summaryrefslogtreecommitdiffstats
path: root/src/video_core/renderer_opengl
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core/renderer_opengl')
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.cpp67
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer.h3
-rw-r--r--src/video_core/renderer_opengl/gl_shader_cache.cpp82
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp10
4 files changed, 67 insertions, 95 deletions
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 0719f2407..725b4c32d 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -184,8 +184,12 @@ void RasterizerOpenGL::SetupVertexBuffer() {
const GPUVAddr start = vertex_array.StartAddress();
const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
- ASSERT(end > start);
- const u64 size = end - start + 1;
+ ASSERT(end >= start);
+ const u64 size = end - start;
+ if (size == 0) {
+ glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
+ continue;
+ }
const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
vertex_array.stride);
@@ -309,8 +313,8 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
const GPUVAddr start = regs.vertex_array[index].StartAddress();
const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
- ASSERT(end > start);
- size += end - start + 1;
+ size += end - start;
+ ASSERT(end >= start);
}
return size;
@@ -342,7 +346,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
texture_cache.GuardRenderTargets(true);
- View depth_surface = texture_cache.GetDepthBufferSurface();
+ View depth_surface = texture_cache.GetDepthBufferSurface(true);
const auto& regs = gpu.regs;
UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
@@ -351,7 +355,7 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
FramebufferCacheKey key;
const auto colors_count = static_cast<std::size_t>(regs.rt_control.count);
for (std::size_t index = 0; index < colors_count; ++index) {
- View color_surface{texture_cache.GetColorBufferSurface(index)};
+ View color_surface{texture_cache.GetColorBufferSurface(index, true)};
if (!color_surface) {
continue;
}
@@ -375,28 +379,52 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer_cache.GetFramebuffer(key));
}
-void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color_fb, bool using_depth_fb,
- bool using_stencil_fb) {
+void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color, bool using_depth_stencil) {
auto& gpu = system.GPU().Maxwell3D();
const auto& regs = gpu.regs;
texture_cache.GuardRenderTargets(true);
View color_surface;
- if (using_color_fb) {
+
+ if (using_color) {
+ // Determine if we have to preserve the contents.
+ // First we have to make sure all clear masks are enabled.
+ bool preserve_contents = !regs.clear_buffers.R || !regs.clear_buffers.G ||
+ !regs.clear_buffers.B || !regs.clear_buffers.A;
const std::size_t index = regs.clear_buffers.RT;
- color_surface = texture_cache.GetColorBufferSurface(index);
+ if (regs.clear_flags.scissor) {
+ // Then we have to confirm scissor testing clears the whole image.
+ const auto& scissor = regs.scissor_test[0];
+ preserve_contents |= scissor.min_x > 0;
+ preserve_contents |= scissor.min_y > 0;
+ preserve_contents |= scissor.max_x < regs.rt[index].width;
+ preserve_contents |= scissor.max_y < regs.rt[index].height;
+ }
+
+ color_surface = texture_cache.GetColorBufferSurface(index, preserve_contents);
texture_cache.MarkColorBufferInUse(index);
}
+
View depth_surface;
- if (using_depth_fb || using_stencil_fb) {
- depth_surface = texture_cache.GetDepthBufferSurface();
+ if (using_depth_stencil) {
+ bool preserve_contents = false;
+ if (regs.clear_flags.scissor) {
+ // For depth stencil clears we only have to confirm scissor test covers the whole image.
+ const auto& scissor = regs.scissor_test[0];
+ preserve_contents |= scissor.min_x > 0;
+ preserve_contents |= scissor.min_y > 0;
+ preserve_contents |= scissor.max_x < regs.zeta_width;
+ preserve_contents |= scissor.max_y < regs.zeta_height;
+ }
+
+ depth_surface = texture_cache.GetDepthBufferSurface(preserve_contents);
texture_cache.MarkDepthBufferInUse();
}
texture_cache.GuardRenderTargets(false);
FramebufferCacheKey key;
- key.colors[0] = color_surface;
- key.zeta = depth_surface;
+ key.colors[0] = std::move(color_surface);
+ key.zeta = std::move(depth_surface);
state_tracker.NotifyFramebuffer();
glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer_cache.GetFramebuffer(key));
@@ -416,8 +444,7 @@ void RasterizerOpenGL::Clear() {
if (regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B ||
regs.clear_buffers.A) {
use_color = true;
- }
- if (use_color) {
+
state_tracker.NotifyColorMask0();
glColorMaski(0, regs.clear_buffers.R != 0, regs.clear_buffers.G != 0,
regs.clear_buffers.B != 0, regs.clear_buffers.A != 0);
@@ -455,7 +482,7 @@ void RasterizerOpenGL::Clear() {
UNIMPLEMENTED_IF(regs.clear_flags.viewport);
- ConfigureClearFramebuffer(use_color, use_depth, use_stencil);
+ ConfigureClearFramebuffer(use_color, use_depth || use_stencil);
if (use_color) {
glClearBufferfv(GL_COLOR, 0, regs.clear_color);
@@ -993,11 +1020,7 @@ void RasterizerOpenGL::SyncDepthClamp() {
}
flags[Dirty::DepthClampEnabled] = false;
- const auto& state = gpu.regs.view_volume_clip_control;
- UNIMPLEMENTED_IF_MSG(state.depth_clamp_far != state.depth_clamp_near,
- "Unimplemented depth clamp separation!");
-
- oglEnable(GL_DEPTH_CLAMP, state.depth_clamp_far || state.depth_clamp_near);
+ oglEnable(GL_DEPTH_CLAMP, gpu.regs.view_volume_clip_control.depth_clamp_disabled == 0);
}
void RasterizerOpenGL::SyncClipEnabled(u32 clip_mask) {
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index ebd2173eb..87249fb6f 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -95,7 +95,8 @@ private:
/// Configures the color and depth framebuffer states.
void ConfigureFramebuffers();
- void ConfigureClearFramebuffer(bool using_color_fb, bool using_depth_fb, bool using_stencil_fb);
+ /// Configures the color and depth framebuffer for clearing.
+ void ConfigureClearFramebuffer(bool using_color, bool using_depth_stencil);
/// Configures the current constbuffers to use for the draw command.
void SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader);
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index f63156b8d..9759a7078 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -10,8 +10,6 @@
#include <thread>
#include <unordered_set>
-#include <boost/functional/hash.hpp>
-
#include "common/alignment.h"
#include "common/assert.h"
#include "common/logging/log.h"
@@ -28,76 +26,26 @@
#include "video_core/renderer_opengl/gl_shader_disk_cache.h"
#include "video_core/renderer_opengl/gl_state_tracker.h"
#include "video_core/renderer_opengl/utils.h"
+#include "video_core/shader/memory_util.h"
#include "video_core/shader/registry.h"
#include "video_core/shader/shader_ir.h"
namespace OpenGL {
using Tegra::Engines::ShaderType;
+using VideoCommon::Shader::GetShaderAddress;
+using VideoCommon::Shader::GetShaderCode;
+using VideoCommon::Shader::GetUniqueIdentifier;
+using VideoCommon::Shader::KERNEL_MAIN_OFFSET;
using VideoCommon::Shader::ProgramCode;
using VideoCommon::Shader::Registry;
using VideoCommon::Shader::ShaderIR;
+using VideoCommon::Shader::STAGE_MAIN_OFFSET;
namespace {
-constexpr u32 STAGE_MAIN_OFFSET = 10;
-constexpr u32 KERNEL_MAIN_OFFSET = 0;
-
constexpr VideoCommon::Shader::CompilerSettings COMPILER_SETTINGS{};
-/// Gets the address for the specified shader stage program
-GPUVAddr GetShaderAddress(Core::System& system, Maxwell::ShaderProgram program) {
- const auto& gpu{system.GPU().Maxwell3D()};
- const auto& shader_config{gpu.regs.shader_config[static_cast<std::size_t>(program)]};
- return gpu.regs.code_address.CodeAddress() + shader_config.offset;
-}
-
-/// Gets if the current instruction offset is a scheduler instruction
-constexpr bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
- // Sched instructions appear once every 4 instructions.
- constexpr std::size_t SchedPeriod = 4;
- const std::size_t absolute_offset = offset - main_offset;
- return (absolute_offset % SchedPeriod) == 0;
-}
-
-/// Calculates the size of a program stream
-std::size_t CalculateProgramSize(const ProgramCode& program) {
- constexpr std::size_t start_offset = 10;
- // This is the encoded version of BRA that jumps to itself. All Nvidia
- // shaders end with one.
- constexpr u64 self_jumping_branch = 0xE2400FFFFF07000FULL;
- constexpr u64 mask = 0xFFFFFFFFFF7FFFFFULL;
- std::size_t offset = start_offset;
- while (offset < program.size()) {
- const u64 instruction = program[offset];
- if (!IsSchedInstruction(offset, start_offset)) {
- if ((instruction & mask) == self_jumping_branch) {
- // End on Maxwell's "nop" instruction
- break;
- }
- if (instruction == 0) {
- break;
- }
- }
- offset++;
- }
- // The last instruction is included in the program size
- return std::min(offset + 1, program.size());
-}
-
-/// Gets the shader program code from memory for the specified address
-ProgramCode GetShaderCode(Tegra::MemoryManager& memory_manager, const GPUVAddr gpu_addr,
- const u8* host_ptr) {
- ProgramCode code(VideoCommon::Shader::MAX_PROGRAM_LENGTH);
- ASSERT_OR_EXECUTE(host_ptr != nullptr, {
- std::fill(code.begin(), code.end(), 0);
- return code;
- });
- memory_manager.ReadBlockUnsafe(gpu_addr, code.data(), code.size() * sizeof(u64));
- code.resize(CalculateProgramSize(code));
- return code;
-}
-
/// Gets the shader type from a Maxwell program type
constexpr GLenum GetGLShaderType(ShaderType shader_type) {
switch (shader_type) {
@@ -114,17 +62,6 @@ constexpr GLenum GetGLShaderType(ShaderType shader_type) {
}
}
-/// Hashes one (or two) program streams
-u64 GetUniqueIdentifier(ShaderType shader_type, bool is_a, const ProgramCode& code,
- const ProgramCode& code_b = {}) {
- u64 unique_identifier = boost::hash_value(code);
- if (is_a) {
- // VertexA programs include two programs
- boost::hash_combine(unique_identifier, boost::hash_value(code_b));
- }
- return unique_identifier;
-}
-
constexpr const char* GetShaderTypeName(ShaderType shader_type) {
switch (shader_type) {
case ShaderType::Vertex:
@@ -456,11 +393,12 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
const auto host_ptr{memory_manager.GetPointer(address)};
// No shader found - create a new one
- ProgramCode code{GetShaderCode(memory_manager, address, host_ptr)};
+ ProgramCode code{GetShaderCode(memory_manager, address, host_ptr, false)};
ProgramCode code_b;
if (program == Maxwell::ShaderProgram::VertexA) {
const GPUVAddr address_b{GetShaderAddress(system, Maxwell::ShaderProgram::VertexB)};
- code_b = GetShaderCode(memory_manager, address_b, memory_manager.GetPointer(address_b));
+ const u8* host_ptr_b = memory_manager.GetPointer(address_b);
+ code_b = GetShaderCode(memory_manager, address_b, host_ptr_b, false);
}
const auto unique_identifier = GetUniqueIdentifier(
@@ -498,7 +436,7 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
const auto host_ptr{memory_manager.GetPointer(code_addr)};
// No kernel found, create a new one
- auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
+ auto code{GetShaderCode(memory_manager, code_addr, host_ptr, true)};
const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
const ShaderParameters params{system, disk_cache, device,
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 1b474337a..99fd4ae2c 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1870,6 +1870,14 @@ private:
return GenerateBinaryInfix(operation, ">=", Type::Bool, type, type);
}
+ Expression LogicalAddCarry(Operation operation) {
+ const std::string carry = code.GenerateTemporary();
+ code.AddLine("uint {};", carry);
+ code.AddLine("uaddCarry({}, {}, {});", VisitOperand(operation, 0).AsUint(),
+ VisitOperand(operation, 1).AsUint(), carry);
+ return {fmt::format("({} != 0)", carry), Type::Bool};
+ }
+
Expression LogicalFIsNan(Operation operation) {
return GenerateUnary(operation, "isnan", Type::Bool, Type::Float);
}
@@ -2439,6 +2447,8 @@ private:
&GLSLDecompiler::LogicalNotEqual<Type::Uint>,
&GLSLDecompiler::LogicalGreaterEqual<Type::Uint>,
+ &GLSLDecompiler::LogicalAddCarry,
+
&GLSLDecompiler::Logical2HLessThan<false>,
&GLSLDecompiler::Logical2HEqual<false>,
&GLSLDecompiler::Logical2HLessEqual<false>,
generated by cgit v1.2.3 (git 2.25.1) at 2024-07-21 13:44:47 +0200