[llvm] r268295 - AMDGPU/SI: Use v_readfirstlane_b32 when restoring SGPRs spilled to scratch

[Tom Stellard via llvm-commits]

Author: tstellar
Date: Mon May  2 15:11:44 2016
New Revision: 268295

URL: http://llvm.org/viewvc/llvm-project?rev=268295&view=rev
Log:
AMDGPU/SI: Use v_readfirstlane_b32 when restoring SGPRs spilled to scratch

We were using v_readlane_b32 with the lane set to zero, but this won't
work if thread 0 is not active.

Differential Revision: http://reviews.llvm.org/D19745

Modified:
    llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
    llvm/trunk/test/CodeGen/AMDGPU/si-spill-sgpr-stack.ll

Modified: llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp?rev=268295&r1=268294&r2=268295&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/SIRegisterInfo.cpp Mon May  2 15:11:44 2016
@@ -589,9 +589,8 @@ void SIRegisterInfo::eliminateFrameIndex
                   .addImm(i * 4)                          // offset
                   .addMemOperand(MMO);
           BuildMI(*MBB, MI, DL,
-                  TII->getMCOpcodeFromPseudo(AMDGPU::V_READLANE_B32), SubReg)
+                  TII->get(AMDGPU::V_READFIRSTLANE_B32), SubReg)
                   .addReg(TmpReg, RegState::Kill)
-                  .addImm(0)
                   .addReg(MI->getOperand(0).getReg(), RegState::ImplicitDefine);
         }
       }

Modified: llvm/trunk/test/CodeGen/AMDGPU/si-spill-sgpr-stack.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/si-spill-sgpr-stack.ll?rev=268295&r1=268294&r2=268295&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/si-spill-sgpr-stack.ll (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/si-spill-sgpr-stack.ll Mon May  2 15:11:44 2016
@@ -5,7 +5,7 @@
 ; Make sure we are handling hazards correctly.
 ; CHECK: buffer_load_dword [[VHI:v[0-9]+]], off, s[{{[0-9]+:[0-9]+}}], s{{[0-9]+}} offset:12
 ; CHECK-NEXT: s_waitcnt vmcnt(0)
-; CHECK-NEXT: v_readlane_b32 s[[HI:[0-9]+]], [[VHI]]
+; CHECK-NEXT: v_readfirstlane_b32 s[[HI:[0-9]+]], [[VHI]]
 ; CHECK-NEXT: s_nop 4
 ; CHECK-NEXT: buffer_store_dword v0, off, s[0:[[HI]]{{\]}}, 0
 ; CHECK: s_endpgm