Mirrors/yuzu
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge pull request #3081 from ReinUsesLisp/fswzadd-shuffles

Browse source 
shader: Implement FSWZADD and reimplement SHFL
This commit is contained in:
Fernando Sahmkow 2019-11-14 10:27:27 -04:00 • committed by GitHub
commit b6f6733131
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
8 changed files with 138 additions and 136 deletions
Download patch file Download diff file Expand all files Collapse all files

2
src/video_core/renderer_opengl/gl_device.cpp
View file

@ -62,6 +62,7 @@ Device::Device() {
max_varyings = GetInteger<u32>(GL_MAX_VARYING_VECTORS);
has_warp_intrinsics = GLAD_GL_NV_gpu_shader5 && GLAD_GL_NV_shader_thread_group &&
GLAD_GL_NV_shader_thread_shuffle;
has_shader_ballot = GLAD_GL_ARB_shader_ballot;
has_vertex_viewport_layer = GLAD_GL_ARB_shader_viewport_layer_array;
has_image_load_formatted = HasExtension(extensions, "GL_EXT_shader_image_load_formatted");
has_variable_aoffi = TestVariableAoffi();
@ -79,6 +80,7 @@ Device::Device(std::nullptr_t) {
max_vertex_attributes = 16;
max_varyings = 15;
has_warp_intrinsics = true;
has_shader_ballot = true;
has_vertex_viewport_layer = true;
has_image_load_formatted = true;
has_variable_aoffi = true;

5
src/video_core/renderer_opengl/gl_device.h
View file

@ -34,6 +34,10 @@ public:
return has_warp_intrinsics;
}
bool HasShaderBallot() const {
return has_shader_ballot;
}
bool HasVertexViewportLayer() const {
return has_vertex_viewport_layer;
}
@ -68,6 +72,7 @@ private:
u32 max_vertex_attributes{};
u32 max_varyings{};
bool has_warp_intrinsics{};
bool has_shader_ballot{};
bool has_vertex_viewport_layer{};
bool has_image_load_formatted{};
bool has_variable_aoffi{};

19
src/video_core/renderer_opengl/gl_shader_cache.cpp
View file

@ -275,16 +275,25 @@ CachedProgram BuildShader(const Device& device, u64 unique_identifier, ProgramTy
std::string source = fmt::format(R"(// {}
#version 430 core
#extension GL_ARB_separate_shader_objects : enable
#extension GL_ARB_shader_viewport_layer_array : enable
#extension GL_EXT_shader_image_load_formatted : enable
#extension GL_NV_gpu_shader5 : enable
#extension GL_NV_shader_thread_group : enable
#extension GL_NV_shader_thread_shuffle : enable
)",
GetShaderId(unique_identifier, program_type));
if (is_compute) {
source += "#extension GL_ARB_compute_variable_group_size : require\n";
}
if (device.HasShaderBallot()) {
source += "#extension GL_ARB_shader_ballot : require\n";
}
if (device.HasVertexViewportLayer()) {
source += "#extension GL_ARB_shader_viewport_layer_array : require\n";
}
if (device.HasImageLoadFormatted()) {
source += "#extension GL_EXT_shader_image_load_formatted : require\n";
}
if (device.HasWarpIntrinsics()) {
source += "#extension GL_NV_gpu_shader5 : require\n"
"#extension GL_NV_shader_thread_group : require\n"
"#extension GL_NV_shader_thread_shuffle : require\n";
}
source += '\n';
if (!is_compute) {

80
src/video_core/renderer_opengl/gl_shader_decompiler.cpp
View file

@ -1379,6 +1379,26 @@ private:
return GenerateUnary(operation, "float", Type::Float, type);
}
Expression FSwizzleAdd(Operation operation) {
const std::string op_a = VisitOperand(operation, 0).AsFloat();
const std::string op_b = VisitOperand(operation, 1).AsFloat();
if (!device.HasShaderBallot()) {
LOG_ERROR(Render_OpenGL, "Shader ballot is unavailable but required by the shader");
return {fmt::format("{} + {}", op_a, op_b), Type::Float};
}
const std::string instr_mask = VisitOperand(operation, 2).AsUint();
const std::string mask = code.GenerateTemporary();
code.AddLine("uint {} = ({} >> ((gl_SubGroupInvocationARB & 3) << 1)) & 3;", mask,
instr_mask);
const std::string modifier_a = fmt::format("fswzadd_modifiers_a[{}]", mask);
const std::string modifier_b = fmt::format("fswzadd_modifiers_b[{}]", mask);
return {fmt::format("(({} * {}) + ({} * {}))", op_a, modifier_a, op_b, modifier_b),
Type::Float};
}
Expression ICastFloat(Operation operation) {
return GenerateUnary(operation, "int", Type::Int, Type::Float);
}
@ -1942,34 +1962,24 @@ private:
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};
Expression ThreadId(Operation operation) {
if (!device.HasShaderBallot()) {
LOG_ERROR(Render_OpenGL, "Shader ballot is unavailable but required by the shader");
return {"0U", Type::Uint};
}
return {"gl_SubGroupInvocationARB", Type::Uint};
}
Expression ShuffleIndexed(Operation operation) {
std::string value = VisitOperand(operation, 0).AsFloat();
if (!device.HasShaderBallot()) {
LOG_ERROR(Render_OpenGL, "Shader ballot is unavailable but required by the shader");
return {std::move(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};
return {fmt::format("readInvocationARB({}, {})", value, index), Type::Float};
}
struct Func final {
@ -1981,11 +1991,6 @@ private:
static constexpr std::string_view Or = "Or";
static constexpr std::string_view Xor = "Xor";
static constexpr std::string_view Exchange = "Exchange";
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 = {
@ -2016,6 +2021,7 @@ private:
&GLSLDecompiler::FTrunc,
&GLSLDecompiler::FCastInteger<Type::Int>,
&GLSLDecompiler::FCastInteger<Type::Uint>,
&GLSLDecompiler::FSwizzleAdd,
&GLSLDecompiler::Add<Type::Int>,
&GLSLDecompiler::Mul<Type::Int>,
@ -2151,15 +2157,8 @@ private:
&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>,
&GLSLDecompiler::ThreadId,
&GLSLDecompiler::ShuffleIndexed,
};
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
@ -2492,6 +2491,9 @@ bvec2 HalfFloatNanComparison(bvec2 comparison, vec2 pair1, vec2 pair2) {
bvec2 is_nan2 = isnan(pair2);
return bvec2(comparison.x || is_nan1.x || is_nan2.x, comparison.y || is_nan1.y || is_nan2.y);
}
const float fswzadd_modifiers_a[] = float[4](-1.0f, 1.0f, -1.0f, 0.0f );
const float fswzadd_modifiers_b[] = float[4](-1.0f, -1.0f, 1.0f, -1.0f );
)";
}