shader/other: Implement BAR.SYNC 0x0
Trivially implement this particular case of BAR. Unless games use OpenCL or CUDA barriers, we shouldn't hit any other case here.
This commit is contained in:
parent
cf4ee279c6
commit
5d0986a53b
2
externals/sirit
vendored
2
externals/sirit
vendored
|
@ -1 +1 @@
|
|||
Subproject commit 414fc4dbd28d8fe48f735a0c389db8a234f733c0
|
||||
Subproject commit a62c5bbc100a5e5a31ea0ccc4a78d8fa6a4167ce
|
|
@ -2321,6 +2321,15 @@ private:
|
|||
return {fmt::format("readInvocationARB({}, {})", value, index), Type::Float};
|
||||
}
|
||||
|
||||
Expression Barrier(Operation) {
|
||||
if (!ir.IsDecompiled()) {
|
||||
LOG_ERROR(Render_OpenGL, "barrier() used but shader is not decompiled");
|
||||
return {};
|
||||
}
|
||||
code.AddLine("barrier();");
|
||||
return {};
|
||||
}
|
||||
|
||||
Expression MemoryBarrierGL(Operation) {
|
||||
code.AddLine("memoryBarrier();");
|
||||
return {};
|
||||
|
@ -2556,6 +2565,7 @@ private:
|
|||
&GLSLDecompiler::ThreadId,
|
||||
&GLSLDecompiler::ShuffleIndexed,
|
||||
|
||||
&GLSLDecompiler::Barrier,
|
||||
&GLSLDecompiler::MemoryBarrierGL,
|
||||
};
|
||||
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
|
||||
|
|
|
@ -2181,6 +2181,22 @@ private:
|
|||
return {OpSubgroupReadInvocationKHR(t_float, value, index), Type::Float};
|
||||
}
|
||||
|
||||
Expression Barrier(Operation) {
|
||||
if (!ir.IsDecompiled()) {
|
||||
LOG_ERROR(Render_Vulkan, "OpBarrier used by shader is not decompiled");
|
||||
return {};
|
||||
}
|
||||
|
||||
const auto scope = spv::Scope::Workgroup;
|
||||
const auto memory = spv::Scope::Workgroup;
|
||||
const auto semantics =
|
||||
spv::MemorySemanticsMask::WorkgroupMemory | spv::MemorySemanticsMask::AcquireRelease;
|
||||
OpControlBarrier(Constant(t_uint, static_cast<u32>(scope)),
|
||||
Constant(t_uint, static_cast<u32>(memory)),
|
||||
Constant(t_uint, static_cast<u32>(semantics)));
|
||||
return {};
|
||||
}
|
||||
|
||||
Expression MemoryBarrierGL(Operation) {
|
||||
const auto scope = spv::Scope::Device;
|
||||
const auto semantics =
|
||||
|
@ -2641,6 +2657,7 @@ private:
|
|||
&SPIRVDecompiler::ThreadId,
|
||||
&SPIRVDecompiler::ShuffleIndexed,
|
||||
|
||||
&SPIRVDecompiler::Barrier,
|
||||
&SPIRVDecompiler::MemoryBarrierGL,
|
||||
};
|
||||
static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
|
||||
|
|
|
@ -272,6 +272,11 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) {
|
|||
SetRegister(bb, instr.gpr0, GetRegister(instr.gpr8));
|
||||
break;
|
||||
}
|
||||
case OpCode::Id::BAR: {
|
||||
UNIMPLEMENTED_IF_MSG(instr.value != 0xF0A81B8000070000ULL, "BAR is not BAR.SYNC 0x0");
|
||||
bb.push_back(Operation(OperationCode::Barrier));
|
||||
break;
|
||||
}
|
||||
case OpCode::Id::MEMBAR: {
|
||||
UNIMPLEMENTED_IF(instr.membar.type != Tegra::Shader::MembarType::GL);
|
||||
UNIMPLEMENTED_IF(instr.membar.unknown != Tegra::Shader::MembarUnknown::Default);
|
||||
|
|
|
@ -228,6 +228,7 @@ enum class OperationCode {
|
|||
ThreadId, /// () -> uint
|
||||
ShuffleIndexed, /// (uint value, uint index) -> uint
|
||||
|
||||
Barrier, /// () -> void
|
||||
MemoryBarrierGL, /// () -> void
|
||||
|
||||
Amount,
|
||||
|
|
Loading…
Reference in a new issue