aboutsummaryrefslogtreecommitdiff
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_shader_cache.cpp3
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp110
-rw-r--r--src/video_core/renderer_opengl/maxwell_to_gl.h4
3 files changed, 96 insertions, 21 deletions
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 909ccb82c..0dbc4c02f 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -214,7 +214,8 @@ CachedProgram SpecializeShader(const std::string& code, const GLShader::ShaderEn
std::string source = "#version 430 core\n"
"#extension GL_ARB_separate_shader_objects : enable\n"
"#extension GL_NV_gpu_shader5 : enable\n"
- "#extension GL_NV_shader_thread_group : enable\n";
+ "#extension GL_NV_shader_thread_group : enable\n"
+ "#extension GL_NV_shader_thread_shuffle : enable\n";
if (entries.shader_viewport_layer_array) {
source += "#extension GL_ARB_shader_viewport_layer_array : enable\n";
}
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 137b23740..76439e7ab 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -325,6 +325,7 @@ public:
DeclareRegisters();
DeclarePredicates();
DeclareLocalMemory();
+ DeclareSharedMemory();
DeclareInternalFlags();
DeclareInputAttributes();
DeclareOutputAttributes();
@@ -499,6 +500,13 @@ private:
code.AddNewLine();
}
+ void DeclareSharedMemory() {
+ if (stage != ProgramType::Compute) {
+ return;
+ }
+ code.AddLine("shared uint {}[];", GetSharedMemory());
+ }
+
void DeclareInternalFlags() {
for (u32 flag = 0; flag < static_cast<u32>(InternalFlag::Amount); flag++) {
const auto flag_code = static_cast<InternalFlag>(flag);
@@ -881,6 +889,12 @@ private:
Type::Uint};
}
+ if (const auto smem = std::get_if<SmemNode>(&*node)) {
+ return {
+ fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
+ Type::Uint};
+ }
+
if (const auto internal_flag = std::get_if<InternalFlagNode>(&*node)) {
return {GetInternalFlag(internal_flag->GetFlag()), Type::Bool};
}
@@ -1007,10 +1021,10 @@ private:
return {std::move(temporary), value.GetType()};
}
- Expression GetOutputAttribute(const AbufNode* abuf) {
+ std::optional<Expression> GetOutputAttribute(const AbufNode* abuf) {
switch (const auto attribute = abuf->GetIndex()) {
case Attribute::Index::Position:
- return {"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float};
+ return {{"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float}};
case Attribute::Index::LayerViewportPointSize:
switch (abuf->GetElement()) {
case 0:
@@ -1020,25 +1034,25 @@ private:
if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
return {};
}
- return {"gl_Layer", Type::Int};
+ return {{"gl_Layer", Type::Int}};
case 2:
if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
return {};
}
- return {"gl_ViewportIndex", Type::Int};
+ return {{"gl_ViewportIndex", Type::Int}};
case 3:
UNIMPLEMENTED_MSG("Requires some state changes for gl_PointSize to work in shader");
- return {"gl_PointSize", Type::Float};
+ return {{"gl_PointSize", Type::Float}};
}
return {};
case Attribute::Index::ClipDistances0123:
- return {fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float};
+ return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float}};
case Attribute::Index::ClipDistances4567:
- return {fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float};
+ return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float}};
default:
if (IsGenericAttribute(attribute)) {
- return {GetOutputAttribute(attribute) + GetSwizzle(abuf->GetElement()),
- Type::Float};
+ return {
+ {GetOutputAttribute(attribute) + GetSwizzle(abuf->GetElement()), Type::Float}};
}
UNIMPLEMENTED_MSG("Unhandled output attribute: {}", static_cast<u32>(attribute));
return {};
@@ -1278,7 +1292,11 @@ private:
target = {GetRegister(gpr->GetIndex()), Type::Float};
} else if (const auto abuf = std::get_if<AbufNode>(&*dest)) {
UNIMPLEMENTED_IF(abuf->IsPhysicalBuffer());
- target = GetOutputAttribute(abuf);
+ auto output = GetOutputAttribute(abuf);
+ if (!output) {
+ return {};
+ }
+ target = std::move(*output);
} else if (const auto lmem = std::get_if<LmemNode>(&*dest)) {
if (stage == ProgramType::Compute) {
LOG_WARNING(Render_OpenGL, "Local memory is stubbed on compute shaders");
@@ -1286,6 +1304,11 @@ private:
target = {
fmt::format("{}[{} >> 2]", GetLocalMemory(), Visit(lmem->GetAddress()).AsUint()),
Type::Uint};
+ } else if (const auto smem = std::get_if<SmemNode>(&*dest)) {
+ ASSERT(stage == ProgramType::Compute);
+ target = {
+ fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
+ Type::Uint};
} else if (const auto gmem = std::get_if<GmemNode>(&*dest)) {
const std::string real = Visit(gmem->GetRealAddress()).AsUint();
const std::string base = Visit(gmem->GetBaseAddress()).AsUint();
@@ -1934,8 +1957,7 @@ private:
Expression BallotThread(Operation operation) {
const std::string value = VisitOperand(operation, 0).AsBool();
if (!device.HasWarpIntrinsics()) {
- LOG_ERROR(Render_OpenGL,
- "Nvidia warp intrinsics are not available and its required by a shader");
+ LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
// Stub on non-Nvidia devices by simulating all threads voting the same as the active
// one.
return {fmt::format("({} ? 0xFFFFFFFFU : 0U)", value), Type::Uint};
@@ -1946,8 +1968,7 @@ private:
Expression Vote(Operation operation, const char* func) {
const std::string value = VisitOperand(operation, 0).AsBool();
if (!device.HasWarpIntrinsics()) {
- LOG_ERROR(Render_OpenGL,
- "Nvidia vote intrinsics are not available and its required by a shader");
+ LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
// Stub with a warp size of one.
return {value, Type::Bool};
}
@@ -1964,15 +1985,54 @@ private:
Expression VoteEqual(Operation operation) {
if (!device.HasWarpIntrinsics()) {
- LOG_ERROR(Render_OpenGL,
- "Nvidia vote intrinsics are not available and its required by a shader");
- // We must return true here since a stub for a theoretical warp size of 1 will always
- // return an equal result for all its votes.
+ LOG_ERROR(Render_OpenGL, "Nvidia vote intrinsics are required by this shader");
+ // We must return true here since a stub for a theoretical warp size of 1.
+ // This will always return an equal result across all votes.
return {"true", Type::Bool};
}
return Vote(operation, "allThreadsEqualNV");
}
+ template <const std::string_view& func>
+ Expression Shuffle(Operation operation) {
+ const std::string value = VisitOperand(operation, 0).AsFloat();
+ if (!device.HasWarpIntrinsics()) {
+ LOG_ERROR(Render_OpenGL, "Nvidia shuffle intrinsics are required by this shader");
+ // On a "single-thread" device we are either on the same thread or out of bounds. Both
+ // cases return the passed value.
+ return {value, Type::Float};
+ }
+
+ const std::string index = VisitOperand(operation, 1).AsUint();
+ const std::string width = VisitOperand(operation, 2).AsUint();
+ return {fmt::format("{}({}, {}, {})", func, value, index, width), Type::Float};
+ }
+
+ template <const std::string_view& func>
+ Expression InRangeShuffle(Operation operation) {
+ const std::string index = VisitOperand(operation, 0).AsUint();
+ const std::string width = VisitOperand(operation, 1).AsUint();
+ if (!device.HasWarpIntrinsics()) {
+ // On a "single-thread" device we are only in bounds when the requested index is 0.
+ return {fmt::format("({} == 0U)", index), Type::Bool};
+ }
+
+ const std::string in_range = code.GenerateTemporary();
+ code.AddLine("bool {};", in_range);
+ code.AddLine("{}(0U, {}, {}, {});", func, index, width, in_range);
+ return {in_range, Type::Bool};
+ }
+
+ struct Func final {
+ Func() = delete;
+ ~Func() = delete;
+
+ static constexpr std::string_view ShuffleIndexed = "shuffleNV";
+ static constexpr std::string_view ShuffleUp = "shuffleUpNV";
+ static constexpr std::string_view ShuffleDown = "shuffleDownNV";
+ static constexpr std::string_view ShuffleButterfly = "shuffleXorNV";
+ };
+
static constexpr std::array operation_decompilers = {
&GLSLDecompiler::Assign,
@@ -2135,6 +2195,16 @@ private:
&GLSLDecompiler::VoteAll,
&GLSLDecompiler::VoteAny,
&GLSLDecompiler::VoteEqual,
+
+ &GLSLDecompiler::Shuffle<Func::ShuffleIndexed>,
+ &GLSLDecompiler::Shuffle<Func::ShuffleUp>,
+ &GLSLDecompiler::Shuffle<Func::ShuffleDown>,
+ &GLSLDecompiler::Shuffle<Func::ShuffleButterfly>,
+
+ &GLSLDecompiler::InRangeShuffle<Func::ShuffleIndexed>,
+ &GLSLDecompiler::InRangeShuffle<Func::ShuffleUp>,
+ &GLSLDecompiler::InRangeShuffle<Func::ShuffleDown>,
+ &GLSLDecompiler::InRangeShuffle<Func::ShuffleButterfly>,
};
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
@@ -2175,6 +2245,10 @@ private:
return "lmem_" + suffix;
}
+ std::string GetSharedMemory() const {
+ return fmt::format("smem_{}", suffix);
+ }
+
std::string GetInternalFlag(InternalFlag flag) const {
constexpr std::array InternalFlagNames = {"zero_flag", "sign_flag", "carry_flag",
"overflow_flag"};
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index ea77dd211..9ed738171 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -145,7 +145,7 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
case Tegra::Texture::TextureMipmapFilter::None:
return GL_LINEAR;
case Tegra::Texture::TextureMipmapFilter::Nearest:
- return GL_NEAREST_MIPMAP_LINEAR;
+ return GL_LINEAR_MIPMAP_NEAREST;
case Tegra::Texture::TextureMipmapFilter::Linear:
return GL_LINEAR_MIPMAP_LINEAR;
}
@@ -157,7 +157,7 @@ inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
case Tegra::Texture::TextureMipmapFilter::Nearest:
return GL_NEAREST_MIPMAP_NEAREST;
case Tegra::Texture::TextureMipmapFilter::Linear:
- return GL_LINEAR_MIPMAP_NEAREST;
+ return GL_NEAREST_MIPMAP_LINEAR;
}
}
}
generated by cgit v1.2.3 (git 2.25.1) at 2025-12-17 09:41:41 +0000