diff options
| author | Matt Arsenault <Matthew.Arsenault@amd.com> | 2017-01-27 17:42:26 +0000 |
|---|---|---|
| committer | Matt Arsenault <Matthew.Arsenault@amd.com> | 2017-01-27 17:42:26 +0000 |
| commit | d8f7ea381f8958b94d3eb82103be427e8a283347 (patch) | |
| tree | af4046f997f9d7608f7fa923e4b59f88c25ea963 /llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp | |
| parent | 3650df13bea0566c6cdaa16812b548d3d6aa61cf (diff) | |
| download | bcm5719-llvm-d8f7ea381f8958b94d3eb82103be427e8a283347.tar.gz bcm5719-llvm-d8f7ea381f8958b94d3eb82103be427e8a283347.zip | |
AMDGPU: Enable FeatureFlatForGlobal on Volcanic Islands
Accomplishes what r292982 was supposed to, which ended up
only really making the necessary test changes.
This should be applied to the 4.0 branch.
Patch by Vedran Miletić <vedran@miletic.net>
llvm-svn: 293310
Diffstat (limited to 'llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp | 8 |
1 files changed, 7 insertions, 1 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp index f19382ea0a9..c85d2159bdb 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp @@ -49,6 +49,13 @@ AMDGPUSubtarget::initializeSubtargetDependencies(const Triple &TT, ParseSubtargetFeatures(GPU, FullFS); + // Unless +-flat-for-global is specified, turn on FlatForGlobal for all OS-es + // on VI and newer hardware to avoid assertion failures due to missing ADDR64 + // variants of MUBUF instructions. + if (!hasAddr64() && !FS.contains("flat-for-global")) { + FlatForGlobal = true; + } + // FIXME: I don't think think Evergreen has any useful support for // denormals, but should be checked. Should we issue a warning somewhere // if someone tries to enable these? @@ -82,7 +89,6 @@ AMDGPUSubtarget::AMDGPUSubtarget(const Triple &TT, StringRef GPU, StringRef FS, FP64FP16Denormals(false), FPExceptions(false), FlatForGlobal(false), - NoAddr64(false), UnalignedScratchAccess(false), UnalignedBufferAccess(false), |
