diff options
| author | Stanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com> | 2017-05-30 16:49:24 +0000 |
|---|---|---|
| committer | Stanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com> | 2017-05-30 16:49:24 +0000 |
| commit | 56ea488d8b8c897599f13c9f9e8058cb36ff808f (patch) | |
| tree | c6091973507e778cdeebd01dc0a57f517fedd56c /llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll | |
| parent | 8c099fe06e9c4efdea698ac82df2685792ea0b6e (diff) | |
| download | bcm5719-llvm-56ea488d8b8c897599f13c9f9e8058cb36ff808f.tar.gz bcm5719-llvm-56ea488d8b8c897599f13c9f9e8058cb36ff808f.zip | |
[AMDGPU] Allow SDWA in instructions with immediates and SGPRs
An encoding does not allow to use SDWA in an instruction with
scalar operands, either literals or SGPRs. That is however possible
to copy these operands into a VGPR first.
Several copies of the value are produced if multiple SDWA conversions
were done. To cleanup MachineLICM (to hoist copies out of loops),
MachineCSE (to remove duplicate copies) and SIFoldOperands (to replace
SGPR to VGPR copy with immediate copy right to the VGPR) runs are added
after the SDWA pass.
Differential Revision: https://reviews.llvm.org/D33583
llvm-svn: 304219
Diffstat (limited to 'llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll')
| -rw-r--r-- | llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll b/llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll index 1e78c4ebcc9..176d1d25f19 100644 --- a/llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll +++ b/llvm/test/CodeGen/AMDGPU/mad24-get-global-id.ll @@ -10,7 +10,7 @@ declare i8 addrspace(2)* @llvm.amdgcn.dispatch.ptr() #0 ; GCN-LABEL: {{^}}get_global_id_0: ; GCN: s_and_b32 [[WGSIZEX:s[0-9]+]], {{s[0-9]+}}, 0xffff ; GCN: v_mov_b32_e32 [[VWGSIZEX:v[0-9]+]], [[WGSIZEX]] -; GCN: v_mad_u32_u24 v{{[0-9]+}}, [[VWGSIZEX]], s8, v0 +; GCN: v_mad_u32_u24 v{{[0-9]+}}, s8, [[VWGSIZEX]], v0 define amdgpu_kernel void @get_global_id_0(i32 addrspace(1)* %out) #1 { %dispatch.ptr = call i8 addrspace(2)* @llvm.amdgcn.dispatch.ptr() %cast.dispatch.ptr = bitcast i8 addrspace(2)* %dispatch.ptr to i32 addrspace(2)* |
