog::Debug_Log("Create new shader for type:%d filter:%d wrap:%d for:%d depth:%d aem:%d color:%d decompression:%d ctx:%d... Ok \n", type, texfilter, texwrap, fog, writedepth, testaem, exactcolor, ps, context);
 
 	GL_REPORT_ERRORD();
 	return program;
@@ -699,6 +690,7 @@ FRAGMENTSHADER* ZZshLoadShadeEffect(int type, int texfilter, int fog, int testae
 		GL_REPORT_ERRORD();
 
 		if( glGetError() != GL_NO_ERROR ) {
+				ZZLog::Check_GL_Error();
 				ZZLog::Error_Log("Failed to load shader %d,%d,%d,%d.", type, fog, texfilter, 4*clamp.wms+clamp.wmt);
 				if (pbFailed != NULL ) *pbFailed = true;
 				return pf;

plugins/zzogl-pg/opengl/ps2hw_gl4.glsl

@@ -110,7 +110,7 @@ out gl_PerVertex {
 };
 
 layout(location = 0) in ivec2 Vert;
-layout(location = 1) in vec4  Color;
+layout(location = 1) in vec4 Color;
 layout(location = 2) in vec4 SecondaryColor;
 layout(location = 3) in vec3 TexCoord;
 
@@ -126,7 +126,6 @@ layout(location = 0) out vertex VSout;
 
 layout(location = 0) in vertex PSin;
 
-// NOTE: Basic s/gl_FragData[X]/FragDataX/ I think
 // FIXME: host only do glDrawBuffers of 1 buffers not 2. I think this is a major bug
 layout(location = 0) out vec4 FragData0;
 layout(location = 1) out vec4 FragData1;
@@ -603,6 +602,7 @@ half4 ps2FinalColor(half4 col)
 half4 ps2FinalColor(half4 col)
 {
 	return col * g_fOneColor.xxxy + g_fOneColor.zzzw;
+	//return vec4(1.0f,0.0f,1.0f, 0.8f);
 }
 #endif
 
@@ -765,7 +765,7 @@ void CRTCTargInterPS() {
 void CRTCTargPS() {
 	float4 c = BilinearFloat16(PSin.tex.xy);
 	// FIXME DEBUG: to validate tex coord on blit
-	//vec4 c = vec4(PSin.tex.x/512.0f, PSin.tex.y/512.0f, 0.0, 1.0);
+	//c = vec4(PSin.tex.x/512.0f, PSin.tex.y/512.0f, 0.0, 1.0);
 	c.w = g_fc0.w * c.w * g_fOneColor.x + g_fOneColor.y;
 	FragData0 = c;
 }
@@ -821,11 +821,12 @@ void ZeroPS() {
 }
 
 void ZeroDebugPS() {
-	FragData0 = vec4(0.0, 1.0, 0.0, 1.0);
+	FragData0 = vec4(0.0, 1.0, 0.0, 0.5);
+	//FragData0 = vec4(PSin.position.x, PSin.position.y, 1.0, 0.5);
 }
 
 void ZeroDebug2PS() {
-	FragData0 = vec4(1.0, 0.0, 0.0, 1.0);
+	FragData0 = vec4(1.0, 0.0, 0.0, 0.5);
 }
 
 void BaseTexturePS() {
@@ -876,30 +877,37 @@ void Convert32to16PS() {
 #ifdef VERTEX_SHADER
 
 float4 OutPosition() {
-	float4 Position;
+	float4 position;
-	Position.xy = Vert.xy * g_fPosXY.xy + g_fPosXY.zw;
+	position.xy = vec2(Vert.xy) * g_fPosXY.xy + g_fPosXY.zw;
-	Position.z = (log(g_fc0.y + dot(g_fZ, SecondaryColor.zyxw)) * g_fZNorm.x + g_fZNorm.y) * g_fZMin.y + dot(g_fZ, SecondaryColor.zyxw) * g_fZMin.x ;
+	position.z = (log(g_fc0.y + dot(g_fZ, SecondaryColor.zyxw)) * g_fZNorm.x + g_fZNorm.y) * g_fZMin.y + dot(g_fZ, SecondaryColor.zyxw) * g_fZMin.x ;
-	Position.w = g_fc0.y;
+	position.w = g_fc0.y;
-	return Position;
+	return position;
 }
 
 // just smooth shadering
 void RegularVS() {
-	gl_Position = OutPosition();
+	float4 position = OutPosition();
+	gl_Position = position;
+
 	VSout.color = Color;
+
 	DOZWRITE(VSout.z = SecondaryColor * g_fZBias.x + g_fZBias.y;)
     DOZWRITE(VSout.z.w = g_fc0.y;)
 }
 
 // diffuse texture mapping
 void TextureVS() {
-	gl_Position = OutPosition();
+	float4 position = OutPosition();
+	gl_Position = position;
+
 	VSout.color = Color;
+
 #ifdef PERSPECTIVE_CORRECT_TEX
 	VSout.tex.xyz = TexCoord.xyz;
 #else
 	VSout.tex.xy = TexCoord.xy/TexCoord.z;
 #endif
+
  	DOZWRITE(VSout.z = SecondaryColor * g_fZBias.x + g_fZBias.y;)
     DOZWRITE(VSout.z.w = g_fc0.y;)
 }
@@ -907,8 +915,11 @@ void TextureVS() {
 void RegularFogVS() {
 	float4 position = OutPosition();
 	gl_Position = position;
+
 	VSout.color = Color;
+
     VSout.fog = position.z * g_fBilinear.w;
+
 	DOZWRITE(VSout.z = SecondaryColor * g_fZBias.x + g_fZBias.y;)
     DOZWRITE(VSout.z.w = g_fc0.y;)
 }
@@ -916,20 +927,24 @@ void RegularFogVS() {
 void TextureFogVS() {
 	float4 position = OutPosition();
 	gl_Position = position;
+
 	VSout.color = Color;
+
 #ifdef PERSPECTIVE_CORRECT_TEX
 	VSout.tex.xyz = TexCoord.xyz;
 #else
 	VSout.tex.xy = TexCoord.xy/TexCoord.z;
 #endif
+
     VSout.fog = position.z * g_fBilinear.w;
+
 	DOZWRITE(VSout.z = SecondaryColor * g_fZBias.x + g_fZBias.y;)
     DOZWRITE(VSout.z.w = g_fc0.y;)
 }
 
 void BitBltVS() {
 	vec4 position;
-	position.xy = Vert.xy * g_fBitBltPos.xy + g_fBitBltPos.zw;
+	position.xy = vec2(Vert.xy) * g_fBitBltPos.xy + g_fBitBltPos.zw;
 	position.zw = g_fc0.xy;
 	gl_Position = position;