AMDGPU: Fix SDWA crash on inline asm

[arsenm]

Summary

This was only searching for explicit defs,
and asserting for any implicit or variadic
instruction defs, like inline asm.

[rampitec]

LGTM

[arsenm]

r319826

lib/Target/AMDGPU/SIPeepholeSDWA.cpp

static MachineOperand *findSingleRegDef(const MachineOperand *Reg,
   if (!DefInstr)
     return nullptr;
 
   for (auto &DefMO : DefInstr->defs()) {
     if (DefMO.isReg() && DefMO.getReg() == Reg->getReg())
       return &DefMO;
   }
 
-  llvm_unreachable("invalid reg");
+  // Ignore implicit defs.
+  return nullptr;
 }
 
 uint64_t SDWASrcOperand::getSrcMods(const SIInstrInfo *TII,
                                     const MachineOperand *SrcOp) const {
   uint64_t Mods = 0;
   const auto *MI = SrcOp->getParent();
   if (TII->getNamedOperand(*MI, AMDGPU::OpName::src0) == SrcOp) {
     if (auto *Mod = TII->getNamedOperand(*MI, AMDGPU::OpName::src0_modifiers)) {

test/CodeGen/AMDGPU/sdwa-peephole.ll

entry:
   %tmp17 = shufflevector <2 x i8> %tmp10, <2 x i8> %tmp12, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %tmp18 = shufflevector <2 x i8> %tmp14, <2 x i8> %tmp16, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
   %tmp19 = shufflevector <4 x i8> %tmp17, <4 x i8> %tmp18, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 
   %arrayidx5 = getelementptr inbounds <8 x i8>, <8 x i8> addrspace(1)* %destValues, i64 %idxprom
   store <8 x i8> %tmp19, <8 x i8> addrspace(1)* %arrayidx5, align 8
   ret void
 }
+
+; GCN-LABEL: {{^}}sdwa_crash_inlineasm_de
+; GCN: s_mov_b32 s{{[0-9]+}}, 0xffff
+; GCN: v_and_b32_e32 v{{[0-9]+}}, s{{[0-9]+}}, v{{[0-9]+}}
+; GCN: v_or_b32_e32 v{{[0-9]+}}, 0x10000,
+define amdgpu_kernel void @sdwa_crash_inlineasm_def() #0 {
+bb:
+  br label %bb1
+
+bb1:                                              ; preds = %bb11, %bb
+  %tmp = phi <2 x i32> [ %tmp12, %bb11 ], [ undef, %bb ]
+  br i1 true, label %bb2, label %bb11
+
+bb2:                                              ; preds = %bb1
+  %tmp3 = call i32 asm "v_and_b32_e32 $0, $1, $2", "=v,s,v"(i32 65535, i32 undef) #1
+  %tmp5 = or i32 %tmp3, 65536
+  %tmp6 = insertelement <2 x i32> %tmp, i32 %tmp5, i64 0
+  br label %bb11
+
+bb11:                                             ; preds = %bb10, %bb2
+  %tmp12 = phi <2 x i32> [ %tmp6, %bb2 ], [ %tmp, %bb1 ]
+  br label %bb1
+}