gsdx-ogl: LINUX-ONLY

* fix vertex shader for HW renderer :) Remains the fragment part...
* add some dumping infrastucture (DUMP_START and DUMP_LENGTH)


git-svn-id: http://pcsx2.googlecode.com/svn/branches/gsdx-ogl@5030 96395faa-99c1-11dd-bbfe-3dabce05a288

plugins/GSdx/GSDeviceOGL.cpp

@@ -53,6 +53,15 @@
 // glUniformBlockBinding(program, block_index, block_binding_point);
 
 //#define LOUD_DEBUGGING
+//#define DUMP_START (380)
+#define DUMP_LENGTH (20)
+
+#ifdef DUMP_START
+static uint32 g_draw_count = 0;
+static uint32 g_frame_count = 0;		
+#endif
+
+
 GSDeviceOGL::GSDeviceOGL()
 	: m_free_window(false)
 	  , m_window(NULL)
@@ -492,11 +501,43 @@ void GSDeviceOGL::Flip()
 	// FIXME: disable it when code is working
 	CheckDebugLog();
 	m_wnd->Flip();
+#ifdef DUMP_START
+	g_frame_count++;
+#endif
 }
 
 void GSDeviceOGL::DrawPrimitive()
 {
 	glDrawArrays(m_state.topology, m_state.vb_state->start, m_state.vb_state->count);
+#ifdef DUMP_START
+	if (g_draw_count > DUMP_START && g_draw_count < (DUMP_START+DUMP_LENGTH)) {
+		for (auto i = 0 ; i < 3 ; i++) {
+			if (m_state.ps_srv[i] != NULL) {
+				m_state.ps_srv[i]->Save(format("/tmp/in_%d__%d.bmp", g_draw_count, i),false);
+			}
+		}
+		if (m_state.rtv != NULL) m_state.rtv->Save(format("/tmp/out_%d.bmp", g_draw_count),false);
+		if (m_state.dsv != NULL) m_state.dsv->Save(format("/tmp/out_%d_ds.bmp", g_draw_count),false);
+
+		string topo;
+		switch (m_state.topology) {
+			case GL_TRIANGLE_STRIP: topo = "triangle_strip"; break;
+			case GL_TRIANGLES: topo = "triangle"; break;
+			case GL_LINES: topo = "line"; break;
+			case GL_POINTS: topo = "point"; break;
+			default: topo = "!!!!";
+		}
+		fprintf(stderr, "Draw %d (Frame %d), %d elem of %s\n", g_draw_count, g_frame_count, m_state.vb_state->count, topo.c_str() );
+		fprintf(stderr, "vs: %d ; gs: %d ; ps: %d\n", m_state.vs, m_state.gs, m_state.ps);
+		fprintf(stderr, "Blend: %d, Depth: %d, Stencil: %d \n",m_state.bs->m_enable, m_state.dss->m_depth_enable, m_state.dss->m_stencil_enable);
+
+		//fprintf(stderr, "type: %d, format: 0x%x\n", m_state.rtv->GetType(), m_state.rtv->GetFormat());
+		fprintf(stderr, "\n");
+
+	}
+
+	g_draw_count++;
+#endif
 }
 
 void GSDeviceOGL::ClearRenderTarget(GSTexture* t, const GSVector4& c)
@@ -941,6 +982,7 @@ void GSDeviceOGL::GSSetShader(GLuint gs)
 	if(m_state.gs != gs)
 	{
 		m_state.gs = gs;
+		// FIXME AMD driver bug !!!!!!!!
 		glUseProgramStages(m_pipeline, GL_GEOMETRY_SHADER_BIT, gs);
 	}
 }
@@ -954,9 +996,7 @@ void GSDeviceOGL::PSSetShaderResources(GSTexture* sr0, GSTexture* sr1)
 
 void GSDeviceOGL::PSSetShaderResource(int i, GSTexture* sr)
 {
-	GSTextureOGL* srv = NULL;
+	GSTextureOGL* srv = static_cast<GSTextureOGL*>(sr);
-
-	if(sr) srv = (GSTextureOGL*)sr;
 
 	if(m_state.ps_srv[i] != srv)
 	{
@@ -1082,24 +1122,11 @@ void GSDeviceOGL::OMSetBlendState(GSBlendStateOGL* bs, float bf)
 
 void GSDeviceOGL::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVector4i* scissor)
 {
-	// attach render&depth  to the FBO
 	// Hum, need to separate 2 case, Render target fbo and render target backbuffer
 	// Or maybe blit final result to the backbuffer
-#if 0
+	m_state.rtv = static_cast<GSTextureOGL*>(rt);
-	ID3D11RenderTargetView* rtv = NULL;
+	m_state.dsv = static_cast<GSTextureOGL*>(ds);
-	ID3D11DepthStencilView* dsv = NULL;
 
-	if(rt) rtv = *(GSTexture11*)rt;
-	if(ds) dsv = *(GSTexture11*)ds;
-
-	if(m_state.rtv != rtv || m_state.dsv != dsv)
-	{
-		m_state.rtv = rtv;
-		m_state.dsv = dsv;
-
-		m_ctx->OMSetRenderTargets(1, &rtv, dsv);
-	}
-#endif
 	if (static_cast<GSTextureOGL*>(rt)->IsBackbuffer()) {
 		OMSetFBO(0);
 
@@ -1108,7 +1135,16 @@ void GSDeviceOGL::OMSetRenderTargets(GSTexture* rt, GSTexture* ds, const GSVecto
 		OMSetFBO(m_fbo);
 
 		assert(rt != NULL); // a render target must exists
-		static_cast<GSTextureOGL*>(rt)->Attach(GL_COLOR_ATTACHMENT0);
+
+		// FIXME DEBUG special case for GL_R16UI
+		if (rt->GetFormat() == GL_R16UI) {
+			glDrawBuffer(GL_COLOR_ATTACHMENT1);
+			static_cast<GSTextureOGL*>(rt)->Attach(GL_COLOR_ATTACHMENT1);
+		} else {
+			glDrawBuffer(GL_COLOR_ATTACHMENT0);
+			static_cast<GSTextureOGL*>(rt)->Attach(GL_COLOR_ATTACHMENT0);
+		}
+
 		if (ds != NULL)
 			static_cast<GSTextureOGL*>(ds)->Attach(GL_DEPTH_STENCIL_ATTACHMENT);
 	}
@@ -1215,7 +1251,8 @@ void GSDeviceOGL::CompileShaderFromSource(const std::string& glsl_file, const st
 	char* log = (char*)malloc(log_length);
 	glGetProgramInfoLog(*program, log_length, NULL, log);
 
-	fprintf(stderr, "%s (%s) :", glsl_file.c_str(), entry.c_str());
+	fprintf(stderr, "%s (entry %s, prog %d) :", glsl_file.c_str(), entry.c_str(), *program);
+	fprintf(stderr, "\n%s", macro_sel.c_str());
 	fprintf(stderr, "%s\n", log);
 	free(log);
 }

plugins/GSdx/GSDeviceOGL.h

@@ -87,10 +87,12 @@ class GSUniformBufferOGL {
 	GLuint buffer;		// data object
 	GLuint index;		// GLSL slot
 	uint   size;	    // size of the data
+	const GLenum target;
 
 public:
 	GSUniformBufferOGL(GLuint index, uint size) : index(index)
-														, size(size)
+												  , size(size)
+												  ,target(GL_UNIFORM_BUFFER)
 	{
 		glGenBuffers(1, &buffer);
 		bind();
@@ -100,25 +102,25 @@ public:
 
 	void bind()
 	{
-		glBindBuffer(GL_UNIFORM_BUFFER, buffer);
+		glBindBuffer(target, buffer);
 	}
 
 	void allocate()
 	{
-		glBufferData(GL_UNIFORM_BUFFER, size, NULL, GL_STREAM_DRAW);
+		glBufferData(target, size, NULL, GL_STREAM_DRAW);
 	}
 
 	void attach()
 	{
-		glBindBufferBase(GL_UNIFORM_BUFFER, index, buffer);
+		glBindBufferBase(target, index, buffer);
 	}
 
 	void upload(const void* src)
 	{
 		uint32 flags = GL_MAP_WRITE_BIT | GL_MAP_INVALIDATE_BUFFER_BIT;
-		uint8* dst = (uint8*) glMapBufferRange(GL_UNIFORM_BUFFER, 0, size, flags);
+		uint8* dst = (uint8*) glMapBufferRange(target, 0, size, flags);
 		memcpy(dst, src, size);
-		glUnmapBuffer(GL_UNIFORM_BUFFER);
+		glUnmapBuffer(target);
 	}
 
 	~GSUniformBufferOGL() {
@@ -178,7 +180,10 @@ struct GSVertexBufferState {
 		for (int i = 0; i < layout_nbr; i++) {
 			// Note this function need both a vertex array object and a GL_ARRAY_BUFFER buffer
 			glEnableVertexAttribArray(layout[i].index);
-			glVertexAttribPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].normalize,  layout[i].stride, layout[i].offset);
+			if (layout[i].type == GL_UNSIGNED_INT || layout[i].type == GL_UNSIGNED_SHORT)
+				glVertexAttribIPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].stride, layout[i].offset);
+			else
+				glVertexAttribPointer(layout[i].index, layout[i].size, layout[i].type, layout[i].normalize,  layout[i].stride, layout[i].offset);
 		}
 	}
 

plugins/GSdx/GSRendererOGL.h

@@ -215,6 +215,7 @@ class GSRendererOGL : public GSRendererHW<GSVertexHW11>
 				}
 			}
 
+			// FIXME Opengl support half pixel center (as dx10). Code could be easier!!!
 			GSDeviceOGL::VSConstantBuffer vs_cb;
 
 			float sx = 2.0f * rtscale.x / (rtsize.x << 4);
@@ -242,6 +243,7 @@ class GSRendererOGL : public GSRendererHW<GSVertexHW11>
 
 			vs_cb.VertexScale  = GSVector4(sx, -sy, ldexpf(1, -32), 0.0f);
 			vs_cb.VertexOffset = GSVector4(ox * sx + ox2 + 1, -(oy * sy + oy2 + 1), 0.0f, -1.0f);
+			// END of FIXME
 
 			// gs
 

plugins/GSdx/GSTextureFXOGL.cpp

@@ -106,10 +106,9 @@ void GSDeviceOGL::SetupVS(VSSelector sel, const VSConstantBuffer* cb)
 	// *************************************************************
 	// Dynamic
 	// *************************************************************
-	if(m_vs_cb_cache.Update(cb))
+	if(m_vs_cb_cache.Update(cb)) {
-	{
 		SetUniformBuffer(m_vs_cb);
-		m_vs_cb->upload(&cb);
+		m_vs_cb->upload(cb);
 	}
 
 	VSSetShader(i->second);
@@ -178,10 +177,9 @@ void GSDeviceOGL::SetupPS(PSSelector sel, const PSConstantBuffer* cb, PSSamplerS
 	// *************************************************************
 	// Dynamic
 	// *************************************************************
-	if(m_ps_cb_cache.Update(cb))
+	if(m_ps_cb_cache.Update(cb)) {
-	{
 		SetUniformBuffer(m_ps_cb);
-		m_ps_cb->upload(&cb);
+		m_ps_cb->upload(cb);
 	}
 
 	GLuint ss0, ss1;

plugins/GSdx/GSTextureOGL.cpp

@@ -146,6 +146,7 @@ GSTextureOGL::GSTextureOGL(int type, int w, int h, bool msaa, int format)
 
 			glBindBuffer(GL_PIXEL_PACK_BUFFER, m_extra_buffer_id);
 			glBufferData(GL_PIXEL_PACK_BUFFER, m_pbo_size, NULL, GL_STREAM_DRAW);
+			glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
 			break;
 		default: break;
 	}
@@ -295,6 +296,8 @@ bool GSTextureOGL::Map(GSMap& m, const GSVector4i* r)
 		m.bits = (uint8*) glMapBufferRange(GL_PIXEL_PACK_BUFFER, 0, m_pbo_size, map_flags);
 		m.pitch = m_size.x;
 
+		glBindBuffer(GL_PIXEL_PACK_BUFFER, 0);
+
 		return true;
 	}
 
@@ -367,20 +370,21 @@ struct BITMAPINFOHEADER
 
 bool GSTextureOGL::Save(const string& fn, bool dds)
 {
-	switch (m_type) {
+	// Code not yet working
-		case GSTexture::DepthStencil:
+	if (IsDss()) return false;
-		case GSTexture::Offscreen:
-			ASSERT(0);
-			break;
-		default: break;
-	}
 
 	// Collect the texture data
-	char* image = (char*)malloc(4 * m_size.x * m_size.y);
+	uint32 pitch = 4 * m_size.x;
+	if (IsDss()) pitch *= 2;
+	char* image = (char*)malloc(pitch * m_size.y);
+
 	if (IsBackbuffer()) {
 		// TODO backbuffer
 		glReadBuffer(GL_BACK);
 		glReadPixels(0, 0, m_size.x, m_size.y, GL_RGBA, GL_UNSIGNED_BYTE, image);
+	} else if(IsDss()) {
+		Attach(GL_DEPTH_STENCIL_ATTACHMENT);
+		glGetTexImage(GL_TEXTURE_2D, 0, GL_DEPTH_STENCIL, GL_FLOAT_32_UNSIGNED_INT_24_8_REV, image);
 	} else {
 		EnableUnit(0);
 		glGetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_UNSIGNED_BYTE, image);
@@ -417,12 +421,29 @@ bool GSTextureOGL::Save(const string& fn, bool dds)
 		fwrite(&bfh, 1, sizeof(bfh), fp);
 		fwrite(&bih, 1, sizeof(bih), fp);
 
-		uint32 pitch = 4 * m_size.x;
 		uint8* data = (uint8*)image + (m_size.y - 1) * pitch;
 
 		for(int h = m_size.y; h > 0; h--, data -= pitch)
 		{
-			fwrite(data, 1, m_size.x << 2, fp); // TODO: swap red-blue?
+			if (IsDss()) {
+				// Only get the depth and convert it to an integer
+				uint8* better_data = data;
+				for (int w = m_size.x; w > 0; w--, better_data += 8) {
+					float* input = (float*)better_data;
+					uint32 depth = (uint32)ldexpf(*input, 32);
+					fwrite(&depth, 1, 4, fp);
+				}
+			} else {
+				// swap red and blue
+				uint8* better_data = data;
+				for (int w = m_size.x; w > 0; w--, better_data += 4) {
+					uint8 red = better_data[2];
+					better_data[2] = better_data[0];
+					better_data[0] = red;
+					fwrite(better_data, 1, 4, fp);
+				}
+			}
+			// fwrite(data, 1, m_size.x << 2, fp); // TODO: swap red-blue?
 		}
 
 		fclose(fp);

plugins/GSdx/GSTextureOGL.h

@@ -45,4 +45,5 @@ class GSTextureOGL : public GSTexture
 		void Attach(GLenum attachment);
 
 		bool IsBackbuffer() { return (m_type == GSTexture::Backbuffer); }
+		bool IsDss() { return (m_type == GSTexture::DepthStencil); }
 };

plugins/GSdx/res/convert.glsl

@@ -39,6 +39,7 @@ layout(location = 0) in vec4 SV_Position;
 layout(location = 1) in vec2 TEXCOORD0;
 
 layout(location = 0) out vec4 SV_Target0;
+layout(location = 1) out uint SV_Target1;
 
 layout(binding = 0) uniform sampler2D TextureSampler;
 
@@ -65,6 +66,17 @@ void ps_main0()
     SV_Target0 = sample_c();
 }
 
+void ps_main1()
+{
+    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
+
+	highp uvec4 i = uvec4(c * vec4(0x001f, 0x03e0, 0x7c00, 0x8000));
+
+    SV_Target1 = (i.x & 0x001f) | (i.y & 0x03e0) | (i.z & 0x7c00) | (i.w & 0x8000);
+}
+
 void ps_main7()
 {
     vec4 c = sample_c();
@@ -93,9 +105,6 @@ void ps_main6() // diagonal
 }
 
 // Avoid to log useless error compilation failure
-void ps_main1()
-{
-}
 void ps_main2()
 {
 }
@@ -106,20 +115,6 @@ void ps_main4()
 {
 }
 
-//void ps_main1()
-//{
-//    vec4 c = sample_c();
-//
-//	c.a *= 256.0f / 127; // hm, 0.5 won't give us 1.0 if we just multiply with 2
-//
-//	uvec4 i = uvec4(c * vec4(0x001f, 0x03e0, 0x7c00, 0x8000));
-//
-//#ifdef DEBUG_FRAG
-//    SV_Target0 = vec4(0.5,0.5,0.5,1.0);
-//#else
-//    SV_Target0 = (i.x & 0x001f) | (i.y & 0x03e0) | (i.z & 0x7c00) | (i.w & 0x8000);
-//#endif
-//}
 
 // Texture2D Texture;
 // SamplerState TextureSampler;

plugins/GSdx/res/tfx.glsl

@@ -86,12 +86,13 @@ void vs_main()
 	// example: 133.0625 (133 + 1/16) should start from line 134, ceil(133.0625 - 0.05) still above 133
 	
 	vec4 p = vec4(i_p, z, 0) - vec4(0.05f, 0.05f, 0, 0); 
+	vec4 final_p = p * VertexScale - VertexOffset;
 
-	OUT.p = p * VertexScale - VertexOffset;
+	OUT.p = final_p;
+    gl_Position = final_p; // NOTE I don't know if it is possible to merge POSITION_OUT and gl_Position
 #if VS_RTCOPY
-	OUT.tp = (p * VertexScale - VertexOffset) * vec4(0.5, -0.5, 0, 0) + 0.5;
+	OUT.tp = final_p * vec4(0.5, -0.5, 0, 0) + 0.5;
 #endif
-    gl_Position = p; // NOTE I don't know if it is possible to merge POSITION_OUT and gl_Position
 
 	if(VS_TME != 0)
 	{
@@ -237,8 +238,13 @@ void gs_main()
 layout(location = 0) in vertex PSin;
 
 // Same buffer but 2 colors for dual source blending
-layout(location = 0, index = 0) out vec4 SV_Target0;
+//FIXME
-layout(location = 0, index = 1) out vec4 SV_Target1;
+#if 1
+	layout(location = 0, index = 0) out vec4 SV_Target0;
+	layout(location = 0, index = 1) out vec4 SV_Target1;
+#else
+	layout(location = 0)		    out vec4 SV_Target;
+#endif
 
 layout(binding = 0) uniform sampler2D TextureSampler;
 layout(binding = 1) uniform sampler2D PaletteSampler;
@@ -583,8 +589,6 @@ vec4 ps_color()
 
 	if(PS_CLR1 != 0) // needed for Cd * (As/Ad/F + 1) blending modes
 	{
-        // FIXME: I'm not sure about the value of others field
-		//c.rgb = 1.0f; 
 		c.rgb = vec3(1.0f, 1.0f, 1.0f); 
 	}
 
@@ -593,6 +597,8 @@ vec4 ps_color()
 
 void ps_main()
 {
+	//FIXME
+#if 1
 	vec4 c = ps_color();
 
     // FIXME: I'm not sure about the value of others field
@@ -611,5 +617,11 @@ void ps_main()
 	}
 
     SV_Target0 = c;
+
+    //SV_Target0 = vec4(1.0f,0.0f,0.0f, 1.0f);
+    //SV_Target1 = vec4(1.0f,0.0f,0.0f, 1.0f);
+#else
+    SV_Target = vec4(1.0f,0.0f,0.0f, 1.0f);
+#endif
 }
 #endif