From 08b2b1080a2e5794c65ebb999c435f6eb9d1aa26 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sat, 2 Nov 2019 23:44:13 -0300 Subject: gl_shader_decompiler: Reimplement shuffles with platform agnostic intrinsics --- src/video_core/renderer_opengl/gl_shader_cache.cpp | 1 + .../renderer_opengl/gl_shader_decompiler.cpp | 48 ++++------------------ 2 files changed, 9 insertions(+), 40 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index f1b89165d..a287b5ee1 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -275,6 +275,7 @@ 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_ballot : enable #extension GL_ARB_shader_viewport_layer_array : enable #extension GL_EXT_shader_image_load_formatted : enable #extension GL_NV_gpu_shader5 : enable diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 92ee8459e..ca4e6e468 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -1942,34 +1942,14 @@ private: return Vote(operation, "allThreadsEqualNV"); } - template - 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}; + Expression ThreadId(Operation operation) { + return {"gl_SubGroupInvocationARB", Type::Uint}; } - template - 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}; + Expression ShuffleIndexed(Operation operation) { + const std::string value = VisitOperand(operation, 0).AsFloat(); + const std::string index = VisitOperand(operation, 1).AsUint(); + return {fmt::format("readInvocationARB({}, {})", value, index), Type::Float}; } struct Func final { @@ -1981,11 +1961,6 @@ 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 = { @@ -2151,15 +2126,8 @@ private: &GLSLDecompiler::VoteAny, &GLSLDecompiler::VoteEqual, - &GLSLDecompiler::Shuffle, - &GLSLDecompiler::Shuffle, - &GLSLDecompiler::Shuffle, - &GLSLDecompiler::Shuffle, - - &GLSLDecompiler::InRangeShuffle, - &GLSLDecompiler::InRangeShuffle, - &GLSLDecompiler::InRangeShuffle, - &GLSLDecompiler::InRangeShuffle, + &GLSLDecompiler::ThreadId, + &GLSLDecompiler::ShuffleIndexed, }; static_assert(operation_decompilers.size() == static_cast(OperationCode::Amount)); -- cgit v1.2.3 From 56e237d1f998a4090afb6763222cd65593b299d7 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Sat, 2 Nov 2019 23:44:46 -0300 Subject: shader_ir/warp: Implement FSWZADD --- .../renderer_opengl/gl_shader_decompiler.cpp | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index ca4e6e468..21c137ec5 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -1379,6 +1379,20 @@ 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(); + const std::string instr_mask = VisitOperand(operation, 2).AsUint(); + + const std::string mask = code.GenerateTemporary(); + code.AddLine("uint {} = {} >> ((gl_SubGroupInvocationARB & 3) << 1);", mask, instr_mask); + + const std::string modifier_a = fmt::format("fswzadd_modifiers_a[{} & 3]", mask); + const std::string modifier_b = fmt::format("fswzadd_modifiers_b[{} & 3]", 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); } @@ -1991,6 +2005,7 @@ private: &GLSLDecompiler::FTrunc, &GLSLDecompiler::FCastInteger, &GLSLDecompiler::FCastInteger, + &GLSLDecompiler::FSwizzleAdd, &GLSLDecompiler::Add, &GLSLDecompiler::Mul, @@ -2460,6 +2475,9 @@ 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 ); )"; } -- cgit v1.2.3 From cd663959445c855ab3d0c4c66b8930d30164cc40 Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Thu, 7 Nov 2019 20:03:50 -0300 Subject: gl_shader_decompiler: Add safe fallbacks when ARB_shader_ballot is not available --- src/video_core/renderer_opengl/gl_device.cpp | 2 ++ src/video_core/renderer_opengl/gl_device.h | 5 +++++ .../renderer_opengl/gl_shader_decompiler.cpp | 26 +++++++++++++++++----- 3 files changed, 28 insertions(+), 5 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_device.cpp b/src/video_core/renderer_opengl/gl_device.cpp index c65b24c69..b30d5be74 100644 --- a/src/video_core/renderer_opengl/gl_device.cpp +++ b/src/video_core/renderer_opengl/gl_device.cpp @@ -62,6 +62,7 @@ Device::Device() { max_varyings = GetInteger(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; diff --git a/src/video_core/renderer_opengl/gl_device.h b/src/video_core/renderer_opengl/gl_device.h index bf35bd0b6..6c86fe207 100644 --- a/src/video_core/renderer_opengl/gl_device.h +++ b/src/video_core/renderer_opengl/gl_device.h @@ -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{}; diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp index 21c137ec5..5d2c38a5e 100644 --- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp +++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp @@ -1382,13 +1382,19 @@ private: Expression FSwizzleAdd(Operation operation) { const std::string op_a = VisitOperand(operation, 0).AsFloat(); const std::string op_b = VisitOperand(operation, 1).AsFloat(); - const std::string instr_mask = VisitOperand(operation, 2).AsUint(); + 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);", mask, instr_mask); + code.AddLine("uint {} = ({} >> ((gl_SubGroupInvocationARB & 3) << 1)) & 3;", mask, + instr_mask); - const std::string modifier_a = fmt::format("fswzadd_modifiers_a[{} & 3]", mask); - const std::string modifier_b = fmt::format("fswzadd_modifiers_b[{} & 3]", 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}; } @@ -1957,11 +1963,21 @@ private: } 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) { - const std::string value = VisitOperand(operation, 0).AsFloat(); + 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(); return {fmt::format("readInvocationARB({}, {})", value, index), Type::Float}; } -- cgit v1.2.3 From 3ab05146985f09994fcb98388b22a31f9fbea5bf Mon Sep 17 00:00:00 2001 From: ReinUsesLisp Date: Thu, 7 Nov 2019 20:07:45 -0300 Subject: gl_shader_cache: Enable extensions only when available Silence GLSL compilation warnings. --- src/video_core/renderer_opengl/gl_shader_cache.cpp | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'src/video_core/renderer_opengl') diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp index a287b5ee1..21bac11ec 100644 --- a/src/video_core/renderer_opengl/gl_shader_cache.cpp +++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp @@ -275,17 +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_ballot : 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) { -- cgit v1.2.3