[AMDGPU] Disable dpp src1 sgpr on gfx11

[ptrojahn]

#67461 enabled SGPRs as src1 by default for all dpp opcodes with manual checks for targets where this is not supported. In that case, isOperandLegal checked if the second operand is legal as src0. #155595 disabled this check by removing the calls to isOperandLegal, which resulted in SGPRs being used as operands for src1 on gfx11. This PR reenables this check and fixes the lit test.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Author: Paul Trojahn (ptrojahn)

Changes

#67461 enabled SGPRs as src1 by default for all opcodes with manual checks for targets where this is not supported. #155595 disabled this check for hasDPPSrc1SGPR, which resulted in SGPRs being used as operands for src1 on gfx11. This PR reenables this check and fixes the lit test.

---

Full diff: https://github.com/llvm/llvm-project/pull/164241.diff

2 Files Affected:

• (modified) llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp (+5-3)
• (modified) llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir (+7-5)

diff --git a/llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp b/llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp
index 464cbec6c46bc..e394096b9fe09 100644
--- a/llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNDPPCombine.cpp
@@ -307,10 +307,12 @@ MachineInstr *GCNDPPCombine::createDPPInst(MachineInstr &OrigMI,
     if (Src1) {
       assert(AMDGPU::hasNamedOperand(DPPOp, AMDGPU::OpName::src1) &&
              "dpp version of instruction missing src1");
-      // If subtarget does not support SGPRs for src1 operand then the
-      // requirements are the same as for src0. We check src0 instead because
-      // pseudos are shared between subtargets and allow SGPR for src1 on all.
       if (!ST->hasDPPSrc1SGPR()) {
+        if (Src1->isReg() && ST->getRegisterInfo()->isSGPRClass(
+                                 MRI->getRegClass(Src1->getReg()))) {
+          Fail = true;
+          break;
+        }
         assert(TII->getOpSize(*DPPInst, Src0Idx) ==
                    TII->getOpSize(*DPPInst, NumOperands) &&
                "Src0 and Src1 operands should have the same size");
diff --git a/llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir b/llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir
index 3725384e885ee..87ec1eabc2033 100644
--- a/llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir
+++ b/llvm/test/CodeGen/AMDGPU/dpp_combine_gfx11.mir
@@ -1,6 +1,6 @@
-# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN
-# RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN
-# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN,GFX1100
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1150 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN,GFX1150
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1200 -mattr=-real-true16 -run-pass=gcn-dpp-combine -verify-machineinstrs -o - %s | FileCheck %s -check-prefixes=GCN,GFX1150
 
 ---
 
@@ -38,8 +38,10 @@ body:             |
 
 # GCN-LABEL: name: vop3_sgpr_src1
 # GCN: %6:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, %1, 0, %2, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
-# GCN: %8:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, %2, 0, %1, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
-# GCN: %10:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, %2, 0, %3, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GFX1100: %8:vgpr_32 = V_MED3_F32_e64 0, %7, 0, %2, 0, %1, 0, 0, implicit $mode, implicit $exec
+# GFX1150: %8:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, %2, 0, %1, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
+# GFX1100: %10:vgpr_32 = V_MED3_F32_e64 0, %9, 0, %2, 0, %3, 0, 0, implicit $mode, implicit $exec
+# GFX1150: %10:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, %2, 0, %3, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
 # GCN: %12:vgpr_32 = V_MED3_F32_e64_dpp %4, 0, %0, 0, 42, 0, %2, 0, 0, 1, 15, 15, 1, implicit $mode, implicit $exec
 # GCN: %14:vgpr_32 = V_MED3_F32_e64 0, %13, 0, 4242, 0, %2, 0, 0, implicit $mode, implicit $exec
 name: vop3_sgpr_src1

[arsenm]

Can you include an IR test that hits this problem? I also don't see this case handled in isOperandLegal

[arsenm]

Preserve this comment? Also, can we start using RegClassByHwMode to avoid this problem?