Skip to content

S
swiftshader
Commit bc1c067b by Ben Clayton
Options

clang-format the src/Pipeline directory

Bug: b/144825072 Change-Id: I869aef91d6318bf6955581e5dad762800bd46296 Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/39655Tested-by: 's avatarBen Clayton <bclayton@google.com> Reviewed-by: 's avatarNicolas Capens <nicolascapens@google.com>
parent fccfc566
Showing with 446 additions and 390 deletions
src/Pipeline/ComputeProgram.cpp
...@@ -26,16 +26,21 @@ ...@@ -26,16 +26,21 @@
namespace { namespace {
enum { X, Y, Z }; enum
{
X,
Y,
Z
};
} // anonymous namespace } // anonymous namespace
namespace sw { namespace sw {
ComputeProgram::ComputeProgram(SpirvShader const *shader, vk::PipelineLayout const *pipelineLayout, const vk::DescriptorSet::Bindings &descriptorSets) ComputeProgram::ComputeProgram(SpirvShader const *shader, vk::PipelineLayout const *pipelineLayout, const vk::DescriptorSet::Bindings &descriptorSets)
: shader(shader), : shader(shader)
pipelineLayout(pipelineLayout), , pipelineLayout(pipelineLayout)
descriptorSets(descriptorSets) , descriptorSets(descriptorSets)
{ {
} }
...@@ -53,10 +58,9 @@ void ComputeProgram::generate() ...@@ -53,10 +58,9 @@ void ComputeProgram::generate()
shader->emitEpilog(&routine); shader->emitEpilog(&routine);
} }
void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routine, Int workgroupID[3]) void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine *routine, Int workgroupID[3])
{ {
routine->setInputBuiltin(shader, spv::BuiltInNumWorkgroups, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInNumWorkgroups, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
auto numWorkgroups = *Pointer<Int4>(data + OFFSET(Data, numWorkgroups)); auto numWorkgroups = *Pointer<Int4>(data + OFFSET(Data, numWorkgroups));
for(uint32_t component = 0; component < builtin.SizeInComponents; component++) for(uint32_t component = 0; component < builtin.SizeInComponents; component++)
{ {
...@@ -65,8 +69,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout ...@@ -65,8 +69,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout
} }
}); });
routine->setInputBuiltin(shader, spv::BuiltInWorkgroupId, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInWorkgroupId, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
for(uint32_t component = 0; component < builtin.SizeInComponents; component++) for(uint32_t component = 0; component < builtin.SizeInComponents; component++)
{ {
value[builtin.FirstComponent + component] = value[builtin.FirstComponent + component] =
...@@ -74,8 +77,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout ...@@ -74,8 +77,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout
} }
}); });
routine->setInputBuiltin(shader, spv::BuiltInWorkgroupSize, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInWorkgroupSize, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
auto workgroupSize = *Pointer<Int4>(data + OFFSET(Data, workgroupSize)); auto workgroupSize = *Pointer<Int4>(data + OFFSET(Data, workgroupSize));
for(uint32_t component = 0; component < builtin.SizeInComponents; component++) for(uint32_t component = 0; component < builtin.SizeInComponents; component++)
{ {
...@@ -84,15 +86,13 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout ...@@ -84,15 +86,13 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout
} }
}); });
routine->setInputBuiltin(shader, spv::BuiltInNumSubgroups, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInNumSubgroups, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
ASSERT(builtin.SizeInComponents == 1); ASSERT(builtin.SizeInComponents == 1);
auto subgroupsPerWorkgroup = *Pointer<Int>(data + OFFSET(Data, subgroupsPerWorkgroup)); auto subgroupsPerWorkgroup = *Pointer<Int>(data + OFFSET(Data, subgroupsPerWorkgroup));
value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(subgroupsPerWorkgroup)); value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(subgroupsPerWorkgroup));
}); });
routine->setInputBuiltin(shader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
ASSERT(builtin.SizeInComponents == 1); ASSERT(builtin.SizeInComponents == 1);
auto invocationsPerSubgroup = *Pointer<Int>(data + OFFSET(Data, invocationsPerSubgroup)); auto invocationsPerSubgroup = *Pointer<Int>(data + OFFSET(Data, invocationsPerSubgroup));
value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(invocationsPerSubgroup)); value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(invocationsPerSubgroup));
...@@ -101,7 +101,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout ...@@ -101,7 +101,7 @@ void ComputeProgram::setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* rout
routine->setImmutableInputBuiltins(shader); routine->setImmutableInputBuiltins(shader);
} }
void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routine, Int workgroupID[3], SIMD::Int localInvocationIndex, Int subgroupIndex) void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine *routine, Int workgroupID[3], SIMD::Int localInvocationIndex, Int subgroupIndex)
{ {
Int4 numWorkgroups = *Pointer<Int4>(data + OFFSET(Data, numWorkgroups)); Int4 numWorkgroups = *Pointer<Int4>(data + OFFSET(Data, numWorkgroups));
Int4 workgroupSize = *Pointer<Int4>(data + OFFSET(Data, workgroupSize)); Int4 workgroupSize = *Pointer<Int4>(data + OFFSET(Data, workgroupSize));
...@@ -120,20 +120,17 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi ...@@ -120,20 +120,17 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi
localInvocationID[X] = idx; localInvocationID[X] = idx;
} }
routine->setInputBuiltin(shader, spv::BuiltInLocalInvocationIndex, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInLocalInvocationIndex, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
ASSERT(builtin.SizeInComponents == 1); ASSERT(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<SIMD::Float>(localInvocationIndex); value[builtin.FirstComponent] = As<SIMD::Float>(localInvocationIndex);
}); });
routine->setInputBuiltin(shader, spv::BuiltInSubgroupId, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInSubgroupId, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
ASSERT(builtin.SizeInComponents == 1); ASSERT(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(subgroupIndex)); value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(subgroupIndex));
}); });
routine->setInputBuiltin(shader, spv::BuiltInLocalInvocationId, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInLocalInvocationId, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
for(uint32_t component = 0; component < builtin.SizeInComponents; component++) for(uint32_t component = 0; component < builtin.SizeInComponents; component++)
{ {
value[builtin.FirstComponent + component] = value[builtin.FirstComponent + component] =
...@@ -141,8 +138,7 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi ...@@ -141,8 +138,7 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi
} }
}); });
routine->setInputBuiltin(shader, spv::BuiltInGlobalInvocationId, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine->setInputBuiltin(shader, spv::BuiltInGlobalInvocationId, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
SIMD::Int wgID = 0; SIMD::Int wgID = 0;
wgID = Insert(wgID, workgroupID[X], X); wgID = Insert(wgID, workgroupID[X], X);
wgID = Insert(wgID, workgroupID[Y], Y); wgID = Insert(wgID, workgroupID[Y], Y);
...@@ -156,7 +152,7 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi ...@@ -156,7 +152,7 @@ void ComputeProgram::setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routi
}); });
} }
void ComputeProgram::emit(SpirvRoutine* routine) void ComputeProgram::emit(SpirvRoutine *routine)
{ {
Pointer<Byte> data = Arg<0>(); Pointer<Byte> data = Arg<0>();
Int workgroupX = Arg<1>(); Int workgroupX = Arg<1>();
...@@ -174,7 +170,7 @@ void ComputeProgram::emit(SpirvRoutine* routine) ...@@ -174,7 +170,7 @@ void ComputeProgram::emit(SpirvRoutine* routine)
Int invocationsPerWorkgroup = *Pointer<Int>(data + OFFSET(Data, invocationsPerWorkgroup)); Int invocationsPerWorkgroup = *Pointer<Int>(data + OFFSET(Data, invocationsPerWorkgroup));
Int workgroupID[3] = {workgroupX, workgroupY, workgroupZ}; Int workgroupID[3] = { workgroupX, workgroupY, workgroupZ };
setWorkgroupBuiltins(data, routine, workgroupID); setWorkgroupBuiltins(data, routine, workgroupID);
For(Int i = 0, i < subgroupCount, i++) For(Int i = 0, i < subgroupCount, i++)
...@@ -231,8 +227,7 @@ void ComputeProgram::run( ...@@ -231,8 +227,7 @@ void ComputeProgram::run(
for(uint32_t batchID = 0; batchID < batchCount && batchID < groupCount; batchID++) for(uint32_t batchID = 0; batchID < batchCount && batchID < groupCount; batchID++)
{ {
wg.add(1); wg.add(1);
marl::schedule([=, &data] marl::schedule([=, &data] {
{
defer(wg.done()); defer(wg.done());
std::vector<uint8_t> workgroupMemory(shader->workgroupMemory.size()); std::vector<uint8_t> workgroupMemory(shader->workgroupMemory.size());
......
src/Pipeline/ComputeProgram.hpp
...@@ -17,13 +17,15 @@ ...@@ -17,13 +17,15 @@
#include "SpirvShader.hpp" #include "SpirvShader.hpp"
#include "Reactor/Coroutine.hpp"
#include "Device/Context.hpp" #include "Device/Context.hpp"
#include "Reactor/Coroutine.hpp"
#include "Vulkan/VkDescriptorSet.hpp" #include "Vulkan/VkDescriptorSet.hpp"
#include <functional> #include <functional>
namespace vk { class PipelineLayout; } namespace vk {
class PipelineLayout;
}
namespace sw { namespace sw {
...@@ -34,11 +36,11 @@ struct Constants; ...@@ -34,11 +36,11 @@ struct Constants;
// ComputeProgram builds a SPIR-V compute shader. // ComputeProgram builds a SPIR-V compute shader.
class ComputeProgram : public Coroutine<SpirvShader::YieldResult( class ComputeProgram : public Coroutine<SpirvShader::YieldResult(
void* data, void *data,
int32_t workgroupX, int32_t workgroupX,
int32_t workgroupY, int32_t workgroupY,
int32_t workgroupZ, int32_t workgroupZ,
void* workgroupMemory, void *workgroupMemory,
int32_t firstSubgroup, int32_t firstSubgroup,
int32_t subgroupCount)> int32_t subgroupCount)>
{ {
...@@ -59,9 +61,9 @@ public: ...@@ -59,9 +61,9 @@ public:
uint32_t groupCountX, uint32_t groupCountY, uint32_t groupCountZ); uint32_t groupCountX, uint32_t groupCountY, uint32_t groupCountZ);
protected: protected:
void emit(SpirvRoutine* routine); void emit(SpirvRoutine *routine);
void setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routine, Int workgroupID[3]); void setWorkgroupBuiltins(Pointer<Byte> data, SpirvRoutine *routine, Int workgroupID[3]);
void setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine* routine, Int workgroupID[3], SIMD::Int localInvocationIndex, Int subgroupIndex); void setSubgroupBuiltins(Pointer<Byte> data, SpirvRoutine *routine, Int workgroupID[3], SIMD::Int localInvocationIndex, Int subgroupIndex);
struct Data struct Data
{ {
...@@ -76,8 +78,8 @@ protected: ...@@ -76,8 +78,8 @@ protected:
const Constants *constants; const Constants *constants;
}; };
SpirvShader const * const shader; SpirvShader const *const shader;
vk::PipelineLayout const * const pipelineLayout; vk::PipelineLayout const *const pipelineLayout;
const vk::DescriptorSet::Bindings &descriptorSets; const vk::DescriptorSet::Bindings &descriptorSets;
}; };
......
src/Pipeline/Constants.hpp
...@@ -15,8 +15,8 @@ ...@@ -15,8 +15,8 @@
#ifndef sw_Constants_hpp #ifndef sw_Constants_hpp
#define sw_Constants_hpp #define sw_Constants_hpp
#include "System/Types.hpp"
#include "System/Math.hpp" #include "System/Math.hpp"
#include "System/Types.hpp"
#include "Vulkan/VkConfig.h" #include "Vulkan/VkConfig.h"
namespace sw { namespace sw {
...@@ -91,10 +91,10 @@ struct Constants ...@@ -91,10 +91,10 @@ struct Constants
// VK_SAMPLE_COUNT_4_BIT // VK_SAMPLE_COUNT_4_BIT
// https://www.khronos.org/registry/vulkan/specs/1.1/html/vkspec.html#primsrast-multisampling // https://www.khronos.org/registry/vulkan/specs/1.1/html/vkspec.html#primsrast-multisampling
static constexpr float VkSampleLocations4[][2] = { static constexpr float VkSampleLocations4[][2] = {
{0.375, 0.125}, { 0.375, 0.125 },
{0.875, 0.375}, { 0.875, 0.375 },
{0.125, 0.625}, { 0.125, 0.625 },
{0.625, 0.875}, { 0.625, 0.875 },
}; };
// Vulkan spec sample positions are relative to 0,0 in top left corner, with Y+ going down. // Vulkan spec sample positions are relative to 0,0 in top left corner, with Y+ going down.
...@@ -141,6 +141,6 @@ struct Constants ...@@ -141,6 +141,6 @@ struct Constants
extern Constants constants; extern Constants constants;
} // namepsace sw } // namespace sw
#endif // sw_Constants_hpp #endif // sw_Constants_hpp
src/Pipeline/PixelProgram.cpp
...@@ -56,48 +56,43 @@ Int4 PixelProgram::maskAny(Int cMask[4], Int sMask[4], Int zMask[4]) const ...@@ -56,48 +56,43 @@ Int4 PixelProgram::maskAny(Int cMask[4], Int sMask[4], Int zMask[4]) const
return mask; return mask;
} }
void PixelProgram::setBuiltins(Int &x, Int &y, Float4(&z)[4], Float4 &w, Int cMask[4]) void PixelProgram::setBuiltins(Int &x, Int &y, Float4 (&z)[4], Float4 &w, Int cMask[4])
{ {
routine.setImmutableInputBuiltins(spirvShader); routine.setImmutableInputBuiltins(spirvShader);
routine.setInputBuiltin(spirvShader, spv::BuiltInViewIndex, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInViewIndex, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 1); assert(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, viewID))))); value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, viewID)))));
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInFragCoord, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInFragCoord, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 4); assert(builtin.SizeInComponents == 4);
value[builtin.FirstComponent+0] = SIMD::Float(Float(x)) + SIMD::Float(0.5f, 1.5f, 0.5f, 1.5f); value[builtin.FirstComponent + 0] = SIMD::Float(Float(x)) + SIMD::Float(0.5f, 1.5f, 0.5f, 1.5f);
value[builtin.FirstComponent+1] = SIMD::Float(Float(y)) + SIMD::Float(0.5f, 0.5f, 1.5f, 1.5f); value[builtin.FirstComponent + 1] = SIMD::Float(Float(y)) + SIMD::Float(0.5f, 0.5f, 1.5f, 1.5f);
value[builtin.FirstComponent+2] = z[0]; // sample 0 value[builtin.FirstComponent + 2] = z[0]; // sample 0
value[builtin.FirstComponent+3] = w; value[builtin.FirstComponent + 3] = w;
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInPointCoord, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInPointCoord, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 2); assert(builtin.SizeInComponents == 2);
value[builtin.FirstComponent+0] = SIMD::Float(0.5f, 1.5f, 0.5f, 1.5f) + value[builtin.FirstComponent + 0] = SIMD::Float(0.5f, 1.5f, 0.5f, 1.5f) +
SIMD::Float(Float(x) - (*Pointer<Float>(primitive + OFFSET(Primitive, pointCoordX)))); SIMD::Float(Float(x) - (*Pointer<Float>(primitive + OFFSET(Primitive, pointCoordX))));
value[builtin.FirstComponent+1] = SIMD::Float(0.5f, 0.5f, 1.5f, 1.5f) + value[builtin.FirstComponent + 1] = SIMD::Float(0.5f, 0.5f, 1.5f, 1.5f) +
SIMD::Float(Float(y) - (*Pointer<Float>(primitive + OFFSET(Primitive, pointCoordY)))); SIMD::Float(Float(y) - (*Pointer<Float>(primitive + OFFSET(Primitive, pointCoordY))));
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 1); assert(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(SIMD::Width)); value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(SIMD::Width));
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInHelperInvocation, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInHelperInvocation, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 1); assert(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<SIMD::Float>(~maskAny(cMask)); value[builtin.FirstComponent] = As<SIMD::Float>(~maskAny(cMask));
}); });
routine.windowSpacePosition[0] = x + SIMD::Int(0,1,0,1); routine.windowSpacePosition[0] = x + SIMD::Int(0, 1, 0, 1);
routine.windowSpacePosition[1] = y + SIMD::Int(0,0,1,1); routine.windowSpacePosition[1] = y + SIMD::Int(0, 0, 1, 1);
routine.viewID = *Pointer<Int>(data + OFFSET(DrawData, viewID)); routine.viewID = *Pointer<Int>(data + OFFSET(DrawData, viewID));
} }
...@@ -169,7 +164,7 @@ void PixelProgram::applyShader(Int cMask[4], Int sMask[4], Int zMask[4]) ...@@ -169,7 +164,7 @@ void PixelProgram::applyShader(Int cMask[4], Int sMask[4], Int zMask[4])
for(auto i = 0u; i < state.multiSample; i++) for(auto i = 0u; i < state.multiSample; i++)
{ {
cMask[i] &= SignMask(CmpNEQ(outputSampleMask & SIMD::Int(1<<i), SIMD::Int(0))); cMask[i] &= SignMask(CmpNEQ(outputSampleMask & SIMD::Int(1 << i), SIMD::Int(0)));
} }
} }
...@@ -313,10 +308,14 @@ void PixelProgram::clampColor(Vector4f oC[RENDERTARGETS]) ...@@ -313,10 +308,14 @@ void PixelProgram::clampColor(Vector4f oC[RENDERTARGETS])
case VK_FORMAT_A8B8G8R8_UNORM_PACK32: case VK_FORMAT_A8B8G8R8_UNORM_PACK32:
case VK_FORMAT_A8B8G8R8_SRGB_PACK32: case VK_FORMAT_A8B8G8R8_SRGB_PACK32:
case VK_FORMAT_A2B10G10R10_UNORM_PACK32: case VK_FORMAT_A2B10G10R10_UNORM_PACK32:
oC[index].x = Max(oC[index].x, Float4(0.0f)); oC[index].x = Min(oC[index].x, Float4(1.0f)); oC[index].x = Max(oC[index].x, Float4(0.0f));
oC[index].y = Max(oC[index].y, Float4(0.0f)); oC[index].y = Min(oC[index].y, Float4(1.0f)); oC[index].x = Min(oC[index].x, Float4(1.0f));
oC[index].z = Max(oC[index].z, Float4(0.0f)); oC[index].z = Min(oC[index].z, Float4(1.0f)); oC[index].y = Max(oC[index].y, Float4(0.0f));
oC[index].w = Max(oC[index].w, Float4(0.0f)); oC[index].w = Min(oC[index].w, Float4(1.0f)); oC[index].y = Min(oC[index].y, Float4(1.0f));
oC[index].z = Max(oC[index].z, Float4(0.0f));
oC[index].z = Min(oC[index].z, Float4(1.0f));
oC[index].w = Max(oC[index].w, Float4(0.0f));
oC[index].w = Min(oC[index].w, Float4(1.0f));
break; break;
case VK_FORMAT_R32_SFLOAT: case VK_FORMAT_R32_SFLOAT:
case VK_FORMAT_R32G32_SFLOAT: case VK_FORMAT_R32G32_SFLOAT:
...@@ -361,4 +360,4 @@ Float4 PixelProgram::linearToSRGB(const Float4 &x) // Approximates x^(1.0/2.2) ...@@ -361,4 +360,4 @@ Float4 PixelProgram::linearToSRGB(const Float4 &x) // Approximates x^(1.0/2.2)
return Min(Max(sRGB, Float4(0.0f)), Float4(1.0f)); return Min(Max(sRGB, Float4(0.0f)), Float4(1.0f));
} }
} // namepsace sw } // namespace sw
src/Pipeline/PixelProgram.hpp
...@@ -26,15 +26,15 @@ public: ...@@ -26,15 +26,15 @@ public:
const PixelProcessor::State &state, const PixelProcessor::State &state,
vk::PipelineLayout const *pipelineLayout, vk::PipelineLayout const *pipelineLayout,
SpirvShader const *spirvShader, SpirvShader const *spirvShader,
const vk::DescriptorSet::Bindings &descriptorSets) : const vk::DescriptorSet::Bindings &descriptorSets)
PixelRoutine(state, pipelineLayout, spirvShader, descriptorSets) : PixelRoutine(state, pipelineLayout, spirvShader, descriptorSets)
{ {
} }
virtual ~PixelProgram() {} virtual ~PixelProgram() {}
protected: protected:
virtual void setBuiltins(Int &x, Int &y, Float4(&z)[4], Float4 &w, Int cMask[4]); virtual void setBuiltins(Int &x, Int &y, Float4 (&z)[4], Float4 &w, Int cMask[4]);
virtual void applyShader(Int cMask[4], Int sMask[4], Int zMask[4]); virtual void applyShader(Int cMask[4], Int sMask[4], Int zMask[4]);
virtual Bool alphaTest(Int cMask[4]); virtual Bool alphaTest(Int cMask[4]);
virtual void rasterOperation(Pointer<Byte> cBuffer[4], Int &x, Int sMask[4], Int zMask[4], Int cMask[4]); virtual void rasterOperation(Pointer<Byte> cBuffer[4], Int &x, Int sMask[4], Int zMask[4], Int cMask[4]);
......
src/Pipeline/PixelRoutine.hpp
...@@ -43,7 +43,7 @@ protected: ...@@ -43,7 +43,7 @@ protected:
// Depth output // Depth output
Float4 oDepth; Float4 oDepth;
virtual void setBuiltins(Int &x, Int &y, Float4(&z)[4], Float4 &w, Int cMask[4]) = 0; virtual void setBuiltins(Int &x, Int &y, Float4 (&z)[4], Float4 &w, Int cMask[4]) = 0;
virtual void applyShader(Int cMask[4], Int sMask[4], Int zMask[4]) = 0; virtual void applyShader(Int cMask[4], Int sMask[4], Int zMask[4]) = 0;
virtual Bool alphaTest(Int cMask[4]) = 0; virtual Bool alphaTest(Int cMask[4]) = 0;
virtual void rasterOperation(Pointer<Byte> cBuffer[4], Int &x, Int sMask[4], Int zMask[4], Int cMask[4]) = 0; virtual void rasterOperation(Pointer<Byte> cBuffer[4], Int &x, Int sMask[4], Int zMask[4], Int cMask[4]) = 0;
...@@ -55,7 +55,7 @@ protected: ...@@ -55,7 +55,7 @@ protected:
// Raster operations // Raster operations
void alphaBlend(int index, const Pointer<Byte> &cBuffer, Vector4s &current, const Int &x); void alphaBlend(int index, const Pointer<Byte> &cBuffer, Vector4s &current, const Int &x);
void writeColor(int index, const Pointer<Byte> &cBuffer, const Int& x, Vector4f& oC, const Int& sMask, const Int& zMask, const Int& cMask); void writeColor(int index, const Pointer<Byte> &cBuffer, const Int &x, Vector4f &oC, const Int &sMask, const Int &zMask, const Int &cMask);
void alphaBlend(int index, const Pointer<Byte> &cBuffer, Vector4f &oC, const Int &x); void alphaBlend(int index, const Pointer<Byte> &cBuffer, Vector4f &oC, const Int &x);
void writeColor(int index, const Pointer<Byte> &cBuffer, const Int &x, Vector4s &current, const Int &sMask, const Int &zMask, const Int &cMask); void writeColor(int index, const Pointer<Byte> &cBuffer, const Int &x, Vector4s &current, const Int &sMask, const Int &zMask, const Int &cMask);
......
src/Pipeline/SamplerCore.hpp
...@@ -21,7 +21,7 @@ ...@@ -21,7 +21,7 @@
#include "Reactor/Reactor.hpp" #include "Reactor/Reactor.hpp"
#ifdef None #ifdef None
#undef None // b/127920555 # undef None // b/127920555
#endif #endif
namespace sw { namespace sw {
...@@ -45,7 +45,9 @@ enum SamplerMethod : uint32_t ...@@ -45,7 +45,9 @@ enum SamplerMethod : uint32_t
struct SamplerFunction struct SamplerFunction
{ {
SamplerFunction(SamplerMethod method, bool offset = false, bool sample = false) SamplerFunction(SamplerMethod method, bool offset = false, bool sample = false)
: method(method), offset(offset), sample(sample) : method(method)
, offset(offset)
, sample(sample)
{} {}
operator SamplerMethod() { return method; } operator SamplerMethod() { return method; }
...@@ -60,20 +62,20 @@ class SamplerCore ...@@ -60,20 +62,20 @@ class SamplerCore
public: public:
SamplerCore(Pointer<Byte> &constants, const Sampler &state); SamplerCore(Pointer<Byte> &constants, const Sampler &state);
Vector4f sampleTexture(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float4 uvw[4], Float4 &q, Float &&lodOrBias, Float4 &dsx, Float4 &dsy, Vector4f &offset, Int4& sampleId, SamplerFunction function); Vector4f sampleTexture(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float4 uvw[4], Float4 &q, Float &&lodOrBias, Float4 &dsx, Float4 &dsy, Vector4f &offset, Int4 &sampleId, SamplerFunction function);
private: private:
Short4 offsetSample(Short4 &uvw, Pointer<Byte> &mipmap, int halfOffset, bool wrap, int count, Float &lod); Short4 offsetSample(Short4 &uvw, Pointer<Byte> &mipmap, int halfOffset, bool wrap, int count, Float &lod);
Vector4s sampleFilter(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, SamplerFunction function); Vector4s sampleFilter(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, SamplerFunction function);
Vector4s sampleAniso(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, bool secondLOD, SamplerFunction function); Vector4s sampleAniso(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, bool secondLOD, SamplerFunction function);
Vector4s sampleQuad(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4s sampleQuad(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Vector4s sampleQuad2D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4s sampleQuad2D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Vector4s sample3D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4s sample3D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Vector4f sampleFloatFilter(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, SamplerFunction function); Vector4f sampleFloatFilter(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, SamplerFunction function);
Vector4f sampleFloatAniso(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, bool secondLOD, SamplerFunction function); Vector4f sampleFloatAniso(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, bool secondLOD, SamplerFunction function);
Vector4f sampleFloat(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4f sampleFloat(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Vector4f sampleFloat2D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4f sampleFloat2D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Float4 &q, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Vector4f sampleFloat3D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4& cubeArrayCoord, const Int4& sampleId, Float &lod, bool secondLOD, SamplerFunction function); Vector4f sampleFloat3D(Pointer<Byte> &texture, Float4 &u, Float4 &v, Float4 &w, Vector4f &offset, const Float4 &cubeArrayCoord, const Int4 &sampleId, Float &lod, bool secondLOD, SamplerFunction function);
Float log2sqrt(Float lod); Float log2sqrt(Float lod);
Float log2(Float lod); Float log2(Float lod);
void computeLod(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, Float4 &u, Float4 &v, Float4 &dsx, Float4 &dsy, SamplerFunction function); void computeLod(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float &lod, Float &anisotropy, Float4 &uDelta, Float4 &vDelta, Float4 &u, Float4 &v, Float4 &dsx, Float4 &dsy, SamplerFunction function);
...@@ -81,15 +83,15 @@ private: ...@@ -81,15 +83,15 @@ private:
void computeLod3D(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float &lod, Float4 &u, Float4 &v, Float4 &w, Float4 &dsx, Float4 &dsy, SamplerFunction function); void computeLod3D(Pointer<Byte> &texture, Pointer<Byte> &sampler, Float &lod, Float4 &u, Float4 &v, Float4 &w, Float4 &dsx, Float4 &dsy, SamplerFunction function);
Int4 cubeFace(Float4 &U, Float4 &V, Float4 &x, Float4 &y, Float4 &z, Float4 &M); Int4 cubeFace(Float4 &U, Float4 &V, Float4 &x, Float4 &y, Float4 &z, Float4 &M);
Short4 applyOffset(Short4 &uvw, Float4 &offset, const Int4 &whd, AddressingMode mode); Short4 applyOffset(Short4 &uvw, Float4 &offset, const Int4 &whd, AddressingMode mode);
void computeIndices(UInt index[4], Short4 uuuu, Short4 vvvv, Short4 wwww, Vector4f &offset, const Pointer<Byte> &mipmap, const Short4& cubeArrayId, const Int4& sampleId, SamplerFunction function); void computeIndices(UInt index[4], Short4 uuuu, Short4 vvvv, Short4 wwww, Vector4f &offset, const Pointer<Byte> &mipmap, const Short4 &cubeArrayId, const Int4 &sampleId, SamplerFunction function);
void computeIndices(UInt index[4], Int4 uuuu, Int4 vvvv, Int4 wwww, Int4 valid, const Pointer<Byte> &mipmap, const Int4& cubeArrayId, const Int4& sampleId, SamplerFunction function); void computeIndices(UInt index[4], Int4 uuuu, Int4 vvvv, Int4 wwww, Int4 valid, const Pointer<Byte> &mipmap, const Int4 &cubeArrayId, const Int4 &sampleId, SamplerFunction function);
Vector4s sampleTexel(Short4 &u, Short4 &v, Short4 &s, Vector4f &offset, Pointer<Byte> &mipmap, const Short4& cubeArrayId, const Int4& sampleId, Pointer<Byte> buffer, SamplerFunction function); Vector4s sampleTexel(Short4 &u, Short4 &v, Short4 &s, Vector4f &offset, Pointer<Byte> &mipmap, const Short4 &cubeArrayId, const Int4 &sampleId, Pointer<Byte> buffer, SamplerFunction function);
Vector4s sampleTexel(UInt index[4], Pointer<Byte> buffer); Vector4s sampleTexel(UInt index[4], Pointer<Byte> buffer);
Vector4f sampleTexel(Int4 &u, Int4 &v, Int4 &s, Float4 &z, Pointer<Byte> &mipmap, const Int4& cubeArrayId, const Int4& sampleId, Pointer<Byte> buffer, SamplerFunction function); Vector4f sampleTexel(Int4 &u, Int4 &v, Int4 &s, Float4 &z, Pointer<Byte> &mipmap, const Int4 &cubeArrayId, const Int4 &sampleId, Pointer<Byte> buffer, SamplerFunction function);
Vector4f replaceBorderTexel(const Vector4f &c, Int4 valid); Vector4f replaceBorderTexel(const Vector4f &c, Int4 valid);
void selectMipmap(const Pointer<Byte> &texture, Pointer<Byte> &mipmap, Pointer<Byte> &buffer, const Float &lod, bool secondLOD); void selectMipmap(const Pointer<Byte> &texture, Pointer<Byte> &mipmap, Pointer<Byte> &buffer, const Float &lod, bool secondLOD);
Short4 address(const Float4 &uw, AddressingMode addressingMode, Pointer<Byte>& mipmap); Short4 address(const Float4 &uw, AddressingMode addressingMode, Pointer<Byte> &mipmap);
void address(const Float4 &uw, Int4& xyz0, Int4& xyz1, Float4& f, Pointer<Byte>& mipmap, Float4 &texOffset, Int4 &filter, int whd, AddressingMode addressingMode, SamplerFunction function); void address(const Float4 &uw, Int4 &xyz0, Int4 &xyz1, Float4 &f, Pointer<Byte> &mipmap, Float4 &texOffset, Int4 &filter, int whd, AddressingMode addressingMode, SamplerFunction function);
Int4 computeFilterOffset(Float &lod); Int4 computeFilterOffset(Float &lod);
void convertSigned15(Float4 &cf, Short4 &ci); void convertSigned15(Float4 &cf, Short4 &ci);
...@@ -120,9 +122,10 @@ private: ...@@ -120,9 +122,10 @@ private:
#ifdef ENABLE_RR_PRINT #ifdef ENABLE_RR_PRINT
namespace rr { namespace rr {
template <> struct PrintValue::Ty<sw::SamplerFunction> template<>
struct PrintValue::Ty<sw::SamplerFunction>
{ {
static std::string fmt(const sw::SamplerFunction& v) static std::string fmt(const sw::SamplerFunction &v)
{ {
return std::string("SamplerFunction[") + return std::string("SamplerFunction[") +
"method: " + std::to_string(v.method) + "method: " + std::to_string(v.method) +
...@@ -131,7 +134,7 @@ template <> struct PrintValue::Ty<sw::SamplerFunction> ...@@ -131,7 +134,7 @@ template <> struct PrintValue::Ty<sw::SamplerFunction>
"]"; "]";
} }
static std::vector<rr::Value*> val(const sw::SamplerFunction& v) { return {}; } static std::vector<rr::Value *> val(const sw::SamplerFunction &v) { return {}; }
}; };
} // namespace rr } // namespace rr
......
src/Pipeline/ShaderCore.hpp
...@@ -19,8 +19,8 @@ ...@@ -19,8 +19,8 @@
#include "Reactor/Reactor.hpp" #include "Reactor/Reactor.hpp"
#include "Vulkan/VkDebug.hpp" #include "Vulkan/VkDebug.hpp"
#include <atomic> // std::memory_order
#include <array> #include <array>
#include <atomic> // std::memory_order
#include <utility> // std::pair #include <utility> // std::pair
namespace sw { namespace sw {
...@@ -87,17 +87,17 @@ struct Pointer ...@@ -87,17 +87,17 @@ struct Pointer
Pointer(rr::Pointer<Byte> base, rr::Int limit, SIMD::Int offset); Pointer(rr::Pointer<Byte> base, rr::Int limit, SIMD::Int offset);
Pointer(rr::Pointer<Byte> base, unsigned int limit, SIMD::Int offset); Pointer(rr::Pointer<Byte> base, unsigned int limit, SIMD::Int offset);
Pointer& operator += (Int i); Pointer &operator+=(Int i);
Pointer& operator *= (Int i); Pointer &operator*=(Int i);
Pointer operator + (SIMD::Int i); Pointer operator+(SIMD::Int i);
Pointer operator * (SIMD::Int i); Pointer operator*(SIMD::Int i);
Pointer& operator += (int i); Pointer &operator+=(int i);
Pointer& operator *= (int i); Pointer &operator*=(int i);
Pointer operator + (int i); Pointer operator+(int i);
Pointer operator * (int i); Pointer operator*(int i);
SIMD::Int offsets() const; SIMD::Int offsets() const;
...@@ -146,10 +146,24 @@ struct Pointer ...@@ -146,10 +146,24 @@ struct Pointer
bool hasDynamicOffsets; // True if any dynamicOffsets are non-zero. bool hasDynamicOffsets; // True if any dynamicOffsets are non-zero.
}; };
template <typename T> struct Element {}; template<typename T>
template <> struct Element<Float> { using type = rr::Float; }; struct Element
template <> struct Element<Int> { using type = rr::Int; }; {};
template <> struct Element<UInt> { using type = rr::UInt; }; template<>
struct Element<Float>
{
using type = rr::Float;
};
template<>
struct Element<Int>
{
using type = rr::Int;
};
template<>
struct Element<UInt>
{
using type = rr::UInt;
};
} // namespace SIMD } // namespace SIMD
...@@ -199,10 +213,10 @@ rr::RValue<rr::Bool> AnyTrue(rr::RValue<sw::SIMD::Int> const &ints); ...@@ -199,10 +213,10 @@ rr::RValue<rr::Bool> AnyTrue(rr::RValue<sw::SIMD::Int> const &ints);
rr::RValue<rr::Bool> AnyFalse(rr::RValue<sw::SIMD::Int> const &ints); rr::RValue<rr::Bool> AnyFalse(rr::RValue<sw::SIMD::Int> const &ints);
template <typename T> template<typename T>
inline rr::RValue<T> AndAll(rr::RValue<T> const &mask); inline rr::RValue<T> AndAll(rr::RValue<T> const &mask);
template <typename T> template<typename T>
inline rr::RValue<T> OrAll(rr::RValue<T> const &mask); inline rr::RValue<T> OrAll(rr::RValue<T> const &mask);
rr::RValue<sw::SIMD::Float> Sign(rr::RValue<sw::SIMD::Float> const &val); rr::RValue<sw::SIMD::Float> Sign(rr::RValue<sw::SIMD::Float> const &val);
...@@ -476,7 +490,7 @@ inline void SIMD::Pointer::Store(RValue<T> val, OutOfBoundsBehavior robustness, ...@@ -476,7 +490,7 @@ inline void SIMD::Pointer::Store(RValue<T> val, OutOfBoundsBehavior robustness,
Store(T(val), robustness, mask, atomic, order); Store(T(val), robustness, mask, atomic, order);
} }
template <typename T> template<typename T>
inline rr::RValue<T> AndAll(rr::RValue<T> const &mask) inline rr::RValue<T> AndAll(rr::RValue<T> const &mask)
{ {
T v1 = mask; // [x] [y] [z] [w] T v1 = mask; // [x] [y] [z] [w]
...@@ -484,7 +498,7 @@ inline rr::RValue<T> AndAll(rr::RValue<T> const &mask) ...@@ -484,7 +498,7 @@ inline rr::RValue<T> AndAll(rr::RValue<T> const &mask)
return v2.xxxx & v2.yyyy; // [xyzw] [xyzw] [xyzw] [xyzw] return v2.xxxx & v2.yyyy; // [xyzw] [xyzw] [xyzw] [xyzw]
} }
template <typename T> template<typename T>
inline rr::RValue<T> OrAll(rr::RValue<T> const &mask) inline rr::RValue<T> OrAll(rr::RValue<T> const &mask)
{ {
T v1 = mask; // [x] [y] [z] [w] T v1 = mask; // [x] [y] [z] [w]
...@@ -496,9 +510,10 @@ inline rr::RValue<T> OrAll(rr::RValue<T> const &mask) ...@@ -496,9 +510,10 @@ inline rr::RValue<T> OrAll(rr::RValue<T> const &mask)
#ifdef ENABLE_RR_PRINT #ifdef ENABLE_RR_PRINT
namespace rr { namespace rr {
template <> struct PrintValue::Ty<sw::Vector4f> template<>
struct PrintValue::Ty<sw::Vector4f>
{ {
static std::string fmt(const sw::Vector4f& v) static std::string fmt(const sw::Vector4f &v)
{ {
return "[x: " + PrintValue::fmt(v.x) + return "[x: " + PrintValue::fmt(v.x) +
", y: " + PrintValue::fmt(v.y) + ", y: " + PrintValue::fmt(v.y) +
...@@ -506,14 +521,15 @@ template <> struct PrintValue::Ty<sw::Vector4f> ...@@ -506,14 +521,15 @@ template <> struct PrintValue::Ty<sw::Vector4f>
", w: " + PrintValue::fmt(v.w) + "]"; ", w: " + PrintValue::fmt(v.w) + "]";
} }
static std::vector<rr::Value*> val(const sw::Vector4f& v) static std::vector<rr::Value *> val(const sw::Vector4f &v)
{ {
return PrintValue::vals(v.x, v.y, v.z, v.w); return PrintValue::vals(v.x, v.y, v.z, v.w);
} }
}; };
template <> struct PrintValue::Ty<sw::Vector4s> template<>
struct PrintValue::Ty<sw::Vector4s>
{ {
static std::string fmt(const sw::Vector4s& v) static std::string fmt(const sw::Vector4s &v)
{ {
return "[x: " + PrintValue::fmt(v.x) + return "[x: " + PrintValue::fmt(v.x) +
", y: " + PrintValue::fmt(v.y) + ", y: " + PrintValue::fmt(v.y) +
...@@ -521,13 +537,13 @@ template <> struct PrintValue::Ty<sw::Vector4s> ...@@ -521,13 +537,13 @@ template <> struct PrintValue::Ty<sw::Vector4s>
", w: " + PrintValue::fmt(v.w) + "]"; ", w: " + PrintValue::fmt(v.w) + "]";
} }
static std::vector<rr::Value*> val(const sw::Vector4s& v) static std::vector<rr::Value *> val(const sw::Vector4s &v)
{ {
return PrintValue::vals(v.x, v.y, v.z, v.w); return PrintValue::vals(v.x, v.y, v.z, v.w);
} }
}; };
} // namespace sw } // namespace rr
#endif // ENABLE_RR_PRINT #endif // ENABLE_RR_PRINT
#endif // sw_ShaderCore_hpp #endif // sw_ShaderCore_hpp
src/Pipeline/SpirvID.hpp
...@@ -15,8 +15,8 @@ ...@@ -15,8 +15,8 @@
#ifndef sw_ID_hpp #ifndef sw_ID_hpp
#define sw_ID_hpp #define sw_ID_hpp
#include <unordered_map>
#include <cstdint> #include <cstdint>
#include <unordered_map>
namespace sw { namespace sw {
...@@ -25,39 +25,43 @@ namespace sw { ...@@ -25,39 +25,43 @@ namespace sw {
// ID; instead it is used to prevent implicit casts between identifiers of // ID; instead it is used to prevent implicit casts between identifiers of
// different T types. // different T types.
// IDs are typically used as a map key to value of type T. // IDs are typically used as a map key to value of type T.
template <typename T> template<typename T>
class SpirvID class SpirvID
{ {
public: public:
SpirvID() : id(0) {} SpirvID()
SpirvID(uint32_t id) : id(id) {} : id(0)
bool operator == (const SpirvID<T>& rhs) const { return id == rhs.id; } {}
bool operator != (const SpirvID<T>& rhs) const { return id != rhs.id; } SpirvID(uint32_t id)
bool operator < (const SpirvID<T>& rhs) const { return id < rhs.id; } : id(id)
{}
bool operator==(const SpirvID<T> &rhs) const { return id == rhs.id; }
bool operator!=(const SpirvID<T> &rhs) const { return id != rhs.id; }
bool operator<(const SpirvID<T> &rhs) const { return id < rhs.id; }
// value returns the numerical value of the identifier. // value returns the numerical value of the identifier.
uint32_t value() const { return id; } uint32_t value() const { return id; }
private: private:
uint32_t id; uint32_t id;
}; };
// HandleMap<T> is an unordered map of SpirvID<T> to T. // HandleMap<T> is an unordered map of SpirvID<T> to T.
template <typename T> template<typename T>
using HandleMap = std::unordered_map<SpirvID<T>, T>; using HandleMap = std::unordered_map<SpirvID<T>, T>;
} } // namespace sw
namespace std namespace std {
{
// std::hash implementation for sw::SpirvID<T> // std::hash implementation for sw::SpirvID<T>
template<typename T> template<typename T>
struct hash< sw::SpirvID<T> > struct hash<sw::SpirvID<T> >
{ {
std::size_t operator()(const sw::SpirvID<T>& id) const noexcept std::size_t operator()(const sw::SpirvID<T> &id) const noexcept
{ {
return std::hash<uint32_t>()(id.value()); return std::hash<uint32_t>()(id.value());
} }
}; };
} // namespace sw } // namespace std
#endif // sw_ID_hpp #endif // sw_ID_hpp
src/Pipeline/SpirvShaderArithmetic.cpp
...@@ -254,8 +254,7 @@ SpirvShader::EmitResult SpirvShader::EmitUnaryOp(InsnIterator insn, EmitState *s ...@@ -254,8 +254,7 @@ SpirvShader::EmitResult SpirvShader::EmitUnaryOp(InsnIterator insn, EmitState *s
break; break;
case spv::OpFwidth: case spv::OpFwidth:
case spv::OpFwidthCoarse: case spv::OpFwidthCoarse:
dst.move(i, SIMD::Float(Abs(Extract(src.Float(i), 1) - Extract(src.Float(i), 0)) dst.move(i, SIMD::Float(Abs(Extract(src.Float(i), 1) - Extract(src.Float(i), 0)) + Abs(Extract(src.Float(i), 2) - Extract(src.Float(i), 0))));
+ Abs(Extract(src.Float(i), 2) - Extract(src.Float(i), 0))));
break; break;
case spv::OpDPdxFine: case spv::OpDPdxFine:
{ {
...@@ -532,7 +531,7 @@ SpirvShader::EmitResult SpirvShader::EmitDot(InsnIterator insn, EmitState *state ...@@ -532,7 +531,7 @@ SpirvShader::EmitResult SpirvShader::EmitDot(InsnIterator insn, EmitState *state
return EmitResult::Continue; return EmitResult::Continue;
} }
SIMD::Float SpirvShader::Dot(unsigned numComponents, GenericValue const & x, GenericValue const & y) const SIMD::Float SpirvShader::Dot(unsigned numComponents, GenericValue const &x, GenericValue const &y) const
{ {
SIMD::Float d = x.Float(0) * y.Float(0); SIMD::Float d = x.Float(0) * y.Float(0);
......
src/Pipeline/SpirvShaderControlFlow.cpp
...@@ -24,7 +24,9 @@ ...@@ -24,7 +24,9 @@
namespace sw { namespace sw {
SpirvShader::Block::Block(InsnIterator begin, InsnIterator end) : begin_(begin), end_(end) SpirvShader::Block::Block(InsnIterator begin, InsnIterator end)
: begin_(begin)
, end_(end)
{ {
// Default to a Simple, this may change later. // Default to a Simple, this may change later.
kind = Block::Simple; kind = Block::Simple;
...@@ -111,7 +113,7 @@ SpirvShader::Block::Block(InsnIterator begin, InsnIterator end) : begin_(begin), ...@@ -111,7 +113,7 @@ SpirvShader::Block::Block(InsnIterator begin, InsnIterator end) : begin_(begin),
} }
} }
void SpirvShader::Function::TraverseReachableBlocks(Block::ID id, SpirvShader::Block::Set& reachable) const void SpirvShader::Function::TraverseReachableBlocks(Block::ID id, SpirvShader::Block::Set &reachable) const
{ {
if(reachable.count(id) == 0) if(reachable.count(id) == 0)
{ {
...@@ -196,7 +198,7 @@ void SpirvShader::EmitState::addOutputActiveLaneMaskEdge(Block::ID to, RValue<SI ...@@ -196,7 +198,7 @@ void SpirvShader::EmitState::addOutputActiveLaneMaskEdge(Block::ID to, RValue<SI
void SpirvShader::EmitState::addActiveLaneMaskEdge(Block::ID from, Block::ID to, RValue<SIMD::Int> mask) void SpirvShader::EmitState::addActiveLaneMaskEdge(Block::ID from, Block::ID to, RValue<SIMD::Int> mask)
{ {
auto edge = Block::Edge{from, to}; auto edge = Block::Edge{ from, to };
auto it = edgeActiveLaneMasks.find(edge); auto it = edgeActiveLaneMasks.find(edge);
if(it == edgeActiveLaneMasks.end()) if(it == edgeActiveLaneMasks.end())
{ {
...@@ -212,7 +214,7 @@ void SpirvShader::EmitState::addActiveLaneMaskEdge(Block::ID from, Block::ID to, ...@@ -212,7 +214,7 @@ void SpirvShader::EmitState::addActiveLaneMaskEdge(Block::ID from, Block::ID to,
RValue<SIMD::Int> SpirvShader::GetActiveLaneMaskEdge(EmitState *state, Block::ID from, Block::ID to) const RValue<SIMD::Int> SpirvShader::GetActiveLaneMaskEdge(EmitState *state, Block::ID from, Block::ID to) const
{ {
auto edge = Block::Edge{from, to}; auto edge = Block::Edge{ from, to };
auto it = state->edgeActiveLaneMasks.find(edge); auto it = state->edgeActiveLaneMasks.find(edge);
ASSERT_MSG(it != state->edgeActiveLaneMasks.end(), "Could not find edge %d -> %d", from.value(), to.value()); ASSERT_MSG(it != state->edgeActiveLaneMasks.end(), "Could not find edge %d -> %d", from.value(), to.value());
return it->second; return it->second;
...@@ -239,8 +241,7 @@ void SpirvShader::EmitBlocks(Block::ID id, EmitState *state, Block::ID ignore /* ...@@ -239,8 +241,7 @@ void SpirvShader::EmitBlocks(Block::ID id, EmitState *state, Block::ID ignore /*
// Ensure all dependency blocks have been generated. // Ensure all dependency blocks have been generated.
auto depsDone = true; auto depsDone = true;
function.ForeachBlockDependency(id, [&](Block::ID dep) function.ForeachBlockDependency(id, [&](Block::ID dep) {
{
if(state->visited.count(dep) == 0) if(state->visited.count(dep) == 0)
{ {
state->pending->push_front(dep); state->pending->push_front(dep);
...@@ -414,7 +415,7 @@ void SpirvShader::EmitLoop(EmitState *state) const ...@@ -414,7 +415,7 @@ void SpirvShader::EmitLoop(EmitState *state) const
// Add active lanes to the merge lane mask. // Add active lanes to the merge lane mask.
for(auto in : function.getBlock(mergeBlockId).ins) for(auto in : function.getBlock(mergeBlockId).ins)
{ {
auto edge = Block::Edge{in, mergeBlockId}; auto edge = Block::Edge{ in, mergeBlockId };
auto it = state->edgeActiveLaneMasks.find(edge); auto it = state->edgeActiveLaneMasks.find(edge);
if(it != state->edgeActiveLaneMasks.end()) if(it != state->edgeActiveLaneMasks.end())
{ {
...@@ -563,9 +564,9 @@ SpirvShader::EmitResult SpirvShader::EmitKill(InsnIterator insn, EmitState *stat ...@@ -563,9 +564,9 @@ SpirvShader::EmitResult SpirvShader::EmitKill(InsnIterator insn, EmitState *stat
SpirvShader::EmitResult SpirvShader::EmitFunctionCall(InsnIterator insn, EmitState *state) const SpirvShader::EmitResult SpirvShader::EmitFunctionCall(InsnIterator insn, EmitState *state) const
{ {
auto functionId = Function::ID(insn.word(3)); auto functionId = Function::ID(insn.word(3));
const auto& functionIt = functions.find(functionId); const auto &functionIt = functions.find(functionId);
ASSERT(functionIt != functions.end()); ASSERT(functionIt != functions.end());
auto& function = functionIt->second; auto &function = functionIt->second;
// TODO(b/141246700): Add full support for spv::OpFunctionCall // TODO(b/141246700): Add full support for spv::OpFunctionCall
// The only supported function is a single OpKill wrapped in a // The only supported function is a single OpKill wrapped in a
...@@ -654,7 +655,7 @@ void SpirvShader::LoadPhi(InsnIterator insn, EmitState *state) const ...@@ -654,7 +655,7 @@ void SpirvShader::LoadPhi(InsnIterator insn, EmitState *state) const
} }
} }
void SpirvShader::StorePhi(Block::ID currentBlock, InsnIterator insn, EmitState *state, std::unordered_set<SpirvShader::Block::ID> const& filter) const void SpirvShader::StorePhi(Block::ID currentBlock, InsnIterator insn, EmitState *state, std::unordered_set<SpirvShader::Block::ID> const &filter) const
{ {
auto typeId = Type::ID(insn.word(1)); auto typeId = Type::ID(insn.word(1));
auto type = getType(typeId); auto type = getType(typeId);
......
src/Pipeline/SpirvShaderEnumNames.cpp
...@@ -14,15 +14,15 @@ ...@@ -14,15 +14,15 @@
// This file contains code used to aid debugging. // This file contains code used to aid debugging.
#include <spirv/unified1/spirv.h>
#include "SpirvShader.hpp" #include "SpirvShader.hpp"
#include <spirv/unified1/spirv.h>
// Prototypes for SPIRV-Tools functions that do not have public headers. // Prototypes for SPIRV-Tools functions that do not have public headers.
// This is a C++ function, so the name is mangled, and signature changes will // This is a C++ function, so the name is mangled, and signature changes will
// result in a linker error instead of runtime signature mismatches. // result in a linker error instead of runtime signature mismatches.
// Gets the name of an instruction, without the "Op" prefix. // Gets the name of an instruction, without the "Op" prefix.
extern const char* spvOpcodeString(const SpvOp opcode); extern const char *spvOpcodeString(const SpvOp opcode);
namespace sw { namespace sw {
......
src/Pipeline/SpirvShaderGLSLstd450.cpp
...@@ -16,12 +16,11 @@ ...@@ -16,12 +16,11 @@
#include "ShaderCore.hpp" #include "ShaderCore.hpp"
#include <spirv/unified1/spirv.hpp>
#include <spirv/unified1/GLSL.std.450.h> #include <spirv/unified1/GLSL.std.450.h>
#include <spirv/unified1/spirv.hpp>
namespace namespace {
{ constexpr float PI = 3.141592653589793f;
constexpr float PI = 3.141592653589793f;
} }
namespace sw { namespace sw {
...@@ -196,7 +195,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -196,7 +195,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
for(auto i = 0u; i < type.sizeInComponents; i++) for(auto i = 0u; i < type.sizeInComponents; i++)
{ {
auto tx = Min(Max((x.Float(i) - edge0.Float(i)) / auto tx = Min(Max((x.Float(i) - edge0.Float(i)) /
(edge1.Float(i) - edge0.Float(i)), SIMD::Float(0.0f)), SIMD::Float(1.0f)); (edge1.Float(i) - edge0.Float(i)),
SIMD::Float(0.0f)),
SIMD::Float(1.0f));
dst.move(i, tx * tx * (Float4(3.0f) - Float4(2.0f) * tx)); dst.move(i, tx * tx * (Float4(3.0f) - Float4(2.0f) * tx));
} }
break; break;
...@@ -394,11 +395,14 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -394,11 +395,14 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
dst.move(0, (SIMD::Int(Round(Min(Max(val.Float(0), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) & dst.move(0, (SIMD::Int(Round(Min(Max(val.Float(0), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) &
SIMD::Int(0xFF)) | SIMD::Int(0xFF)) |
((SIMD::Int(Round(Min(Max(val.Float(1), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) & ((SIMD::Int(Round(Min(Max(val.Float(1), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) &
SIMD::Int(0xFF)) << 8) | SIMD::Int(0xFF))
<< 8) |
((SIMD::Int(Round(Min(Max(val.Float(2), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) & ((SIMD::Int(Round(Min(Max(val.Float(2), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) &
SIMD::Int(0xFF)) << 16) | SIMD::Int(0xFF))
<< 16) |
((SIMD::Int(Round(Min(Max(val.Float(3), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) & ((SIMD::Int(Round(Min(Max(val.Float(3), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(127.0f))) &
SIMD::Int(0xFF)) << 24)); SIMD::Int(0xFF))
<< 24));
break; break;
} }
case GLSLstd450PackUnorm4x8: case GLSLstd450PackUnorm4x8:
...@@ -416,7 +420,8 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -416,7 +420,8 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
dst.move(0, (SIMD::Int(Round(Min(Max(val.Float(0), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(32767.0f))) & dst.move(0, (SIMD::Int(Round(Min(Max(val.Float(0), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(32767.0f))) &
SIMD::Int(0xFFFF)) | SIMD::Int(0xFFFF)) |
((SIMD::Int(Round(Min(Max(val.Float(1), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(32767.0f))) & ((SIMD::Int(Round(Min(Max(val.Float(1), SIMD::Float(-1.0f)), SIMD::Float(1.0f)) * SIMD::Float(32767.0f))) &
SIMD::Int(0xFFFF)) << 16)); SIMD::Int(0xFFFF))
<< 16));
break; break;
} }
case GLSLstd450PackUnorm2x16: case GLSLstd450PackUnorm2x16:
...@@ -425,7 +430,8 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -425,7 +430,8 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
dst.move(0, (SIMD::UInt(Round(Min(Max(val.Float(0), SIMD::Float(0.0f)), SIMD::Float(1.0f)) * SIMD::Float(65535.0f))) & dst.move(0, (SIMD::UInt(Round(Min(Max(val.Float(0), SIMD::Float(0.0f)), SIMD::Float(1.0f)) * SIMD::Float(65535.0f))) &
SIMD::UInt(0xFFFF)) | SIMD::UInt(0xFFFF)) |
((SIMD::UInt(Round(Min(Max(val.Float(1), SIMD::Float(0.0f)), SIMD::Float(1.0f)) * SIMD::Float(65535.0f))) & ((SIMD::UInt(Round(Min(Max(val.Float(1), SIMD::Float(0.0f)), SIMD::Float(1.0f)) * SIMD::Float(65535.0f))) &
SIMD::UInt(0xFFFF)) << 16)); SIMD::UInt(0xFFFF))
<< 16));
break; break;
} }
case GLSLstd450PackHalf2x16: case GLSLstd450PackHalf2x16:
...@@ -437,9 +443,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -437,9 +443,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
case GLSLstd450UnpackSnorm4x8: case GLSLstd450UnpackSnorm4x8:
{ {
auto val = GenericValue(this, state, insn.word(5)); auto val = GenericValue(this, state, insn.word(5));
dst.move(0, Min(Max(SIMD::Float(((val.Int(0)<<24) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f))); dst.move(0, Min(Max(SIMD::Float(((val.Int(0) << 24) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f)));
dst.move(1, Min(Max(SIMD::Float(((val.Int(0)<<16) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f))); dst.move(1, Min(Max(SIMD::Float(((val.Int(0) << 16) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f)));
dst.move(2, Min(Max(SIMD::Float(((val.Int(0)<<8) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f))); dst.move(2, Min(Max(SIMD::Float(((val.Int(0) << 8) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f)));
dst.move(3, Min(Max(SIMD::Float(((val.Int(0)) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f))); dst.move(3, Min(Max(SIMD::Float(((val.Int(0)) & SIMD::Int(0xFF000000))) * SIMD::Float(1.0f / float(0x7f000000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f)));
break; break;
} }
...@@ -447,9 +453,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -447,9 +453,9 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
{ {
auto val = GenericValue(this, state, insn.word(5)); auto val = GenericValue(this, state, insn.word(5));
dst.move(0, SIMD::Float((val.UInt(0) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f)); dst.move(0, SIMD::Float((val.UInt(0) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f));
dst.move(1, SIMD::Float(((val.UInt(0)>>8) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f)); dst.move(1, SIMD::Float(((val.UInt(0) >> 8) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f));
dst.move(2, SIMD::Float(((val.UInt(0)>>16) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f)); dst.move(2, SIMD::Float(((val.UInt(0) >> 16) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f));
dst.move(3, SIMD::Float(((val.UInt(0)>>24) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f)); dst.move(3, SIMD::Float(((val.UInt(0) >> 24) & SIMD::UInt(0xFF))) * SIMD::Float(1.0f / 255.f));
break; break;
} }
case GLSLstd450UnpackSnorm2x16: case GLSLstd450UnpackSnorm2x16:
...@@ -457,9 +463,12 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn, ...@@ -457,9 +463,12 @@ SpirvShader::EmitResult SpirvShader::EmitExtendedInstruction(InsnIterator insn,
auto val = GenericValue(this, state, insn.word(5)); auto val = GenericValue(this, state, insn.word(5));
// clamp(f / 32767.0, -1.0, 1.0) // clamp(f / 32767.0, -1.0, 1.0)
dst.move(0, Min(Max(SIMD::Float(As<SIMD::Int>((val.UInt(0) & SIMD::UInt(0x0000FFFF)) << 16)) * dst.move(0, Min(Max(SIMD::Float(As<SIMD::Int>((val.UInt(0) & SIMD::UInt(0x0000FFFF)) << 16)) *
SIMD::Float(1.0f / float(0x7FFF0000)), SIMD::Float(-1.0f)), SIMD::Float(1.0f))); SIMD::Float(1.0f / float(0x7FFF0000)),
SIMD::Float(-1.0f)),
SIMD::Float(1.0f)));
dst.move(1, Min(Max(SIMD::Float(As<SIMD::Int>(val.UInt(0) & SIMD::UInt(0xFFFF0000))) * SIMD::Float(1.0f / float(0x7FFF0000)), dst.move(1, Min(Max(SIMD::Float(As<SIMD::Int>(val.UInt(0) & SIMD::UInt(0xFFFF0000))) * SIMD::Float(1.0f / float(0x7FFF0000)),
SIMD::Float(-1.0f)), SIMD::Float(1.0f))); SIMD::Float(-1.0f)),
SIMD::Float(1.0f)));
break; break;
} }
case GLSLstd450UnpackUnorm2x16: case GLSLstd450UnpackUnorm2x16:
......
src/Pipeline/SpirvShaderGroup.cpp
...@@ -18,20 +18,21 @@ ...@@ -18,20 +18,21 @@
namespace sw { namespace sw {
struct SpirvShader::GroupOps { struct SpirvShader::GroupOps
{
// Template function to perform a binary operation. // Template function to perform a binary operation.
// |TYPE| should be the type of the identity value (as an SIMD::<Type>). // |TYPE| should be the type of the identity value (as an SIMD::<Type>).
// |APPLY| should be a callable object that takes two RValue<TYPE> parameters // |APPLY| should be a callable object that takes two RValue<TYPE> parameters
// and returns a new RValue<TYPE> corresponding to the operation's result. // and returns a new RValue<TYPE> corresponding to the operation's result.
template <typename TYPE, typename APPLY> template<typename TYPE, typename APPLY>
static void BinaryOperation( static void BinaryOperation(
const SpirvShader* shader, const SpirvShader *shader,
const SpirvShader::InsnIterator& insn, const SpirvShader::InsnIterator &insn,
const SpirvShader::EmitState* state, const SpirvShader::EmitState *state,
Intermediate& dst, Intermediate &dst,
const TYPE& identity, const TYPE &identity,
APPLY&& apply) APPLY &&apply)
{ {
SpirvShader::GenericValue value(shader, state, insn.word(5)); SpirvShader::GenericValue value(shader, state, insn.word(5));
auto &type = shader->getType(SpirvShader::Type::ID(insn.word(1))); auto &type = shader->getType(SpirvShader::Type::ID(insn.word(1)));
...@@ -72,10 +73,8 @@ struct SpirvShader::GroupOps { ...@@ -72,10 +73,8 @@ struct SpirvShader::GroupOps {
} }
} }
} }
}; };
SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, EmitState *state) const SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, EmitState *state) const
{ {
static_assert(SIMD::Width == 4, "EmitGroupNonUniform makes many assumptions that the SIMD vector width is 4"); static_assert(SIMD::Width == 4, "EmitGroupNonUniform makes many assumptions that the SIMD vector width is 4");
...@@ -322,8 +321,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -322,8 +321,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ return a + b; } [](RValue<Type> a, RValue<Type> b) { return a + b; });
);
break; break;
} }
...@@ -333,8 +331,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -333,8 +331,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0.), Type(0.),
[](RValue<Type>a, RValue<Type>b){ return a + b; } [](RValue<Type> a, RValue<Type> b) { return a + b; });
);
break; break;
} }
...@@ -344,8 +341,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -344,8 +341,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(1), Type(1),
[](RValue<Type>a, RValue<Type>b){ return a * b; } [](RValue<Type> a, RValue<Type> b) { return a * b; });
);
break; break;
} }
...@@ -355,8 +351,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -355,8 +351,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(1.), Type(1.),
[](RValue<Type>a, RValue<Type>b){ return a * b; } [](RValue<Type> a, RValue<Type> b) { return a * b; });
);
break; break;
} }
...@@ -366,8 +361,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -366,8 +361,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(~0u), Type(~0u),
[](RValue<Type>a, RValue<Type>b){ return a & b; } [](RValue<Type> a, RValue<Type> b) { return a & b; });
);
break; break;
} }
...@@ -377,8 +371,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -377,8 +371,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ return a | b; } [](RValue<Type> a, RValue<Type> b) { return a | b; });
);
break; break;
} }
...@@ -388,8 +381,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -388,8 +381,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ return a ^ b; } [](RValue<Type> a, RValue<Type> b) { return a ^ b; });
);
break; break;
} }
...@@ -399,8 +391,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -399,8 +391,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(INT32_MAX), Type(INT32_MAX),
[](RValue<Type>a, RValue<Type>b){ return Min(a, b); } [](RValue<Type> a, RValue<Type> b) { return Min(a, b); });
);
break; break;
} }
...@@ -410,8 +401,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -410,8 +401,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(~0u), Type(~0u),
[](RValue<Type>a, RValue<Type>b){ return Min(a, b); } [](RValue<Type> a, RValue<Type> b) { return Min(a, b); });
);
break; break;
} }
...@@ -421,8 +411,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -421,8 +411,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type::positive_inf(), Type::positive_inf(),
[](RValue<Type>a, RValue<Type>b){ return NMin(a, b); } [](RValue<Type> a, RValue<Type> b) { return NMin(a, b); });
);
break; break;
} }
...@@ -432,8 +421,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -432,8 +421,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(INT32_MIN), Type(INT32_MIN),
[](RValue<Type>a, RValue<Type>b){ return Max(a, b); } [](RValue<Type> a, RValue<Type> b) { return Max(a, b); });
);
break; break;
} }
...@@ -443,8 +431,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -443,8 +431,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ return Max(a, b); } [](RValue<Type> a, RValue<Type> b) { return Max(a, b); });
);
break; break;
} }
...@@ -454,8 +441,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -454,8 +441,7 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type::negative_inf(), Type::negative_inf(),
[](RValue<Type>a, RValue<Type>b){ return NMax(a, b); } [](RValue<Type> a, RValue<Type> b) { return NMax(a, b); });
);
break; break;
} }
...@@ -465,11 +451,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -465,11 +451,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(~0u), Type(~0u),
[](RValue<Type>a, RValue<Type>b){ [](RValue<Type> a, RValue<Type> b) {
SIMD::UInt zero = SIMD::UInt(0); SIMD::UInt zero = SIMD::UInt(0);
return CmpNEQ(a, zero) & CmpNEQ(b, zero); return CmpNEQ(a, zero) & CmpNEQ(b, zero);
} });
);
break; break;
} }
...@@ -479,11 +464,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -479,11 +464,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ [](RValue<Type> a, RValue<Type> b) {
SIMD::UInt zero = SIMD::UInt(0); SIMD::UInt zero = SIMD::UInt(0);
return CmpNEQ(a, zero) | CmpNEQ(b, zero); return CmpNEQ(a, zero) | CmpNEQ(b, zero);
} });
);
break; break;
} }
...@@ -493,11 +477,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit ...@@ -493,11 +477,10 @@ SpirvShader::EmitResult SpirvShader::EmitGroupNonUniform(InsnIterator insn, Emit
SpirvShader::GroupOps::BinaryOperation( SpirvShader::GroupOps::BinaryOperation(
this, insn, state, dst, this, insn, state, dst,
Type(0), Type(0),
[](RValue<Type>a, RValue<Type>b){ [](RValue<Type> a, RValue<Type> b) {
SIMD::UInt zero = SIMD::UInt(0); SIMD::UInt zero = SIMD::UInt(0);
return CmpNEQ(a, zero) ^ CmpNEQ(b, zero); return CmpNEQ(a, zero) ^ CmpNEQ(b, zero);
} });
);
break; break;
} }
......
src/Pipeline/SpirvShaderMemory.cpp
...@@ -57,8 +57,7 @@ SpirvShader::EmitResult SpirvShader::EmitLoad(InsnIterator insn, EmitState *stat ...@@ -57,8 +57,7 @@ SpirvShader::EmitResult SpirvShader::EmitLoad(InsnIterator insn, EmitState *stat
auto &dst = state->createIntermediate(resultId, resultTy.sizeInComponents); auto &dst = state->createIntermediate(resultId, resultTy.sizeInComponents);
auto robustness = state->getOutOfBoundsBehavior(pointerTy.storageClass); auto robustness = state->getOutOfBoundsBehavior(pointerTy.storageClass);
VisitMemoryObject(pointerId, [&](const MemoryElement& el) VisitMemoryObject(pointerId, [&](const MemoryElement &el) {
{
auto p = ptr + el.offset; auto p = ptr + el.offset;
if(interleavedByLane) { p = InterleaveByLane(p); } // TODO: Interleave once, then add offset? if(interleavedByLane) { p = InterleaveByLane(p); } // TODO: Interleave once, then add offset?
dst.move(el.index, p.Load<SIMD::Float>(robustness, state->activeLaneMask(), atomic, memoryOrder)); dst.move(el.index, p.Load<SIMD::Float>(robustness, state->activeLaneMask(), atomic, memoryOrder));
...@@ -101,8 +100,7 @@ SpirvShader::EmitResult SpirvShader::EmitStore(InsnIterator insn, EmitState *sta ...@@ -101,8 +100,7 @@ SpirvShader::EmitResult SpirvShader::EmitStore(InsnIterator insn, EmitState *sta
{ {
// Constant source data. // Constant source data.
const uint32_t *src = object.constantValue.get(); const uint32_t *src = object.constantValue.get();
VisitMemoryObject(pointerId, [&](const MemoryElement& el) VisitMemoryObject(pointerId, [&](const MemoryElement &el) {
{
auto p = ptr + el.offset; auto p = ptr + el.offset;
if(interleavedByLane) { p = InterleaveByLane(p); } if(interleavedByLane) { p = InterleaveByLane(p); }
p.Store(SIMD::Int(src[el.index]), robustness, mask, atomic, memoryOrder); p.Store(SIMD::Int(src[el.index]), robustness, mask, atomic, memoryOrder);
...@@ -112,8 +110,7 @@ SpirvShader::EmitResult SpirvShader::EmitStore(InsnIterator insn, EmitState *sta ...@@ -112,8 +110,7 @@ SpirvShader::EmitResult SpirvShader::EmitStore(InsnIterator insn, EmitState *sta
{ {
// Intermediate source data. // Intermediate source data.
auto &src = state->getIntermediate(objectId); auto &src = state->getIntermediate(objectId);
VisitMemoryObject(pointerId, [&](const MemoryElement& el) VisitMemoryObject(pointerId, [&](const MemoryElement &el) {
{
auto p = ptr + el.offset; auto p = ptr + el.offset;
if(interleavedByLane) { p = InterleaveByLane(p); } if(interleavedByLane) { p = InterleaveByLane(p); }
p.Store(src.Float(el.index), robustness, mask, atomic, memoryOrder); p.Store(src.Float(el.index), robustness, mask, atomic, memoryOrder);
...@@ -239,8 +236,7 @@ SpirvShader::EmitResult SpirvShader::EmitVariable(InsnIterator insn, EmitState * ...@@ -239,8 +236,7 @@ SpirvShader::EmitResult SpirvShader::EmitVariable(InsnIterator insn, EmitState *
bool interleavedByLane = IsStorageInterleavedByLane(objectTy.storageClass); bool interleavedByLane = IsStorageInterleavedByLane(objectTy.storageClass);
auto ptr = GetPointerToData(resultId, 0, state); auto ptr = GetPointerToData(resultId, 0, state);
GenericValue initialValue(this, state, initializerId); GenericValue initialValue(this, state, initializerId);
VisitMemoryObject(resultId, [&](const MemoryElement& el) VisitMemoryObject(resultId, [&](const MemoryElement &el) {
{
auto p = ptr + el.offset; auto p = ptr + el.offset;
if(interleavedByLane) { p = InterleaveByLane(p); } if(interleavedByLane) { p = InterleaveByLane(p); }
auto robustness = OutOfBoundsBehavior::UndefinedBehavior; // Local variables are always within bounds. auto robustness = OutOfBoundsBehavior::UndefinedBehavior; // Local variables are always within bounds.
...@@ -271,10 +267,9 @@ SpirvShader::EmitResult SpirvShader::EmitCopyMemory(InsnIterator insn, EmitState ...@@ -271,10 +267,9 @@ SpirvShader::EmitResult SpirvShader::EmitCopyMemory(InsnIterator insn, EmitState
std::unordered_map<uint32_t, uint32_t> srcOffsets; std::unordered_map<uint32_t, uint32_t> srcOffsets;
VisitMemoryObject(srcPtrId, [&](const MemoryElement& el) { srcOffsets[el.index] = el.offset; }); VisitMemoryObject(srcPtrId, [&](const MemoryElement &el) { srcOffsets[el.index] = el.offset; });
VisitMemoryObject(dstPtrId, [&](const MemoryElement& el) VisitMemoryObject(dstPtrId, [&](const MemoryElement &el) {
{
auto it = srcOffsets.find(el.index); auto it = srcOffsets.find(el.index);
ASSERT(it != srcOffsets.end()); ASSERT(it != srcOffsets.end());
auto srcOffset = it->second; auto srcOffset = it->second;
...@@ -303,7 +298,7 @@ SpirvShader::EmitResult SpirvShader::EmitMemoryBarrier(InsnIterator insn, EmitSt ...@@ -303,7 +298,7 @@ SpirvShader::EmitResult SpirvShader::EmitMemoryBarrier(InsnIterator insn, EmitSt
return EmitResult::Continue; return EmitResult::Continue;
} }
void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::SpirvShader::Decorations d, uint32_t& index, uint32_t offset, const MemoryVisitor& f) const void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::SpirvShader::Decorations d, uint32_t &index, uint32_t offset, const MemoryVisitor &f) const
{ {
ApplyDecorationsForId(&d, id); ApplyDecorationsForId(&d, id);
auto const &type = getType(id); auto const &type = getType(id);
...@@ -322,7 +317,7 @@ void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::Spirv ...@@ -322,7 +317,7 @@ void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::Spirv
case spv::OpTypeInt: case spv::OpTypeInt:
case spv::OpTypeFloat: case spv::OpTypeFloat:
case spv::OpTypeRuntimeArray: case spv::OpTypeRuntimeArray:
f(MemoryElement{index++, offset, type}); f(MemoryElement{ index++, offset, type });
break; break;
case spv::OpTypeVector: case spv::OpTypeVector:
{ {
...@@ -366,10 +361,10 @@ void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::Spirv ...@@ -366,10 +361,10 @@ void SpirvShader::VisitMemoryObjectInner(sw::SpirvShader::Type::ID id, sw::Spirv
} }
} }
void SpirvShader::VisitMemoryObject(sw::SpirvShader::Object::ID id, const MemoryVisitor& f) const void SpirvShader::VisitMemoryObject(sw::SpirvShader::Object::ID id, const MemoryVisitor &f) const
{ {
auto typeId = getObject(id).type; auto typeId = getObject(id).type;
auto const & type = getType(typeId); auto const &type = getType(typeId);
if(IsExplicitLayout(type.storageClass)) if(IsExplicitLayout(type.storageClass))
{ {
Decorations d{}; Decorations d{};
...@@ -384,7 +379,7 @@ void SpirvShader::VisitMemoryObject(sw::SpirvShader::Object::ID id, const Memory ...@@ -384,7 +379,7 @@ void SpirvShader::VisitMemoryObject(sw::SpirvShader::Object::ID id, const Memory
for(auto index = 0u; index < elType.sizeInComponents; index++) for(auto index = 0u; index < elType.sizeInComponents; index++)
{ {
auto offset = static_cast<uint32_t>(index * sizeof(float)); auto offset = static_cast<uint32_t>(index * sizeof(float));
f({index, offset, elType}); f({ index, offset, elType });
} }
} }
} }
...@@ -442,8 +437,7 @@ std::memory_order SpirvShader::MemoryOrder(spv::MemorySemanticsMask memorySemant ...@@ -442,8 +437,7 @@ std::memory_order SpirvShader::MemoryOrder(spv::MemorySemanticsMask memorySemant
spv::MemorySemanticsAcquireMask | spv::MemorySemanticsAcquireMask |
spv::MemorySemanticsReleaseMask | spv::MemorySemanticsReleaseMask |
spv::MemorySemanticsAcquireReleaseMask | spv::MemorySemanticsAcquireReleaseMask |
spv::MemorySemanticsSequentiallyConsistentMask spv::MemorySemanticsSequentiallyConsistentMask);
);
switch(control) switch(control)
{ {
case spv::MemorySemanticsMaskNone: return std::memory_order_relaxed; case spv::MemorySemanticsMaskNone: return std::memory_order_relaxed;
......
src/Pipeline/SpirvShaderSampling.cpp
...@@ -15,13 +15,13 @@ ...@@ -15,13 +15,13 @@
#include "SpirvShader.hpp" #include "SpirvShader.hpp"
#include "SamplerCore.hpp" // TODO: Figure out what's needed. #include "SamplerCore.hpp" // TODO: Figure out what's needed.
#include "Device/Config.hpp"
#include "System/Math.hpp" #include "System/Math.hpp"
#include "Vulkan/VkDebug.hpp" #include "Vulkan/VkDebug.hpp"
#include "Vulkan/VkDescriptorSetLayout.hpp" #include "Vulkan/VkDescriptorSetLayout.hpp"
#include "Vulkan/VkDevice.hpp" #include "Vulkan/VkDevice.hpp"
#include "Vulkan/VkImageView.hpp" #include "Vulkan/VkImageView.hpp"
#include "Vulkan/VkSampler.hpp" #include "Vulkan/VkSampler.hpp"
#include "Device/Config.hpp"
#include <spirv/unified1/spirv.hpp> #include <spirv/unified1/spirv.hpp>
...@@ -36,22 +36,22 @@ SpirvShader::ImageSampler *SpirvShader::getImageSampler(uint32_t inst, vk::Sampl ...@@ -36,22 +36,22 @@ SpirvShader::ImageSampler *SpirvShader::getImageSampler(uint32_t inst, vk::Sampl
const auto samplerId = sampler ? sampler->id : 0; const auto samplerId = sampler ? sampler->id : 0;
ASSERT(imageDescriptor->imageViewId != 0 && (samplerId != 0 || instruction.samplerMethod == Fetch)); ASSERT(imageDescriptor->imageViewId != 0 && (samplerId != 0 || instruction.samplerMethod == Fetch));
vk::Device::SamplingRoutineCache::Key key = {inst, imageDescriptor->imageViewId, samplerId}; vk::Device::SamplingRoutineCache::Key key = { inst, imageDescriptor->imageViewId, samplerId };
ASSERT(imageDescriptor->device); ASSERT(imageDescriptor->device);
if(auto routine = imageDescriptor->device->findInConstCache(key)) if(auto routine = imageDescriptor->device->findInConstCache(key))
{ {
return (ImageSampler*)(routine->getEntry()); return (ImageSampler *)(routine->getEntry());
} }
std::unique_lock<std::mutex> lock(imageDescriptor->device->getSamplingRoutineCacheMutex()); std::unique_lock<std::mutex> lock(imageDescriptor->device->getSamplingRoutineCacheMutex());
vk::Device::SamplingRoutineCache* cache = imageDescriptor->device->getSamplingRoutineCache(); vk::Device::SamplingRoutineCache *cache = imageDescriptor->device->getSamplingRoutineCache();
auto routine = cache->query(key); auto routine = cache->query(key);
if(routine) if(routine)
{ {
return (ImageSampler*)(routine->getEntry()); return (ImageSampler *)(routine->getEntry());
} }
auto type = imageDescriptor->type; auto type = imageDescriptor->type;
...@@ -100,7 +100,7 @@ SpirvShader::ImageSampler *SpirvShader::getImageSampler(uint32_t inst, vk::Sampl ...@@ -100,7 +100,7 @@ SpirvShader::ImageSampler *SpirvShader::getImageSampler(uint32_t inst, vk::Sampl
routine = emitSamplerRoutine(instruction, samplerState); routine = emitSamplerRoutine(instruction, samplerState);
cache->add(key, routine); cache->add(key, routine);
return (ImageSampler*)(routine->getEntry()); return (ImageSampler *)(routine->getEntry());
} }
std::shared_ptr<rr::Routine> SpirvShader::emitSamplerRoutine(ImageInstruction instruction, const Sampler &samplerState) std::shared_ptr<rr::Routine> SpirvShader::emitSamplerRoutine(ImageInstruction instruction, const Sampler &samplerState)
...@@ -114,17 +114,17 @@ std::shared_ptr<rr::Routine> SpirvShader::emitSamplerRoutine(ImageInstruction in ...@@ -114,17 +114,17 @@ std::shared_ptr<rr::Routine> SpirvShader::emitSamplerRoutine(ImageInstruction in
Pointer<SIMD::Float> out = function.Arg<3>(); Pointer<SIMD::Float> out = function.Arg<3>();
Pointer<Byte> constants = function.Arg<4>(); Pointer<Byte> constants = function.Arg<4>();
SIMD::Float uvw[4] = {0, 0, 0, 0}; SIMD::Float uvw[4] = { 0, 0, 0, 0 };
SIMD::Float q = 0; SIMD::Float q = 0;
SIMD::Float lodOrBias = 0; // Explicit level-of-detail, or bias added to the implicit level-of-detail (depending on samplerMethod). SIMD::Float lodOrBias = 0; // Explicit level-of-detail, or bias added to the implicit level-of-detail (depending on samplerMethod).
Vector4f dsx = {0, 0, 0, 0}; Vector4f dsx = { 0, 0, 0, 0 };
Vector4f dsy = {0, 0, 0, 0}; Vector4f dsy = { 0, 0, 0, 0 };
Vector4f offset = {0, 0, 0, 0}; Vector4f offset = { 0, 0, 0, 0 };
SIMD::Int sampleId = 0; SIMD::Int sampleId = 0;
SamplerFunction samplerFunction = instruction.getSamplerFunction(); SamplerFunction samplerFunction = instruction.getSamplerFunction();
uint32_t i = 0; uint32_t i = 0;
for( ; i < instruction.coordinates; i++) for(; i < instruction.coordinates; i++)
{ {
uvw[i] = in[i]; uvw[i] = in[i];
} }
......
src/Pipeline/VertexProgram.cpp
...@@ -29,26 +29,23 @@ VertexProgram::VertexProgram( ...@@ -29,26 +29,23 @@ VertexProgram::VertexProgram(
vk::PipelineLayout const *pipelineLayout, vk::PipelineLayout const *pipelineLayout,
SpirvShader const *spirvShader, SpirvShader const *spirvShader,
const vk::DescriptorSet::Bindings &descriptorSets) const vk::DescriptorSet::Bindings &descriptorSets)
: VertexRoutine(state, pipelineLayout, spirvShader), : VertexRoutine(state, pipelineLayout, spirvShader)
descriptorSets(descriptorSets) , descriptorSets(descriptorSets)
{ {
routine.setImmutableInputBuiltins(spirvShader); routine.setImmutableInputBuiltins(spirvShader);
routine.setInputBuiltin(spirvShader, spv::BuiltInViewIndex, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInViewIndex, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
assert(builtin.SizeInComponents == 1); assert(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, viewID))))); value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, viewID)))));
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInInstanceIndex, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInInstanceIndex, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
// TODO: we could do better here; we know InstanceIndex is uniform across all lanes // TODO: we could do better here; we know InstanceIndex is uniform across all lanes
assert(builtin.SizeInComponents == 1); assert(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, instanceID))))); value[builtin.FirstComponent] = As<Float4>(Int4((*Pointer<Int>(data + OFFSET(DrawData, instanceID)))));
}); });
routine.setInputBuiltin(spirvShader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping& builtin, Array<SIMD::Float>& value) routine.setInputBuiltin(spirvShader, spv::BuiltInSubgroupSize, [&](const SpirvShader::BuiltinMapping &builtin, Array<SIMD::Float> &value) {
{
ASSERT(builtin.SizeInComponents == 1); ASSERT(builtin.SizeInComponents == 1);
value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(SIMD::Width)); value[builtin.FirstComponent] = As<SIMD::Float>(SIMD::Int(SIMD::Width));
}); });
...@@ -63,7 +60,7 @@ VertexProgram::~VertexProgram() ...@@ -63,7 +60,7 @@ VertexProgram::~VertexProgram()
{ {
} }
void VertexProgram::program(Pointer<UInt> &batch, UInt& vertexCount) void VertexProgram::program(Pointer<UInt> &batch, UInt &vertexCount)
{ {
auto it = spirvShader->inputBuiltins.find(spv::BuiltInVertexIndex); auto it = spirvShader->inputBuiltins.find(spv::BuiltInVertexIndex);
if(it != spirvShader->inputBuiltins.end()) if(it != spirvShader->inputBuiltins.end())
...@@ -82,4 +79,4 @@ void VertexProgram::program(Pointer<UInt> &batch, UInt& vertexCount) ...@@ -82,4 +79,4 @@ void VertexProgram::program(Pointer<UInt> &batch, UInt& vertexCount)
spirvShader->emitEpilog(&routine); spirvShader->emitEpilog(&routine);
} }
} // namepsace sw } // namespace sw
src/Pipeline/VertexProgram.hpp
...@@ -15,8 +15,8 @@ ...@@ -15,8 +15,8 @@
#ifndef sw_VertexProgram_hpp #ifndef sw_VertexProgram_hpp
#define sw_VertexProgram_hpp #define sw_VertexProgram_hpp
#include "VertexRoutine.hpp"
#include "ShaderCore.hpp" #include "ShaderCore.hpp"
#include "VertexRoutine.hpp"
namespace sw { namespace sw {
...@@ -34,11 +34,11 @@ public: ...@@ -34,11 +34,11 @@ public:
virtual ~VertexProgram(); virtual ~VertexProgram();
private: private:
void program(Pointer<UInt> &batch, UInt& vertexCount) override; void program(Pointer<UInt> &batch, UInt &vertexCount) override;
const vk::DescriptorSet::Bindings &descriptorSets; const vk::DescriptorSet::Bindings &descriptorSets;
}; };
} // namepsace sw } // namespace sw
#endif // sw_VertexProgram_hpp #endif // sw_VertexProgram_hpp
src/Pipeline/VertexRoutine.hpp
...@@ -20,14 +20,21 @@ ...@@ -20,14 +20,21 @@
#include "Device/Color.hpp" #include "Device/Color.hpp"
#include "Device/VertexProcessor.hpp" #include "Device/VertexProcessor.hpp"
namespace vk { class PipelineLayout; } namespace vk {
class PipelineLayout;
}
namespace sw { namespace sw {
class VertexRoutinePrototype : public VertexRoutineFunction class VertexRoutinePrototype : public VertexRoutineFunction
{ {
public: public:
VertexRoutinePrototype() : vertex(Arg<0>()), batch(Arg<1>()), task(Arg<2>()), data(Arg<3>()) {} VertexRoutinePrototype()
: vertex(Arg<0>())
, batch(Arg<1>())
, task(Arg<2>())
, data(Arg<3>())
{}
virtual ~VertexRoutinePrototype() {} virtual ~VertexRoutinePrototype() {}
protected: protected:
...@@ -57,15 +64,15 @@ protected: ...@@ -57,15 +64,15 @@ protected:
SpirvRoutine routine; SpirvRoutine routine;
const VertexProcessor::State &state; const VertexProcessor::State &state;
SpirvShader const * const spirvShader; SpirvShader const *const spirvShader;
private: private:
virtual void program(Pointer<UInt> &batch, UInt& vertexCount) = 0; virtual void program(Pointer<UInt> &batch, UInt &vertexCount) = 0;
typedef VertexProcessor::State::Input Stream; typedef VertexProcessor::State::Input Stream;
Vector4f readStream(Pointer<Byte> &buffer, UInt &stride, const Stream &stream, Pointer<UInt> &batch, Vector4f readStream(Pointer<Byte> &buffer, UInt &stride, const Stream &stream, Pointer<UInt> &batch,
bool robustBufferAccess, UInt& robustnessSize, Int baseVertex); bool robustBufferAccess, UInt &robustnessSize, Int baseVertex);
void readInput(Pointer<UInt> &batch); void readInput(Pointer<UInt> &batch);
void computeClipFlags(); void computeClipFlags();
void computeCullMask(); void computeCullMask();
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment