Revert "Revert "[SelectionDAG] Handle more opcodes in canCreateUndefOrPoison (#84921)" and more..."

This reverts commit 16bd10a.

Re-applies:
    b3c55b7 - "[SelectionDAG] Handle more opcodes in canCreateUndefOrPoison (#84921)"
    8e2f649 - "[DAGCombiner] Do not always fold FREEZE over BUILD_VECTOR (#85932)"
    73472c5 - "[SelectionDAG] Treat CopyFromReg as freezing the value (#85932)"

with a fix in DAGCombiner::visitFREEZE.

Author: bjope

llvm/include/llvm/CodeGen/ISDOpcodes.h

@@ -205,6 +205,7 @@ enum NodeType {
   /// CopyFromReg - This node indicates that the input value is a virtual or
   /// physical register that is defined outside of the scope of this
   /// SelectionDAG.  The register is available from the RegisterSDNode object.
+  /// Note that CopyFromReg is considered as also freezing the value.
   CopyFromReg,
 
   /// UNDEF - An undefined node.

llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

@@ -15459,6 +15459,12 @@ SDValue DAGCombiner::visitFREEZE(SDNode *N) {
   if (DAG.isGuaranteedNotToBeUndefOrPoison(N0, /*PoisonOnly*/ false))
     return N0;
 
+  // We currently avoid folding freeze over SRA/SRL, due to the problems seen
+  // with (freeze (assert ext)) blocking simplifications of SRA/SRL. See for
+  // example https://reviews.llvm.org/D136529#4120959.
+  if (N0.getOpcode() == ISD::SRA || N0.getOpcode() == ISD::SRL)
+    return SDValue();
+
   // Fold freeze(op(x, ...)) -> op(freeze(x), ...).
   // Try to push freeze through instructions that propagate but don't produce
   // poison as far as possible. If an operand of freeze follows three
@@ -15475,6 +15481,26 @@ SDValue DAGCombiner::visitFREEZE(SDNode *N) {
                                           N0.getOpcode() == ISD::BUILD_PAIR ||
                                           N0.getOpcode() == ISD::CONCAT_VECTORS;
 
+  // Avoid turning a BUILD_VECTOR that can be recognized as "all zeros", "all
+  // ones" or "constant" into something that depends on FrozenUndef. We can
+  // instead pick undef values to keep those properties, while at the same time
+  // folding away the freeze.
+  // If we implement a more general solution for folding away freeze(undef) in
+  // the future, then this special handling can be removed.
+  if (N0.getOpcode() == ISD::BUILD_VECTOR) {
+    SDLoc DL(N0);
+    EVT VT = N0.getValueType();
+    if (llvm::ISD::isBuildVectorAllOnes(N0.getNode()))
+      return DAG.getAllOnesConstant(DL, VT);
+    if (llvm::ISD::isBuildVectorOfConstantSDNodes(N0.getNode())) {
+      SmallVector<SDValue, 8> NewVecC;
+      for (const SDValue &Op : N0->op_values())
+        NewVecC.push_back(
+            Op.isUndef() ? DAG.getConstant(0, DL, Op.getValueType()) : Op);
+      return DAG.getBuildVector(VT, DL, NewVecC);
+    }
+  }
+
   SmallSetVector<SDValue, 8> MaybePoisonOperands;
   for (SDValue Op : N0->ops()) {
     if (DAG.isGuaranteedNotToBeUndefOrPoison(Op, /*PoisonOnly*/ false,

llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp

@@ -5063,6 +5063,7 @@ bool SelectionDAG::isGuaranteedNotToBeUndefOrPoison(SDValue Op,
   case ISD::VALUETYPE:
   case ISD::FrameIndex:
   case ISD::TargetFrameIndex:
+  case ISD::CopyFromReg:
     return true;
 
   case ISD::UNDEF:
@@ -5136,6 +5137,16 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
   case ISD::FREEZE:
   case ISD::CONCAT_VECTORS:
   case ISD::INSERT_SUBVECTOR:
+  case ISD::SADDSAT:
+  case ISD::UADDSAT:
+  case ISD::SSUBSAT:
+  case ISD::USUBSAT:
+  case ISD::MULHU:
+  case ISD::MULHS:
+  case ISD::SMIN:
+  case ISD::SMAX:
+  case ISD::UMIN:
+  case ISD::UMAX:
   case ISD::AND:
   case ISD::XOR:
   case ISD::ROTL:
@@ -5156,6 +5167,7 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
   case ISD::BUILD_PAIR:
     return false;
 
+  case ISD::SELECT_CC:
   case ISD::SETCC: {
     // Integer setcc cannot create undef or poison.
     if (Op.getOperand(0).getValueType().isInteger())
@@ -5165,7 +5177,8 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
     // based on options and flags. The options and flags also cause special
     // nonan condition codes to be used. Those condition codes may be preserved
     // even if the nonan flag is dropped somewhere.
-    ISD::CondCode CCCode = cast<CondCodeSDNode>(Op.getOperand(2))->get();
+    unsigned CCOp = Opcode == ISD::SETCC ? 2 : 4;
+    ISD::CondCode CCCode = cast<CondCodeSDNode>(Op.getOperand(CCOp))->get();
     if (((unsigned)CCCode & 0x10U))
       return true;
 
@@ -5182,6 +5195,8 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
     return false;
 
   case ISD::SHL:
+  case ISD::SRL:
+  case ISD::SRA:
     // If the max shift amount isn't in range, then the shift can create poison.
     return !getValidMaximumShiftAmountConstant(Op, DemandedElts);
 

llvm/test/CodeGen/AArch64/combine-mul.ll

@@ -44,8 +44,7 @@ define <4 x i1> @PR48683_vec_undef(<4 x i32> %x) {
 define i64 @combine_mul_self_demandedbits(i64 %x) {
 ; CHECK-LABEL: combine_mul_self_demandedbits:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mul x8, x0, x0
-; CHECK-NEXT:    and x0, x8, #0xfffffffffffffffd
+; CHECK-NEXT:    mul x0, x0, x0
 ; CHECK-NEXT:    ret
   %1 = mul i64 %x, %x
   %2 = and i64 %1, -3
@@ -77,7 +76,7 @@ define i8 @one_demanded_bit(i8 %x) {
 define <2 x i64> @one_demanded_bit_splat(<2 x i64> %x) {
 ; CHECK-LABEL: one_demanded_bit_splat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov w8, #32
+; CHECK-NEXT:    mov w8, #32 // =0x20
 ; CHECK-NEXT:    shl v0.2d, v0.2d, #5
 ; CHECK-NEXT:    dup v1.2d, x8
 ; CHECK-NEXT:    and v0.16b, v0.16b, v1.16b
@@ -131,7 +130,7 @@ define i32 @squared_demanded_2_low_bits(i32 %x) {
 define <2 x i64> @squared_demanded_2_low_bits_splat(<2 x i64> %x) {
 ; CHECK-LABEL: squared_demanded_2_low_bits_splat:
 ; CHECK:       // %bb.0:
-; CHECK-NEXT:    mov x8, #-2
+; CHECK-NEXT:    mov x8, #-2 // =0xfffffffffffffffe
 ; CHECK-NEXT:    dup v1.2d, x8
 ; CHECK-NEXT:    orr v0.16b, v0.16b, v1.16b
 ; CHECK-NEXT:    ret

llvm/test/CodeGen/AMDGPU/div_i128.ll

@@ -282,21 +282,21 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    ; kill: def $vgpr15 killed $vgpr15 def $vgpr15_vgpr16 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v16, v1
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v15
-; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v16
+; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v16
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v13
-; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v14
+; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v14
 ; GFX9-O0-NEXT:    v_sub_co_u32_e32 v9, vcc, v9, v4
-; GFX9-O0-NEXT:    v_subb_co_u32_e32 v1, vcc, v1, v6, vcc
-; GFX9-O0-NEXT:    v_subb_co_u32_e32 v13, vcc, v10, v4, vcc
 ; GFX9-O0-NEXT:    v_subb_co_u32_e32 v5, vcc, v5, v6, vcc
+; GFX9-O0-NEXT:    v_subb_co_u32_e32 v13, vcc, v10, v4, vcc
+; GFX9-O0-NEXT:    v_subb_co_u32_e32 v1, vcc, v1, v6, vcc
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
-; GFX9-O0-NEXT:    ; kill: def $vgpr13 killed $vgpr13 def $vgpr13_vgpr14 killed $exec
-; GFX9-O0-NEXT:    v_mov_b32_e32 v14, v5
+; GFX9-O0-NEXT:    ; kill: def $vgpr9 killed $vgpr9 def $vgpr9_vgpr10 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v5
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
-; GFX9-O0-NEXT:    ; kill: def $vgpr9 killed $vgpr9 def $vgpr9_vgpr10 killed $exec
-; GFX9-O0-NEXT:    v_mov_b32_e32 v10, v1
+; GFX9-O0-NEXT:    ; kill: def $vgpr13 killed $vgpr13 def $vgpr13_vgpr14 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v14, v1
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v12
 ; GFX9-O0-NEXT:    v_xor_b32_e64 v1, v5, v1
@@ -312,21 +312,21 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    ; kill: def $vgpr7 killed $vgpr7 def $vgpr7_vgpr8 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v1
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v1, v7
-; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v8
-; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v11
+; GFX9-O0-NEXT:    ; kill: def $vgpr8 killed $vgpr8 killed $vgpr7_vgpr8 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v11
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v12
 ; GFX9-O0-NEXT:    v_sub_co_u32_e32 v1, vcc, v1, v3
-; GFX9-O0-NEXT:    v_subb_co_u32_e32 v7, vcc, v7, v5, vcc
-; GFX9-O0-NEXT:    v_subb_co_u32_e32 v11, vcc, v8, v3, vcc
-; GFX9-O0-NEXT:    v_subb_co_u32_e32 v2, vcc, v2, v5, vcc
+; GFX9-O0-NEXT:    v_subb_co_u32_e32 v8, vcc, v8, v5, vcc
+; GFX9-O0-NEXT:    v_subb_co_u32_e32 v11, vcc, v7, v3, vcc
+; GFX9-O0-NEXT:    v_subb_co_u32_e32 v7, vcc, v2, v5, vcc
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
-; GFX9-O0-NEXT:    ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
-; GFX9-O0-NEXT:    v_mov_b32_e32 v12, v2
+; GFX9-O0-NEXT:    ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v8
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
 ; GFX9-O0-NEXT:    ; implicit-def: $sgpr4
-; GFX9-O0-NEXT:    ; kill: def $vgpr1 killed $vgpr1 def $vgpr1_vgpr2 killed $exec
-; GFX9-O0-NEXT:    v_mov_b32_e32 v2, v7
+; GFX9-O0-NEXT:    ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
+; GFX9-O0-NEXT:    v_mov_b32_e32 v12, v7
 ; GFX9-O0-NEXT:    v_xor_b32_e64 v5, v5, v6
 ; GFX9-O0-NEXT:    v_xor_b32_e64 v3, v3, v4
 ; GFX9-O0-NEXT:    ; kill: def $vgpr3 killed $vgpr3 def $vgpr3_vgpr4 killed $exec
@@ -339,18 +339,26 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:68 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:72 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v1, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v11
+; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v12
+; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:60 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v2, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v11, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:64 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v2
+; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v1
+; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:52 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v12, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v9, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:56 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v13
+; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v14
+; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:44 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v10, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
-; GFX9-O0-NEXT:    buffer_store_dword v13, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:48 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    v_mov_b32_e32 v3, v9
+; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v10
+; GFX9-O0-NEXT:    buffer_store_dword v3, off, s[0:3], s32 offset:36 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-O0-NEXT:    buffer_store_dword v14, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
+; GFX9-O0-NEXT:    buffer_store_dword v4, off, s[0:3], s32 offset:40 ; 4-byte Folded Spill
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v12
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v8, v2
 ; GFX9-O0-NEXT:    v_or_b32_e64 v3, v8, v7
@@ -403,7 +411,8 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    ; kill: def $vgpr8 killed $vgpr8 def $vgpr8_vgpr9 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v9, v5
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v9
-; GFX9-O0-NEXT:    v_cmp_ne_u64_e64 s[12:13], v[11:12], s[6:7]
+; GFX9-O0-NEXT:    s_mov_b64 s[12:13], s[6:7]
+; GFX9-O0-NEXT:    v_cmp_ne_u64_e64 s[12:13], v[11:12], s[12:13]
 ; GFX9-O0-NEXT:    v_cndmask_b32_e64 v5, v5, v10, s[12:13]
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v6
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v8
@@ -439,7 +448,8 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    ; kill: def $vgpr11 killed $vgpr11 def $vgpr11_vgpr12 killed $exec
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v12, v5
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v5, v12
-; GFX9-O0-NEXT:    v_cmp_ne_u64_e64 s[8:9], v[13:14], s[6:7]
+; GFX9-O0-NEXT:    s_mov_b64 s[8:9], s[6:7]
+; GFX9-O0-NEXT:    v_cmp_ne_u64_e64 s[8:9], v[13:14], s[8:9]
 ; GFX9-O0-NEXT:    v_cndmask_b32_e64 v5, v5, v8, s[8:9]
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v7, v6
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v6, v11
@@ -690,10 +700,10 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    buffer_load_dword v28, off, s[0:3], s32 offset:260 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v25, off, s[0:3], s32 offset:264 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v26, off, s[0:3], s32 offset:268 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v19, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v20, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v21, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v22, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v19, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v20, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v21, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v22, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v14, off, s[0:3], s32 offset:272 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v15, off, s[0:3], s32 offset:276 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v11, off, s[0:3], s32 offset:280 ; 4-byte Folded Reload
@@ -903,14 +913,14 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    s_or_saveexec_b64 s[18:19], -1
 ; GFX9-O0-NEXT:    buffer_load_dword v16, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT:    buffer_load_dword v17, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v18, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v13, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v14, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v19, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v20, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v21, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v22, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v17, off, s[0:3], s32 offset:60 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v18, off, s[0:3], s32 offset:64 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v13, off, s[0:3], s32 offset:52 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v14, off, s[0:3], s32 offset:56 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v19, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v20, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v21, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v22, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(9)
 ; GFX9-O0-NEXT:    v_mov_b32_e32 v4, v10
 ; GFX9-O0-NEXT:    s_waitcnt vmcnt(0)
@@ -1028,10 +1038,10 @@ define i128 @v_sdiv_i128_vv(i128 %lhs, i128 %rhs) {
 ; GFX9-O0-NEXT:    s_or_saveexec_b64 s[18:19], -1
 ; GFX9-O0-NEXT:    buffer_load_dword v0, off, s[0:3], s32 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    s_mov_b64 exec, s[18:19]
-; GFX9-O0-NEXT:    buffer_load_dword v7, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v8, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v11, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
-; GFX9-O0-NEXT:    buffer_load_dword v12, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v7, off, s[0:3], s32 offset:36 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v8, off, s[0:3], s32 offset:40 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v11, off, s[0:3], s32 offset:44 ; 4-byte Folded Reload
+; GFX9-O0-NEXT:    buffer_load_dword v12, off, s[0:3], s32 offset:48 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v5, off, s[0:3], s32 offset:28 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v6, off, s[0:3], s32 offset:32 ; 4-byte Folded Reload
 ; GFX9-O0-NEXT:    buffer_load_dword v1, off, s[0:3], s32 offset:20 ; 4-byte Folded Reload