[llvm] r375030 - [AMDGPU] Do not combine dpp with physreg def

[Stanislav Mekhanoshin via llvm-commits]

Author: rampitec
Date: Wed Oct 16 11:48:54 2019
New Revision: 375030

URL: http://llvm.org/viewvc/llvm-project?rev=375030&view=rev
Log:
[AMDGPU] Do not combine dpp with physreg def

We will remove dpp mov along with the physreg def otherwise.

Differential Revision: https://reviews.llvm.org/D69063

Modified:
    llvm/trunk/lib/Target/AMDGPU/GCNDPPCombine.cpp
    llvm/trunk/test/CodeGen/AMDGPU/dpp_combine.mir

Modified: llvm/trunk/lib/Target/AMDGPU/GCNDPPCombine.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AMDGPU/GCNDPPCombine.cpp?rev=375030&r1=375029&r2=375030&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AMDGPU/GCNDPPCombine.cpp (original)
+++ llvm/trunk/lib/Target/AMDGPU/GCNDPPCombine.cpp Wed Oct 16 11:48:54 2019
@@ -353,6 +353,10 @@ bool GCNDPPCombine::combineDPPMov(Machin
   auto *DstOpnd = TII->getNamedOperand(MovMI, AMDGPU::OpName::vdst);
   assert(DstOpnd && DstOpnd->isReg());
   auto DPPMovReg = DstOpnd->getReg();
+  if (DPPMovReg.isPhysical()) {
+    LLVM_DEBUG(dbgs() << "  failed: dpp move writes physreg\n");
+    return false;
+  }
   if (execMayBeModifiedBeforeAnyUse(*MRI, DPPMovReg, MovMI)) {
     LLVM_DEBUG(dbgs() << "  failed: EXEC mask should remain the same"
                          " for all uses\n");

Modified: llvm/trunk/test/CodeGen/AMDGPU/dpp_combine.mir
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AMDGPU/dpp_combine.mir?rev=375030&r1=375029&r2=375030&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/AMDGPU/dpp_combine.mir (original)
+++ llvm/trunk/test/CodeGen/AMDGPU/dpp_combine.mir Wed Oct 16 11:48:54 2019
@@ -563,6 +563,18 @@ body: |
     %3:vgpr_32 = V_CEIL_F32_e32 %2, implicit $exec
 ...
 
+# Do not combine a dpp mov which writes a physreg.
+# GCN-LABEL: name: phys_dpp_mov_dst
+# GCN: $vgpr0 = V_MOV_B32_dpp undef %0:vgpr_32, undef %1:vgpr_32, 1, 15, 15, 1, implicit $exec
+# GCN: %2:vgpr_32 = V_CEIL_F32_e32 $vgpr0, implicit $exec
+name: phys_dpp_mov_dst
+tracksRegLiveness: true
+body: |
+  bb.0:
+    $vgpr0 = V_MOV_B32_dpp undef %1:vgpr_32, undef %0:vgpr_32, 1, 15, 15, 1, implicit $exec
+    %2:vgpr_32 = V_CEIL_F32_e32 $vgpr0, implicit $exec
+...
+
 # GCN-LABEL: name: dpp_reg_sequence_both_combined
 # GCN: %0:vreg_64 = COPY $vgpr0_vgpr1
 # GCN: %1:vreg_64 = COPY $vgpr2_vgpr3