diff options
| author | Tom Stellard <thomas.stellard@amd.com> | 2015-12-15 23:15:25 +0000 |
|---|---|---|
| committer | Tom Stellard <thomas.stellard@amd.com> | 2015-12-15 23:15:25 +0000 |
| commit | 7750f4ed9e223394cc6a552c9129f619b1ed3bb2 (patch) | |
| tree | a22deff371ade965bedef35f839a1fd227683c45 /llvm | |
| parent | 3eb47e223ed928f9eb65378a99c578fc47bbaa89 (diff) | |
| download | bcm5719-llvm-7750f4ed9e223394cc6a552c9129f619b1ed3bb2.tar.gz bcm5719-llvm-7750f4ed9e223394cc6a552c9129f619b1ed3bb2.zip | |
AMDGPU/SI: Set the code object work group segment size when targeting HSA
Reviewers: arsenm
Subscribers: arsenm, llvm-commits
Differential Revision: http://reviews.llvm.org/D15493
llvm-svn: 255702
Diffstat (limited to 'llvm')
| -rw-r--r-- | llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp | 1 | ||||
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/private-memory.ll | 5 |
2 files changed, 6 insertions, 0 deletions
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp index be0df57e9f0..ac8f568da18 100644 --- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp +++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp @@ -620,6 +620,7 @@ void AMDGPUAsmPrinter::EmitAmdKernelCodeT(const MachineFunction &MF, header.wavefront_sgpr_count = KernelInfo.NumSGPR; header.workitem_vgpr_count = KernelInfo.NumVGPR; header.workitem_private_segment_byte_size = KernelInfo.ScratchSize; + header.workgroup_group_segment_byte_size = KernelInfo.LDSSize; AMDGPUTargetStreamer *TS = static_cast<AMDGPUTargetStreamer *>(OutStreamer->getTargetStreamer()); diff --git a/llvm/test/CodeGen/AMDGPU/private-memory.ll b/llvm/test/CodeGen/AMDGPU/private-memory.ll index 628b8a21b81..79778eebd80 100644 --- a/llvm/test/CodeGen/AMDGPU/private-memory.ll +++ b/llvm/test/CodeGen/AMDGPU/private-memory.ll @@ -1,5 +1,6 @@ ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck %s -check-prefix=R600 -check-prefix=FUNC ; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=SI < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC +; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -mtriple=amdgcn--amdhsa -mcpu=kaveri < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC -check-prefix=HSA-PROMOTE ; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=SI < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC ; RUN: llc -show-mc-encoding -mattr=-promote-alloca -verify-machineinstrs -mtriple=amdgcn-amdhsa -mcpu=kaveri < %s | FileCheck %s -check-prefix=SI-ALLOCA -check-prefix=SI -check-prefix=FUNC -check-prefix=HSA-ALLOCA ; RUN: llc -show-mc-encoding -mattr=+promote-alloca -verify-machineinstrs -march=amdgcn -mcpu=tonga < %s | FileCheck %s -check-prefix=SI-PROMOTE -check-prefix=SI -check-prefix=FUNC @@ -14,6 +15,10 @@ declare i32 @llvm.r600.read.tidig.x() nounwind readnone ; R600: LDS_READ ; R600: LDS_READ +; HSA-PROMOTE: .amd_kernel_code_t +; HSA-PROMOTE: workgroup_group_segment_byte_size = 5120 +; HSA-PROMOTE: .end_amd_kernel_code_t + ; SI-PROMOTE: ds_write_b32 ; SI-PROMOTE: ds_write_b32 ; SI-PROMOTE: ds_read_b32 |
