SpirvShader: Implement OpBranch
Bug: b/128527271
Change-Id: I367ed0d578e36a56baf4b8c4c2256ee4de1297cc
Reviewed-on: https://swiftshader-review.googlesource.com/c/SwiftShader/+/27097
Presubmit-Ready: Ben Clayton <bclayton@google.com>
Reviewed-by: Nicolas Capens <nicolascapens@google.com>
Reviewed-by: Chris Forbes <chrisforbes@google.com>
Tested-by: Ben Clayton <bclayton@google.com>
Kokoro-Presubmit: kokoro <noreply+kokoro@google.com>
diff --git a/src/Pipeline/SpirvShader.cpp b/src/Pipeline/SpirvShader.cpp
index 9f2b2d5..794f671 100644
--- a/src/Pipeline/SpirvShader.cpp
+++ b/src/Pipeline/SpirvShader.cpp
@@ -1185,6 +1185,10 @@
EmitAll(insn, routine);
break;
+ case spv::OpBranch:
+ EmitBranch(insn, routine);
+ break;
+
default:
UNIMPLEMENTED(OpcodeName(insn.opcode()).c_str());
break;
@@ -2148,6 +2152,12 @@
dst.emplace(0, result);
}
+ void SpirvShader::EmitBranch(InsnIterator insn, SpirvRoutine *routine) const
+ {
+ auto blockId = Block::ID(insn.word(1));
+ EmitBlock(routine, getBlock(blockId));
+ }
+
void SpirvShader::emitEpilog(SpirvRoutine *routine) const
{
for (auto insn : *this)
diff --git a/src/Pipeline/SpirvShader.hpp b/src/Pipeline/SpirvShader.hpp
index 34c7951..ecfa91c 100644
--- a/src/Pipeline/SpirvShader.hpp
+++ b/src/Pipeline/SpirvShader.hpp
@@ -490,6 +490,7 @@
void EmitExtendedInstruction(InsnIterator insn, SpirvRoutine *routine) const;
void EmitAny(InsnIterator insn, SpirvRoutine *routine) const;
void EmitAll(InsnIterator insn, SpirvRoutine *routine) const;
+ void EmitBranch(InsnIterator insn, SpirvRoutine *routine) const;
// OpcodeName returns the name of the opcode op.
// If NDEBUG is defined, then OpcodeName will only return the numerical code.
diff --git a/tests/VulkanUnitTests/unittests.cpp b/tests/VulkanUnitTests/unittests.cpp
index 78eb489..bc57e41 100644
--- a/tests/VulkanUnitTests/unittests.cpp
+++ b/tests/VulkanUnitTests/unittests.cpp
@@ -149,82 +149,41 @@
struct ComputeParams
{
+ size_t numElements;
int localSizeX;
int localSizeY;
int localSizeZ;
+
+ friend std::ostream& operator<<(std::ostream& os, const ComputeParams& params) {
+ return os << "ComputeParams{" <<
+ "numElements: " << params.numElements << ", " <<
+ "localSizeX: " << params.localSizeX << ", " <<
+ "localSizeY: " << params.localSizeY << ", " <<
+ "localSizeZ: " << params.localSizeZ <<
+ "}";
+ }
};
-class SwiftShaderVulkanComputeTest : public testing::TestWithParam<ComputeParams> {};
-
-INSTANTIATE_TEST_CASE_P(ComputeParams, SwiftShaderVulkanComputeTest, testing::Values(
- ComputeParams{1, 1, 1},
- ComputeParams{2, 1, 1},
- ComputeParams{4, 1, 1},
- ComputeParams{8, 1, 1},
- ComputeParams{16, 1, 1},
- ComputeParams{32, 1, 1}
-));
-
-TEST_P(SwiftShaderVulkanComputeTest, Memcpy)
+// Base class for compute tests that read from an input buffer and write to an
+// output buffer of same length.
+class SwiftShaderVulkanBufferToBufferComputeTest : public testing::TestWithParam<ComputeParams>
{
+public:
+ void test(const std::string& shader,
+ std::function<uint32_t(uint32_t idx)> input,
+ std::function<uint32_t(uint32_t idx)> expected);
+};
+
+void SwiftShaderVulkanBufferToBufferComputeTest::test(
+ const std::string& shader,
+ std::function<uint32_t(uint32_t idx)> input,
+ std::function<uint32_t(uint32_t idx)> expected)
+{
+ auto code = compileSpirv(shader.c_str());
+
Driver driver;
ASSERT_TRUE(driver.loadSwiftShader());
- auto params = GetParam();
-
- std::stringstream src;
- src <<
- "OpCapability Shader\n"
- "OpMemoryModel Logical GLSL450\n"
- "OpEntryPoint GLCompute %1 \"main\" %2\n"
- "OpExecutionMode %1 LocalSize " <<
- params.localSizeX << " " <<
- params.localSizeY << " " <<
- params.localSizeZ << "\n" <<
- "OpDecorate %3 ArrayStride 4\n"
- "OpMemberDecorate %4 0 Offset 0\n"
- "OpDecorate %4 BufferBlock\n"
- "OpDecorate %5 DescriptorSet 0\n"
- "OpDecorate %5 Binding 1\n"
- "OpDecorate %2 BuiltIn GlobalInvocationId\n"
- "OpDecorate %6 ArrayStride 4\n"
- "OpMemberDecorate %7 0 Offset 0\n"
- "OpDecorate %7 BufferBlock\n"
- "OpDecorate %8 DescriptorSet 0\n"
- "OpDecorate %8 Binding 0\n"
- "%9 = OpTypeVoid\n"
- "%10 = OpTypeFunction %9\n"
- "%11 = OpTypeInt 32 1\n"
- "%3 = OpTypeRuntimeArray %11\n"
- "%4 = OpTypeStruct %3\n"
- "%12 = OpTypePointer Uniform %4\n"
- "%5 = OpVariable %12 Uniform\n"
- "%13 = OpConstant %11 0\n"
- "%14 = OpTypeInt 32 0\n"
- "%15 = OpTypeVector %14 3\n"
- "%16 = OpTypePointer Input %15\n"
- "%2 = OpVariable %16 Input\n"
- "%17 = OpConstant %14 0\n"
- "%18 = OpTypePointer Input %14\n"
- "%6 = OpTypeRuntimeArray %11\n"
- "%7 = OpTypeStruct %6\n"
- "%19 = OpTypePointer Uniform %7\n"
- "%8 = OpVariable %19 Uniform\n"
- "%20 = OpTypePointer Uniform %11\n"
- "%21 = OpConstant %11 1\n"
- "%1 = OpFunction %9 None %10\n"
- "%22 = OpLabel\n"
- "%23 = OpAccessChain %18 %2 %17\n"
- "%24 = OpLoad %14 %23\n"
- "%25 = OpAccessChain %20 %8 %13 %24\n"
- "%26 = OpLoad %11 %25\n"
- "%27 = OpAccessChain %20 %5 %13 %24\n"
- "OpStore %27 %26\n"
- "OpReturn\n"
- "OpFunctionEnd\n";
-
- auto code = compileSpirv(src.str().c_str());
-
const VkInstanceCreateInfo createInfo = {
VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO, // sType
nullptr, // pNext
@@ -245,48 +204,57 @@
VK_ASSERT(Device::CreateComputeDevice(&driver, instance, &device));
ASSERT_TRUE(device.IsValid());
- constexpr int NUM_ELEMENTS = 256;
-
- struct Buffers
- {
- uint32_t magic0;
- uint32_t in[NUM_ELEMENTS];
- uint32_t magic1;
- uint32_t out[NUM_ELEMENTS];
- uint32_t magic2;
- };
-
- constexpr uint32_t magic0 = 0x01234567;
- constexpr uint32_t magic1 = 0x89abcdef;
- constexpr uint32_t magic2 = 0xfedcba99;
+ // struct Buffers
+ // {
+ // uint32_t magic0;
+ // uint32_t in[NUM_ELEMENTS];
+ // uint32_t magic1;
+ // uint32_t out[NUM_ELEMENTS];
+ // uint32_t magic2;
+ // };
+ static constexpr uint32_t magic0 = 0x01234567;
+ static constexpr uint32_t magic1 = 0x89abcdef;
+ static constexpr uint32_t magic2 = 0xfedcba99;
+ size_t numElements = GetParam().numElements;
+ size_t magic0Offset = 0;
+ size_t inOffset = 1 + magic0Offset;
+ size_t magic1Offset = numElements + inOffset;
+ size_t outOffset = 1 + magic1Offset;
+ size_t magic2Offset = numElements + outOffset;
+ size_t buffersTotalElements = 1 + magic2Offset;
+ size_t buffersSize = sizeof(uint32_t) * buffersTotalElements;
VkDeviceMemory memory;
- VK_ASSERT(device.AllocateMemory(sizeof(Buffers),
+ VK_ASSERT(device.AllocateMemory(buffersSize,
VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT | VK_MEMORY_PROPERTY_HOST_COHERENT_BIT,
&memory));
- Buffers* buffers;
- VK_ASSERT(device.MapMemory(memory, 0, sizeof(Buffers), 0, (void**)&buffers));
+ uint32_t* buffers;
+ VK_ASSERT(device.MapMemory(memory, 0, buffersSize, 0, (void**)&buffers));
- memset(buffers, 0, sizeof(Buffers));
+ buffers[magic0Offset] = magic0;
+ buffers[magic1Offset] = magic1;
+ buffers[magic2Offset] = magic2;
- buffers->magic0 = magic0;
- buffers->magic1 = magic1;
- buffers->magic2 = magic2;
-
- for(int i = 0; i < NUM_ELEMENTS; i++)
+ for(size_t i = 0; i < numElements; i++)
{
- buffers->in[i] = (uint32_t)i;
+ buffers[inOffset + i] = input(i);
}
device.UnmapMemory(memory);
buffers = nullptr;
VkBuffer bufferIn;
- VK_ASSERT(device.CreateStorageBuffer(memory, sizeof(Buffers::in), offsetof(Buffers, in), &bufferIn));
+ VK_ASSERT(device.CreateStorageBuffer(memory,
+ sizeof(uint32_t) * numElements,
+ sizeof(uint32_t) * inOffset,
+ &bufferIn));
VkBuffer bufferOut;
- VK_ASSERT(device.CreateStorageBuffer(memory, sizeof(Buffers::out), offsetof(Buffers, out), &bufferOut));
+ VK_ASSERT(device.CreateStorageBuffer(memory,
+ sizeof(uint32_t) * numElements,
+ sizeof(uint32_t) * outOffset,
+ &bufferOut));
VkShaderModule shaderModule;
VK_ASSERT(device.CreateShaderModule(code, &shaderModule));
@@ -352,24 +320,206 @@
driver.vkCmdBindDescriptorSets(commandBuffer, VK_PIPELINE_BIND_POINT_COMPUTE, pipelineLayout, 0, 1, &descriptorSet,
0, nullptr);
- driver.vkCmdDispatch(commandBuffer, NUM_ELEMENTS / params.localSizeX, 1, 1);
+ driver.vkCmdDispatch(commandBuffer, numElements / GetParam().localSizeX, 1, 1);
VK_ASSERT(driver.vkEndCommandBuffer(commandBuffer));
VK_ASSERT(device.QueueSubmitAndWait(commandBuffer));
- VK_ASSERT(device.MapMemory(memory, 0, sizeof(Buffers), 0, (void**)&buffers));
+ VK_ASSERT(device.MapMemory(memory, 0, buffersSize, 0, (void**)&buffers));
- for (int i = 0; i < NUM_ELEMENTS; ++i)
+ for (size_t i = 0; i < numElements; ++i)
{
- EXPECT_EQ(buffers->in[i], buffers->out[i]) << "Unexpected output at " << i;
+ auto got = buffers[i + outOffset];
+ EXPECT_EQ(expected(i), got) << "Unexpected output at " << i;
}
// Check for writes outside of bounds.
- EXPECT_EQ(buffers->magic0, magic0);
- EXPECT_EQ(buffers->magic1, magic1);
- EXPECT_EQ(buffers->magic2, magic2);
+ EXPECT_EQ(buffers[magic0Offset], magic0);
+ EXPECT_EQ(buffers[magic1Offset], magic1);
+ EXPECT_EQ(buffers[magic2Offset], magic2);
device.UnmapMemory(memory);
buffers = nullptr;
}
+
+INSTANTIATE_TEST_CASE_P(ComputeParams, SwiftShaderVulkanBufferToBufferComputeTest, testing::Values(
+ ComputeParams{512, 1, 1, 1},
+ ComputeParams{512, 2, 1, 1},
+ ComputeParams{512, 4, 1, 1},
+ ComputeParams{512, 8, 1, 1},
+ ComputeParams{512, 16, 1, 1},
+ ComputeParams{512, 32, 1, 1},
+
+ // Non-multiple of SIMD-lane.
+ ComputeParams{3, 1, 1, 1},
+ ComputeParams{2, 1, 1, 1}
+));
+
+TEST_P(SwiftShaderVulkanBufferToBufferComputeTest, Memcpy)
+{
+ std::stringstream src;
+ src <<
+ "OpCapability Shader\n"
+ "OpMemoryModel Logical GLSL450\n"
+ "OpEntryPoint GLCompute %1 \"main\" %2\n"
+ "OpExecutionMode %1 LocalSize " <<
+ GetParam().localSizeX << " " <<
+ GetParam().localSizeY << " " <<
+ GetParam().localSizeZ << "\n" <<
+ "OpDecorate %3 ArrayStride 4\n"
+ "OpMemberDecorate %4 0 Offset 0\n"
+ "OpDecorate %4 BufferBlock\n"
+ "OpDecorate %5 DescriptorSet 0\n"
+ "OpDecorate %5 Binding 1\n"
+ "OpDecorate %2 BuiltIn GlobalInvocationId\n"
+ "OpDecorate %6 DescriptorSet 0\n"
+ "OpDecorate %6 Binding 0\n"
+ "%7 = OpTypeVoid\n"
+ "%8 = OpTypeFunction %7\n" // void()
+ "%9 = OpTypeInt 32 1\n" // int32
+ "%10 = OpTypeInt 32 0\n" // uint32
+ "%3 = OpTypeRuntimeArray %9\n" // int32[]
+ "%4 = OpTypeStruct %3\n" // struct{ int32[] }
+ "%11 = OpTypePointer Uniform %4\n" // struct{ int32[] }*
+ "%5 = OpVariable %11 Uniform\n" // struct{ int32[] }* in
+ "%12 = OpConstant %9 0\n" // int32(0)
+ "%13 = OpConstant %10 0\n" // uint32(0)
+ "%14 = OpTypeVector %10 3\n" // vec4<int32>
+ "%15 = OpTypePointer Input %14\n" // vec4<int32>*
+ "%2 = OpVariable %15 Input\n" // gl_GlobalInvocationId
+ "%16 = OpTypePointer Input %10\n" // uint32*
+ "%6 = OpVariable %11 Uniform\n" // struct{ int32[] }* out
+ "%17 = OpTypePointer Uniform %9\n" // int32*
+ "%1 = OpFunction %7 None %8\n" // -- Function begin --
+ "%18 = OpLabel\n"
+ "%19 = OpAccessChain %16 %2 %13\n" // &gl_GlobalInvocationId.x
+ "%20 = OpLoad %10 %19\n" // gl_GlobalInvocationId.x
+ "%21 = OpAccessChain %17 %6 %12 %20\n" // &in.arr[gl_GlobalInvocationId.x]
+ "%22 = OpLoad %9 %21\n" // out.arr[gl_GlobalInvocationId.x]
+ "%23 = OpAccessChain %17 %5 %12 %20\n" // &out.arr[gl_GlobalInvocationId.x]
+ "OpStore %23 %22\n" // out.arr[gl_GlobalInvocationId.x] = in[gl_GlobalInvocationId.x]
+ "OpReturn\n"
+ "OpFunctionEnd\n";
+
+ test(src.str(), [](uint32_t i) { return i; }, [](uint32_t i) { return i; });
+}
+
+TEST_P(SwiftShaderVulkanBufferToBufferComputeTest, BranchSimple)
+{
+ std::stringstream src;
+ src <<
+ "OpCapability Shader\n"
+ "OpMemoryModel Logical GLSL450\n"
+ "OpEntryPoint GLCompute %1 \"main\" %2\n"
+ "OpExecutionMode %1 LocalSize " <<
+ GetParam().localSizeX << " " <<
+ GetParam().localSizeY << " " <<
+ GetParam().localSizeZ << "\n" <<
+ "OpDecorate %3 ArrayStride 4\n"
+ "OpMemberDecorate %4 0 Offset 0\n"
+ "OpDecorate %4 BufferBlock\n"
+ "OpDecorate %5 DescriptorSet 0\n"
+ "OpDecorate %5 Binding 1\n"
+ "OpDecorate %2 BuiltIn GlobalInvocationId\n"
+ "OpDecorate %6 DescriptorSet 0\n"
+ "OpDecorate %6 Binding 0\n"
+ "%7 = OpTypeVoid\n"
+ "%8 = OpTypeFunction %7\n" // void()
+ "%9 = OpTypeInt 32 1\n" // int32
+ "%10 = OpTypeInt 32 0\n" // uint32
+ "%3 = OpTypeRuntimeArray %9\n" // int32[]
+ "%4 = OpTypeStruct %3\n" // struct{ int32[] }
+ "%11 = OpTypePointer Uniform %4\n" // struct{ int32[] }*
+ "%5 = OpVariable %11 Uniform\n" // struct{ int32[] }* in
+ "%12 = OpConstant %9 0\n" // int32(0)
+ "%13 = OpConstant %10 0\n" // uint32(0)
+ "%14 = OpTypeVector %10 3\n" // vec4<int32>
+ "%15 = OpTypePointer Input %14\n" // vec4<int32>*
+ "%2 = OpVariable %15 Input\n" // gl_GlobalInvocationId
+ "%16 = OpTypePointer Input %10\n" // uint32*
+ "%6 = OpVariable %11 Uniform\n" // struct{ int32[] }* out
+ "%17 = OpTypePointer Uniform %9\n" // int32*
+ "%1 = OpFunction %7 None %8\n" // -- Function begin --
+ "%18 = OpLabel\n"
+ "%19 = OpAccessChain %16 %2 %13\n" // &gl_GlobalInvocationId.x
+ "%20 = OpLoad %10 %19\n" // gl_GlobalInvocationId.x
+ "%21 = OpAccessChain %17 %6 %12 %20\n" // &in.arr[gl_GlobalInvocationId.x]
+ "%22 = OpLoad %9 %21\n" // in.arr[gl_GlobalInvocationId.x]
+ "%23 = OpAccessChain %17 %5 %12 %20\n" // &out.arr[gl_GlobalInvocationId.x]
+ // Start of branch logic
+ // %22 = in value
+ "OpBranch %24\n"
+ "%24 = OpLabel\n"
+ "OpBranch %25\n"
+ "%25 = OpLabel\n"
+ "OpBranch %26\n"
+ "%26 = OpLabel\n"
+ // %22 = out value
+ // End of branch logic
+ "OpStore %23 %22\n"
+ "OpReturn\n"
+ "OpFunctionEnd\n";
+
+ test(src.str(), [](uint32_t i) { return i; }, [](uint32_t i) { return i; });
+}
+
+TEST_P(SwiftShaderVulkanBufferToBufferComputeTest, BranchDeclareSSA)
+{
+ std::stringstream src;
+ src <<
+ "OpCapability Shader\n"
+ "OpMemoryModel Logical GLSL450\n"
+ "OpEntryPoint GLCompute %1 \"main\" %2\n"
+ "OpExecutionMode %1 LocalSize " <<
+ GetParam().localSizeX << " " <<
+ GetParam().localSizeY << " " <<
+ GetParam().localSizeZ << "\n" <<
+ "OpDecorate %3 ArrayStride 4\n"
+ "OpMemberDecorate %4 0 Offset 0\n"
+ "OpDecorate %4 BufferBlock\n"
+ "OpDecorate %5 DescriptorSet 0\n"
+ "OpDecorate %5 Binding 1\n"
+ "OpDecorate %2 BuiltIn GlobalInvocationId\n"
+ "OpDecorate %6 DescriptorSet 0\n"
+ "OpDecorate %6 Binding 0\n"
+ "%7 = OpTypeVoid\n"
+ "%8 = OpTypeFunction %7\n" // void()
+ "%9 = OpTypeInt 32 1\n" // int32
+ "%10 = OpTypeInt 32 0\n" // uint32
+ "%3 = OpTypeRuntimeArray %9\n" // int32[]
+ "%4 = OpTypeStruct %3\n" // struct{ int32[] }
+ "%11 = OpTypePointer Uniform %4\n" // struct{ int32[] }*
+ "%5 = OpVariable %11 Uniform\n" // struct{ int32[] }* in
+ "%12 = OpConstant %9 0\n" // int32(0)
+ "%13 = OpConstant %10 0\n" // uint32(0)
+ "%14 = OpTypeVector %10 3\n" // vec4<int32>
+ "%15 = OpTypePointer Input %14\n" // vec4<int32>*
+ "%2 = OpVariable %15 Input\n" // gl_GlobalInvocationId
+ "%16 = OpTypePointer Input %10\n" // uint32*
+ "%6 = OpVariable %11 Uniform\n" // struct{ int32[] }* out
+ "%17 = OpTypePointer Uniform %9\n" // int32*
+ "%1 = OpFunction %7 None %8\n" // -- Function begin --
+ "%18 = OpLabel\n"
+ "%19 = OpAccessChain %16 %2 %13\n" // &gl_GlobalInvocationId.x
+ "%20 = OpLoad %10 %19\n" // gl_GlobalInvocationId.x
+ "%21 = OpAccessChain %17 %6 %12 %20\n" // &in.arr[gl_GlobalInvocationId.x]
+ "%22 = OpLoad %9 %21\n" // in.arr[gl_GlobalInvocationId.x]
+ "%23 = OpAccessChain %17 %5 %12 %20\n" // &out.arr[gl_GlobalInvocationId.x]
+ // Start of branch logic
+ // %22 = in value
+ "OpBranch %24\n"
+ "%24 = OpLabel\n"
+ "%25 = OpIAdd %9 %22 %22\n" // %25 = in*2
+ "OpBranch %26\n"
+ "%26 = OpLabel\n"
+ "OpBranch %27\n"
+ "%27 = OpLabel\n"
+ // %25 = out value
+ // End of branch logic
+ "OpStore %23 %25\n" // use SSA value from previous block
+ "OpReturn\n"
+ "OpFunctionEnd\n";
+
+ test(src.str(), [](uint32_t i) { return i; }, [](uint32_t i) { return i * 2; });
+}
\ No newline at end of file
