RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-02-04 03:06:28 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

AMDGPU: Fix default device handling

Browse Source 
When no device name is specified, default to kaveri
for HSA since SI is not supported and it woud fail.

Default to "tahiti" instead of "SI" since these are
effectively the same, and tahiti is an actual device.

Move default device handling to the TargetMachine
rather than the AMDGPUSubtarget. The module ISA version
is computed from the device name provided with the target
machine, so the attributes printed by the AsmPrinter were
inconsistent with those computed in the subtarget.

Also remove DevName field from subtarget since it's redundant
with getCPU() in the superclass.

llvm-svn: 258901
This commit is contained in:
Matt Arsenault 2016-01-27 02:17:49 +00:00
parent b49dfe295d
commit 06feef3a96
4 changed files with 28 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/Target/AMDGPU/AMDGPUSubtarget.cpp
View File

@ -49,9 +49,6 @@ AMDGPUSubtarget::initializeSubtargetDependencies(const Triple &TT,
FullFS += "+flat-for-global,";
FullFS += FS;
if (GPU == "" && TT.getArch() == Triple::amdgcn)
GPU = "SI";
ParseSubtargetFeatures(GPU, FullFS);
// FIXME: I don't think think Evergreen has any useful support for
@ -66,7 +63,7 @@ AMDGPUSubtarget::initializeSubtargetDependencies(const Triple &TT,
AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
TargetMachine &TM)
: AMDGPUGenSubtargetInfo(TT, GPU, FS), DevName(GPU),
: AMDGPUGenSubtargetInfo(TT, GPU, FS),
DumpCode(false), R600ALUInst(false), HasVertexCache(false),
TexVTXClauseSize(0), Gen(AMDGPUSubtarget::R600), FP64(false),
FP64Denormals(false), FP32Denormals(false), FastFMAF32(false),

5
lib/Target/AMDGPU/AMDGPUSubtarget.h
View File

@ -58,7 +58,6 @@ public:
};
private:
std::string DevName;
bool DumpCode;
bool R600ALUInst;
bool HasVertexCache;
@ -274,10 +273,6 @@ public:
return false;
}
StringRef getDeviceName() const {
return DevName;
}
bool enableHugeScratchBuffer() const {
return EnableHugeScratchBuffer;
}

18
lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
View File

@ -88,14 +88,28 @@ static std::string computeDataLayout(const Triple &TT) {
return Ret;
}
LLVM_READNONE
static StringRef getGPUOrDefault(const Triple &TT, StringRef GPU) {
if (!GPU.empty())
return GPU;
// HSA only supports CI+, so change the default GPU to a CI for HSA.
if (TT.getArch() == Triple::amdgcn)
return (TT.getOS() == Triple::AMDHSA) ? "kaveri" : "tahiti";
return "";
}
AMDGPUTargetMachine::AMDGPUTargetMachine(const Target &T, const Triple &TT,
StringRef CPU, StringRef FS,
TargetOptions Options, Reloc::Model RM,
CodeModel::Model CM,
CodeGenOpt::Level OptLevel)
: LLVMTargetMachine(T, computeDataLayout(TT), TT, CPU, FS, Options, RM, CM,
: LLVMTargetMachine(T, computeDataLayout(TT), TT,
getGPUOrDefault(TT, CPU), FS, Options, RM, CM,
OptLevel),
TLOF(createTLOF(getTargetTriple())), Subtarget(TT, CPU, FS, *this),
TLOF(createTLOF(getTargetTriple())),
Subtarget(TT, getTargetCPU(), FS, *this),
IntrinsicInfo() {
setRequiresStructuredCFG(true);
initAsmInfo();

11
test/CodeGen/AMDGPU/hsa-default-device.ll Normal file
View File

@ -0,0 +1,11 @@
; RUN: llc -march=amdgcn -mtriple=amdgcn-unknown-amdhsa < %s | FileCheck %s
; Make sure that with an HSA triple, we don't default to an
; unsupported device.
; CHECK: .hsa_code_object_isa 7,0,0,"AMD","AMDGPU"
define void @test_kernel(float addrspace(1)* %out0, double addrspace(1)* %out1) nounwind {
store float 0.0, float addrspace(1)* %out0
ret void
}