diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-05-24 16:52:35 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2019-05-24 16:52:35 +0000 |
| commit | 0ff901fba0ae5ca950a23c3b993240b62877f42f (patch) | |
| tree | 81bbdfe13608c149cc2fdf1d5a4f319da25deda0 /llvm/lib/Target | |
| parent | 6f7734a1255975125c1ceb9b14adfd17ee7be177 (diff) | |
| download | bcm5719-llvm-0ff901fba0ae5ca950a23c3b993240b62877f42f.tar.gz bcm5719-llvm-0ff901fba0ae5ca950a23c3b993240b62877f42f.zip | |
AMDGPU: Boost inline threshold with addrspacecasted alloca arguments
This was skipping GetUnderlyingObject for nonprivate addresses, but an
alloca could also be found through an addrspacecast if it's flat.
llvm-svn: 361649
Diffstat (limited to 'llvm/lib/Target')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUInline.cpp | 7 |
1 files changed, 4 insertions, 3 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp b/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp index 84419a147c8..22c7c0d51f4 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUInline.cpp @@ -123,10 +123,11 @@ unsigned AMDGPUInliner::getInlineThreshold(CallSite CS) const { uint64_t AllocaSize = 0; SmallPtrSet<const AllocaInst *, 8> AIVisited; for (Value *PtrArg : CS.args()) { - Type *Ty = PtrArg->getType(); - if (!Ty->isPointerTy() || - Ty->getPointerAddressSpace() != AMDGPUAS::PRIVATE_ADDRESS) + PointerType *Ty = dyn_cast<PointerType>(PtrArg->getType()); + if (!Ty || (Ty->getAddressSpace() != AMDGPUAS::PRIVATE_ADDRESS && + Ty->getAddressSpace() != AMDGPUAS::FLAT_ADDRESS)) continue; + PtrArg = GetUnderlyingObject(PtrArg, DL); if (const AllocaInst *AI = dyn_cast<AllocaInst>(PtrArg)) { if (!AI->isStaticAlloca() || !AIVisited.insert(AI).second) |
