Implemented fragment constants loading (OpenGL renderer)

Fixed nv308a::color
Minor improvements

rpcs3/Emu/RSX/Common/VertexProgramDecompiler.cpp

@@ -136,13 +136,13 @@ void VertexProgramDecompiler::SetDST(bool is_sca, std::string value)
 
 	std::string mask = GetMask(is_sca);
 
-	value += mask;
+	if (is_sca)
-
-	if (is_sca && d0.vec_result)
 	{
-		//value = "vec4(" + value + ")";
+		value = getFloatTypeName(4) + "(" + value + ")";
 	}
 
+	value += mask;
+
 	if (d0.staturate)
 	{
 		value = "clamp(" + value + ", 0.0, 1.0)";
@@ -309,7 +309,7 @@ void VertexProgramDecompiler::AddCodeCond(const std::string& dst, const std::str
 
 	if (dst_var.swizzles[0].length() == 1)
 	{
-		AddCode("if (" + cond + ".x) " + dst + " = " + getFloatTypeName(4) + "(" + src + ".xxxx).x;");
+		AddCode("if (" + cond + ".x) " + dst + " = " + src + ";");
 	}
 	else
 	{
@@ -559,7 +559,7 @@ std::string VertexProgramDecompiler::Decompile()
 		case RSX_SCA_OPCODE_MOV: SetDSTSca("$s"); break;
 		case RSX_SCA_OPCODE_RCP: SetDSTSca("(1.0 / $s)"); break;
 		case RSX_SCA_OPCODE_RCC: SetDSTSca("clamp(1.0 / $s, 5.42101e-20, 1.884467e19)"); break;
-		case RSX_SCA_OPCODE_RSQ: SetDSTSca("1.f / sqrt($s)"); break;
+		case RSX_SCA_OPCODE_RSQ: SetDSTSca("(1.f / sqrt($s))"); break;
 		case RSX_SCA_OPCODE_EXP: SetDSTSca("exp($s)"); break;
 		case RSX_SCA_OPCODE_LOG: SetDSTSca("log($s)"); break;
 		case RSX_SCA_OPCODE_LIT: SetDSTSca(getFloatTypeName(4) + "(1.0, $s.x, ($s.x > 0.0 ? exp($s.w * log2($s.y)) : 0.0), 1.0)"); break;

rpcs3/Emu/RSX/GCM.h

@@ -498,9 +498,9 @@ struct CellGcmSurface
 
 struct CellGcmReportData
 {
-	u64 timer;
+	be_t<u64> timer;
-	u32 value;
+	be_t<u32> value;
-	u32 padding;
+	be_t<u32> padding;
 };
 
 struct CellGcmZcullInfo

rpcs3/Emu/RSX/GL/GLGSRender.cpp

@@ -578,9 +578,9 @@ void GLGSRender::begin()
 
 	u32 color_mask = rsx::method_registers[NV4097_SET_COLOR_MASK];
 	bool color_mask_b = color_mask & 0xff;
-	bool color_mask_g = color_mask >> 8;
+	bool color_mask_g = (color_mask >> 8) & 0xff;
-	bool color_mask_r = color_mask >> 16;
+	bool color_mask_r = (color_mask >> 16) & 0xff;
-	bool color_mask_a = color_mask >> 24;
+	bool color_mask_a = (color_mask >> 24) & 0xff;
 
 	__glcheck glColorMask(color_mask_r, color_mask_g, color_mask_b, color_mask_a);
 	__glcheck glDepthMask(rsx::method_registers[NV4097_SET_DEPTH_MASK]);
@@ -1035,7 +1035,7 @@ void GLGSRender::end()
 
 	if (vertex_index_array.empty())
 	{
-		glDrawArrays(draw_mode - 1, 0, vertex_draw_count);
+		draw_fbo.draw_arrays(gl::draw_mode(draw_mode - 1), vertex_draw_count);\
 	}
 	else
 	{
@@ -1158,11 +1158,7 @@ void nv4097_clear_surface(u32 arg, GLGSRender* renderer)
 	}
 
 	renderer->clear_surface_buffers = (gl::buffers)mask;
-	//renderer->init_buffers();
+	renderer->draw_fbo.clear((gl::buffers)mask);
-	//renderer->draw_fbo.draw_buffer(renderer->draw_fbo.color[0]);
-	//renderer->draw_fbo.clear(gl::buffers(mask));
-	//renderer->draw_fbo.draw_arrays(gl::draw_mode::lines, 0);
-	//renderer->write_buffers();
 }
 
 using rsx_method_impl_t = void(*)(u32, GLGSRender*);
@@ -1291,6 +1287,18 @@ bool GLGSRender::load_program()
 		__glcheck m_program.uniforms["vc[" + std::to_string(constant.first) + "]"] = constant.second;
 	}
 
+	for (u32 constant_offset : m_prog_buffer.getFragmentConstantOffsetsCache(&fragment_program))
+	{
+		be_t<u32> *data = vm::get_ptr<be_t<u32>>(fragment_program.addr + constant_offset);
+
+		u32 c0 = (data[0] >> 16 | data[0] << 16);
+		u32 c1 = (data[1] >> 16 | data[1] << 16);
+		u32 c2 = (data[2] >> 16 | data[2] << 16);
+		u32 c3 = (data[3] >> 16 | data[3] << 16);
+
+		m_program.uniforms["fc" + std::to_string(constant_offset)] = color4f{ (f32&)c0, (f32&)c1, (f32&)c2, (f32&)c3 };
+	}
+
 	return true;
 }
 
@@ -1461,11 +1469,11 @@ void GLGSRender::init_buffers()
 	case CELL_GCM_SURFACE_TARGET_NONE: break;
 
 	case CELL_GCM_SURFACE_TARGET_0:
-		__glcheck draw_fbo.draw_buffers({ draw_fbo.color[0] });
+		__glcheck draw_fbo.draw_buffer(draw_fbo.color[0]);
 		break;
 
 	case CELL_GCM_SURFACE_TARGET_1:
-		__glcheck draw_fbo.draw_buffers({ draw_fbo.color[1] });
+		__glcheck draw_fbo.draw_buffer(draw_fbo.color[1] );
 		break;
 
 	case CELL_GCM_SURFACE_TARGET_MRT1:
@@ -1487,7 +1495,7 @@ void GLGSRender::init_buffers()
 
 	if (clear_surface_buffers != gl::buffers::none)
 	{
-		draw_fbo.clear(clear_surface_buffers);
+		//draw_fbo.clear(clear_surface_buffers);
 
 		clear_surface_buffers = gl::buffers::none;
 	}
@@ -1707,6 +1715,8 @@ void GLGSRender::flip(int buffer)
 
 	if (draw_fbo && !Ini.GSDumpColorBuffers.GetValue())
 	{
+		skip_read = true;
+		/*
 		for (uint i = 0; i < rsx::limits::color_buffers_count; ++i)
 		{
 			u32 color_address = rsx::get_address(rsx::method_registers[mr_color_offset[i]], rsx::method_registers[mr_color_dma[i]]);
@@ -1718,6 +1728,7 @@ void GLGSRender::flip(int buffer)
 				break;
 			}
 		}
+		*/
 	}
 
 	if (!skip_read)

rpcs3/Emu/RSX/RSXThread.cpp

@@ -47,13 +47,13 @@ namespace rsx
 		__forceinline void semaphore_acquire(thread* rsx, u32 arg)
 		{
 			//TODO: dma
-			/*while (vm::read32(rsx->label_addr + method_registers[NV406E_SEMAPHORE_OFFSET]) != arg)
+			while (vm::read32(rsx->label_addr + method_registers[NV406E_SEMAPHORE_OFFSET]) != arg)
 			{
 				if (Emu.IsStopped())
 					break;
 
 				std::this_thread::sleep_for(std::chrono::milliseconds(1));
-			}*/
+			}
 		}
 
 		__forceinline void semaphore_release(thread* rsx, u32 arg)
@@ -246,7 +246,11 @@ namespace rsx
 			u8 type = arg >> 24;
 			u32 offset = arg & 0xffffff;
 
-			u32 value;
+			//TODO: use DMA
+			vm::ptr<CellGcmReportData> result = { rsx->local_mem_addr + offset };
+
+			result->timer = rsx->timestamp();
+
 			switch (type)
 			{
 			case CELL_GCM_ZPASS_PIXEL_CNT:
@@ -254,24 +258,17 @@ namespace rsx
 			case CELL_GCM_ZCULL_STATS1:
 			case CELL_GCM_ZCULL_STATS2:
 			case CELL_GCM_ZCULL_STATS3:
-				value = 0;
+				result->value = 0;
 				LOG_WARNING(RSX, "NV4097_GET_REPORT: Unimplemented type %d", type);
 				break;
 
 			default:
-				value = 0;
+				result->value = 0;
 				LOG_ERROR(RSX, "NV4097_GET_REPORT: Bad type %d", type);
 				break;
 			}
 
-			// NOTE: DMA broken, implement proper lpar mapping (sys_rsx)
+			//result->padding = 0;
-			//dma_write64(dma_report, offset + 0x0, rsx->timestamp());
-			//dma_write32(dma_report, offset + 0x8, value);
-			//dma_write32(dma_report, offset + 0xc, 0);
-
-			vm::write64(rsx->local_mem_addr + offset + 0x0, rsx->timestamp());
-			vm::write32(rsx->local_mem_addr + offset + 0x8, value);
-			vm::write32(rsx->local_mem_addr + offset + 0xc, 0);
 		}
 
 		__forceinline void clear_report_value(thread* rsx, u32 arg)
@@ -305,7 +302,7 @@ namespace rsx
 				LOG_ERROR(RSX, "%s: y is not null (0x%x)", __FUNCTION__, y);
 			}
 
-			u32 address = get_address(method_registers[NV3062_SET_OFFSET_DESTIN] + (x << 2) + index, method_registers[NV3062_SET_CONTEXT_DMA_IMAGE_DESTIN]);
+			u32 address = get_address(method_registers[NV3062_SET_OFFSET_DESTIN] + (x << 2) + index * 4, method_registers[NV3062_SET_CONTEXT_DMA_IMAGE_DESTIN]);
 			vm::write32(address, arg);
 		}
 	}
@@ -916,7 +913,6 @@ namespace rsx
 		for (auto &vertex_array : vertex_arrays)
 			vertex_array.clear();
 
-		fragment_constants.clear();
 		transform_constants.clear();
 	}
 

rpcs3/Emu/RSX/RSXThread.h

@@ -159,7 +159,6 @@ namespace rsx
 		std::vector<u8> vertex_index_array;
 		u32 vertex_draw_count = 0;
 
-		std::unordered_map<u32, color4_base<f32>> fragment_constants;
 		std::unordered_map<u32, color4_base<f32>> transform_constants;
 
 		u32 transform_program[512 * 4] = {};