[GlobalISel] Combine out-of-range shifts to undef

[arsenm]

llvm/include/llvm/CodeGen/GlobalISel/CombinerHelper.h

793

matchOutOfBoundsShifts?

[paquette]

I think this seems reasonable?

llvm/include/llvm/CodeGen/GlobalISel/CombinerHelper.h

792

[aemerson]

llvm/include/llvm/CodeGen/GlobalISel/CombinerHelper.h

793

I dunno, I like the childish naïveté of "too big"

llvm/include/llvm/CodeGen/GlobalISel/CombinerHelper.h

public:
   ///   (X + Y) == X -> Y == 0
   ///   (X - Y) == X -> Y == 0
   ///   (X ^ Y) == X -> Y == 0
   ///   (X + Y) != X -> Y != 0
   ///   (X - Y) != X -> Y != 0
   ///   (X ^ Y) != X -> Y != 0
   bool matchRedundantBinOpInEquality(MachineInstr &MI, BuildFnTy &MatchInfo);
 
+  /// Match shifts greater or equal to the bitwidth of the operation.
+  bool matchShiftsTooBig(MachineInstr &MI);
+
 private:
   /// Given a non-indexed load or store instruction \p MI, find an offset that
   /// can be usefully and legally folded into it as a post-indexing operation.
   ///
   /// \returns true if a candidate is found.
   bool findPostIndexCandidate(MachineInstr &MI, Register &Addr, Register &Base,
                               Register &Offset);
 

llvm/include/llvm/Target/GlobalISel/Combine.td

 
 def ptr_add_immed_matchdata : GIDefMatchData<"PtrAddChain">;
 def ptr_add_immed_chain : GICombineRule<
   (defs root:$d, ptr_add_immed_matchdata:$matchinfo),
   (match (wip_match_opcode G_PTR_ADD):$d,
          [{ return Helper.matchPtrAddImmedChain(*${d}, ${matchinfo}); }]),
   (apply [{ Helper.applyPtrAddImmedChain(*${d}, ${matchinfo}); }])>;
 
+def shifts_too_big : GICombineRule<
+  (defs root:$root),
+  (match (wip_match_opcode G_SHL, G_ASHR, G_LSHR):$root,
+         [{ return Helper.matchShiftsTooBig(*${root}); }]),
+  (apply [{ Helper.replaceInstWithUndef(*${root}); }])>;
+
 // Fold shift (shift base x), y -> shift base, (x+y), if shifts are same
 def shift_immed_matchdata : GIDefMatchData<"RegisterImmPair">;
 def shift_immed_chain : GICombineRule<
   (defs root:$d, shift_immed_matchdata:$matchinfo),
   (match (wip_match_opcode G_SHL, G_ASHR, G_LSHR, G_SSHLSAT, G_USHLSAT):$d,
          [{ return Helper.matchShiftImmedChain(*${d}, ${matchinfo}); }]),
   (apply [{ Helper.applyShiftImmedChain(*${d}, ${matchinfo}); }])>;
 
def fma_combines : GICombineGroup<[combine_fadd_fmul_to_fmad_or_fma,
   combine_fadd_fpext_fmul_to_fmad_or_fma, combine_fadd_fma_fmul_to_fmad_or_fma,
   combine_fadd_fpext_fma_fmul_to_fmad_or_fma, combine_fsub_fmul_to_fmad_or_fma,
   combine_fsub_fneg_fmul_to_fmad_or_fma, combine_fsub_fpext_fmul_to_fmad_or_fma,
   combine_fsub_fpext_fneg_fmul_to_fmad_or_fma]>;
 
 def all_combines : GICombineGroup<[trivial_combines, insert_vec_elt_combines,
     extract_vec_elt_combines, combines_for_extload,
     combine_indexed_load_store, undef_combines, identity_combines, phi_combines,
-    simplify_add_to_sub, hoist_logic_op_with_same_opcode_hands,
+    simplify_add_to_sub, hoist_logic_op_with_same_opcode_hands, shifts_too_big,
     reassocs, ptr_add_immed_chain,
     shl_ashr_to_sext_inreg, sext_inreg_of_load,
     width_reduction_combines, select_combines,
     known_bits_simplifications, ext_ext_fold,
     not_cmp_fold, opt_brcond_by_inverting_cond,
     unmerge_merge, unmerge_cst, unmerge_dead_to_trunc,
     unmerge_zext_to_zext, merge_unmerge, trunc_ext_fold, trunc_shift,
     const_combines, xor_of_and_with_same_reg, ptr_add_with_zero,

llvm/lib/CodeGen/GlobalISel/CombinerHelper.cpp

bool CombinerHelper::matchRedundantBinOpInEquality(MachineInstr &MI,
   }
   MatchInfo = [=](MachineIRBuilder &B) {
     auto Zero = B.buildConstant(MRI.getType(Y), 0);
     B.buildICmp(Pred, Dst, Y, Zero);
   };
   return CmpInst::isEquality(Pred) && Y.isValid();
 }
 
+bool CombinerHelper::matchShiftsTooBig(MachineInstr &MI) {
+  Register ShiftReg = MI.getOperand(2).getReg();
+  LLT ResTy = MRI.getType(MI.getOperand(0).getReg());
+  auto IsShiftTooBig = [&](const Constant *C) {
+    auto *CI = dyn_cast<ConstantInt>(C);
+    return CI && CI->uge(ResTy.getScalarSizeInBits());
+  };
+  return matchUnaryPredicate(MRI, ShiftReg, IsShiftTooBig);
+}
+
 bool CombinerHelper::tryCombine(MachineInstr &MI) {
   if (tryCombineCopy(MI))
     return true;
   if (tryCombineExtendingLoads(MI))
     return true;
   if (tryCombineIndexedLoadStore(MI))
     return true;
   return false;
 }

llvm/test/CodeGen/AArch64/GlobalISel/combine-shifts-undef.mir

+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple aarch64 -run-pass=aarch64-prelegalizer-combiner -verify-machineinstrs %s -o - | FileCheck %s
+---
+name:            shl_by_ge_bw
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$w0' }
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: shl_by_ge_bw
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[DEF]](s16)
+    ; CHECK-NEXT: $w0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %1:_(s32) = COPY $w0
+    %0:_(s16) = G_TRUNC %1(s32)
+    %2:_(s16) = G_CONSTANT i16 20
+    %3:_(s16) = G_SHL %0, %2(s16)
+    %4:_(s32) = G_ANYEXT %3(s16)
+    $w0 = COPY %4(s32)
+    RET_ReallyLR implicit $w0
+
+...
+---
+name:            lshr_by_ge_bw
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$w0' }
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: lshr_by_ge_bw
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[DEF]](s16)
+    ; CHECK-NEXT: $w0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %1:_(s32) = COPY $w0
+    %0:_(s16) = G_TRUNC %1(s32)
+    %2:_(s16) = G_CONSTANT i16 16
+    %3:_(s16) = G_LSHR %0, %2(s16)
+    %4:_(s32) = G_ANYEXT %3(s16)
+    $w0 = COPY %4(s32)
+    RET_ReallyLR implicit $w0
+
+...
+---
+name:            ashr_by_ge_bw
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$w0' }
+body:             |
+  bb.1:
+    liveins: $w0
+
+    ; CHECK-LABEL: name: ashr_by_ge_bw
+    ; CHECK: liveins: $w0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s16) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[DEF]](s16)
+    ; CHECK-NEXT: $w0 = COPY [[ANYEXT]](s32)
+    ; CHECK-NEXT: RET_ReallyLR implicit $w0
+    %1:_(s32) = COPY $w0
+    %0:_(s16) = G_TRUNC %1(s32)
+    %2:_(s16) = G_CONSTANT i16 20
+    %3:_(s16) = G_ASHR %0, %2(s16)
+    %4:_(s32) = G_ANYEXT %3(s16)
+    $w0 = COPY %4(s32)
+    RET_ReallyLR implicit $w0
+
+...
+---
+name:            shl_by_ge_bw_vector
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$q0' }
+body:             |
+  bb.1:
+    liveins: $q0
+
+    ; CHECK-LABEL: name: shl_by_ge_bw_vector
+    ; CHECK: liveins: $q0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: %shl:_(<4 x s32>) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $q0 = COPY %shl(<4 x s32>)
+    ; CHECK-NEXT: RET_ReallyLR implicit $q0
+    %1:_(<4 x s32>) = COPY $q0
+    %0:_(s32) = G_CONSTANT i32 32
+    %bv:_(<4 x s32>) = G_BUILD_VECTOR %0, %0, %0, %0
+    %shl:_(<4 x s32>) = G_SHL %1, %bv(<4 x s32>)
+    $q0 = COPY %shl(<4 x s32>)
+    RET_ReallyLR implicit $q0
+
+...
+---
+name:            shl_by_ge_bw_vector_partial
+alignment:       4
+tracksRegLiveness: true
+liveins:
+  - { reg: '$q0' }
+body:             |
+  bb.1:
+    liveins: $q0
+
+    ; CHECK-LABEL: name: shl_by_ge_bw_vector_partial
+    ; CHECK: liveins: $q0
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(<4 x s32>) = COPY $q0
+    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 32
+    ; CHECK-NEXT: %small:_(s32) = G_CONSTANT i32 4
+    ; CHECK-NEXT: %bv:_(<4 x s32>) = G_BUILD_VECTOR [[C]](s32), [[C]](s32), [[C]](s32), %small(s32)
+    ; CHECK-NEXT: %shl:_(<4 x s32>) = G_SHL [[COPY]], %bv(<4 x s32>)
+    ; CHECK-NEXT: $q0 = COPY %shl(<4 x s32>)
+    ; CHECK-NEXT: RET_ReallyLR implicit $q0
+    %1:_(<4 x s32>) = COPY $q0
+    %0:_(s32) = G_CONSTANT i32 32
+    %small:_(s32) = G_CONSTANT i32 4
+    %bv:_(<4 x s32>) = G_BUILD_VECTOR %0, %0, %0, %small
+    %shl:_(<4 x s32>) = G_SHL %1, %bv(<4 x s32>)
+    $q0 = COPY %shl(<4 x s32>)
+    RET_ReallyLR implicit $q0
+...

llvm/test/CodeGen/AMDGPU/GlobalISel/combine-amdgpu-cvt-f32-ubyte.mir

 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0
 
     ; CHECK-LABEL: name: cvt_f32_ubyte0_zext_lshr_16
     ; CHECK: liveins: $vgpr0
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: %arg:_(s32) = COPY $vgpr0
-    ; CHECK-NEXT: %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE2 %arg
+    ; CHECK-NEXT: %shift:_(s16) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: %zext:_(s32) = G_ZEXT %shift(s16)
+    ; CHECK-NEXT: %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE0 %zext
     ; CHECK-NEXT: $vgpr0 = COPY %result(s32)
     %arg:_(s32) = COPY $vgpr0
     %trunc:_(s16) = G_TRUNC %arg
     %shiftamt:_(s32) = G_CONSTANT i32 16
     %shift:_(s16) = G_LSHR %trunc, %shiftamt
     %zext:_(s32) = G_ZEXT %shift
     %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE0 %zext
     $vgpr0 = COPY %result
 ...
 
 ---
 name: cvt_f32_ubyte0_zext_lshr_24
 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0
 
     ; CHECK-LABEL: name: cvt_f32_ubyte0_zext_lshr_24
     ; CHECK: liveins: $vgpr0
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: %arg:_(s32) = COPY $vgpr0
-    ; CHECK-NEXT: %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE3 %arg
+    ; CHECK-NEXT: %shift:_(s16) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: %zext:_(s32) = G_ZEXT %shift(s16)
+    ; CHECK-NEXT: %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE0 %zext
     ; CHECK-NEXT: $vgpr0 = COPY %result(s32)
     %arg:_(s32) = COPY $vgpr0
     %trunc:_(s16) = G_TRUNC %arg
     %shiftamt:_(s32) = G_CONSTANT i32 24
     %shift:_(s16) = G_LSHR %trunc, %shiftamt
     %zext:_(s32) = G_ZEXT %shift
     %result:_(s32) = G_AMDGPU_CVT_F32_UBYTE0 %zext
     $vgpr0 = COPY %result

llvm/test/CodeGen/AMDGPU/GlobalISel/combine-ashr-narrow.mir

 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_ashr_s64_64
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 64
-    ; CHECK-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[ASHR]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 64
     %2:_(s64) = G_ASHR %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_ashr_s64_65
 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_ashr_s64_65
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65
-    ; CHECK-NEXT: [[ASHR:%[0-9]+]]:_(s64) = G_ASHR [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[ASHR]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 65
     %2:_(s64) = G_ASHR %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_ashr_s32_16

llvm/test/CodeGen/AMDGPU/GlobalISel/combine-lshr-narrow.mir

 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_lshr_s64_64
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 64
-    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[LSHR]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 64
     %2:_(s64) = G_LSHR %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_lshr_s64_65
 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_lshr_s64_65
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65
-    ; CHECK-NEXT: [[LSHR:%[0-9]+]]:_(s64) = G_LSHR [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[LSHR]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 65
     %2:_(s64) = G_LSHR %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_lshr_s32_16

llvm/test/CodeGen/AMDGPU/GlobalISel/combine-shl-narrow.mir

 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_shl_s64_64
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 64
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[SHL]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 64
     %2:_(s64) = G_SHL %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_shl_s64_65
 tracksRegLiveness: true
 body:             |
   bb.0:
     liveins: $vgpr0_vgpr1
 
     ; CHECK-LABEL: name: narrow_shl_s64_65
     ; CHECK: liveins: $vgpr0_vgpr1
     ; CHECK-NEXT: {{  $}}
-    ; CHECK-NEXT: [[COPY:%[0-9]+]]:_(s64) = COPY $vgpr0_vgpr1
-    ; CHECK-NEXT: [[C:%[0-9]+]]:_(s32) = G_CONSTANT i32 65
-    ; CHECK-NEXT: [[SHL:%[0-9]+]]:_(s64) = G_SHL [[COPY]], [[C]](s32)
-    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[SHL]](s64)
+    ; CHECK-NEXT: [[DEF:%[0-9]+]]:_(s64) = G_IMPLICIT_DEF
+    ; CHECK-NEXT: $vgpr0_vgpr1 = COPY [[DEF]](s64)
     %0:_(s64) = COPY $vgpr0_vgpr1
     %1:_(s32) = G_CONSTANT i32 65
     %2:_(s64) = G_SHL %0, %1
     $vgpr0_vgpr1 = COPY %2
 ...
 
 ---
 name:            narrow_shl_s32_16

llvm/test/CodeGen/AMDGPU/GlobalISel/sext_inreg.ll

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 28
 ; GFX6-NEXT:    s_lshl_b32 s1, s2, 28
 ; GFX6-NEXT:    s_and_b32 s0, s0, 0xffff
 ; GFX6-NEXT:    s_and_b32 s1, s1, 0xffff
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_sext_inreg_v4i16_14:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s0, 0, 0x100000
+; GFX8-NEXT:    s_bfe_u32 s0, -1, 0x100000
 ; GFX8-NEXT:    s_mov_b32 s1, s0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_sext_inreg_v4i16_14:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s2, s0, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 0xe000e
 ; GFX9-NEXT:    s_lshl_b32 s2, s2, 14

 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v3
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_sext_inreg_v8i16_11:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    s_bfe_u32 s4, 0, 0x100000
+; GFX8-NEXT:    s_bfe_u32 s4, -1, 0x100000
 ; GFX8-NEXT:    v_mov_b32_e32 v0, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX8-NEXT:    v_mov_b32_e32 v3, s4
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_sext_inreg_v8i16_11:
 ; GFX9:       ; %bb.0: