[GlobalISelEmitter] handle operand without MVT/class

[shchenz]

Summary

There are some patterns in td files without MVT/class set for some operands in target
pattern that are from the source pattern. This prevents GlobalISelEmitter from adding
them as a valid rule, because the target child operand is an unsupported kind operand.
For now, for a leaf child, only IntInit and DefInit are handled in GlobalISelEmitter.

This issue can be workaround by adding MVT/class to the patterns in the td files, like
the workarounds for patterns anyext and setcc in PPCInstrInfo.td in D140878.

To avoid adding the same workarounds for other patterns in td files, this patch tries to
handle the UnsetInit case in GlobalISelEmitter.

Adding the new handling allows us to remove the workarounds in the td files and also
generates many selection rules for PPC target.

[shchenz]

gentle ping.
We really want this to move forward on PPC. More patterns on PPC tds are without MVTs or register classes. Thank you!

[arsenm]

Can you add a tablegen test for this?

llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f16.ll

673–674

This is an improvement

1283

Improvement

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll

34–35

This is a regression cc @jrbyrnes

llvm/utils/TableGen/GlobalISelEmitter.cpp

4647–4648

Where is the type check performed?

[shchenz]

add a new case GlobalISelEmitter-notypeoperand.ll

[shchenz]

llvm/utils/TableGen/GlobalISelEmitter.cpp

4647–4648

Do you mean check the type for the output pattern's operands? I am not familiar here, do we need type check for the output pattern operands? And the input operands type check should already be handled in previous logic. See the new introduced rule in the new tblgen case GlobalISelEmitter-notypeoperand.ll:

constexpr static int64_t MatchTable0[] = {
  GIM_Try, /*On fail goto*//*Label 0*/ 74, // Rule ID 0 //
    GIM_CheckNumOperands, /*MI*/0, /*Expected*/2,
    GIM_CheckOpcode, /*MI*/0, TargetOpcode::G_ANYEXT,
    // MIs[0] dst
    GIM_CheckType, /*MI*/0, /*Op*/0, /*Type*/GILLT_s32,
    GIM_CheckRegBankForClass, /*MI*/0, /*Op*/0, /*RC*/MyTarget::GPR32RegClassID,
    // MIs[0] in
    GIM_CheckType, /*MI*/0, /*Op*/1, /*Type*/GILLT_s8,
    // (anyext:{ *:[i32] } i8:{ *:[i8] }:$in)  =>  (SELECT_I4:{ *:[i32] } ?:{ *:[i8] }:$in, (LI:{ *:[i32] } 1:{ *:[i32] }), (LI:{ *:[i32] } 0:{ *:[i32] }))
    GIR_MakeTempReg, /*TempRegID*/0, /*TypeID*/GILLT_s32,
    GIR_MakeTempReg, /*TempRegID*/1, /*TypeID*/GILLT_s32,
    GIR_BuildMI, /*InsnID*/2, /*Opcode*/MyTarget::LI,
    GIR_AddTempRegister, /*InsnID*/2, /*TempRegID*/1, /*TempRegFlags*/RegState::Define,
    GIR_AddImm, /*InsnID*/2, /*Imm*/0,
    GIR_ConstrainSelectedInstOperands, /*InsnID*/2,
    GIR_BuildMI, /*InsnID*/1, /*Opcode*/MyTarget::LI,
    GIR_AddTempRegister, /*InsnID*/1, /*TempRegID*/0, /*TempRegFlags*/RegState::Define,
    GIR_AddImm, /*InsnID*/1, /*Imm*/1,
    GIR_ConstrainSelectedInstOperands, /*InsnID*/1,
    GIR_BuildMI, /*InsnID*/0, /*Opcode*/MyTarget::SELECT_I4,
    GIR_Copy, /*NewInsnID*/0, /*OldInsnID*/0, /*OpIdx*/0, // dst
    GIR_Copy, /*NewInsnID*/0, /*OldInsnID*/0, /*OpIdx*/1, // in
    GIR_AddTempRegister, /*InsnID*/0, /*TempRegID*/0, /*TempRegFlags*/0,
    GIR_AddTempRegister, /*InsnID*/0, /*TempRegID*/1, /*TempRegFlags*/0,
    GIR_EraseFromParent, /*InsnID*/0,
    GIR_ConstrainSelectedInstOperands, /*InsnID*/0,
    // GIR_Coverage, 0,
    GIR_Done,
  // Label 0: @74
  GIM_Reject,
  };

[arsenm]

llvm/test/TableGen/GlobalISelEmitter-notypeoperand.ll

1 ↗

(On Diff #503316)

This should be a .td file, not .ll

7 ↗

(On Diff #503316)

Should check the actual checks performed, not just there is a matcher

llvm/utils/TableGen/GlobalISelEmitter.cpp

4647–4648

I meant more where that gets emitted in the emitter code, not the output

[shchenz]

fix the table-gen tests

[shchenz]

llvm/utils/TableGen/GlobalISelEmitter.cpp

4647–4648

Hope I understand your comments right. The type check for the input pattern are added before the renders handling in GlobalISelEmitter::runOnPattern(). To be specific, type check are added in createAndImportSelDAGMatcher() where type checks are added to source pattern results and inputs.

Will this impact how this issue is handled?

Thank you!

[arsenm]

LGTM

[shchenz]

@arsenm Hi Matt, thanks for confirmation for the direction. I will continue to fix all the LIT failures and the crashes in the AMDGPU LIT cases.

[shchenz]

Hi @arsenm , I created a github issue for the crashes this patch hits, would you please help to have a look? https://github.com/llvm/llvm-project/issues/61430. Thanks!

[shchenz]

Change this to formal review

[shchenz]

Request formal review. I saw improvement and degradations for AMDGPU changes, but in general I think we should try to reuse the td file patterns as more as possible?

llvm/test/CodeGen/AArch64/arm64-vcvt.ll

405 ↗

(On Diff #512067)

Before the change:

remark: <unknown>:0:0: cannot select: %1:gpr(s64) = G_INTRINSIC intrinsic(@llvm.aarch64.neon.fcvtzs), %2:gpr(s64) (in function: fcvtzs_1d_intrinsic)
warning: Instruction selection used fallback path for fcvtzs_1d_intrinsic

After the change:

	fcvtzs	x8, d0
	fmov	d0, x8
	ret

This should be expected as now global-isel are reusing more patterns defined in TD files. But here there is an opportunity to remove the fmov?

[shchenz]

gentle ping

llvm/lib/Target/PowerPC/PPCInstrInfo.td

 def : Pat<(i64 (zext i1:$in)),
           (SELECT_I8 $in, (LI8 1), (LI8 0))>;
 def : Pat<(i64 (sext i1:$in)),
           (SELECT_I8 $in, (LI8 -1), (LI8 0))>;
 
 // FIXME: We should choose either a zext or a sext based on other constants
 // already around.
 def : Pat<(i32 (anyext i1:$in)),
-          (SELECT_I4 crbitrc:$in, (LI 1), (LI 0))>;
+          (SELECT_I4 $in, (LI 1), (LI 0))>;
 def : Pat<(i64 (anyext i1:$in)),
-          (SELECT_I8 crbitrc:$in, (LI8 1), (LI8 0))>;
+          (SELECT_I8 $in, (LI8 1), (LI8 0))>;
 
 // match setcc on i1 variables.
 // CRANDC is:
 //   1 1 : F
 //   1 0 : T
 //   0 1 : F
 //   0 0 : F
 //

 defm : CRNotPat<(i1 (setcc i64:$s1, i64:$s2, SETLE)),
                 (EXTRACT_SUBREG (CMPD $s1, $s2), sub_gt)>;
 defm : CRNotPat<(i1 (setcc i64:$s1, i64:$s2, SETNE)),
                 (EXTRACT_SUBREG (CMPD $s1, $s2), sub_eq)>;
 }
 
 multiclass FSetCCPat<SDPatternOperator SetCC, ValueType Ty, I FCmp> {
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETUGE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_lt)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_lt)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETGE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_lt)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_lt)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETULE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_gt)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_gt)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETLE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_gt)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_gt)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETUNE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_eq)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_eq)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETNE)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_eq)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_eq)>;
   defm : CRNotPat<(i1 (SetCC Ty:$s1, Ty:$s2, SETO)),
-                  (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_un)>;
+                  (EXTRACT_SUBREG (FCmp $s1, $s2), sub_un)>;
 
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETOLT)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_lt)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_lt)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETLT)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_lt)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_lt)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETOGT)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_gt)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_gt)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETGT)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_gt)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_gt)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETOEQ)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_eq)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_eq)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETEQ)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_eq)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_eq)>;
   def : Pat<(i1 (SetCC Ty:$s1, Ty:$s2, SETUO)),
-            (EXTRACT_SUBREG (FCmp Ty:$s1, Ty:$s2), sub_un)>;
+            (EXTRACT_SUBREG (FCmp $s1, $s2), sub_un)>;
 }
 
 let Predicates = [HasFPU] in {
 // FCMPU: If either of the operands is a Signaling NaN, then VXSNAN is set.
 // SETCC for f32.
 defm : FSetCCPat<any_fsetcc, f32, FCMPUS>;
 
 // SETCC for f64.

llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f16.ll

 ; GFX8-NEXT:    v_rcp_f32_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v7, v6
 ; GFX8-NEXT:    v_rcp_f32_e32 v5, v5
 ; GFX8-NEXT:    v_mul_f32_e32 v2, v3, v2
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_f32_e32 v3, v7, v5
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; GFX8-NEXT:    v_div_fixup_f16 v0, v2, v1, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
 ; GFX8-NEXT:    v_div_fixup_f16 v1, v3, v4, v6
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 16, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v5, v4

 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fdiv_v2f16_afn:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX8-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_mul_f16_e32 v2, v0, v2
-; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX8-NEXT:    v_mov_b32_e32 v1, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16_afn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX9-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_mul_f16_e32 v2, v0, v2

 ; GFX8-NEXT:    v_rcp_f32_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v7, v6
 ; GFX8-NEXT:    v_rcp_f32_e32 v5, v5
 ; GFX8-NEXT:    v_mul_f32_e32 v2, v3, v2
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_f32_e32 v3, v7, v5
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; GFX8-NEXT:    v_div_fixup_f16 v0, v2, v1, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
 ; GFX8-NEXT:    v_div_fixup_f16 v1, v3, v4, v6
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16_ulp25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 16, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v5, v4

 ; GFX8-NEXT:    v_rcp_f32_e32 v1, v1
 ; GFX8-NEXT:    v_rcp_f32_e32 v3, v3
 ; GFX8-NEXT:    v_mul_f32_e32 v1, v4, v1
 ; GFX8-NEXT:    v_mul_f32_e32 v3, v4, v3
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; GFX8-NEXT:    v_div_fixup_f16 v0, v1, v0, 1.0
 ; GFX8-NEXT:    v_div_fixup_f16 v1, v3, v2, 1.0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_rcp_v2f16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v1, v0
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v2

 ; GFX8-NEXT:    v_rcp_f32_e32 v1, v1
 ; GFX8-NEXT:    v_rcp_f32_e32 v3, v3
 ; GFX8-NEXT:    v_mul_f32_e32 v1, v4, v1
 ; GFX8-NEXT:    v_mul_f32_e32 v3, v4, v3
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; GFX8-NEXT:    v_div_fixup_f16 v0, v1, v0, 1.0
 ; GFX8-NEXT:    v_div_fixup_f16 v1, v3, v2, 1.0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_rcp_v2f16_arcp:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v1, v0
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v2

 ; GFX6-NEXT:    v_mul_f32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX6-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_rcp_v2f16_arcp_afn:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_rcp_f16_sdwa v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
-; GFX8-NEXT:    v_rcp_f16_e32 v0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_rcp_f16_e32 v1, v0
+; GFX8-NEXT:    v_rcp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_rcp_v2f16_arcp_afn:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_rcp_f16_e32 v1, v0
 ; GFX9-NEXT:    v_rcp_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_pack_b32_f16 v0, v1, v0

 ; GFX6-FLUSH-NEXT:    v_div_fmas_f32 v2, v2, v3, v6
 ; GFX6-FLUSH-NEXT:    v_div_fixup_f32 v1, v2, v1, v4
 ; GFX6-FLUSH-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; GFX6-FLUSH-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_rcp_v2f16_ulp25:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    v_rcp_f16_sdwa v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
-; GFX8-NEXT:    v_rcp_f16_e32 v0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_rcp_f16_e32 v1, v0
+; GFX8-NEXT:    v_rcp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_rcp_v2f16_ulp25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_rcp_f16_e32 v1, v0
 ; GFX9-NEXT:    v_rcp_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_pack_b32_f16 v0, v1, v0

 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fdiv_v2f16_afn_ulp25:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX8-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_mul_f16_e32 v2, v0, v2
-; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX8-NEXT:    v_mov_b32_e32 v1, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16_afn_ulp25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX9-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_mul_f16_e32 v2, v0, v2

 ; GFX8-NEXT:    v_rcp_f32_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v7, v6
 ; GFX8-NEXT:    v_rcp_f32_e32 v5, v5
 ; GFX8-NEXT:    v_mul_f32_e32 v2, v3, v2
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX8-NEXT:    v_mul_f32_e32 v3, v7, v5
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; GFX8-NEXT:    v_div_fixup_f16 v0, v2, v1, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
 ; GFX8-NEXT:    v_div_fixup_f16 v1, v3, v4, v6
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16_arcp_ulp25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 16, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v5, v4

 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fdiv_v2f16_arcp_afn_ulp25:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX8-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_mul_f16_e32 v2, v0, v2
-; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX8-NEXT:    v_mov_b32_e32 v1, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_mul_f16_sdwa v0, v0, v1 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fdiv_v2f16_arcp_afn_ulp25:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_rcp_f16_e32 v2, v1
 ; GFX9-NEXT:    v_rcp_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_mul_f16_e32 v2, v0, v2

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

 ; GFX8-NEXT:    v_log_f16_e32 v2, v0
 ; GFX8-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX8-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_exp_f16_e32 v0, v0
+; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX8-NEXT:    v_exp_f16_e32 v1, v1
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_exp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_pow_v2f16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_log_f16_e32 v2, v0
 ; GFX9-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX9-NEXT:    s_mov_b32 s4, 0x5040100
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v1, v2
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX9-NEXT:    v_exp_f16_e32 v1, v1
 ; GFX9-NEXT:    v_exp_f16_e32 v0, v0
-; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
+; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_pow_v2f16:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_log_f16_e32 v2, v0
 ; GFX10-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX10-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX10-NEXT:    v_cvt_f16_f32_e32 v1, v2
+; GFX10-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX10-NEXT:    v_exp_f16_e32 v1, v1
+; GFX10-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX10-NEXT:    v_exp_f16_e32 v0, v0
-; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x5040100
+; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_pow_v2f16:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-NEXT:    v_log_f16_e32 v0, v0
+; GFX11-NEXT:    v_log_f16_e32 v2, v0
+; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v1, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_log_f16_e32 v2, v2
+; GFX11-NEXT:    v_log_f16_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v2, v2
+; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v0, v0, v1 :: v_dual_mul_dx9_zero_f32 v1, v2, v3
-; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
 ; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT:    v_exp_f16_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX11-NEXT:    v_exp_f16_e32 v1, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_exp_f16_e32 v0, v0
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
+; GFX11-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x, <2 x half> %y)
   ret <2 x half> %pow
 }
 
 define <2 x half> @v_pow_v2f16_fneg_lhs(<2 x half> %x, <2 x half> %y) {
 ; GFX6-LABEL: v_pow_v2f16_fneg_lhs:
 ; GFX6:       ; %bb.0:

 ; GFX8-NEXT:    v_log_f16_e32 v2, v0
 ; GFX8-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX8-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_exp_f16_e32 v0, v0
+; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX8-NEXT:    v_exp_f16_e32 v1, v1
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_exp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_pow_v2f16_fneg_lhs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX9-NEXT:    v_log_f16_e32 v2, v0
 ; GFX9-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX9-NEXT:    s_mov_b32 s4, 0x5040100
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v1, v2
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX9-NEXT:    v_exp_f16_e32 v1, v1
 ; GFX9-NEXT:    v_exp_f16_e32 v0, v0
-; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
+; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_pow_v2f16_fneg_lhs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX10-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_log_f16_e32 v2, v0
 ; GFX10-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX10-NEXT:    v_cvt_f16_f32_e32 v1, v2
+; GFX10-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX10-NEXT:    v_exp_f16_e32 v1, v1
+; GFX10-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX10-NEXT:    v_exp_f16_e32 v0, v0
-; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x5040100
+; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_pow_v2f16_fneg_lhs:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v1, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-NEXT:    v_log_f16_e32 v0, v0
+; GFX11-NEXT:    v_log_f16_e32 v2, v0
+; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_3) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_log_f16_e32 v2, v2
+; GFX11-NEXT:    v_log_f16_e32 v0, v0
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v2, v2
-; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v0, v0, v1 :: v_dual_mul_dx9_zero_f32 v1, v2, v3
+; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
+; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_exp_f16_e32 v0, v0
 ; GFX11-NEXT:    v_exp_f16_e32 v1, v1
+; GFX11-NEXT:    v_exp_f16_e32 v0, v0
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
+; GFX11-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %x.fneg = fneg <2 x half> %x
   %pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x.fneg, <2 x half> %y)
   ret <2 x half> %pow
 }
 
 define <2 x half> @v_pow_v2f16_fneg_rhs(<2 x half> %x, <2 x half> %y) {
 ; GFX6-LABEL: v_pow_v2f16_fneg_rhs:

 ;
 ; GFX8-LABEL: v_pow_v2f16_fneg_rhs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_log_f16_e32 v2, v0
 ; GFX8-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v3, v1
+; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
-; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
+; GFX8-NEXT:    v_cvt_f16_f32_e32 v2, v2
+; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_exp_f16_e32 v0, v0
-; GFX8-NEXT:    v_exp_f16_e32 v1, v1
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_exp_f16_e32 v1, v2
+; GFX8-NEXT:    v_exp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_pow_v2f16_fneg_rhs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_log_f16_e32 v2, v0
 ; GFX9-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX9-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX9-NEXT:    s_mov_b32 s4, 0x5040100
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX9-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX9-NEXT:    v_exp_f16_e32 v0, v0
-; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
+; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_pow_v2f16_fneg_rhs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_log_f16_e32 v2, v0
-; GFX10-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
+; GFX10-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX10-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX10-NEXT:    v_cvt_f16_f32_e32 v1, v2
+; GFX10-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX10-NEXT:    v_exp_f16_e32 v1, v1
+; GFX10-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX10-NEXT:    v_exp_f16_e32 v0, v0
-; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x5040100
+; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_pow_v2f16_fneg_rhs:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-NEXT:    v_log_f16_e32 v0, v0
+; GFX11-NEXT:    v_log_f16_e32 v2, v0
+; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
 ; GFX11-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_log_f16_e32 v2, v2
+; GFX11-NEXT:    v_log_f16_e32 v0, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v1, v1
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX11-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v2, v2
+; GFX11-NEXT:    v_cvt_f32_f16_e32 v3, v3
+; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v0, v0, v1 :: v_dual_mul_dx9_zero_f32 v1, v2, v3
-; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
 ; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT:    v_exp_f16_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX11-NEXT:    v_exp_f16_e32 v1, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_exp_f16_e32 v0, v0
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
+; GFX11-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %y.fneg = fneg <2 x half> %y
   %pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x, <2 x half> %y.fneg)
   ret <2 x half> %pow
 }
 
 define <2 x half> @v_pow_v2f16_fneg_lhs_rhs(<2 x half> %x, <2 x half> %y) {
 ; GFX6-LABEL: v_pow_v2f16_fneg_lhs_rhs:

 ; GFX8-LABEL: v_pow_v2f16_fneg_lhs_rhs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX8-NEXT:    v_log_f16_e32 v2, v0
 ; GFX8-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v3, v1
+; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX8-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX8-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX8-NEXT:    v_cvt_f32_f16_e32 v2, v2
-; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
+; GFX8-NEXT:    v_cvt_f16_f32_e32 v2, v2
+; GFX8-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX8-NEXT:    v_cvt_f16_f32_e32 v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_exp_f16_e32 v0, v0
-; GFX8-NEXT:    v_exp_f16_e32 v1, v1
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_exp_f16_e32 v1, v2
+; GFX8-NEXT:    v_exp_f16_sdwa v0, v0 dst_sel:WORD_1 dst_unused:UNUSED_PAD src0_sel:DWORD
+; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_pow_v2f16_fneg_lhs_rhs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX9-NEXT:    v_log_f16_e32 v2, v0
 ; GFX9-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX9-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX9-NEXT:    v_cvt_f32_f16_e32 v0, v0
-; GFX9-NEXT:    s_mov_b32 s4, 0x5040100
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX9-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX9-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX9-NEXT:    v_exp_f16_e32 v0, v0
-; GFX9-NEXT:    v_perm_b32 v0, v0, v1, s4
+; GFX9-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_pow_v2f16_fneg_lhs_rhs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX10-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX10-NEXT:    v_log_f16_e32 v2, v0
 ; GFX10-NEXT:    v_log_f16_sdwa v0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v3, v1
 ; GFX10-NEXT:    v_cvt_f32_f16_sdwa v1, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v2, v2, v3
 ; GFX10-NEXT:    v_mul_legacy_f32_e32 v0, v0, v1
-; GFX10-NEXT:    v_cvt_f16_f32_e32 v1, v2
+; GFX10-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; GFX10-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX10-NEXT:    v_exp_f16_e32 v1, v1
+; GFX10-NEXT:    v_exp_f16_e32 v1, v2
 ; GFX10-NEXT:    v_exp_f16_e32 v0, v0
-; GFX10-NEXT:    v_perm_b32 v0, v0, v1, 0x5040100
+; GFX10-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_pow_v2f16_fneg_lhs_rhs:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_xor_b32_e32 v0, 0x80008000, v0
 ; GFX11-NEXT:    v_xor_b32_e32 v1, 0x80008000, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
-; GFX11-NEXT:    v_log_f16_e32 v0, v0
+; GFX11-NEXT:    v_log_f16_e32 v2, v0
+; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 16, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 16, v1
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v1, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    v_log_f16_e32 v2, v2
+; GFX11-NEXT:    v_log_f16_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v3, v3
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_f32_f16_e32 v2, v2
+; GFX11-NEXT:    v_cvt_f32_f16_e32 v0, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v0, v0, v1 :: v_dual_mul_dx9_zero_f32 v1, v2, v3
-; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_dual_mul_dx9_zero_f32 v1, v2, v1 :: v_dual_mul_dx9_zero_f32 v0, v0, v3
 ; GFX11-NEXT:    v_cvt_f16_f32_e32 v1, v1
-; GFX11-NEXT:    v_exp_f16_e32 v0, v0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_cvt_f16_f32_e32 v0, v0
 ; GFX11-NEXT:    v_exp_f16_e32 v1, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_2) | instid1(VALU_DEP_1)
+; GFX11-NEXT:    v_exp_f16_e32 v0, v0
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
-; GFX11-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
+; GFX11-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX11-NEXT:    v_lshl_or_b32 v0, v0, 16, v1
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %x.fneg = fneg <2 x half> %x
   %y.fneg = fneg <2 x half> %y
   %pow = call <2 x half> @llvm.pow.v2f16(<2 x half> %x.fneg, <2 x half> %y.fneg)
   ret <2 x half> %pow
 }
 
 ; FIXME

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

 ; CI-NEXT:    v_fma_f32 v6, -v3, v5, 1.0
 ; CI-NEXT:    v_fma_f32 v5, v6, v5, v5
 ; CI-NEXT:    v_mul_f32_e32 v6, v4, v5
 ; CI-NEXT:    v_fma_f32 v7, -v3, v6, v4
 ; CI-NEXT:    v_fma_f32 v6, v7, v5, v6
 ; CI-NEXT:    v_fma_f32 v3, -v3, v6, v4
 ; CI-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_MODE, 4, 2), 0
 ; CI-NEXT:    v_div_fmas_f32 v3, v3, v5, v6
-; CI-NEXT:    v_bfe_u32 v0, v0, 0, 16
 ; CI-NEXT:    v_div_fixup_f32 v3, v3, v2, v1
 ; CI-NEXT:    v_trunc_f32_e32 v3, v3
 ; CI-NEXT:    v_fma_f32 v1, -v3, v2, v1
 ; CI-NEXT:    v_cvt_f16_f32_e32 v1, v1
-; CI-NEXT:    v_bfe_u32 v1, v1, 0, 16
 ; CI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; CI-NEXT:    v_or_b32_e32 v2, v0, v1
 ; CI-NEXT:    v_mov_b32_e32 v0, s4
 ; CI-NEXT:    v_mov_b32_e32 v1, s5
 ; CI-NEXT:    flat_store_dword v[0:1], v2
 ; CI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: frem_v2f16:

 ; VI-NEXT:    v_trunc_f16_e32 v0, v0
 ; VI-NEXT:    v_fma_f16 v0, -v0, v1, s2
 ; VI-NEXT:    v_cvt_f32_f16_e32 v1, s1
 ; VI-NEXT:    v_mul_f32_e32 v1, v1, v3
 ; VI-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; VI-NEXT:    v_div_fixup_f16 v1, v1, v2, s1
 ; VI-NEXT:    v_trunc_f16_e32 v1, v1
 ; VI-NEXT:    v_fma_f16 v1, -v1, v2, s1
-; VI-NEXT:    v_mov_b32_e32 v2, 16
-; VI-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT:    v_or_b32_sdwa v2, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT:    v_or_b32_e32 v2, v0, v1
 ; VI-NEXT:    v_mov_b32_e32 v0, s4
 ; VI-NEXT:    v_mov_b32_e32 v1, s5
 ; VI-NEXT:    flat_store_dword v[0:1], v2
 ; VI-NEXT:    s_endpgm
    %gep2 = getelementptr <2 x half>, ptr addrspace(1) %in2, i32 4
    %r0 = load <2 x half>, ptr addrspace(1) %in1, align 8
    %r1 = load <2 x half>, ptr addrspace(1) %gep2, align 8
    %r2 = frem <2 x half> %r0, %r1

 ; CI-NEXT:    v_fma_f32 v8, -v5, v7, 1.0
 ; CI-NEXT:    v_fma_f32 v7, v8, v7, v7
 ; CI-NEXT:    v_mul_f32_e32 v8, v6, v7
 ; CI-NEXT:    v_fma_f32 v9, -v5, v8, v6
 ; CI-NEXT:    v_fma_f32 v8, v9, v7, v8
 ; CI-NEXT:    v_fma_f32 v5, -v5, v8, v6
 ; CI-NEXT:    s_setreg_imm32_b32 hwreg(HW_REG_MODE, 4, 2), 0
 ; CI-NEXT:    v_div_fmas_f32 v5, v5, v7, v8
-; CI-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; CI-NEXT:    v_bfe_u32 v0, v0, 0, 16
 ; CI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; CI-NEXT:    v_or_b32_e32 v0, v0, v1
-; CI-NEXT:    v_bfe_u32 v1, v2, 0, 16
 ; CI-NEXT:    v_div_fixup_f32 v5, v5, v4, v3
 ; CI-NEXT:    v_trunc_f32_e32 v5, v5
 ; CI-NEXT:    v_fma_f32 v3, -v5, v4, v3
 ; CI-NEXT:    v_cvt_f16_f32_e32 v3, v3
-; CI-NEXT:    v_bfe_u32 v2, v3, 0, 16
-; CI-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
-; CI-NEXT:    v_or_b32_e32 v1, v1, v2
+; CI-NEXT:    v_lshlrev_b32_e32 v1, 16, v3
+; CI-NEXT:    v_or_b32_e32 v1, v2, v1
 ; CI-NEXT:    v_mov_b32_e32 v2, s4
 ; CI-NEXT:    v_mov_b32_e32 v3, s5
 ; CI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; CI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: frem_v4f16:
 ; VI:       ; %bb.0:
 ; VI-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x24

 ; VI-NEXT:    v_mul_f32_e32 v1, v1, v3
 ; VI-NEXT:    v_cvt_f16_f32_e32 v1, v1
 ; VI-NEXT:    v_mov_b32_e32 v3, s1
 ; VI-NEXT:    v_rcp_f32_e32 v5, v5
 ; VI-NEXT:    v_div_fixup_f16 v1, v1, v2, s6
 ; VI-NEXT:    v_trunc_f16_e32 v1, v1
 ; VI-NEXT:    v_fma_f16 v1, -v1, v2, s6
 ; VI-NEXT:    v_cvt_f32_f16_e32 v2, s3
+; VI-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; VI-NEXT:    v_or_b32_e32 v0, v0, v1
 ; VI-NEXT:    v_mul_f32_e32 v2, v2, v4
 ; VI-NEXT:    v_cvt_f16_f32_e32 v2, v2
 ; VI-NEXT:    v_mov_b32_e32 v4, s9
 ; VI-NEXT:    v_div_fixup_f16 v2, v2, v3, s3
 ; VI-NEXT:    v_trunc_f16_e32 v2, v2
 ; VI-NEXT:    v_fma_f16 v2, -v2, v3, s3
 ; VI-NEXT:    v_cvt_f32_f16_e32 v3, s7
 ; VI-NEXT:    v_mul_f32_e32 v3, v3, v5
 ; VI-NEXT:    v_cvt_f16_f32_e32 v3, v3
 ; VI-NEXT:    v_div_fixup_f16 v3, v3, v4, s7
 ; VI-NEXT:    v_trunc_f16_e32 v3, v3
 ; VI-NEXT:    v_fma_f16 v3, -v3, v4, s7
-; VI-NEXT:    v_mov_b32_e32 v4, 16
-; VI-NEXT:    v_lshlrev_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; VI-NEXT:    v_lshlrev_b32_sdwa v1, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; VI-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; VI-NEXT:    v_lshlrev_b32_e32 v1, 16, v3
+; VI-NEXT:    v_or_b32_e32 v1, v2, v1
 ; VI-NEXT:    v_mov_b32_e32 v2, s4
 ; VI-NEXT:    v_mov_b32_e32 v3, s5
 ; VI-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
 ; VI-NEXT:    s_endpgm
    %gep2 = getelementptr <4 x half>, ptr addrspace(1) %in2, i32 4
    %r0 = load <4 x half>, ptr addrspace(1) %in1, align 16
    %r1 = load <4 x half>, ptr addrspace(1) %gep2, align 16
    %r2 = frem <4 x half> %r0, %r1

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

 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i7:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, 7
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX8-NEXT:    s_and_b32 s2, s2, 0x7f
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0x7f
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX8-NEXT:    v_mul_lo_u32 v1, v0, -7
 ; GFX8-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v1
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GFX8-NEXT:    v_mul_lo_u32 v0, v0, 7

 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i7:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, 7
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX9-NEXT:    s_and_b32 s2, s2, 0x7f
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0x7f
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX9-NEXT:    v_mul_lo_u32 v1, v0, -7
 ; GFX9-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX9-NEXT:    v_add_u32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_mul_hi_u32 v0, s2, v0
 ; GFX9-NEXT:    v_mul_lo_u32 v0, v0, 7

 ; GFX9-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i7:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v0, 7
 ; GFX10-NEXT:    s_and_b32 s2, s2, 0x7f
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0x7f
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX10-NEXT:    v_mul_lo_u32 v1, v0, -7
 ; GFX10-NEXT:    v_mul_hi_u32 v1, v0, v1
 ; GFX10-NEXT:    v_add_nc_u32_e32 v0, v0, v1
 ; GFX10-NEXT:    v_mul_hi_u32 v0, s2, v0

 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i7:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v0, 7
 ; GFX11-NEXT:    s_and_b32 s2, s2, 0x7f
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0x7f
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
 ; GFX11-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX11-NEXT:    v_mul_lo_u32 v1, v0, -7
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX6-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_and_b32 s3, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_and_b32 s3, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_and_b32 s3, s2, 7
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_and_b32 s3, s2, 7
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshl.i8(i8 %lhs, i8 %rhs, i8 %amt)
   ret i8 %result
 }
 
 define i8 @v_fshl_i8(i8 %lhs, i8 %rhs, i8 %amt) {
 ; GFX6-LABEL: v_fshl_i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 1, 7
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v3, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v3, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v3, 1
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, v2, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, v3, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v3, 1
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v1, v2, v1
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT:    v_xor_b32_e32 v3, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v3, v2
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX10-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX10-NEXT:    v_lshrrev_b16 v1, 1, v1
 ; GFX10-NEXT:    v_lshlrev_b16 v0, v2, v0
 ; GFX10-NEXT:    v_lshrrev_b16 v1, v3, v1
 ; GFX10-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshl_i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT:    v_xor_b32_e32 v3, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v3, v2
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX11-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_3)
 ; GFX11-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX11-NEXT:    v_lshrrev_b16 v1, 1, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_lshlrev_b16 v0, v2, v0
 ; GFX11-NEXT:    v_lshrrev_b16 v1, v3, v1

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x40004
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i8_4:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i8_4:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i8_4:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i8_4:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshl.i8(i8 %lhs, i8 %rhs, i8 4)
   ret i8 %result
 }
 

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 5
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x50003
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i8_5:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 5
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 3
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i8_5:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 5
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 3
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i8_5:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 5
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 3
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i8_5:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 5
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 3
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshl.i8(i8 %lhs, i8 %rhs, i8 5)
   ret i8 %result
 }
 

 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v2i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX8-NEXT:    s_and_b32 s6, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX8-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    s_and_b32 s1, s5, 7
 ; GFX8-NEXT:    s_lshl_b32 s1, s3, s1
 ; GFX8-NEXT:    s_and_b32 s3, s4, 0xff
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s5
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
 ; GFX8-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX8-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v2i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX9-NEXT:    s_and_b32 s6, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX9-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    s_and_b32 s1, s5, 7
 ; GFX9-NEXT:    s_lshl_b32 s1, s3, s1
 ; GFX9-NEXT:    s_and_b32 s3, s4, 0xff
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s5
 ; GFX9-NEXT:    s_lshr_b32 s3, s3, 1
 ; GFX9-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX9-NEXT:    s_or_b32 s1, s1, s2
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX9-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX9-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX9-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_v2i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX10-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX10-NEXT:    s_and_b32 s6, s2, 7
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX10-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX10-NEXT:    s_lshr_b32 s3, s0, 8
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX10-NEXT:    s_and_b32 s6, s5, 7
 ; GFX10-NEXT:    s_andn2_b32 s5, 7, s5
 ; GFX10-NEXT:    s_lshr_b32 s4, s4, 1
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, s6
 ; GFX10-NEXT:    s_lshr_b32 s4, s4, s5
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX10-NEXT:    s_or_b32 s2, s3, s4
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    s_and_b32 s1, s2, 0xff
-; GFX10-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX10-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX10-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX10-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_v2i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX11-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX11-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX11-NEXT:    s_and_b32 s6, s2, 7
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX11-NEXT:    s_lshr_b32 s3, s0, 8
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX11-NEXT:    s_and_b32 s6, s5, 7
 ; GFX11-NEXT:    s_and_not1_b32 s5, 7, s5
 ; GFX11-NEXT:    s_lshr_b32 s4, s4, 1
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, s6
 ; GFX11-NEXT:    s_lshr_b32 s4, s4, s5
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX11-NEXT:    s_or_b32 s2, s3, s4
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    s_and_b32 s1, s2, 0xff
-; GFX11-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX11-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX11-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX11-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %lhs = bitcast i16 %lhs.arg to <2 x i8>
   %rhs = bitcast i16 %rhs.arg to <2 x i8>
   %amt = bitcast i16 %amt.arg to <2 x i8>
   %result = call <2 x i8> @llvm.fshl.v2i8(<2 x i8> %lhs, <2 x i8> %rhs, <2 x i8> %amt)
   %cast.result = bitcast <2 x i8> %result to i16
   ret i16 %cast.result
 }
 
 define i16 @v_fshl_v2i8(i16 %lhs.arg, i16 %rhs.arg, i16 %amt.arg) {
 ; GFX6-LABEL: v_fshl_v2i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 8, v2
 ; GFX6-NEXT:    v_and_b32_e32 v5, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v5, v0
 ; GFX6-NEXT:    v_bfe_u32 v5, v1, 1, 7
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v2, v5
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v4
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX6-NEXT:    v_not_b32_e32 v4, v4
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 8, 8
 ; GFX6-NEXT:    v_and_b32_e32 v4, 7, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v4, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 8, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v2i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_and_b32_e32 v6, 7, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v6, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v6, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, v2, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_and_b32_e32 v1, 7, v5
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v5
+; GFX8-NEXT:    v_not_b32_e32 v2, v5
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, v1, v3
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v3, v6, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v2, v3
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX8-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, 8, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v2i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX9-NEXT:    v_and_b32_e32 v6, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    s_mov_b32 s4, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, s4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, v6, v0
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v1, v2, v1
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_and_b32_e32 v1, 7, v5
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v5
+; GFX9-NEXT:    v_not_b32_e32 v2, v5
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v1, v1, v3
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v3, s4, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v2, v2, v3
 ; GFX9-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX9-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v1, 8, v1
 ; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_v2i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 8, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 8, v0
-; GFX10-NEXT:    v_xor_b32_e32 v7, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v7, v2
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v1
-; GFX10-NEXT:    v_xor_b32_e32 v6, -1, v3
+; GFX10-NEXT:    v_not_b32_e32 v6, v3
 ; GFX10-NEXT:    v_and_b32_e32 v4, 0xff, v4
 ; GFX10-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX10-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX10-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX10-NEXT:    v_lshrrev_b16 v4, 1, v4
 ; GFX10-NEXT:    v_lshrrev_b16 v1, 1, v1
 ; GFX10-NEXT:    v_lshlrev_b16 v3, v3, v5

 ;
 ; GFX11-LABEL: v_fshl_v2i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 8, v2
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v5, 8, v0
-; GFX11-NEXT:    v_xor_b32_e32 v7, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v7, v2
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xff, v1
-; GFX11-NEXT:    v_xor_b32_e32 v6, -1, v3
+; GFX11-NEXT:    v_not_b32_e32 v6, v3
 ; GFX11-NEXT:    v_and_b32_e32 v4, 0xff, v4
 ; GFX11-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX11-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX11-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX11-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX11-NEXT:    v_lshrrev_b16 v4, 1, v4
 ; GFX11-NEXT:    v_lshrrev_b16 v1, 1, v1
 ; GFX11-NEXT:    v_lshlrev_b16 v3, v3, v5

 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v4i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_lshr_b32 s6, s1, 8
 ; GFX8-NEXT:    s_lshr_b32 s7, s1, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s1, 24
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX8-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX8-NEXT:    s_and_b32 s12, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX8-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX8-NEXT:    s_lshr_b32 s4, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s5, s0, 24
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s12
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    s_and_b32 s1, s9, 7
 ; GFX8-NEXT:    s_lshl_b32 s1, s3, s1
 ; GFX8-NEXT:    s_and_b32 s3, s6, 0xff
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s9
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
 ; GFX8-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
 ; GFX8-NEXT:    s_and_b32 s2, s10, 7
 ; GFX8-NEXT:    s_lshl_b32 s2, s4, s2
 ; GFX8-NEXT:    s_and_b32 s4, s7, 0xff
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_andn2_b32 s3, 7, s10
 ; GFX8-NEXT:    s_lshr_b32 s4, s4, 1
 ; GFX8-NEXT:    s_lshr_b32 s3, s4, s3
 ; GFX8-NEXT:    s_or_b32 s2, s2, s3
 ; GFX8-NEXT:    s_and_b32 s3, s11, 7
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_andn2_b32 s4, 7, s11
 ; GFX8-NEXT:    s_lshl_b32 s3, s5, s3

 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v4i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s6, s1, 8
 ; GFX9-NEXT:    s_lshr_b32 s7, s1, 16
 ; GFX9-NEXT:    s_lshr_b32 s8, s1, 24
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX9-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX9-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX9-NEXT:    s_and_b32 s12, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX9-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX9-NEXT:    s_lshr_b32 s4, s0, 16
 ; GFX9-NEXT:    s_lshr_b32 s5, s0, 24
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s12
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    s_and_b32 s1, s9, 7
 ; GFX9-NEXT:    s_lshl_b32 s1, s3, s1
 ; GFX9-NEXT:    s_and_b32 s3, s6, 0xff
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s9
 ; GFX9-NEXT:    s_lshr_b32 s3, s3, 1
 ; GFX9-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX9-NEXT:    s_or_b32 s1, s1, s2
 ; GFX9-NEXT:    s_and_b32 s2, s10, 7
 ; GFX9-NEXT:    s_lshl_b32 s2, s4, s2
 ; GFX9-NEXT:    s_and_b32 s4, s7, 0xff
-; GFX9-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX9-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX9-NEXT:    s_andn2_b32 s3, 7, s10
 ; GFX9-NEXT:    s_lshr_b32 s4, s4, 1
 ; GFX9-NEXT:    s_lshr_b32 s3, s4, s3
 ; GFX9-NEXT:    s_or_b32 s2, s2, s3
 ; GFX9-NEXT:    s_and_b32 s3, s11, 7
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_andn2_b32 s4, 7, s11
 ; GFX9-NEXT:    s_lshl_b32 s3, s5, s3

 ;
 ; GFX10-LABEL: s_fshl_v4i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_lshr_b32 s6, s1, 8
 ; GFX10-NEXT:    s_lshr_b32 s7, s1, 16
 ; GFX10-NEXT:    s_lshr_b32 s8, s1, 24
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshr_b32 s9, s2, 8
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX10-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX10-NEXT:    s_and_b32 s12, s2, 7
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX10-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX10-NEXT:    s_and_b32 s2, s6, 0xff
 ; GFX10-NEXT:    s_and_b32 s6, s9, 7
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX10-NEXT:    s_andn2_b32 s9, 7, s9
 ; GFX10-NEXT:    s_lshr_b32 s2, s2, 1
 ; GFX10-NEXT:    s_lshr_b32 s4, s0, 16
 ; GFX10-NEXT:    s_lshr_b32 s5, s0, 24
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s12
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, s6
 ; GFX10-NEXT:    s_lshr_b32 s2, s2, s9
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    s_or_b32 s1, s3, s2
 ; GFX10-NEXT:    s_and_b32 s2, s7, 0xff
 ; GFX10-NEXT:    s_and_b32 s3, s10, 7
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX10-NEXT:    s_andn2_b32 s6, 7, s10
 ; GFX10-NEXT:    s_lshr_b32 s2, s2, 1
 ; GFX10-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX10-NEXT:    s_lshr_b32 s2, s2, s6
 ; GFX10-NEXT:    s_and_b32 s4, s11, 7
 ; GFX10-NEXT:    s_andn2_b32 s6, 7, s11
 ; GFX10-NEXT:    s_lshr_b32 s7, s8, 1
 ; GFX10-NEXT:    s_lshl_b32 s4, s5, s4

 ;
 ; GFX11-LABEL: s_fshl_v4i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_lshr_b32 s6, s1, 8
 ; GFX11-NEXT:    s_lshr_b32 s7, s1, 16
 ; GFX11-NEXT:    s_lshr_b32 s8, s1, 24
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshr_b32 s9, s2, 8
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX11-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX11-NEXT:    s_and_b32 s12, s2, 7
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX11-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX11-NEXT:    s_and_b32 s2, s6, 0xff
 ; GFX11-NEXT:    s_and_b32 s6, s9, 7
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX11-NEXT:    s_and_not1_b32 s9, 7, s9
 ; GFX11-NEXT:    s_lshr_b32 s2, s2, 1
 ; GFX11-NEXT:    s_lshr_b32 s4, s0, 16
 ; GFX11-NEXT:    s_lshr_b32 s5, s0, 24
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s12
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, s6
 ; GFX11-NEXT:    s_lshr_b32 s2, s2, s9
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    s_or_b32 s1, s3, s2
 ; GFX11-NEXT:    s_and_b32 s2, s7, 0xff
 ; GFX11-NEXT:    s_and_b32 s3, s10, 7
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX11-NEXT:    s_and_not1_b32 s6, 7, s10
 ; GFX11-NEXT:    s_lshr_b32 s2, s2, 1
 ; GFX11-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX11-NEXT:    s_lshr_b32 s2, s2, s6
 ; GFX11-NEXT:    s_and_b32 s4, s11, 7
 ; GFX11-NEXT:    s_and_not1_b32 s6, 7, s11
 ; GFX11-NEXT:    s_lshr_b32 s7, s8, 1
 ; GFX11-NEXT:    s_lshl_b32 s4, s5, s4

 define i32 @v_fshl_v4i8(i32 %lhs.arg, i32 %rhs.arg, i32 %amt.arg) {
 ; GFX6-LABEL: v_fshl_v4i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v6, 8, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v7, 16, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v8, 24, v2
 ; GFX6-NEXT:    v_and_b32_e32 v9, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v9, v0
 ; GFX6-NEXT:    v_bfe_u32 v9, v1, 1, 7
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v2, v9
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v6
-; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX6-NEXT:    v_not_b32_e32 v6, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    v_bfe_u32 v3, v1, 8, 8
 ; GFX6-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v6, v3
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    v_and_b32_e32 v3, 7, v7
-; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v7
+; GFX6-NEXT:    v_not_b32_e32 v6, v7
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, v3, v4
 ; GFX6-NEXT:    v_bfe_u32 v4, v1, 16, 8
 ; GFX6-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 1, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v6, v4
-; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v8
+; GFX6-NEXT:    v_not_b32_e32 v6, v8
 ; GFX6-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 7, v8
 ; GFX6-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 25, v1
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, v4, v5
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v6, v1
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xff, v0

 ;
 ; GFX8-LABEL: v_fshl_v4i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v6, 16, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v7, 24, v2
 ; GFX8-NEXT:    v_and_b32_e32 v8, 7, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_mov_b32_e32 v10, 1
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v11, v10, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v8, v8, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v2, v11
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX8-NEXT:    v_or_b32_e32 v2, v8, v2
 ; GFX8-NEXT:    v_and_b32_e32 v8, 7, v5
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX8-NEXT:    v_not_b32_e32 v5, v5
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v4, v10, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_mov_b32_e32 v9, 0xff
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, v8, v3
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v4, v5, v4
 ; GFX8-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX8-NEXT:    v_and_b32_e32 v4, 7, v6
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v6
+; GFX8-NEXT:    v_not_b32_e32 v5, v6
 ; GFX8-NEXT:    v_and_b32_sdwa v6, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v6, 1, v6
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v5, v5, v6
 ; GFX8-NEXT:    v_or_b32_e32 v4, v4, v5
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v7
-; GFX8-NEXT:    v_xor_b32_e32 v6, -1, v7
+; GFX8-NEXT:    v_not_b32_e32 v6, v7
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX8-NEXT:    v_mov_b32_e32 v5, 1
 ; GFX8-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, v6, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 8
 ; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX8-NEXT:    v_and_b32_e32 v2, 0xff, v4
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX8-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 24, v0
 ; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v4i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v6, 16, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v7, 24, v2
 ; GFX9-NEXT:    v_and_b32_e32 v8, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    s_mov_b32 s5, 1
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v10, s5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v8, v8, v0
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v2, v2, v10
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX9-NEXT:    v_or_b32_e32 v2, v8, v2
 ; GFX9-NEXT:    v_and_b32_e32 v8, 7, v5
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX9-NEXT:    v_not_b32_e32 v5, v5
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v4, s5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_mov_b32_e32 v9, 0xff
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v3, v8, v3
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v4, v5, v4
 ; GFX9-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX9-NEXT:    v_and_b32_e32 v4, 7, v6
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v6
+; GFX9-NEXT:    v_not_b32_e32 v5, v6
 ; GFX9-NEXT:    v_and_b32_sdwa v6, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v6, 1, v6
 ; GFX9-NEXT:    v_lshlrev_b16_sdwa v4, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v5, v5, v6
 ; GFX9-NEXT:    v_or_b32_e32 v4, v4, v5
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v7
-; GFX9-NEXT:    v_xor_b32_e32 v6, -1, v7
+; GFX9-NEXT:    v_not_b32_e32 v6, v7
 ; GFX9-NEXT:    v_lshlrev_b16_sdwa v0, v5, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX9-NEXT:    v_mov_b32_e32 v5, 1
 ; GFX9-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v1, v6, v1
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 8
 ; GFX9-NEXT:    s_movk_i32 s4, 0xff

 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v8, 8, v2
 ; GFX10-NEXT:    v_and_b32_e32 v10, 7, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 8, v1
-; GFX10-NEXT:    v_xor_b32_e32 v9, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v9, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 16, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 24, v2
 ; GFX10-NEXT:    v_lshlrev_b16 v0, v10, v0
-; GFX10-NEXT:    v_xor_b32_e32 v10, -1, v8
+; GFX10-NEXT:    v_not_b32_e32 v10, v8
 ; GFX10-NEXT:    v_and_b32_e32 v8, 7, v8
 ; GFX10-NEXT:    v_mov_b32_e32 v13, 0xff
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 24, v1
 ; GFX10-NEXT:    v_and_b32_e32 v12, 0xff, v1
 ; GFX10-NEXT:    v_and_b32_e32 v6, 0xff, v6
 ; GFX10-NEXT:    v_lshlrev_b16 v3, v8, v3
-; GFX10-NEXT:    v_xor_b32_e32 v8, -1, v11
+; GFX10-NEXT:    v_not_b32_e32 v8, v11
 ; GFX10-NEXT:    v_and_b32_sdwa v1, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX10-NEXT:    v_xor_b32_e32 v13, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v13, v2
 ; GFX10-NEXT:    v_and_b32_e32 v10, 7, v10
 ; GFX10-NEXT:    v_lshrrev_b16 v6, 1, v6
 ; GFX10-NEXT:    v_and_b32_e32 v11, 7, v11
 ; GFX10-NEXT:    v_and_b32_e32 v8, 7, v8
 ; GFX10-NEXT:    v_lshrrev_b16 v1, 1, v1
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX10-NEXT:    v_and_b32_e32 v13, 7, v13
 ; GFX10-NEXT:    v_lshrrev_b16 v7, 1, v7

 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v6, 8, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v9, 8, v2
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v7, 16, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v10, 16, v2
 ; GFX11-NEXT:    v_and_b32_e32 v6, 0xff, v6
-; GFX11-NEXT:    v_xor_b32_e32 v13, -1, v9
+; GFX11-NEXT:    v_not_b32_e32 v13, v9
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v11, 24, v2
 ; GFX11-NEXT:    v_and_b32_e32 v9, 7, v9
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v8, 24, v1
 ; GFX11-NEXT:    v_lshrrev_b16 v6, 1, v6
 ; GFX11-NEXT:    v_and_b32_e32 v13, 7, v13
 ; GFX11-NEXT:    v_and_b32_e32 v7, 0xff, v7
 ; GFX11-NEXT:    v_lshlrev_b16 v3, v9, v3
-; GFX11-NEXT:    v_xor_b32_e32 v9, -1, v10
+; GFX11-NEXT:    v_not_b32_e32 v9, v10
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
 ; GFX11-NEXT:    v_lshrrev_b16 v6, v13, v6
-; GFX11-NEXT:    v_xor_b32_e32 v13, -1, v11
+; GFX11-NEXT:    v_not_b32_e32 v13, v11
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
 ; GFX11-NEXT:    v_and_b32_e32 v12, 7, v2
-; GFX11-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v2, v2
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX11-NEXT:    v_and_b32_e32 v10, 7, v10
 ; GFX11-NEXT:    v_and_b32_e32 v9, 7, v9
 ; GFX11-NEXT:    v_lshrrev_b16 v7, 1, v7
 ; GFX11-NEXT:    v_and_b32_e32 v11, 7, v11
 ; GFX11-NEXT:    v_and_b32_e32 v13, 7, v13
 ; GFX11-NEXT:    v_lshrrev_b16 v8, 1, v8
 ; GFX11-NEXT:    v_and_b32_e32 v2, 7, v2

 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX6-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX6-NEXT:    s_lshr_b32 s7, s0, 24
 ; GFX6-NEXT:    s_and_b32 s9, s0, 0xff
 ; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x80008
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 8
 ; GFX6-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX6-NEXT:    s_or_b32 s0, s9, s0
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX6-NEXT:    s_lshr_b32 s8, s1, 8
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX6-NEXT:    s_or_b32 s0, s0, s6
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX6-NEXT:    s_and_b32 s6, s8, 0xff
 ; GFX6-NEXT:    s_or_b32 s1, s7, s1
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    v_mov_b32_e32 v1, 0xffffffe8
 ; GFX6-NEXT:    s_or_b32 s1, s1, s6
 ; GFX6-NEXT:    s_lshr_b32 s6, s2, 16
 ; GFX6-NEXT:    s_lshr_b32 s7, s2, 24
 ; GFX6-NEXT:    s_and_b32 s9, s2, 0xff
 ; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x80008
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, v1
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 8
 ; GFX6-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX6-NEXT:    s_or_b32 s2, s9, s2
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    s_lshr_b32 s8, s3, 8
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX6-NEXT:    s_or_b32 s2, s2, s6
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX6-NEXT:    s_and_b32 s6, s8, 0xff
 ; GFX6-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX6-NEXT:    s_or_b32 s3, s7, s3
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    s_or_b32 s3, s3, s6
 ; GFX6-NEXT:    s_lshr_b32 s6, s4, 16
 ; GFX6-NEXT:    s_lshr_b32 s7, s4, 24
 ; GFX6-NEXT:    s_and_b32 s9, s4, 0xff
 ; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x80008
 ; GFX6-NEXT:    s_lshl_b32 s4, s4, 8
 ; GFX6-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
 ; GFX6-NEXT:    s_or_b32 s4, s9, s4
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    s_or_b32 s4, s4, s6
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX6-NEXT:    s_lshr_b32 s8, s5, 8
 ; GFX6-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX6-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX6-NEXT:    v_mul_lo_u32 v1, v2, v1
 ; GFX6-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    v_subrev_i32_e32 v3, vcc, 24, v0
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v2, v1
 ; GFX6-NEXT:    s_and_b32 s6, s8, 0xff
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX6-NEXT:    s_or_b32 s5, s7, s5
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    v_subrev_i32_e32 v3, vcc, 24, v0
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX6-NEXT:    s_or_b32 s5, s5, s6
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX6-NEXT:    v_sub_i32_e32 v3, vcc, 23, v0
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xffffff, v0

 ; GFX6-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX6-NEXT:    v_readfirstlane_b32 s1, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v2i24:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_lshr_b32 s6, s0, 8
 ; GFX8-NEXT:    s_and_b32 s6, s6, 0xff
-; GFX8-NEXT:    s_bfe_u32 s10, 8, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s0, 24
 ; GFX8-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s10
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX8-NEXT:    s_or_b32 s0, s0, s6
 ; GFX8-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX8-NEXT:    s_lshr_b32 s9, s1, 8
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_or_b32 s0, s0, s6
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s10
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX8-NEXT:    s_and_b32 s6, s9, 0xff
 ; GFX8-NEXT:    s_or_b32 s1, s8, s1
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    s_or_b32 s1, s1, s6
 ; GFX8-NEXT:    s_lshr_b32 s6, s2, 8
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX8-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX8-NEXT:    s_lshr_b32 s7, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s2, 24
 ; GFX8-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s10
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX8-NEXT:    s_or_b32 s2, s2, s6
 ; GFX8-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0xffffffe8
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX8-NEXT:    v_mul_lo_u32 v2, v0, v1
 ; GFX8-NEXT:    s_lshr_b32 s9, s3, 8
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX8-NEXT:    s_or_b32 s2, s2, s6
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s10
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX8-NEXT:    s_and_b32 s6, s9, 0xff
 ; GFX8-NEXT:    s_or_b32 s3, s8, s3
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX8-NEXT:    s_or_b32 s3, s3, s6
 ; GFX8-NEXT:    s_lshr_b32 s6, s4, 8
 ; GFX8-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX8-NEXT:    s_lshr_b32 s7, s4, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s4, 24
 ; GFX8-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s10
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX8-NEXT:    s_or_b32 s4, s4, s6
 ; GFX8-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    s_or_b32 s4, s4, s6
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX8-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX8-NEXT:    s_lshr_b32 s9, s5, 8
 ; GFX8-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX8-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX8-NEXT:    v_mul_lo_u32 v1, v2, v1
-; GFX8-NEXT:    s_lshl_b32 s5, s5, s10
+; GFX8-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    v_subrev_u32_e32 v3, vcc, 24, v0
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v1, v2, v1
 ; GFX8-NEXT:    s_and_b32 s6, s9, 0xff
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX8-NEXT:    s_or_b32 s5, s8, s5
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX8-NEXT:    v_subrev_u32_e32 v3, vcc, 24, v0
-; GFX8-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX8-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX8-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX8-NEXT:    s_or_b32 s5, s5, s6
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX8-NEXT:    v_add_u32_e32 v1, vcc, v2, v1
 ; GFX8-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX8-NEXT:    v_sub_u32_e32 v3, vcc, 23, v0
 ; GFX8-NEXT:    v_and_b32_e32 v0, 0xffffff, v0

 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v2i24:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 8
 ; GFX9-NEXT:    s_and_b32 s7, s7, 0xff
-; GFX9-NEXT:    s_bfe_u32 s12, 8, 0x100000
+; GFX9-NEXT:    s_lshr_b32 s9, s0, 16
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_lshr_b32 s9, s0, 16
 ; GFX9-NEXT:    s_lshr_b32 s10, s0, 24
 ; GFX9-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX9-NEXT:    s_lshl_b32 s7, s7, s12
+; GFX9-NEXT:    s_lshl_b32 s7, s7, 8
 ; GFX9-NEXT:    s_or_b32 s0, s0, s7
 ; GFX9-NEXT:    s_and_b32 s7, s9, 0xff
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0xffffffe8
 ; GFX9-NEXT:    s_lshr_b32 s11, s1, 8
-; GFX9-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    v_mul_lo_u32 v2, v0, v1
 ; GFX9-NEXT:    s_or_b32 s0, s0, s7
-; GFX9-NEXT:    s_lshl_b32 s1, s1, s12
+; GFX9-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX9-NEXT:    s_and_b32 s7, s11, 0xff
 ; GFX9-NEXT:    s_or_b32 s1, s10, s1
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    s_or_b32 s1, s1, s7
 ; GFX9-NEXT:    s_lshr_b32 s7, s2, 8
 ; GFX9-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX9-NEXT:    s_and_b32 s7, s7, 0xff
 ; GFX9-NEXT:    s_lshr_b32 s9, s2, 16
 ; GFX9-NEXT:    s_lshr_b32 s10, s2, 24
 ; GFX9-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX9-NEXT:    s_lshl_b32 s7, s7, s12
+; GFX9-NEXT:    s_lshl_b32 s7, s7, 8
 ; GFX9-NEXT:    s_or_b32 s2, s2, s7
 ; GFX9-NEXT:    s_and_b32 s7, s9, 0xff
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX9-NEXT:    v_add_u32_e32 v0, v0, v2
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
 ; GFX9-NEXT:    s_lshr_b32 s11, s3, 8
-; GFX9-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v2, v2
 ; GFX9-NEXT:    s_or_b32 s2, s2, s7
-; GFX9-NEXT:    s_lshl_b32 s3, s3, s12
+; GFX9-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX9-NEXT:    s_and_b32 s7, s11, 0xff
 ; GFX9-NEXT:    s_or_b32 s3, s10, s3
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    s_or_b32 s3, s3, s7
 ; GFX9-NEXT:    s_lshr_b32 s7, s4, 8
 ; GFX9-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX9-NEXT:    s_and_b32 s7, s7, 0xff
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX9-NEXT:    s_lshr_b32 s9, s4, 16
 ; GFX9-NEXT:    s_lshr_b32 s10, s4, 24
 ; GFX9-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX9-NEXT:    s_lshl_b32 s7, s7, s12
+; GFX9-NEXT:    s_lshl_b32 s7, s7, 8
 ; GFX9-NEXT:    s_or_b32 s4, s4, s7
 ; GFX9-NEXT:    s_and_b32 s7, s9, 0xff
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
+; GFX9-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    v_mul_lo_u32 v1, v2, v1
 ; GFX9-NEXT:    s_or_b32 s4, s4, s7
 ; GFX9-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX9-NEXT:    s_lshr_b32 s11, s5, 8
 ; GFX9-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX9-NEXT:    v_mul_hi_u32 v1, v2, v1
-; GFX9-NEXT:    s_lshl_b32 s5, s5, s12
+; GFX9-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX9-NEXT:    s_and_b32 s7, s11, 0xff
 ; GFX9-NEXT:    s_or_b32 s5, s10, s5
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX9-NEXT:    v_mul_lo_u32 v0, v0, 24
-; GFX9-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX9-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX9-NEXT:    s_or_b32 s5, s5, s7
 ; GFX9-NEXT:    v_add_u32_e32 v1, v2, v1
 ; GFX9-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX9-NEXT:    v_sub_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    v_subrev_u32_e32 v3, 24, v0
 ; GFX9-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc

 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v1, 24
 ; GFX10-NEXT:    s_lshr_b32 s6, s0, 8
 ; GFX10-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX10-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v1, v1
-; GFX10-NEXT:    s_bfe_u32 s10, 8, 0x100000
 ; GFX10-NEXT:    s_lshr_b32 s8, s0, 24
 ; GFX10-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX10-NEXT:    s_lshl_b32 s6, s6, s10
+; GFX10-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX10-NEXT:    s_and_b32 s7, s7, 0xff
 ; GFX10-NEXT:    s_or_b32 s0, s0, s6
-; GFX10-NEXT:    s_bfe_u32 s6, s7, 0x100000
+; GFX10-NEXT:    s_and_b32 s6, 0xffff, s7
 ; GFX10-NEXT:    s_lshr_b32 s7, s4, 8
+; GFX10-NEXT:    s_lshr_b32 s10, s4, 16
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
 ; GFX10-NEXT:    s_and_b32 s7, s7, 0xff
-; GFX10-NEXT:    s_lshr_b32 s11, s4, 16
-; GFX10-NEXT:    s_lshr_b32 s12, s4, 24
+; GFX10-NEXT:    s_lshr_b32 s11, s4, 24
+; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v1, v1
-; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX10-NEXT:    s_lshl_b32 s7, s7, s10
-; GFX10-NEXT:    s_lshr_b32 s13, s5, 8
+; GFX10-NEXT:    s_lshl_b32 s7, s7, 8
+; GFX10-NEXT:    s_lshr_b32 s12, s5, 8
+; GFX10-NEXT:    s_or_b32 s4, s4, s7
 ; GFX10-NEXT:    v_mul_lo_u32 v2, 0xffffffe8, v0
 ; GFX10-NEXT:    v_mul_lo_u32 v3, 0xffffffe8, v1
-; GFX10-NEXT:    s_or_b32 s4, s4, s7
-; GFX10-NEXT:    s_and_b32 s7, s11, 0xff
-; GFX10-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX10-NEXT:    s_and_b32 s7, s10, 0xff
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX10-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX10-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s7, s7, 16
+; GFX10-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX10-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX10-NEXT:    v_mul_hi_u32 v3, v1, v3
-; GFX10-NEXT:    s_lshl_b32 s5, s5, s10
 ; GFX10-NEXT:    s_or_b32 s4, s4, s7
-; GFX10-NEXT:    s_and_b32 s7, s13, 0xff
-; GFX10-NEXT:    s_or_b32 s5, s12, s5
-; GFX10-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX10-NEXT:    s_and_b32 s7, s12, 0xff
+; GFX10-NEXT:    s_or_b32 s5, s11, s5
+; GFX10-NEXT:    s_and_b32 s7, 0xffff, s7
+; GFX10-NEXT:    s_and_b32 s5, 0xffff, s5
+; GFX10-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX10-NEXT:    v_add_nc_u32_e32 v0, v0, v2
 ; GFX10-NEXT:    v_add_nc_u32_e32 v1, v1, v3
-; GFX10-NEXT:    s_lshl_b32 s7, s7, 16
-; GFX10-NEXT:    s_lshr_b32 s9, s1, 8
 ; GFX10-NEXT:    s_or_b32 s5, s5, s7
+; GFX10-NEXT:    s_lshr_b32 s9, s1, 8
+; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX10-NEXT:    v_mul_hi_u32 v1, s5, v1
-; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
+; GFX10-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX10-NEXT:    s_and_b32 s7, s9, 0xff
-; GFX10-NEXT:    s_lshl_b32 s1, s1, s10
-; GFX10-NEXT:    s_lshr_b32 s9, s2, 16
 ; GFX10-NEXT:    s_or_b32 s1, s8, s1
 ; GFX10-NEXT:    s_lshr_b32 s8, s2, 8
+; GFX10-NEXT:    s_lshr_b32 s9, s2, 16
+; GFX10-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX10-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX10-NEXT:    v_mul_lo_u32 v1, v1, 24
-; GFX10-NEXT:    s_and_b32 s8, s8, 0xff
-; GFX10-NEXT:    s_lshr_b32 s11, s2, 24
+; GFX10-NEXT:    s_lshr_b32 s10, s2, 24
 ; GFX10-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX10-NEXT:    s_lshl_b32 s8, s8, s10
-; GFX10-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX10-NEXT:    s_lshl_b32 s8, s8, 8
+; GFX10-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX10-NEXT:    s_or_b32 s2, s2, s8
+; GFX10-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v0, s4, v0
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v1, s5, v1
 ; GFX10-NEXT:    s_lshr_b32 s4, s3, 8
 ; GFX10-NEXT:    s_and_b32 s5, s9, 0xff
 ; GFX10-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v3, 24, v1
-; GFX10-NEXT:    s_bfe_u32 s5, s5, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s3, s3, s10
+; GFX10-NEXT:    s_and_b32 s5, 0xffff, s5
+; GFX10-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX10-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX10-NEXT:    s_lshl_b32 s5, s5, 16
-; GFX10-NEXT:    s_or_b32 s3, s11, s3
+; GFX10-NEXT:    s_or_b32 s3, s10, s3
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX10-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc_lo
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX10-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX10-NEXT:    s_or_b32 s2, s2, s5
-; GFX10-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX10-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v3, 24, v1
 ; GFX10-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
 ; GFX10-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-NEXT:    s_lshr_b32 s2, s2, 1
 ; GFX10-NEXT:    s_or_b32 s3, s3, s4
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v2, 23, v0
 ; GFX10-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc_lo
 ; GFX10-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
-; GFX10-NEXT:    s_bfe_u32 s0, s0, 0x100000
 ; GFX10-NEXT:    s_lshl_b32 s6, s6, 16
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v3, 23, v1
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX10-NEXT:    s_lshl_b32 s7, s7, 16
+; GFX10-NEXT:    s_or_b32 s0, s0, s6
 ; GFX10-NEXT:    v_lshrrev_b32_e64 v2, v2, s2
 ; GFX10-NEXT:    v_and_b32_e32 v3, 0xffffff, v3
 ; GFX10-NEXT:    s_lshr_b32 s2, s3, 1
-; GFX10-NEXT:    s_or_b32 s0, s0, s6
 ; GFX10-NEXT:    s_or_b32 s1, s1, s7
 ; GFX10-NEXT:    v_lshl_or_b32 v0, s0, v0, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e64 v3, v3, s2
 ; GFX10-NEXT:    s_mov_b32 s0, 8
 ; GFX10-NEXT:    v_lshlrev_b32_sdwa v2, s0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1
 ; GFX10-NEXT:    v_lshl_or_b32 v1, s1, v1, v3
 ; GFX10-NEXT:    s_mov_b32 s0, 16
 ; GFX10-NEXT:    v_and_or_b32 v2, v0, 0xff, v2
 ; GFX10-NEXT:    v_and_b32_e32 v3, 0xff, v1
 ; GFX10-NEXT:    v_lshlrev_b32_sdwa v0, s0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_2
 ; GFX10-NEXT:    v_bfe_u32 v4, v1, 8, 8
 ; GFX10-NEXT:    v_bfe_u32 v1, v1, 16, 8
 ; GFX10-NEXT:    v_lshlrev_b32_e32 v3, 24, v3
 ; GFX10-NEXT:    v_lshl_or_b32 v1, v1, 8, v4
 ; GFX10-NEXT:    v_or3_b32 v0, v2, v0, v3
 ; GFX10-NEXT:    v_readfirstlane_b32 s1, v1
 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_v2i24:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v1, 24
 ; GFX11-NEXT:    s_lshr_b32 s6, s0, 8
-; GFX11-NEXT:    s_bfe_u32 s9, 8, 0x100000
+; GFX11-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX11-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX11-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX11-NEXT:    v_rcp_iflag_f32_e32 v1, v1
-; GFX11-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX11-NEXT:    s_lshr_b32 s8, s0, 24
 ; GFX11-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX11-NEXT:    s_lshl_b32 s6, s6, s9
-; GFX11-NEXT:    s_lshr_b32 s11, s4, 24
+; GFX11-NEXT:    s_lshl_b32 s6, s6, 8
+; GFX11-NEXT:    s_lshr_b32 s10, s4, 24
 ; GFX11-NEXT:    s_or_b32 s0, s0, s6
 ; GFX11-NEXT:    s_and_b32 s6, s7, 0xff
-; GFX11-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX11-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
 ; GFX11-NEXT:    v_dual_mul_f32 v0, 0x4f7ffffe, v0 :: v_dual_mul_f32 v1, 0x4f7ffffe, v1
-; GFX11-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX11-NEXT:    s_lshr_b32 s7, s4, 16
 ; GFX11-NEXT:    s_lshl_b32 s6, s6, 16
+; GFX11-NEXT:    s_lshr_b32 s7, s4, 16
+; GFX11-NEXT:    s_or_b32 s0, s0, s6
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_u32_f32_e32 v1, v1
-; GFX11-NEXT:    s_or_b32 s0, s0, s6
 ; GFX11-NEXT:    s_lshr_b32 s6, s4, 8
 ; GFX11-NEXT:    s_and_b32 s4, s4, 0xff
+; GFX11-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX11-NEXT:    v_mul_lo_u32 v2, 0xffffffe8, v0
 ; GFX11-NEXT:    v_mul_lo_u32 v3, 0xffffffe8, v1
-; GFX11-NEXT:    s_and_b32 s6, s6, 0xff
+; GFX11-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX11-NEXT:    s_and_b32 s7, s7, 0xff
-; GFX11-NEXT:    s_lshl_b32 s6, s6, s9
-; GFX11-NEXT:    s_lshr_b32 s12, s5, 8
 ; GFX11-NEXT:    s_or_b32 s4, s4, s6
-; GFX11-NEXT:    s_bfe_u32 s6, s7, 0x100000
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s7
+; GFX11-NEXT:    s_lshr_b32 s11, s5, 8
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX11-NEXT:    v_mul_hi_u32 v2, v0, v2
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
 ; GFX11-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX11-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX11-NEXT:    s_or_b32 s4, s4, s6
-; GFX11-NEXT:    s_lshl_b32 s5, s5, s9
-; GFX11-NEXT:    s_and_b32 s6, s12, 0xff
-; GFX11-NEXT:    s_or_b32 s5, s11, s5
+; GFX11-NEXT:    s_lshl_b32 s5, s5, 8
+; GFX11-NEXT:    s_and_b32 s6, s11, 0xff
+; GFX11-NEXT:    s_or_b32 s5, s10, s5
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX11-NEXT:    v_add_nc_u32_e32 v0, v0, v2
 ; GFX11-NEXT:    v_mul_hi_u32 v2, v1, v3
-; GFX11-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX11-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX11-NEXT:    s_lshl_b32 s6, s6, 16
+; GFX11-NEXT:    s_lshr_b32 s9, s1, 8
 ; GFX11-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX11-NEXT:    s_or_b32 s5, s5, s6
-; GFX11-NEXT:    s_lshr_b32 s10, s1, 8
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
+; GFX11-NEXT:    s_and_b32 s7, s9, 0xff
 ; GFX11-NEXT:    v_add_nc_u32_e32 v1, v1, v2
-; GFX11-NEXT:    s_and_b32 s7, s10, 0xff
-; GFX11-NEXT:    s_lshl_b32 s1, s1, s9
-; GFX11-NEXT:    s_bfe_u32 s6, s7, 0x100000
+; GFX11-NEXT:    s_lshl_b32 s1, s1, 8
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s7
+; GFX11-NEXT:    s_lshr_b32 s7, s2, 8
 ; GFX11-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX11-NEXT:    v_mul_hi_u32 v1, s5, v1
-; GFX11-NEXT:    s_lshr_b32 s7, s2, 8
 ; GFX11-NEXT:    s_or_b32 s1, s8, s1
 ; GFX11-NEXT:    s_lshr_b32 s8, s2, 16
 ; GFX11-NEXT:    s_and_b32 s7, s7, 0xff
-; GFX11-NEXT:    s_lshr_b32 s10, s3, 8
-; GFX11-NEXT:    s_lshl_b32 s7, s7, s9
+; GFX11-NEXT:    s_lshr_b32 s9, s3, 8
+; GFX11-NEXT:    s_lshl_b32 s7, s7, 8
+; GFX11-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v0, s4, v0
 ; GFX11-NEXT:    v_mul_lo_u32 v1, v1, 24
 ; GFX11-NEXT:    s_lshr_b32 s4, s2, 24
 ; GFX11-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX11-NEXT:    s_and_b32 s3, s3, 0xff
+; GFX11-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
 ; GFX11-NEXT:    s_or_b32 s2, s2, s7
-; GFX11-NEXT:    s_lshl_b32 s3, s3, s9
+; GFX11-NEXT:    s_or_b32 s3, s4, s3
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v1, s5, v1
 ; GFX11-NEXT:    s_and_b32 s5, s8, 0xff
 ; GFX11-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
-; GFX11-NEXT:    s_bfe_u32 s5, s5, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX11-NEXT:    s_and_b32 s5, 0xffff, s5
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v2, 24, v1
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v3, 24, v0
 ; GFX11-NEXT:    s_lshl_b32 s5, s5, 16
-; GFX11-NEXT:    s_or_b32 s3, s4, s3
+; GFX11-NEXT:    s_and_b32 s4, s9, 0xff
 ; GFX11-NEXT:    s_or_b32 s2, s2, s5
 ; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc_lo
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX11-NEXT:    s_and_b32 s4, s10, 0xff
 ; GFX11-NEXT:    s_lshr_b32 s2, s2, 1
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v2, 24, v1
 ; GFX11-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc_lo
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX11-NEXT:    s_lshl_b32 s6, s6, 16
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v3, 23, v0
 ; GFX11-NEXT:    v_dual_cndmask_b32 v1, v1, v2 :: v_dual_and_b32 v0, 0xffffff, v0
 ; GFX11-NEXT:    v_and_b32_e32 v2, 0xffffff, v3
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_3)
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v3, 23, v1
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e64 v2, v2, s2
-; GFX11-NEXT:    s_bfe_u32 s2, s3, 0x100000
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s3
 ; GFX11-NEXT:    s_lshl_b32 s3, s4, 16
 ; GFX11-NEXT:    v_and_b32_e32 v3, 0xffffff, v3
 ; GFX11-NEXT:    s_or_b32 s2, s2, s3
 ; GFX11-NEXT:    v_lshl_or_b32 v0, s0, v0, v2
 ; GFX11-NEXT:    s_lshr_b32 s0, s2, 1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_lshrrev_b32_e64 v2, v3, s0
 ; GFX11-NEXT:    s_or_b32 s0, s1, s6

 }
 
 define i32 @v_fshl_i32(i32 %lhs, i32 %rhs, i32 %amt) {
 ; GFX6-LABEL: v_fshl_i32:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_alignbit_b32 v1, v0, v1, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_alignbit_b32 v0, v0, v1, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_i32:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_alignbit_b32 v1, v0, v1, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_alignbit_b32 v0, v0, v1, v2
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_alignbit_b32 v1, v0, v1, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    v_alignbit_b32 v0, v0, v1, v2
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_alignbit_b32 v1, v0, v1, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX10-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v2, v2
 ; GFX10-NEXT:    v_alignbit_b32 v0, v0, v1, v2
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshl_i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_alignbit_b32 v1, v0, v1, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX11-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v2, v2
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_alignbit_b32 v0, v0, v1, v2
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call i32 @llvm.fshl.i32(i32 %lhs, i32 %rhs, i32 %amt)
   ret i32 %result
 }
 
 define i32 @v_fshl_i32_5(i32 %lhs, i32 %rhs) {

 }
 
 define amdgpu_ps float @v_fshl_i32_ssv(i32 inreg %lhs, i32 inreg %rhs, i32 %amt) {
 ; GFX6-LABEL: v_fshl_i32_ssv:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX6-NEXT:    v_alignbit_b32 v1, s0, v1, 1
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, 1
-; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
+; GFX6-NEXT:    v_not_b32_e32 v0, v0
 ; GFX6-NEXT:    v_alignbit_b32 v0, s0, v1, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_i32_ssv:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX8-NEXT:    v_alignbit_b32 v1, s0, v1, 1
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
-; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
+; GFX8-NEXT:    v_not_b32_e32 v0, v0
 ; GFX8-NEXT:    v_alignbit_b32 v0, s0, v1, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_i32_ssv:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX9-NEXT:    v_alignbit_b32 v1, s0, v1, 1
 ; GFX9-NEXT:    s_lshr_b32 s0, s0, 1
-; GFX9-NEXT:    v_xor_b32_e32 v0, -1, v0
+; GFX9-NEXT:    v_not_b32_e32 v0, v0
 ; GFX9-NEXT:    v_alignbit_b32 v0, s0, v1, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshl_i32_ssv:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_alignbit_b32 v1, s0, s1, 1
-; GFX10-NEXT:    v_xor_b32_e32 v0, -1, v0
+; GFX10-NEXT:    v_not_b32_e32 v0, v0
 ; GFX10-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX10-NEXT:    v_alignbit_b32 v0, s0, v1, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshl_i32_ssv:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_alignbit_b32 v1, s0, s1, 1
-; GFX11-NEXT:    v_xor_b32_e32 v0, -1, v0
+; GFX11-NEXT:    v_not_b32_e32 v0, v0
 ; GFX11-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_alignbit_b32 v0, s0, v1, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i32 @llvm.fshl.i32(i32 %lhs, i32 %rhs, i32 %amt)
   %cast.result = bitcast i32 %result to float
   ret float %cast.result
 }

 }
 
 define <2 x i32> @v_fshl_v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt) {
 ; GFX6-LABEL: v_fshl_v2i32:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_alignbit_b32 v2, v0, v2, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX6-NEXT:    v_not_b32_e32 v4, v4
 ; GFX6-NEXT:    v_alignbit_b32 v0, v0, v2, v4
 ; GFX6-NEXT:    v_alignbit_b32 v2, v1, v3, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX6-NEXT:    v_xor_b32_e32 v3, -1, v5
+; GFX6-NEXT:    v_not_b32_e32 v3, v5
 ; GFX6-NEXT:    v_alignbit_b32 v1, v1, v2, v3
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v2i32:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_alignbit_b32 v2, v0, v2, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX8-NEXT:    v_not_b32_e32 v4, v4
 ; GFX8-NEXT:    v_alignbit_b32 v0, v0, v2, v4
 ; GFX8-NEXT:    v_alignbit_b32 v2, v1, v3, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX8-NEXT:    v_xor_b32_e32 v3, -1, v5
+; GFX8-NEXT:    v_not_b32_e32 v3, v5
 ; GFX8-NEXT:    v_alignbit_b32 v1, v1, v2, v3
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v2i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_alignbit_b32 v2, v0, v2, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX9-NEXT:    v_not_b32_e32 v4, v4
 ; GFX9-NEXT:    v_alignbit_b32 v0, v0, v2, v4
 ; GFX9-NEXT:    v_alignbit_b32 v2, v1, v3, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX9-NEXT:    v_xor_b32_e32 v3, -1, v5
+; GFX9-NEXT:    v_not_b32_e32 v3, v5
 ; GFX9-NEXT:    v_alignbit_b32 v1, v1, v2, v3
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_v2i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_alignbit_b32 v2, v0, v2, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX10-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX10-NEXT:    v_not_b32_e32 v4, v4
 ; GFX10-NEXT:    v_alignbit_b32 v3, v1, v3, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX10-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX10-NEXT:    v_not_b32_e32 v5, v5
 ; GFX10-NEXT:    v_alignbit_b32 v0, v0, v2, v4
 ; GFX10-NEXT:    v_alignbit_b32 v1, v1, v3, v5
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshl_v2i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_alignbit_b32 v2, v0, v2, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX11-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX11-NEXT:    v_not_b32_e32 v4, v4
 ; GFX11-NEXT:    v_alignbit_b32 v3, v1, v3, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX11-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX11-NEXT:    v_not_b32_e32 v5, v5
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_alignbit_b32 v0, v0, v2, v4
 ; GFX11-NEXT:    v_alignbit_b32 v1, v1, v3, v5
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %lhs, <2 x i32> %rhs, <2 x i32> %amt)
   ret <2 x i32> %result
 }
 
 define <3 x i32> @v_fshl_v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt) {
 ; GFX6-LABEL: v_fshl_v3i32:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_alignbit_b32 v3, v0, v3, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX6-NEXT:    v_not_b32_e32 v6, v6
 ; GFX6-NEXT:    v_alignbit_b32 v0, v0, v3, v6
 ; GFX6-NEXT:    v_alignbit_b32 v3, v1, v4, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v7
+; GFX6-NEXT:    v_not_b32_e32 v4, v7
 ; GFX6-NEXT:    v_alignbit_b32 v1, v1, v3, v4
 ; GFX6-NEXT:    v_alignbit_b32 v3, v2, v5, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v8
+; GFX6-NEXT:    v_not_b32_e32 v4, v8
 ; GFX6-NEXT:    v_alignbit_b32 v2, v2, v3, v4
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v3i32:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_alignbit_b32 v3, v0, v3, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX8-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX8-NEXT:    v_not_b32_e32 v6, v6
 ; GFX8-NEXT:    v_alignbit_b32 v0, v0, v3, v6
 ; GFX8-NEXT:    v_alignbit_b32 v3, v1, v4, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v7
+; GFX8-NEXT:    v_not_b32_e32 v4, v7
 ; GFX8-NEXT:    v_alignbit_b32 v1, v1, v3, v4
 ; GFX8-NEXT:    v_alignbit_b32 v3, v2, v5, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v8
+; GFX8-NEXT:    v_not_b32_e32 v4, v8
 ; GFX8-NEXT:    v_alignbit_b32 v2, v2, v3, v4
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v3i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_alignbit_b32 v3, v0, v3, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX9-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX9-NEXT:    v_not_b32_e32 v6, v6
 ; GFX9-NEXT:    v_alignbit_b32 v0, v0, v3, v6
 ; GFX9-NEXT:    v_alignbit_b32 v3, v1, v4, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v7
+; GFX9-NEXT:    v_not_b32_e32 v4, v7
 ; GFX9-NEXT:    v_alignbit_b32 v1, v1, v3, v4
 ; GFX9-NEXT:    v_alignbit_b32 v3, v2, v5, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v8
+; GFX9-NEXT:    v_not_b32_e32 v4, v8
 ; GFX9-NEXT:    v_alignbit_b32 v2, v2, v3, v4
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_v3i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_alignbit_b32 v3, v0, v3, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX10-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX10-NEXT:    v_not_b32_e32 v6, v6
 ; GFX10-NEXT:    v_alignbit_b32 v4, v1, v4, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX10-NEXT:    v_xor_b32_e32 v7, -1, v7
+; GFX10-NEXT:    v_not_b32_e32 v7, v7
 ; GFX10-NEXT:    v_alignbit_b32 v5, v2, v5, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX10-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX10-NEXT:    v_not_b32_e32 v8, v8
 ; GFX10-NEXT:    v_alignbit_b32 v0, v0, v3, v6
 ; GFX10-NEXT:    v_alignbit_b32 v1, v1, v4, v7
 ; GFX10-NEXT:    v_alignbit_b32 v2, v2, v5, v8
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshl_v3i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_alignbit_b32 v3, v0, v3, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX11-NEXT:    v_xor_b32_e32 v6, -1, v6
+; GFX11-NEXT:    v_not_b32_e32 v6, v6
 ; GFX11-NEXT:    v_alignbit_b32 v4, v1, v4, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX11-NEXT:    v_xor_b32_e32 v7, -1, v7
+; GFX11-NEXT:    v_not_b32_e32 v7, v7
 ; GFX11-NEXT:    v_alignbit_b32 v5, v2, v5, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX11-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX11-NEXT:    v_not_b32_e32 v8, v8
 ; GFX11-NEXT:    v_alignbit_b32 v0, v0, v3, v6
 ; GFX11-NEXT:    v_alignbit_b32 v1, v1, v4, v7
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_3)
 ; GFX11-NEXT:    v_alignbit_b32 v2, v2, v5, v8
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call <3 x i32> @llvm.fshl.v3i32(<3 x i32> %lhs, <3 x i32> %rhs, <3 x i32> %amt)
   ret <3 x i32> %result
 }
 
 define <4 x i32> @v_fshl_v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt) {
 ; GFX6-LABEL: v_fshl_v4i32:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_alignbit_b32 v4, v0, v4, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX6-NEXT:    v_not_b32_e32 v8, v8
 ; GFX6-NEXT:    v_alignbit_b32 v0, v0, v4, v8
 ; GFX6-NEXT:    v_alignbit_b32 v4, v1, v5, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX6-NEXT:    v_xor_b32_e32 v5, -1, v9
+; GFX6-NEXT:    v_not_b32_e32 v5, v9
 ; GFX6-NEXT:    v_alignbit_b32 v1, v1, v4, v5
 ; GFX6-NEXT:    v_alignbit_b32 v4, v2, v6, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX6-NEXT:    v_xor_b32_e32 v5, -1, v10
+; GFX6-NEXT:    v_not_b32_e32 v5, v10
 ; GFX6-NEXT:    v_alignbit_b32 v2, v2, v4, v5
 ; GFX6-NEXT:    v_alignbit_b32 v4, v3, v7, 1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
-; GFX6-NEXT:    v_xor_b32_e32 v5, -1, v11
+; GFX6-NEXT:    v_not_b32_e32 v5, v11
 ; GFX6-NEXT:    v_alignbit_b32 v3, v3, v4, v5
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v4i32:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_alignbit_b32 v4, v0, v4, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX8-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX8-NEXT:    v_not_b32_e32 v8, v8
 ; GFX8-NEXT:    v_alignbit_b32 v0, v0, v4, v8
 ; GFX8-NEXT:    v_alignbit_b32 v4, v1, v5, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v9
+; GFX8-NEXT:    v_not_b32_e32 v5, v9
 ; GFX8-NEXT:    v_alignbit_b32 v1, v1, v4, v5
 ; GFX8-NEXT:    v_alignbit_b32 v4, v2, v6, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v10
+; GFX8-NEXT:    v_not_b32_e32 v5, v10
 ; GFX8-NEXT:    v_alignbit_b32 v2, v2, v4, v5
 ; GFX8-NEXT:    v_alignbit_b32 v4, v3, v7, 1
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v11
+; GFX8-NEXT:    v_not_b32_e32 v5, v11
 ; GFX8-NEXT:    v_alignbit_b32 v3, v3, v4, v5
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v4i32:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_alignbit_b32 v4, v0, v4, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX9-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX9-NEXT:    v_not_b32_e32 v8, v8
 ; GFX9-NEXT:    v_alignbit_b32 v0, v0, v4, v8
 ; GFX9-NEXT:    v_alignbit_b32 v4, v1, v5, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v9
+; GFX9-NEXT:    v_not_b32_e32 v5, v9
 ; GFX9-NEXT:    v_alignbit_b32 v1, v1, v4, v5
 ; GFX9-NEXT:    v_alignbit_b32 v4, v2, v6, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v10
+; GFX9-NEXT:    v_not_b32_e32 v5, v10
 ; GFX9-NEXT:    v_alignbit_b32 v2, v2, v4, v5
 ; GFX9-NEXT:    v_alignbit_b32 v4, v3, v7, 1
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v11
+; GFX9-NEXT:    v_not_b32_e32 v5, v11
 ; GFX9-NEXT:    v_alignbit_b32 v3, v3, v4, v5
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshl_v4i32:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_alignbit_b32 v4, v0, v4, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX10-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX10-NEXT:    v_not_b32_e32 v8, v8
 ; GFX10-NEXT:    v_alignbit_b32 v5, v1, v5, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX10-NEXT:    v_xor_b32_e32 v9, -1, v9
+; GFX10-NEXT:    v_not_b32_e32 v9, v9
 ; GFX10-NEXT:    v_alignbit_b32 v6, v2, v6, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX10-NEXT:    v_xor_b32_e32 v10, -1, v10
+; GFX10-NEXT:    v_not_b32_e32 v10, v10
 ; GFX10-NEXT:    v_alignbit_b32 v7, v3, v7, 1
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
-; GFX10-NEXT:    v_xor_b32_e32 v11, -1, v11
+; GFX10-NEXT:    v_not_b32_e32 v11, v11
 ; GFX10-NEXT:    v_alignbit_b32 v0, v0, v4, v8
 ; GFX10-NEXT:    v_alignbit_b32 v1, v1, v5, v9
 ; GFX10-NEXT:    v_alignbit_b32 v2, v2, v6, v10
 ; GFX10-NEXT:    v_alignbit_b32 v3, v3, v7, v11
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshl_v4i32:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_alignbit_b32 v4, v0, v4, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v0, 1, v0
-; GFX11-NEXT:    v_xor_b32_e32 v8, -1, v8
+; GFX11-NEXT:    v_not_b32_e32 v8, v8
 ; GFX11-NEXT:    v_alignbit_b32 v5, v1, v5, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v1, 1, v1
-; GFX11-NEXT:    v_xor_b32_e32 v9, -1, v9
+; GFX11-NEXT:    v_not_b32_e32 v9, v9
 ; GFX11-NEXT:    v_alignbit_b32 v6, v2, v6, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v2, 1, v2
-; GFX11-NEXT:    v_xor_b32_e32 v10, -1, v10
+; GFX11-NEXT:    v_not_b32_e32 v10, v10
 ; GFX11-NEXT:    v_alignbit_b32 v7, v3, v7, 1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 1, v3
-; GFX11-NEXT:    v_xor_b32_e32 v11, -1, v11
+; GFX11-NEXT:    v_not_b32_e32 v11, v11
 ; GFX11-NEXT:    v_alignbit_b32 v0, v0, v4, v8
 ; GFX11-NEXT:    v_alignbit_b32 v1, v1, v5, v9
 ; GFX11-NEXT:    v_alignbit_b32 v2, v2, v6, v10
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4)
 ; GFX11-NEXT:    v_alignbit_b32 v3, v3, v7, v11
 ; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %lhs, <4 x i32> %rhs, <4 x i32> %amt)
   ret <4 x i32> %result
 }
 
 define amdgpu_ps i16 @s_fshl_i16(i16 inreg %lhs, i16 inreg %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: s_fshl_i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s3, s2, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX6-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s3, s2, 15
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
 ; GFX8-NEXT:    s_andn2_b32 s2, 15, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s3
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s3, s2, 15
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
 ; GFX9-NEXT:    s_andn2_b32 s2, 15, s2
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX9-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s1, s1, s3
-; GFX9-NEXT:    s_bfe_u32 s2, s2, 0x100000
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i16:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s3, s2, 15
 ; GFX10-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, s3, 0x100000
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s4
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i16:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s3, s2, 15
 ; GFX11-NEXT:    s_and_not1_b32 s2, 15, s2
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, s3, 0x100000
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s4
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s3
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 %amt)
   ret i16 %result
 }
 
 define amdgpu_ps i16 @s_fshl_i16_4(i16 inreg %lhs, i16 inreg %rhs) {
 ; GFX6-LABEL: s_fshl_i16_4:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x4000c
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i16_4:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 4
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 12
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i16_4:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 4
+; GFX9-NEXT:    s_lshr_b32 s1, s1, 12
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i16_4:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, 12, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 4
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 12
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i16_4:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, 12, 0x100000
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 4
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 12
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 4)
   ret i16 %result
 }
 
 define amdgpu_ps i16 @s_fshl_i16_5(i16 inreg %lhs, i16 inreg %rhs) {
 ; GFX6-LABEL: s_fshl_i16_5:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 5
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x5000b
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_i16_5:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 5
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 11
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_i16_5:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 5
+; GFX9-NEXT:    s_lshr_b32 s1, s1, 11
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshl_i16_5:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, 11, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 5
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 11
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshl_i16_5:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, 11, 0x100000
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 5
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 11
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 5)
   ret i16 %result
 }
 
 define i16 @v_fshl_i16(i16 %lhs, i16 %rhs, i16 %amt) {
 ; GFX6-LABEL: v_fshl_i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v3, 15, v2
 ; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v2
-; GFX6-NEXT:    v_bfe_u32 v3, v3, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v3
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v3, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 }
 
 define amdgpu_ps half @v_fshl_i16_ssv(i16 inreg %lhs, i16 inreg %rhs, i16 %amt) {
 ; GFX6-LABEL: v_fshl_i16_ssv:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX6-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    v_lshl_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    s_bfe_u32 s0, s1, 0xf0001
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshr_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_i16_ssv:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v1, v1, s0
-; GFX8-NEXT:    s_bfe_u32 s0, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, 1, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s1
 ; GFX8-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v0, v0, s0
 ; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_i16_ssv:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX9-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX9-NEXT:    v_lshlrev_b16_e64 v1, v1, s0
-; GFX9-NEXT:    s_bfe_u32 s0, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s1, 1, 0x100000
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s1
 ; GFX9-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX9-NEXT:    s_lshr_b32 s0, s0, s1
+; GFX9-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX9-NEXT:    v_lshrrev_b16_e64 v0, v0, s0
 ; GFX9-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshl_i16_ssv:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_xor_b32_e32 v1, -1, v0
 ; GFX10-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX10-NEXT:    v_and_b32_e32 v1, 15, v1
 ; GFX10-NEXT:    v_lshlrev_b16 v0, v0, s0
 ; GFX10-NEXT:    v_lshrrev_b16 v1, v1, s1
 ; GFX10-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshl_i16_ssv:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_xor_b32_e32 v1, -1, v0
 ; GFX11-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 1
 ; GFX11-NEXT:    v_and_b32_e32 v1, 15, v1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_lshlrev_b16 v0, v0, s0
 ; GFX11-NEXT:    v_lshrrev_b16 v1, v1, s1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps half @v_fshl_i16_svs(i16 inreg %lhs, i16 %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: v_fshl_i16_svs:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s2, s1, 15
 ; GFX6-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    v_bfe_u32 v0, v0, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, s1, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_i16_svs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s2, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, 1, v0
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, s1, v0
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_i16_svs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 15
 ; GFX9-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX9-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v0, 1, v0
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v0, s1, v0
 ; GFX9-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshl_i16_svs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_lshrrev_b16 v0, 1, v0
 ; GFX10-NEXT:    s_andn2_b32 s2, 15, s1
 ; GFX10-NEXT:    s_and_b32 s1, s1, 15
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    v_lshrrev_b16 v0, s2, v0
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX10-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshl_i16_svs:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_lshrrev_b16 v0, 1, v0
 ; GFX11-NEXT:    s_and_not1_b32 s2, 15, s1
 ; GFX11-NEXT:    s_and_b32 s1, s1, 15
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    v_lshrrev_b16 v0, s2, v0
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps half @v_fshl_i16_vss(i16 %lhs, i16 inreg %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: v_fshl_i16_vss:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s2, s1, 15
 ; GFX6-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_bfe_u32 s0, s0, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s2, v0
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_i16_vss:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s2, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, s2, v0
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_i16_vss:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 15
 ; GFX9-NEXT:    s_andn2_b32 s1, 15, s1
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX9-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, s2, v0
-; GFX9-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX9-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX9-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshl_i16_vss:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s2, s1, 15
 ; GFX10-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX10-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, 1, 0x100000
+; GFX10-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX10-NEXT:    v_lshlrev_b16 v0, s2, v0
-; GFX10-NEXT:    s_lshr_b32 s0, s0, s3
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX10-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshl_i16_vss:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s2, s1, 15
 ; GFX11-NEXT:    s_and_not1_b32 s1, 15, s1
-; GFX11-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, 1, 0x100000
+; GFX11-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX11-NEXT:    v_lshlrev_b16 v0, s2, v0
-; GFX11-NEXT:    s_lshr_b32 s0, s0, s3
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshl.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps i32 @s_fshl_v2i16(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: s_fshl_v2i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s6, s4, 15
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    s_bfe_u32 s2, s2, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX6-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX6-NEXT:    s_or_b32 s0, s0, s2
 ; GFX6-NEXT:    s_and_b32 s2, s5, 15
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s5
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s2
 ; GFX6-NEXT:    s_bfe_u32 s2, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s3, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s4
 ; GFX6-NEXT:    s_lshr_b32 s2, s2, s3
 ; GFX6-NEXT:    s_or_b32 s1, s1, s2
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v2i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_and_b32 s6, s2, 15
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s3, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s4, s1, 16
 ; GFX8-NEXT:    s_lshr_b32 s5, s2, 16
+; GFX8-NEXT:    s_and_b32 s6, s2, 15
 ; GFX8-NEXT:    s_andn2_b32 s2, 15, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_lshr_b32 s3, s0, 16
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s6
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s6, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s6
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    s_and_b32 s1, s5, 15
 ; GFX8-NEXT:    s_andn2_b32 s2, 15, s5
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s1, s3, s1
-; GFX8-NEXT:    s_lshr_b32 s3, s4, s6
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s3, s4, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v2i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s3, s2, 0xf000f
 ; GFX9-NEXT:    s_lshr_b32 s4, s0, 16

 
 define <2 x i16> @v_fshl_v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt) {
 ; GFX6-LABEL: v_fshl_v2i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v4
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v6, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_bfe_u32 v2, v2, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v6, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v4, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v5
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v5
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_bfe_u32 v2, v3, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v3, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v3, v2
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v2i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 16, v2
 ; GFX8-NEXT:    v_and_b32_e32 v4, 15, v2
 ; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v2
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v5, 1, v1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v4, v4, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v2, v5
 ; GFX8-NEXT:    v_or_b32_e32 v2, v4, v2
 ; GFX8-NEXT:    v_and_b32_e32 v4, 15, v3
 ; GFX8-NEXT:    v_xor_b32_e32 v3, -1, v3
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_mov_b32_e32 v4, 1
 ; GFX8-NEXT:    v_and_b32_e32 v3, 15, v3
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, v3, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v1, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v2i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, 0xf000f, v2
 ; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call <2 x i16> @llvm.fshl.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
   ret <2 x i16> %result
 }
 
 define <2 x i16> @v_fshl_v2i16_4_8(<2 x i16> %lhs, <2 x i16> %rhs) {
 ; GFX6-LABEL: v_fshl_v2i16_4_8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX6-NEXT:    s_bfe_u32 s4, 4, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
 ; GFX6-NEXT:    v_bfe_u32 v2, v2, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s4, 11, 0x100000
-; GFX6-NEXT:    v_lshrrev_b32_e32 v2, s4, v2
-; GFX6-NEXT:    s_bfe_u32 s4, 8, 0x100000
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 11, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s4, v1
 ; GFX6-NEXT:    v_bfe_u32 v2, v3, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s4, 7, 0x100000
-; GFX6-NEXT:    v_lshrrev_b32_e32 v2, s4, v2
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 8, v1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v2i16_4_8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, 4, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, 12, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v3, 8
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v2, 8, v2
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v2i16_4_8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0x80004
 ; GFX9-NEXT:    v_pk_lshlrev_b16 v0, v2, v0

 }
 
 define amdgpu_ps float @v_fshl_v2i16_ssv(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> %amt) {
 ; GFX6-LABEL: v_fshl_v2i16_ssv:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v0
 ; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX6-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s0, v2
 ; GFX6-NEXT:    s_bfe_u32 s0, s2, 0xf0001
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshr_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v1
 ; GFX6-NEXT:    v_xor_b32_e32 v1, -1, v1
 ; GFX6-NEXT:    v_and_b32_e32 v1, 15, v1
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    s_bfe_u32 s0, s3, 0xf0001
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s1, v2
 ; GFX6-NEXT:    v_lshr_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_v2i16_ssv:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v0
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
-; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
 ; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v2, v2, s0
-; GFX8-NEXT:    s_bfe_u32 s0, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, 1, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s1
 ; GFX8-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v0, v0, s0
+; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
 ; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v1
 ; GFX8-NEXT:    v_xor_b32_e32 v1, -1, v1
 ; GFX8-NEXT:    v_and_b32_e32 v1, 15, v1
-; GFX8-NEXT:    s_lshr_b32 s0, s3, s1
+; GFX8-NEXT:    s_lshr_b32 s0, s3, 1
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v2, v2, s2
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v1, v1, s0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_v2i16_ssv:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_and_b32_e32 v1, 0xf000f, v0
 ; GFX9-NEXT:    v_pk_lshlrev_b16 v1, v1, s0
 ; GFX9-NEXT:    s_lshr_b32 s0, s1, 16
; GFX11-NEXT:    ; return to shader part epilog
   ret float %cast
 }
 
 define amdgpu_ps float @v_fshl_v2i16_svs(<2 x i16> inreg %lhs, <2 x i16> %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: v_fshl_v2i16_svs:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s4, s2, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    v_bfe_u32 v0, v0, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, s2, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    s_and_b32 s0, s3, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s3
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s0, s1, s0
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, s1, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, s0, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_v2i16_svs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s4, s1, 15
 ; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, 1, v0
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s4
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, s1, v1
 ; GFX8-NEXT:    v_or_b32_e32 v1, s0, v1
 ; GFX8-NEXT:    s_and_b32 s0, s3, 15
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 1
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s3
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    s_lshl_b32 s0, s2, s0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, s1, v0
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_v2i16_svs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 0xf000f
 ; GFX9-NEXT:    s_lshr_b32 s3, s0, 16
 ; GFX9-NEXT:    s_lshr_b32 s4, s2, 16
; GFX11-NEXT:    ; return to shader part epilog
   ret float %cast
 }
 
 define amdgpu_ps float @v_fshl_v2i16_vss(<2 x i16> %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: v_fshl_v2i16_vss:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s4, s2, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_bfe_u32 s0, s0, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s2
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    s_and_b32 s0, s3, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s3
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    s_bfe_u32 s0, s1, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX6-NEXT:    v_or_b32_e32 v1, s0, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshl_v2i16_vss:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_and_b32 s4, s1, 15
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
+; GFX8-NEXT:    s_and_b32 s4, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, s4, v0
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s4
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX8-NEXT:    v_or_b32_e32 v1, s0, v1
 ; GFX8-NEXT:    s_and_b32 s0, s3, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s3
 ; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    s_lshr_b32 s0, s2, s4
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s0, s2, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshl_v2i16_vss:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 0xf000f
 ; GFX9-NEXT:    v_pk_lshlrev_b16 v0, s2, v0
 ; GFX9-NEXT:    s_lshr_b32 s2, s0, 16

 }
 
 
 define amdgpu_ps i48 @s_fshl_v3i16(<3 x i16> inreg %lhs, <3 x i16> inreg %rhs, <3 x i16> inreg %amt) {
 ; GFX6-LABEL: s_fshl_v3i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s9, s6, 15
 ; GFX6-NEXT:    s_andn2_b32 s6, 15, s6
-; GFX6-NEXT:    s_bfe_u32 s9, s9, 0x100000
+; GFX6-NEXT:    s_and_b32 s9, 0xffff, s9
 ; GFX6-NEXT:    s_bfe_u32 s3, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s9
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s6
 ; GFX6-NEXT:    s_or_b32 s0, s0, s3
 ; GFX6-NEXT:    s_and_b32 s3, s7, 15
 ; GFX6-NEXT:    s_andn2_b32 s6, 15, s7
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s3
 ; GFX6-NEXT:    s_bfe_u32 s3, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s6
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s4
 ; GFX6-NEXT:    s_or_b32 s1, s1, s3
 ; GFX6-NEXT:    s_and_b32 s3, s8, 15
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s8
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, s3
 ; GFX6-NEXT:    s_bfe_u32 s3, s5, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s4
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_or_b32 s2, s2, s3
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v3i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_and_b32 s9, s4, 15
-; GFX8-NEXT:    s_bfe_u32 s9, s9, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s7, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s4, 16
+; GFX8-NEXT:    s_and_b32 s9, s4, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s4
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s9, 0xffff, s9
+; GFX8-NEXT:    s_lshr_b32 s2, s2, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s9
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s9, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s2, s2, s9
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
 ; GFX8-NEXT:    s_and_b32 s2, s8, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s8
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s2, s6, s2
-; GFX8-NEXT:    s_lshr_b32 s6, s7, s9
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s6, s7, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshr_b32 s4, s6, s4
 ; GFX8-NEXT:    s_or_b32 s2, s2, s4
 ; GFX8-NEXT:    s_and_b32 s4, s5, 15
 ; GFX8-NEXT:    s_andn2_b32 s5, 15, s5
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, s4
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s9
-; GFX8-NEXT:    s_bfe_u32 s4, s5, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s5
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s3
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v3i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s6, s4, 0xf000f
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX9-NEXT:    s_lshr_b32 s8, s6, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s6

 
 define <3 x half> @v_fshl_v3i16(<3 x i16> %lhs, <3 x i16> %rhs, <3 x i16> %amt) {
 ; GFX6-LABEL: v_fshl_v3i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v9, 15, v6
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v9, v9, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v9, 0xffff, v9
 ; GFX6-NEXT:    v_bfe_u32 v3, v3, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v6, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v9, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v6, v3
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v3
 ; GFX6-NEXT:    v_and_b32_e32 v3, 15, v7
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v7
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v3, v3, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v3
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_bfe_u32 v3, v4, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v3
 ; GFX6-NEXT:    v_and_b32_e32 v3, 15, v8
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v8
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v3, v3, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v3
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v3, v2
 ; GFX6-NEXT:    v_bfe_u32 v3, v5, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v3i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v6, 16, v4

 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v6, v2
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v5
 ; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v5
 ; GFX8-NEXT:    v_and_b32_e32 v5, 15, v5
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, v2, v1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, 1, v3
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v5, v2
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX8-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v3i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v6, 0xf000f, v4
 ; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX9-NEXT:    v_and_b32_e32 v4, 0xf000f, v4
; GFX11-NEXT:    s_setpc_b64 s[30:31]
   ret <3 x half> %cast.result
 }
 
 define amdgpu_ps <2 x i32> @s_fshl_v4i16(<4 x i16> inreg %lhs, <4 x i16> inreg %rhs, <4 x i16> inreg %amt) {
 ; GFX6-LABEL: s_fshl_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s12, s8, 15
 ; GFX6-NEXT:    s_andn2_b32 s8, 15, s8
-; GFX6-NEXT:    s_bfe_u32 s12, s12, 0x100000
+; GFX6-NEXT:    s_and_b32 s12, 0xffff, s12
 ; GFX6-NEXT:    s_bfe_u32 s4, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s12
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s8
 ; GFX6-NEXT:    s_or_b32 s0, s0, s4
 ; GFX6-NEXT:    s_and_b32 s4, s9, 15
 ; GFX6-NEXT:    s_andn2_b32 s8, 15, s9
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s4
 ; GFX6-NEXT:    s_bfe_u32 s4, s5, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s8
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s5
 ; GFX6-NEXT:    s_or_b32 s1, s1, s4
 ; GFX6-NEXT:    s_and_b32 s4, s10, 15
 ; GFX6-NEXT:    s_andn2_b32 s5, 15, s10
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, s4
 ; GFX6-NEXT:    s_bfe_u32 s4, s6, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s5
 ; GFX6-NEXT:    s_or_b32 s2, s2, s4
 ; GFX6-NEXT:    s_and_b32 s4, s11, 15
 ; GFX6-NEXT:    s_andn2_b32 s5, 15, s11
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, s4
 ; GFX6-NEXT:    s_bfe_u32 s4, s7, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s5
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_or_b32 s3, s3, s4
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s2, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX6-NEXT:    s_or_b32 s1, s1, s2
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshl_v4i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_and_b32 s12, s4, 15
-; GFX8-NEXT:    s_bfe_u32 s12, s12, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s8, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s10, s4, 16
+; GFX8-NEXT:    s_and_b32 s12, s4, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s4
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s12, 0xffff, s12
+; GFX8-NEXT:    s_lshr_b32 s2, s2, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s12
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s12, 1, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s2, s2, s12
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
 ; GFX8-NEXT:    s_and_b32 s2, s10, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s10
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s2, s6, s2
-; GFX8-NEXT:    s_lshr_b32 s6, s8, s12
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s6, s8, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshr_b32 s4, s6, s4
 ; GFX8-NEXT:    s_or_b32 s2, s2, s4
 ; GFX8-NEXT:    s_and_b32 s4, s5, 15
 ; GFX8-NEXT:    s_lshr_b32 s9, s3, 16
 ; GFX8-NEXT:    s_lshr_b32 s11, s5, 16
 ; GFX8-NEXT:    s_andn2_b32 s5, 15, s5
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshr_b32 s7, s1, 16
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, s4
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s12
-; GFX8-NEXT:    s_bfe_u32 s4, s5, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s5
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, s4
 ; GFX8-NEXT:    s_or_b32 s1, s1, s3
 ; GFX8-NEXT:    s_and_b32 s3, s11, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s11
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s5, s9, s12
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX8-NEXT:    s_lshr_b32 s5, s9, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshl_b32 s3, s7, s3
 ; GFX8-NEXT:    s_lshr_b32 s4, s5, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_or_b32 s3, s3, s4
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s2, s3, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s3
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshl_v4i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s6, s4, 0xf000f
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 16

 
 define <4 x half> @v_fshl_v4i16(<4 x i16> %lhs, <4 x i16> %rhs, <4 x i16> %amt) {
 ; GFX6-LABEL: v_fshl_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v12, 15, v8
 ; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v8
 ; GFX6-NEXT:    v_and_b32_e32 v8, 15, v8
-; GFX6-NEXT:    v_bfe_u32 v12, v12, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v12, 0xffff, v12
 ; GFX6-NEXT:    v_bfe_u32 v4, v4, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v8, v8, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v8, 0xffff, v8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v12, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v8, v4
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v9
 ; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v9
 ; GFX6-NEXT:    v_and_b32_e32 v8, 15, v8
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v4, v1
 ; GFX6-NEXT:    v_bfe_u32 v4, v5, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v5, v8, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v8
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v5, v4
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v10
 ; GFX6-NEXT:    v_xor_b32_e32 v5, -1, v10
 ; GFX6-NEXT:    v_and_b32_e32 v5, 15, v5
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v4, v2
 ; GFX6-NEXT:    v_bfe_u32 v4, v6, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v5, v5, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v5
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v5, v4
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v11
 ; GFX6-NEXT:    v_xor_b32_e32 v5, -1, v11
 ; GFX6-NEXT:    v_and_b32_e32 v5, 15, v5
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_bfe_u32 v4, v7, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v5, v5, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v5
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v5, v4
 ; GFX6-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshl_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v6, 16, v4

 ; GFX8-NEXT:    v_and_b32_e32 v5, 15, v7
 ; GFX8-NEXT:    v_xor_b32_e32 v6, -1, v7
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_mov_b32_e32 v5, 1
 ; GFX8-NEXT:    v_and_b32_e32 v6, 15, v6
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v3, v5, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, v6, v3
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v3
-; GFX8-NEXT:    v_mov_b32_e32 v3, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshl_v4i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v6, 0xf000f, v4

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

 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_and_b32 s3, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_and_b32 s3, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_and_b32 s3, s2, 7
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_and_b32 s3, s2, 7
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 %amt)
   ret i8 %result
 }
 
 define i8 @v_fshr_i8(i8 %lhs, i8 %rhs, i8 %amt) {
 ; GFX6-LABEL: v_fshr_i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v2, v0
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, v2, v0
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshr_i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT:    v_xor_b32_e32 v3, -1, v2
+; GFX10-NEXT:    v_not_b32_e32 v3, v2
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX10-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX10-NEXT:    v_lshrrev_b16 v1, v2, v1
 ; GFX10-NEXT:    v_lshlrev_b16 v0, v3, v0
 ; GFX10-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshr_i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX11-NEXT:    v_xor_b32_e32 v3, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v3, v2
 ; GFX11-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX11-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_4) | instskip(NEXT) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX11-NEXT:    v_lshrrev_b16 v1, v2, v1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_lshlrev_b16 v0, v3, v0

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x40004
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i8_4:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i8_4:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 4
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i8_4:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i8_4:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 4
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 4)
   ret i8 %result
 }
 

 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 3
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x30005
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i8_5:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 3
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i8_5:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 3
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i8_5:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 3
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i8_5:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 3
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i8 @llvm.fshr.i8(i8 %lhs, i8 %rhs, i8 5)
   ret i8 %result
 }
 

 ; GFX8-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX8-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX8-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX8-NEXT:    s_and_b32 s6, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s5
 ; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s6
 ; GFX8-NEXT:    s_lshl_b32 s2, s3, s2
 ; GFX8-NEXT:    s_and_b32 s3, s4, 0xff
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    s_and_b32 s1, s5, 7
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshr_b32 s1, s3, s1
 ; GFX8-NEXT:    s_or_b32 s1, s2, s1
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX8-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX8-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_v2i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX9-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX9-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX9-NEXT:    s_and_b32 s6, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s5
 ; GFX9-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s6
 ; GFX9-NEXT:    s_lshl_b32 s2, s3, s2
 ; GFX9-NEXT:    s_and_b32 s3, s4, 0xff
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    s_and_b32 s1, s5, 7
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_lshr_b32 s1, s3, s1
 ; GFX9-NEXT:    s_or_b32 s1, s2, s1
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
-; GFX9-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX9-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX9-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX9-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_v2i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX10-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX10-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX10-NEXT:    s_and_b32 s6, s2, 7
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX10-NEXT:    s_and_b32 s2, s5, 7
 ; GFX10-NEXT:    s_andn2_b32 s5, 7, s5
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, 1
-; GFX10-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, s5
 ; GFX10-NEXT:    s_lshr_b32 s2, s4, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s6
 ; GFX10-NEXT:    s_or_b32 s2, s3, s2
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    s_and_b32 s1, s2, 0xff
-; GFX10-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX10-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX10-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX10-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_v2i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_lshr_b32 s4, s1, 8
 ; GFX11-NEXT:    s_lshr_b32 s3, s0, 8
 ; GFX11-NEXT:    s_lshr_b32 s5, s2, 8
 ; GFX11-NEXT:    s_and_b32 s6, s2, 7
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX11-NEXT:    s_and_b32 s4, s4, 0xff
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX11-NEXT:    s_and_b32 s2, s5, 7
 ; GFX11-NEXT:    s_and_not1_b32 s5, 7, s5
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, 1
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, s5
 ; GFX11-NEXT:    s_lshr_b32 s2, s4, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s6
 ; GFX11-NEXT:    s_or_b32 s2, s3, s2
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    s_and_b32 s1, s2, 0xff
-; GFX11-NEXT:    s_bfe_u32 s2, 8, 0x100000
 ; GFX11-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX11-NEXT:    s_lshl_b32 s1, s1, s2
+; GFX11-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %lhs = bitcast i16 %lhs.arg to <2 x i8>
   %rhs = bitcast i16 %rhs.arg to <2 x i8>
   %amt = bitcast i16 %amt.arg to <2 x i8>
   %result = call <2 x i8> @llvm.fshr.v2i8(<2 x i8> %lhs, <2 x i8> %rhs, <2 x i8> %amt)
   %cast.result = bitcast <2 x i8> %result to i16
   ret i16 %cast.result
 }
 
 define i16 @v_fshr_v2i8(i16 %lhs.arg, i16 %rhs.arg, i16 %amt.arg) {
 ; GFX6-LABEL: v_fshr_v2i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 8, v2
 ; GFX6-NEXT:    v_and_b32_e32 v5, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xff, v1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v5, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v4
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX6-NEXT:    v_not_b32_e32 v4, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 7, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v3
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 8, 8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 8, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_v2i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX8-NEXT:    v_and_b32_e32 v6, 7, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v2, v0
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v5
+; GFX8-NEXT:    v_not_b32_e32 v2, v5
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_and_b32_e32 v1, 7, v5
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v2, v2, v3
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v2, v1
 ; GFX8-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, 8, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v2i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX9-NEXT:    v_and_b32_e32 v6, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, v2, v0
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v5
+; GFX9-NEXT:    v_not_b32_e32 v2, v5
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_and_b32_e32 v1, 7, v5
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v2, v2, v3
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v1, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_or_b32_e32 v1, v2, v1
 ; GFX9-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v1, 8, v1
 ; GFX9-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshr_v2i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 8, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v4, 8, v0
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 8, v1
 ; GFX10-NEXT:    v_and_b32_e32 v7, 7, v2
-; GFX10-NEXT:    v_xor_b32_e32 v2, -1, v2
-; GFX10-NEXT:    v_xor_b32_e32 v6, -1, v3
+; GFX10-NEXT:    v_not_b32_e32 v2, v2
+; GFX10-NEXT:    v_not_b32_e32 v6, v3
 ; GFX10-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX10-NEXT:    v_lshlrev_b16 v4, 1, v4
 ; GFX10-NEXT:    v_and_b32_e32 v5, 0xff, v5
 ; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX10-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX10-NEXT:    v_lshrrev_b16 v3, v3, v5

 ; GFX11-LABEL: v_fshr_v2i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 8, v2
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 8, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v5, 8, v1
 ; GFX11-NEXT:    v_and_b32_e32 v7, 7, v2
-; GFX11-NEXT:    v_xor_b32_e32 v2, -1, v2
-; GFX11-NEXT:    v_xor_b32_e32 v6, -1, v3
+; GFX11-NEXT:    v_not_b32_e32 v2, v2
+; GFX11-NEXT:    v_not_b32_e32 v6, v3
 ; GFX11-NEXT:    v_and_b32_e32 v3, 7, v3
 ; GFX11-NEXT:    v_lshlrev_b16 v4, 1, v4
 ; GFX11-NEXT:    v_and_b32_e32 v5, 0xff, v5
 ; GFX11-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX11-NEXT:    v_and_b32_e32 v6, 7, v6
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX11-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX11-NEXT:    v_lshrrev_b16 v3, v3, v5

 ; GFX8-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX8-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX8-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX8-NEXT:    s_and_b32 s12, s2, 7
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_andn2_b32 s2, 7, s9
 ; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s12
 ; GFX8-NEXT:    s_lshl_b32 s2, s3, s2
 ; GFX8-NEXT:    s_and_b32 s3, s6, 0xff
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    s_and_b32 s1, s9, 7
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshr_b32 s1, s3, s1
 ; GFX8-NEXT:    s_andn2_b32 s3, 7, s10
 ; GFX8-NEXT:    s_lshl_b32 s4, s4, 1
 ; GFX8-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX8-NEXT:    s_and_b32 s4, s7, 0xff
 ; GFX8-NEXT:    s_or_b32 s1, s2, s1
 ; GFX8-NEXT:    s_and_b32 s2, s10, 7
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshr_b32 s2, s4, s2
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_or_b32 s2, s3, s2
 ; GFX8-NEXT:    s_and_b32 s3, s11, 7
 ; GFX8-NEXT:    s_andn2_b32 s4, 7, s11
 ; GFX8-NEXT:    s_lshl_b32 s5, s5, 1
 ; GFX8-NEXT:    s_and_b32 s0, s0, 0xff
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, 8

 ; GFX9-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX9-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX9-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX9-NEXT:    s_and_b32 s12, s2, 7
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_andn2_b32 s2, 7, s9
 ; GFX9-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s12
 ; GFX9-NEXT:    s_lshl_b32 s2, s3, s2
 ; GFX9-NEXT:    s_and_b32 s3, s6, 0xff
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    s_and_b32 s1, s9, 7
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_lshr_b32 s1, s3, s1
 ; GFX9-NEXT:    s_andn2_b32 s3, 7, s10
 ; GFX9-NEXT:    s_lshl_b32 s4, s4, 1
 ; GFX9-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX9-NEXT:    s_and_b32 s4, s7, 0xff
 ; GFX9-NEXT:    s_or_b32 s1, s2, s1
 ; GFX9-NEXT:    s_and_b32 s2, s10, 7
-; GFX9-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX9-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX9-NEXT:    s_lshr_b32 s2, s4, s2
 ; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_or_b32 s2, s3, s2
 ; GFX9-NEXT:    s_and_b32 s3, s11, 7
 ; GFX9-NEXT:    s_andn2_b32 s4, 7, s11
 ; GFX9-NEXT:    s_lshl_b32 s5, s5, 1
 ; GFX9-NEXT:    s_and_b32 s0, s0, 0xff
 ; GFX9-NEXT:    s_lshl_b32 s1, s1, 8

 ; GFX10-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX10-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX10-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX10-NEXT:    s_and_b32 s12, s2, 7
 ; GFX10-NEXT:    s_andn2_b32 s2, 7, s2
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX10-NEXT:    s_and_b32 s6, s6, 0xff
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX10-NEXT:    s_and_b32 s2, s9, 7
 ; GFX10-NEXT:    s_andn2_b32 s9, 7, s9
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, 1
-; GFX10-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX10-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s12
 ; GFX10-NEXT:    s_lshl_b32 s3, s3, s9
 ; GFX10-NEXT:    s_lshr_b32 s2, s6, s2
 ; GFX10-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    s_or_b32 s1, s3, s2
 ; GFX10-NEXT:    s_and_b32 s2, s10, 7
 ; GFX10-NEXT:    s_andn2_b32 s3, 7, s10
 ; GFX10-NEXT:    s_lshl_b32 s4, s4, 1
-; GFX10-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX10-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX10-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX10-NEXT:    s_lshr_b32 s2, s6, s2
 ; GFX10-NEXT:    s_andn2_b32 s4, 7, s11
 ; GFX10-NEXT:    s_lshl_b32 s5, s5, 1
 ; GFX10-NEXT:    s_and_b32 s6, s11, 7
 ; GFX10-NEXT:    s_lshl_b32 s4, s5, s4
 ; GFX10-NEXT:    s_lshr_b32 s5, s8, s6
 ; GFX10-NEXT:    s_or_b32 s2, s3, s2

 ; GFX11-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX11-NEXT:    s_lshr_b32 s10, s2, 16
 ; GFX11-NEXT:    s_lshr_b32 s11, s2, 24
 ; GFX11-NEXT:    s_and_b32 s12, s2, 7
 ; GFX11-NEXT:    s_and_not1_b32 s2, 7, s2
 ; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX11-NEXT:    s_and_b32 s6, s6, 0xff
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX11-NEXT:    s_and_b32 s2, s9, 7
 ; GFX11-NEXT:    s_and_not1_b32 s9, 7, s9
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, 1
-; GFX11-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s12
 ; GFX11-NEXT:    s_lshl_b32 s3, s3, s9
 ; GFX11-NEXT:    s_lshr_b32 s2, s6, s2
 ; GFX11-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    s_or_b32 s1, s3, s2
 ; GFX11-NEXT:    s_and_b32 s2, s10, 7
 ; GFX11-NEXT:    s_and_not1_b32 s3, 7, s10
 ; GFX11-NEXT:    s_lshl_b32 s4, s4, 1
-; GFX11-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX11-NEXT:    s_lshl_b32 s3, s4, s3
 ; GFX11-NEXT:    s_lshr_b32 s2, s6, s2
 ; GFX11-NEXT:    s_and_not1_b32 s4, 7, s11
 ; GFX11-NEXT:    s_lshl_b32 s5, s5, 1
 ; GFX11-NEXT:    s_and_b32 s6, s11, 7
 ; GFX11-NEXT:    s_lshl_b32 s4, s5, s4
 ; GFX11-NEXT:    s_lshr_b32 s5, s8, s6
 ; GFX11-NEXT:    s_or_b32 s2, s3, s2

 define i32 @v_fshr_v4i8(i32 %lhs.arg, i32 %rhs.arg, i32 %amt.arg) {
 ; GFX6-LABEL: v_fshr_v4i8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v7, 8, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v8, 16, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v9, 24, v2
 ; GFX6-NEXT:    v_and_b32_e32 v10, 7, v2
-; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX6-NEXT:    v_not_b32_e32 v2, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xff, v1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v10, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 7, v7
-; GFX6-NEXT:    v_xor_b32_e32 v7, -1, v7
+; GFX6-NEXT:    v_not_b32_e32 v7, v7
 ; GFX6-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v3
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, v7, v3
 ; GFX6-NEXT:    v_bfe_u32 v7, v1, 8, 8
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v2, v7
-; GFX6-NEXT:    v_xor_b32_e32 v7, -1, v8
+; GFX6-NEXT:    v_not_b32_e32 v7, v8
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v6, 24, v1
 ; GFX6-NEXT:    v_or_b32_e32 v2, v3, v2
 ; GFX6-NEXT:    v_and_b32_e32 v3, 7, v8
 ; GFX6-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, 1, v4
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 16, 8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, v7, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v4, v1
-; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v9
+; GFX6-NEXT:    v_not_b32_e32 v4, v9
 ; GFX6-NEXT:    v_and_b32_e32 v3, 7, v9
 ; GFX6-NEXT:    v_and_b32_e32 v4, 7, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v5, 1, v5
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, v4, v5
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v3, v6
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 8, v2

 ;
 ; GFX8-LABEL: v_fshr_v4i8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v6, 16, v2
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v7, 24, v2
 ; GFX8-NEXT:    v_and_b32_e32 v8, 7, v2
-; GFX8-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX8-NEXT:    v_not_b32_e32 v2, v2
 ; GFX8-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v9, 1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v2, v2, v9
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v8, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX8-NEXT:    v_or_b32_e32 v2, v2, v8
 ; GFX8-NEXT:    v_and_b32_e32 v8, 7, v5
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX8-NEXT:    v_not_b32_e32 v5, v5
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, v5, v3
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v4, v8, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX8-NEXT:    v_and_b32_e32 v4, 7, v6
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v6
+; GFX8-NEXT:    v_not_b32_e32 v5, v6
 ; GFX8-NEXT:    v_mov_b32_e32 v6, 1
-; GFX8-NEXT:    v_mov_b32_e32 v9, 0xff
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v8, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v5, v5, v8
-; GFX8-NEXT:    v_and_b32_sdwa v8, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX8-NEXT:    v_mov_b32_e32 v8, 0xff
+; GFX8-NEXT:    v_and_b32_sdwa v8, v1, v8 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v4, v4, v8
 ; GFX8-NEXT:    v_or_b32_e32 v4, v5, v4
 ; GFX8-NEXT:    v_and_b32_e32 v5, 7, v7
-; GFX8-NEXT:    v_xor_b32_e32 v7, -1, v7
+; GFX8-NEXT:    v_not_b32_e32 v7, v7
 ; GFX8-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v7, v0
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 8
 ; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX8-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:BYTE_0 src1_sel:DWORD
 ; GFX8-NEXT:    v_and_b32_e32 v2, 0xff, v4
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX8-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 24, v0
 ; GFX8-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v4i8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v6, 16, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v7, 24, v2
 ; GFX9-NEXT:    v_and_b32_e32 v8, 7, v2
-; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
+; GFX9-NEXT:    v_not_b32_e32 v2, v2
 ; GFX9-NEXT:    v_and_b32_e32 v2, 7, v2
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v9, 1, v0
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v2, v2, v9
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v8, v8, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX9-NEXT:    v_or_b32_e32 v2, v2, v8
 ; GFX9-NEXT:    v_and_b32_e32 v8, 7, v5
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v5
+; GFX9-NEXT:    v_not_b32_e32 v5, v5
 ; GFX9-NEXT:    v_lshrrev_b32_e32 v4, 8, v1
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v3, v5, v3
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v4, v8, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX9-NEXT:    v_and_b32_e32 v4, 7, v6
-; GFX9-NEXT:    v_xor_b32_e32 v5, -1, v6
+; GFX9-NEXT:    v_not_b32_e32 v5, v6
 ; GFX9-NEXT:    v_mov_b32_e32 v6, 1
-; GFX9-NEXT:    v_mov_b32_e32 v9, 0xff
+; GFX9-NEXT:    s_movk_i32 s4, 0xff
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v5
 ; GFX9-NEXT:    v_lshlrev_b16_sdwa v8, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v5, v5, v8
-; GFX9-NEXT:    v_and_b32_sdwa v8, v1, v9 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
+; GFX9-NEXT:    v_and_b32_sdwa v8, v1, s4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v4, v4, v8
 ; GFX9-NEXT:    v_or_b32_e32 v4, v5, v4
 ; GFX9-NEXT:    v_and_b32_e32 v5, 7, v7
-; GFX9-NEXT:    v_xor_b32_e32 v7, -1, v7
+; GFX9-NEXT:    v_not_b32_e32 v7, v7
 ; GFX9-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX9-NEXT:    v_lshlrev_b16_sdwa v0, v6, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, v7, v0
 ; GFX9-NEXT:    v_lshrrev_b16_sdwa v1, v5, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_3
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 8
-; GFX9-NEXT:    s_movk_i32 s4, 0xff
 ; GFX9-NEXT:    v_lshlrev_b32_sdwa v1, v1, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX9-NEXT:    v_and_or_b32 v1, v2, s4, v1
 ; GFX9-NEXT:    v_and_b32_e32 v2, 0xff, v4
 ; GFX9-NEXT:    v_and_b32_e32 v0, 0xff, v0
 ; GFX9-NEXT:    v_lshlrev_b32_e32 v2, 16, v2
 ; GFX9-NEXT:    v_lshlrev_b32_e32 v0, 24, v0
 ; GFX9-NEXT:    v_or3_b32 v0, v1, v2, v0
 ; GFX9-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_fshr_v4i8:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX10-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 8, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 8, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
-; GFX10-NEXT:    v_xor_b32_e32 v8, -1, v2
-; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 16, v2
-; GFX10-NEXT:    v_lshrrev_b32_e32 v12, 24, v2
-; GFX10-NEXT:    v_xor_b32_e32 v10, -1, v5
+; GFX10-NEXT:    v_not_b32_e32 v8, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v10, 16, v2
+; GFX10-NEXT:    v_lshrrev_b32_e32 v11, 24, v2
+; GFX10-NEXT:    v_not_b32_e32 v12, v7
 ; GFX10-NEXT:    v_lshlrev_b16 v3, 1, v3
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
-; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 24, v0
-; GFX10-NEXT:    v_lshrrev_b32_e32 v7, 8, v1
-; GFX10-NEXT:    v_and_b32_e32 v10, 7, v10
-; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
+; GFX10-NEXT:    v_lshrrev_b32_e32 v6, 8, v1
+; GFX10-NEXT:    v_and_b32_e32 v12, 7, v12
 ; GFX10-NEXT:    v_and_b32_e32 v8, 7, v8
-; GFX10-NEXT:    v_mov_b32_e32 v13, 0xff
-; GFX10-NEXT:    v_xor_b32_e32 v14, -1, v12
-; GFX10-NEXT:    v_lshlrev_b16 v3, v10, v3
-; GFX10-NEXT:    v_xor_b32_e32 v10, -1, v11
+; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
+; GFX10-NEXT:    v_not_b32_e32 v13, v10
+; GFX10-NEXT:    s_movk_i32 s4, 0xff
+; GFX10-NEXT:    v_lshlrev_b16 v3, v12, v3
+; GFX10-NEXT:    v_not_b32_e32 v12, v11
 ; GFX10-NEXT:    v_lshrrev_b32_e32 v9, 24, v1
 ; GFX10-NEXT:    v_lshlrev_b16 v0, v8, v0
 ; GFX10-NEXT:    v_and_b32_e32 v8, 0xff, v1
-; GFX10-NEXT:    v_and_b32_e32 v5, 7, v5
-; GFX10-NEXT:    v_and_b32_e32 v7, 0xff, v7
-; GFX10-NEXT:    v_and_b32_e32 v11, 7, v11
+; GFX10-NEXT:    v_and_b32_e32 v7, 7, v7
+; GFX10-NEXT:    v_and_b32_e32 v6, 0xff, v6
 ; GFX10-NEXT:    v_and_b32_e32 v10, 7, v10
+; GFX10-NEXT:    v_and_b32_e32 v13, 7, v13
 ; GFX10-NEXT:    v_lshlrev_b16 v4, 1, v4
-; GFX10-NEXT:    v_and_b32_sdwa v1, v1, v13 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
-; GFX10-NEXT:    v_and_b32_e32 v13, 7, v14
-; GFX10-NEXT:    v_lshlrev_b16 v6, 1, v6
+; GFX10-NEXT:    v_and_b32_sdwa v1, v1, s4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_1 src1_sel:DWORD
 ; GFX10-NEXT:    v_and_b32_e32 v12, 7, v12
+; GFX10-NEXT:    v_lshlrev_b16 v5, 1, v5
+; GFX10-NEXT:    v_and_b32_e32 v11, 7, v11
 ; GFX10-NEXT:    v_and_b32_e32 v2, 7, v2
-; GFX10-NEXT:    v_lshrrev_b16 v5, v5, v7
-; GFX10-NEXT:    v_lshlrev_b16 v4, v10, v4
-; GFX10-NEXT:    v_lshrrev_b16 v1, v11, v1
-; GFX10-NEXT:    v_lshlrev_b16 v6, v13, v6
-; GFX10-NEXT:    v_lshrrev_b16 v7, v12, v9
+; GFX10-NEXT:    v_lshrrev_b16 v6, v7, v6
+; GFX10-NEXT:    v_lshlrev_b16 v4, v13, v4
+; GFX10-NEXT:    v_lshrrev_b16 v1, v10, v1
+; GFX10-NEXT:    v_lshlrev_b16 v5, v12, v5
+; GFX10-NEXT:    v_lshrrev_b16 v7, v11, v9
 ; GFX10-NEXT:    v_lshrrev_b16 v2, v2, v8
-; GFX10-NEXT:    v_or_b32_e32 v3, v3, v5
-; GFX10-NEXT:    v_mov_b32_e32 v5, 8
+; GFX10-NEXT:    v_or_b32_e32 v3, v3, v6
+; GFX10-NEXT:    v_mov_b32_e32 v6, 8
 ; GFX10-NEXT:    v_or_b32_e32 v1, v4, v1
-; GFX10-NEXT:    v_or_b32_e32 v4, v6, v7
+; GFX10-NEXT:    v_or_b32_e32 v4, v5, v7
 ; GFX10-NEXT:    v_or_b32_e32 v0, v0, v2
-; GFX10-NEXT:    v_lshlrev_b32_sdwa v2, v5, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
+; GFX10-NEXT:    v_lshlrev_b32_sdwa v2, v6, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_0
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xff, v1
 ; GFX10-NEXT:    v_and_b32_e32 v3, 0xff, v4
 ; GFX10-NEXT:    v_and_or_b32 v0, v0, 0xff, v2
 ; GFX10-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX10-NEXT:    v_lshlrev_b32_e32 v2, 24, v3
 ; GFX10-NEXT:    v_or3_b32 v0, v0, v1, v2
 ; GFX10-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX11-LABEL: v_fshr_v4i8:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v6, 8, v1
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v7, 8, v2
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v3, 8, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v11, 16, v2
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v13, 24, v2
 ; GFX11-NEXT:    v_and_b32_e32 v6, 0xff, v6
-; GFX11-NEXT:    v_xor_b32_e32 v12, -1, v7
+; GFX11-NEXT:    v_not_b32_e32 v12, v7
 ; GFX11-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v4, 16, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v5, 24, v0
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v8, 16, v1
 ; GFX11-NEXT:    v_and_b32_e32 v12, 7, v12
 ; GFX11-NEXT:    v_lshlrev_b16 v3, 1, v3
-; GFX11-NEXT:    v_xor_b32_e32 v14, -1, v11
+; GFX11-NEXT:    v_not_b32_e32 v14, v11
 ; GFX11-NEXT:    v_lshrrev_b16 v6, v7, v6
-; GFX11-NEXT:    v_xor_b32_e32 v7, -1, v13
+; GFX11-NEXT:    v_not_b32_e32 v7, v13
 ; GFX11-NEXT:    v_lshrrev_b32_e32 v9, 24, v1
-; GFX11-NEXT:    v_xor_b32_e32 v10, -1, v2
+; GFX11-NEXT:    v_not_b32_e32 v10, v2
 ; GFX11-NEXT:    v_lshlrev_b16 v3, v12, v3
 ; GFX11-NEXT:    v_and_b32_e32 v11, 7, v11
 ; GFX11-NEXT:    v_and_b32_e32 v12, 7, v14
 ; GFX11-NEXT:    v_lshlrev_b16 v4, 1, v4
 ; GFX11-NEXT:    v_and_b32_e32 v8, 0xff, v8
 ; GFX11-NEXT:    v_and_b32_e32 v7, 7, v7
 ; GFX11-NEXT:    v_lshlrev_b16 v5, 1, v5
 ; GFX11-NEXT:    v_and_b32_e32 v13, 7, v13

 ; GFX6-NEXT:    s_lshr_b32 s8, s2, 16
 ; GFX6-NEXT:    s_lshr_b32 s9, s2, 24
 ; GFX6-NEXT:    s_and_b32 s11, s2, 0xff
 ; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x80008
 ; GFX6-NEXT:    v_mul_lo_u32 v2, v0, v1
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 8
 ; GFX6-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX6-NEXT:    s_or_b32 s2, s11, s2
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX6-NEXT:    s_lshr_b32 s10, s3, 8
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX6-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX6-NEXT:    s_or_b32 s2, s2, s8
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 8
 ; GFX6-NEXT:    s_and_b32 s8, s10, 0xff
 ; GFX6-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX6-NEXT:    s_or_b32 s3, s9, s3
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX6-NEXT:    s_or_b32 s3, s3, s8
 ; GFX6-NEXT:    s_lshr_b32 s8, s4, 16
 ; GFX6-NEXT:    s_lshr_b32 s9, s4, 24
 ; GFX6-NEXT:    s_and_b32 s11, s4, 0xff
 ; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x80008
 ; GFX6-NEXT:    s_lshl_b32 s4, s4, 8
 ; GFX6-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX6-NEXT:    v_add_i32_e32 v0, vcc, v0, v2
 ; GFX6-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
 ; GFX6-NEXT:    s_or_b32 s4, s11, s4
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX6-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX6-NEXT:    s_or_b32 s4, s4, s8
 ; GFX6-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX6-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX6-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX6-NEXT:    s_lshr_b32 s10, s5, 8
 ; GFX6-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX6-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX6-NEXT:    v_mul_lo_u32 v1, v2, v1
 ; GFX6-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX6-NEXT:    v_sub_i32_e32 v0, vcc, s4, v0
 ; GFX6-NEXT:    v_subrev_i32_e32 v3, vcc, 24, v0
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX6-NEXT:    v_mul_hi_u32 v1, v2, v1
 ; GFX6-NEXT:    s_and_b32 s8, s10, 0xff
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX6-NEXT:    s_or_b32 s5, s9, s5
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX6-NEXT:    v_subrev_i32_e32 v3, vcc, 24, v0
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX6-NEXT:    s_or_b32 s5, s5, s8
 ; GFX6-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX6-NEXT:    v_add_i32_e32 v1, vcc, v2, v1
 ; GFX6-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX6-NEXT:    s_and_b32 s6, s6, 0xff
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    v_mul_lo_u32 v1, v1, 24
 ; GFX6-NEXT:    v_sub_i32_e32 v3, vcc, 23, v0
 ; GFX6-NEXT:    s_lshl_b32 s4, s6, 17
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX6-NEXT:    s_or_b32 s0, s4, s0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xffffff, v3
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s0, v2
 ; GFX6-NEXT:    v_lshr_b32_e32 v0, s2, v0
 ; GFX6-NEXT:    v_sub_i32_e32 v1, vcc, s5, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, 24, v1
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
 ; GFX6-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX6-NEXT:    v_subrev_i32_e32 v2, vcc, 24, v1
 ; GFX6-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX6-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX6-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX6-NEXT:    v_sub_i32_e32 v2, vcc, 23, v1
 ; GFX6-NEXT:    s_lshl_b32 s0, s7, 17
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s0, v2

 ; GFX6-NEXT:    v_readfirstlane_b32 s1, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_v2i24:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX8-NEXT:    s_lshr_b32 s9, s1, 8
-; GFX8-NEXT:    s_bfe_u32 s10, 8, 0x100000
 ; GFX8-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX8-NEXT:    s_lshr_b32 s6, s0, 8
 ; GFX8-NEXT:    s_lshr_b32 s8, s0, 24
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s10
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX8-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX8-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX8-NEXT:    s_or_b32 s1, s8, s1
 ; GFX8-NEXT:    s_lshr_b32 s8, s2, 8
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX8-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX8-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s10
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX8-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX8-NEXT:    s_or_b32 s0, s0, s6
 ; GFX8-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX8-NEXT:    s_and_b32 s7, s9, 0xff
 ; GFX8-NEXT:    s_lshr_b32 s9, s2, 16
-; GFX8-NEXT:    s_lshr_b32 s11, s2, 24
+; GFX8-NEXT:    s_lshr_b32 s10, s2, 24
 ; GFX8-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX8-NEXT:    s_lshl_b32 s8, s8, s10
+; GFX8-NEXT:    s_lshl_b32 s8, s8, 8
 ; GFX8-NEXT:    s_or_b32 s2, s2, s8
 ; GFX8-NEXT:    s_and_b32 s8, s9, 0xff
 ; GFX8-NEXT:    v_mov_b32_e32 v1, 0xffffffe8
-; GFX8-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX8-NEXT:    v_mul_lo_u32 v2, v0, v1
-; GFX8-NEXT:    s_lshr_b32 s12, s3, 8
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_lshr_b32 s11, s3, 8
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX8-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX8-NEXT:    s_or_b32 s2, s2, s8
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s10
-; GFX8-NEXT:    s_and_b32 s8, s12, 0xff
-; GFX8-NEXT:    s_or_b32 s3, s11, s3
-; GFX8-NEXT:    s_bfe_u32 s8, s8, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 8
+; GFX8-NEXT:    s_and_b32 s8, s11, 0xff
+; GFX8-NEXT:    s_or_b32 s3, s10, s3
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s8
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX8-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX8-NEXT:    s_or_b32 s3, s3, s8
 ; GFX8-NEXT:    s_lshr_b32 s8, s4, 8
 ; GFX8-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX8-NEXT:    s_lshr_b32 s9, s4, 16
-; GFX8-NEXT:    s_lshr_b32 s11, s4, 24
+; GFX8-NEXT:    s_lshr_b32 s10, s4, 24
 ; GFX8-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX8-NEXT:    s_lshl_b32 s8, s8, s10
+; GFX8-NEXT:    s_lshl_b32 s8, s8, 8
 ; GFX8-NEXT:    s_or_b32 s4, s4, s8
 ; GFX8-NEXT:    s_and_b32 s8, s9, 0xff
 ; GFX8-NEXT:    v_add_u32_e32 v0, vcc, v0, v2
 ; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
-; GFX8-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX8-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX8-NEXT:    s_or_b32 s4, s4, s8
 ; GFX8-NEXT:    v_mul_hi_u32 v0, s4, v0
 ; GFX8-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX8-NEXT:    v_cvt_u32_f32_e32 v2, v2
-; GFX8-NEXT:    s_lshr_b32 s12, s5, 8
+; GFX8-NEXT:    s_lshr_b32 s11, s5, 8
 ; GFX8-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX8-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX8-NEXT:    v_mul_lo_u32 v1, v2, v1
-; GFX8-NEXT:    s_lshl_b32 s5, s5, s10
+; GFX8-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX8-NEXT:    v_sub_u32_e32 v0, vcc, s4, v0
 ; GFX8-NEXT:    v_subrev_u32_e32 v3, vcc, 24, v0
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX8-NEXT:    v_mul_hi_u32 v1, v2, v1
-; GFX8-NEXT:    s_and_b32 s8, s12, 0xff
+; GFX8-NEXT:    s_and_b32 s8, s11, 0xff
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
-; GFX8-NEXT:    s_or_b32 s5, s11, s5
-; GFX8-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX8-NEXT:    s_or_b32 s5, s10, s5
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX8-NEXT:    v_subrev_u32_e32 v3, vcc, 24, v0
-; GFX8-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX8-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX8-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX8-NEXT:    s_or_b32 s5, s5, s8
 ; GFX8-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX8-NEXT:    v_add_u32_e32 v1, vcc, v2, v1
 ; GFX8-NEXT:    v_mul_hi_u32 v1, s5, v1
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX8-NEXT:    v_sub_u32_e32 v3, vcc, 23, v0
 ; GFX8-NEXT:    v_mul_lo_u32 v1, v1, 24
 ; GFX8-NEXT:    s_lshl_b32 s4, s6, 17
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    s_or_b32 s0, s4, s0
 ; GFX8-NEXT:    v_and_b32_e32 v2, 0xffffff, v3
 ; GFX8-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
 ; GFX8-NEXT:    v_lshlrev_b32_e64 v2, v2, s0
 ; GFX8-NEXT:    v_lshrrev_b32_e64 v0, v0, s2
 ; GFX8-NEXT:    v_sub_u32_e32 v1, vcc, s5, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    v_subrev_u32_e32 v2, vcc, 24, v1
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX8-NEXT:    v_subrev_u32_e32 v2, vcc, 24, v1
 ; GFX8-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX8-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX8-NEXT:    v_sub_u32_e32 v2, vcc, 23, v1
 ; GFX8-NEXT:    s_lshl_b32 s0, s7, 17
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GFX8-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX8-NEXT:    v_lshlrev_b32_e64 v2, v2, s0

 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_v2i24:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX9-NEXT:    v_mov_b32_e32 v1, 0xffffffe8
 ; GFX9-NEXT:    s_lshr_b32 s11, s1, 8
-; GFX9-NEXT:    s_bfe_u32 s12, 8, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v0, v0
-; GFX9-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 8
 ; GFX9-NEXT:    s_lshr_b32 s10, s0, 24
+; GFX9-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX9-NEXT:    v_mul_lo_u32 v2, v0, v1
-; GFX9-NEXT:    s_lshl_b32 s1, s1, s12
 ; GFX9-NEXT:    s_and_b32 s7, s7, 0xff
 ; GFX9-NEXT:    s_or_b32 s1, s10, s1
-; GFX9-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX9-NEXT:    s_lshr_b32 s10, s2, 8
+; GFX9-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX9-NEXT:    s_lshr_b32 s9, s0, 16
 ; GFX9-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX9-NEXT:    s_lshl_b32 s7, s7, s12
+; GFX9-NEXT:    s_lshl_b32 s7, s7, 8
 ; GFX9-NEXT:    s_and_b32 s10, s10, 0xff
 ; GFX9-NEXT:    s_or_b32 s0, s0, s7
 ; GFX9-NEXT:    s_and_b32 s7, s9, 0xff
 ; GFX9-NEXT:    s_and_b32 s9, s11, 0xff
 ; GFX9-NEXT:    s_lshr_b32 s11, s2, 16
-; GFX9-NEXT:    s_lshr_b32 s13, s2, 24
+; GFX9-NEXT:    s_lshr_b32 s12, s2, 24
 ; GFX9-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX9-NEXT:    s_lshl_b32 s10, s10, s12
+; GFX9-NEXT:    s_lshl_b32 s10, s10, 8
 ; GFX9-NEXT:    s_or_b32 s2, s2, s10
 ; GFX9-NEXT:    s_and_b32 s10, s11, 0xff
 ; GFX9-NEXT:    v_add_u32_e32 v0, v0, v2
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v2, 24
-; GFX9-NEXT:    s_bfe_u32 s10, s10, 0x100000
+; GFX9-NEXT:    s_and_b32 s10, 0xffff, s10
 ; GFX9-NEXT:    v_rcp_iflag_f32_e32 v2, v2
-; GFX9-NEXT:    s_lshr_b32 s14, s3, 8
-; GFX9-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX9-NEXT:    s_lshr_b32 s13, s3, 8
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX9-NEXT:    s_lshl_b32 s10, s10, 16
 ; GFX9-NEXT:    s_and_b32 s3, s3, 0xff
 ; GFX9-NEXT:    s_or_b32 s2, s2, s10
-; GFX9-NEXT:    s_lshl_b32 s3, s3, s12
-; GFX9-NEXT:    s_and_b32 s10, s14, 0xff
-; GFX9-NEXT:    s_or_b32 s3, s13, s3
-; GFX9-NEXT:    s_bfe_u32 s10, s10, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX9-NEXT:    s_lshl_b32 s3, s3, 8
+; GFX9-NEXT:    s_and_b32 s10, s13, 0xff
+; GFX9-NEXT:    s_or_b32 s3, s12, s3
+; GFX9-NEXT:    s_and_b32 s10, 0xffff, s10
+; GFX9-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX9-NEXT:    s_lshl_b32 s10, s10, 16
 ; GFX9-NEXT:    v_mul_f32_e32 v2, 0x4f7ffffe, v2
 ; GFX9-NEXT:    s_or_b32 s3, s3, s10
 ; GFX9-NEXT:    s_lshr_b32 s10, s4, 8
 ; GFX9-NEXT:    v_cvt_u32_f32_e32 v2, v2
 ; GFX9-NEXT:    s_and_b32 s10, s10, 0xff
 ; GFX9-NEXT:    s_lshr_b32 s11, s4, 16
-; GFX9-NEXT:    s_lshr_b32 s13, s4, 24
+; GFX9-NEXT:    s_lshr_b32 s12, s4, 24
 ; GFX9-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX9-NEXT:    s_lshl_b32 s10, s10, s12
+; GFX9-NEXT:    s_lshl_b32 s10, s10, 8
 ; GFX9-NEXT:    s_or_b32 s4, s4, s10
 ; GFX9-NEXT:    s_and_b32 s10, s11, 0xff
-; GFX9-NEXT:    s_bfe_u32 s10, s10, 0x100000
+; GFX9-NEXT:    s_and_b32 s10, 0xffff, s10
 ; GFX9-NEXT:    v_mul_lo_u32 v1, v2, v1
-; GFX9-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX9-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX9-NEXT:    s_lshl_b32 s10, s10, 16
 ; GFX9-NEXT:    s_or_b32 s4, s4, s10
 ; GFX9-NEXT:    v_mul_hi_u32 v0, s4, v0
-; GFX9-NEXT:    s_lshr_b32 s14, s5, 8
+; GFX9-NEXT:    s_lshr_b32 s13, s5, 8
 ; GFX9-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX9-NEXT:    v_mul_hi_u32 v1, v2, v1
-; GFX9-NEXT:    s_lshl_b32 s5, s5, s12
-; GFX9-NEXT:    s_and_b32 s10, s14, 0xff
-; GFX9-NEXT:    s_or_b32 s5, s13, s5
-; GFX9-NEXT:    s_bfe_u32 s10, s10, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX9-NEXT:    s_lshl_b32 s5, s5, 8
+; GFX9-NEXT:    s_and_b32 s10, s13, 0xff
+; GFX9-NEXT:    s_or_b32 s5, s12, s5
+; GFX9-NEXT:    s_and_b32 s10, 0xffff, s10
+; GFX9-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX9-NEXT:    s_lshl_b32 s10, s10, 16
 ; GFX9-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX9-NEXT:    s_or_b32 s5, s5, s10
 ; GFX9-NEXT:    v_add_u32_e32 v1, v2, v1
 ; GFX9-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX9-NEXT:    v_sub_u32_e32 v0, s4, v0
 ; GFX9-NEXT:    v_subrev_u32_e32 v3, 24, v0
 ; GFX9-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
 ; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX9-NEXT:    v_mul_lo_u32 v1, v1, 24
 ; GFX9-NEXT:    v_subrev_u32_e32 v3, 24, v0
 ; GFX9-NEXT:    v_cmp_le_u32_e32 vcc, 24, v0
-; GFX9-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX9-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX9-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
 ; GFX9-NEXT:    v_sub_u32_e32 v3, 23, v0
 ; GFX9-NEXT:    s_lshl_b32 s4, s7, 17
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX9-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
 ; GFX9-NEXT:    s_or_b32 s0, s4, s0
 ; GFX9-NEXT:    v_and_b32_e32 v2, 0xffffff, v3
 ; GFX9-NEXT:    v_lshrrev_b32_e64 v0, v0, s2
 ; GFX9-NEXT:    v_sub_u32_e32 v1, s5, v1
 ; GFX9-NEXT:    v_lshl_or_b32 v0, s0, v2, v0
 ; GFX9-NEXT:    v_subrev_u32_e32 v2, 24, v1
 ; GFX9-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX9-NEXT:    v_subrev_u32_e32 v2, 24, v1
 ; GFX9-NEXT:    v_cmp_le_u32_e32 vcc, 24, v1
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s9, s9, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_and_b32 s9, 0xffff, s9
 ; GFX9-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc
 ; GFX9-NEXT:    v_sub_u32_e32 v2, 23, v1
 ; GFX9-NEXT:    s_lshl_b32 s0, s9, 17
 ; GFX9-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX9-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GFX9-NEXT:    v_lshrrev_b32_e64 v1, v1, s3

 ; GFX9-NEXT:    v_readfirstlane_b32 s1, v1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_v2i24:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX10-NEXT:    v_cvt_f32_ubyte0_e32 v1, 24
 ; GFX10-NEXT:    s_lshr_b32 s9, s1, 8
-; GFX10-NEXT:    s_bfe_u32 s10, 8, 0x100000
 ; GFX10-NEXT:    s_and_b32 s1, s1, 0xff
+; GFX10-NEXT:    s_lshr_b32 s6, s0, 8
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX10-NEXT:    v_rcp_iflag_f32_e32 v1, v1
-; GFX10-NEXT:    s_lshr_b32 s6, s0, 8
 ; GFX10-NEXT:    s_lshr_b32 s8, s0, 24
-; GFX10-NEXT:    s_lshl_b32 s1, s1, s10
+; GFX10-NEXT:    s_lshl_b32 s1, s1, 8
 ; GFX10-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX10-NEXT:    s_or_b32 s1, s8, s1
 ; GFX10-NEXT:    s_lshr_b32 s8, s4, 8
 ; GFX10-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX10-NEXT:    s_and_b32 s0, s0, 0xff
+; GFX10-NEXT:    s_lshl_b32 s6, s6, 8
 ; GFX10-NEXT:    v_mul_f32_e32 v0, 0x4f7ffffe, v0
 ; GFX10-NEXT:    v_mul_f32_e32 v1, 0x4f7ffffe, v1
-; GFX10-NEXT:    s_lshl_b32 s6, s6, s10
 ; GFX10-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX10-NEXT:    s_or_b32 s0, s0, s6
+; GFX10-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX10-NEXT:    v_cvt_u32_f32_e32 v1, v1
-; GFX10-NEXT:    s_and_b32 s6, s7, 0xff
 ; GFX10-NEXT:    s_and_b32 s7, s9, 0xff
 ; GFX10-NEXT:    s_lshr_b32 s9, s4, 16
+; GFX10-NEXT:    s_lshr_b32 s10, s4, 24
 ; GFX10-NEXT:    v_mul_lo_u32 v2, 0xffffffe8, v0
 ; GFX10-NEXT:    v_mul_lo_u32 v3, 0xffffffe8, v1
-; GFX10-NEXT:    s_lshr_b32 s11, s4, 24
 ; GFX10-NEXT:    s_and_b32 s4, s4, 0xff
-; GFX10-NEXT:    s_lshl_b32 s8, s8, s10
-; GFX10-NEXT:    s_lshr_b32 s12, s5, 8
+; GFX10-NEXT:    s_lshl_b32 s8, s8, 8
+; GFX10-NEXT:    s_lshr_b32 s11, s5, 8
 ; GFX10-NEXT:    s_or_b32 s4, s4, s8
 ; GFX10-NEXT:    s_and_b32 s8, s9, 0xff
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX10-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX10-NEXT:    v_mul_hi_u32 v3, v1, v3
-; GFX10-NEXT:    s_bfe_u32 s8, s8, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX10-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX10-NEXT:    s_and_b32 s5, s5, 0xff
 ; GFX10-NEXT:    s_lshl_b32 s8, s8, 16
-; GFX10-NEXT:    s_lshl_b32 s5, s5, s10
+; GFX10-NEXT:    s_lshl_b32 s5, s5, 8
 ; GFX10-NEXT:    s_or_b32 s4, s4, s8
+; GFX10-NEXT:    s_and_b32 s8, s11, 0xff
 ; GFX10-NEXT:    v_add_nc_u32_e32 v0, v0, v2
-; GFX10-NEXT:    s_and_b32 s8, s12, 0xff
-; GFX10-NEXT:    s_or_b32 s5, s11, s5
-; GFX10-NEXT:    s_bfe_u32 s8, s8, 0x100000
+; GFX10-NEXT:    s_or_b32 s5, s10, s5
+; GFX10-NEXT:    s_and_b32 s8, 0xffff, s8
 ; GFX10-NEXT:    v_add_nc_u32_e32 v1, v1, v3
+; GFX10-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX10-NEXT:    v_mul_hi_u32 v0, s4, v0
-; GFX10-NEXT:    s_bfe_u32 s5, s5, 0x100000
 ; GFX10-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX10-NEXT:    s_lshr_b32 s9, s2, 8
 ; GFX10-NEXT:    s_or_b32 s5, s5, s8
 ; GFX10-NEXT:    s_lshr_b32 s8, s2, 16
 ; GFX10-NEXT:    v_mul_hi_u32 v1, s5, v1
 ; GFX10-NEXT:    s_and_b32 s9, s9, 0xff
+; GFX10-NEXT:    s_lshr_b32 s10, s2, 24
 ; GFX10-NEXT:    v_mul_lo_u32 v0, v0, 24
-; GFX10-NEXT:    s_lshr_b32 s11, s2, 24
-; GFX10-NEXT:    s_lshr_b32 s12, s3, 8
+; GFX10-NEXT:    s_lshr_b32 s11, s3, 8
 ; GFX10-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX10-NEXT:    s_lshl_b32 s9, s9, s10
+; GFX10-NEXT:    s_lshl_b32 s9, s9, 8
 ; GFX10-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX10-NEXT:    v_mul_lo_u32 v1, v1, 24
 ; GFX10-NEXT:    s_and_b32 s3, s3, 0xff
-; GFX10-NEXT:    v_sub_nc_u32_e32 v0, s4, v0
 ; GFX10-NEXT:    s_or_b32 s2, s2, s9
-; GFX10-NEXT:    s_bfe_u32 s4, s8, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s3, s3, s10
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX10-NEXT:    v_sub_nc_u32_e32 v0, s4, v0
+; GFX10-NEXT:    s_and_b32 s4, 0xffff, s8
+; GFX10-NEXT:    s_lshl_b32 s3, s3, 8
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX10-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v1, s5, v1
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX10-NEXT:    s_and_b32 s5, s12, 0xff
-; GFX10-NEXT:    s_lshl_b32 s4, s4, 16
-; GFX10-NEXT:    s_or_b32 s3, s11, s3
-; GFX10-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX10-NEXT:    s_and_b32 s5, s11, 0xff
+; GFX10-NEXT:    s_or_b32 s3, s10, s3
+; GFX10-NEXT:    s_and_b32 s5, 0xffff, s5
+; GFX10-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX10-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v2, 24, v1
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
-; GFX10-NEXT:    s_bfe_u32 s3, s3, 0x100000
 ; GFX10-NEXT:    s_lshl_b32 s5, s5, 16
-; GFX10-NEXT:    v_subrev_nc_u32_e32 v3, 24, v0
 ; GFX10-NEXT:    s_or_b32 s2, s2, s4
+; GFX10-NEXT:    v_subrev_nc_u32_e32 v3, 24, v0
+; GFX10-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX10-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc_lo
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX10-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX10-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX10-NEXT:    s_or_b32 s3, s3, s5
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    v_subrev_nc_u32_e32 v2, 24, v1
 ; GFX10-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc_lo
 ; GFX10-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX10-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX10-NEXT:    s_lshl_b32 s4, s6, 17
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v3, 23, v0
 ; GFX10-NEXT:    v_cndmask_b32_e32 v1, v1, v2, vcc_lo
 ; GFX10-NEXT:    v_and_b32_e32 v0, 0xffffff, v0
-; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX10-NEXT:    s_or_b32 s0, s4, s0
 ; GFX10-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX10-NEXT:    v_and_b32_e32 v3, 0xffffff, v3
 ; GFX10-NEXT:    v_sub_nc_u32_e32 v2, 23, v1
 ; GFX10-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX10-NEXT:    v_lshrrev_b32_e64 v0, v0, s2
-; GFX10-NEXT:    s_or_b32 s0, s4, s0
 ; GFX10-NEXT:    s_lshl_b32 s2, s7, 17
 ; GFX10-NEXT:    v_and_b32_e32 v2, 0xffffff, v2
 ; GFX10-NEXT:    v_lshrrev_b32_e64 v1, v1, s3
 ; GFX10-NEXT:    v_lshl_or_b32 v0, s0, v3, v0
 ; GFX10-NEXT:    s_or_b32 s0, s2, s1
 ; GFX10-NEXT:    v_lshl_or_b32 v1, s0, v2, v1
 ; GFX10-NEXT:    s_mov_b32 s0, 8
 ; GFX10-NEXT:    v_lshlrev_b32_sdwa v2, s0, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:BYTE_1

 ; GFX10-NEXT:    v_readfirstlane_b32 s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_v2i24:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v0, 24
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v1, 24
 ; GFX11-NEXT:    s_lshr_b32 s6, s0, 8
-; GFX11-NEXT:    s_bfe_u32 s9, 8, 0x100000
+; GFX11-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX11-NEXT:    s_and_b32 s6, s6, 0xff
 ; GFX11-NEXT:    v_rcp_iflag_f32_e32 v0, v0
 ; GFX11-NEXT:    v_rcp_iflag_f32_e32 v1, v1
-; GFX11-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX11-NEXT:    s_lshr_b32 s8, s0, 24
 ; GFX11-NEXT:    s_and_b32 s0, s0, 0xff
-; GFX11-NEXT:    s_lshl_b32 s6, s6, s9
-; GFX11-NEXT:    s_lshr_b32 s10, s1, 8
+; GFX11-NEXT:    s_lshl_b32 s6, s6, 8
+; GFX11-NEXT:    s_lshr_b32 s9, s1, 8
 ; GFX11-NEXT:    s_or_b32 s0, s0, s6
 ; GFX11-NEXT:    s_and_b32 s6, s7, 0xff
-; GFX11-NEXT:    s_and_b32 s7, s10, 0xff
+; GFX11-NEXT:    s_and_b32 s7, s9, 0xff
+; GFX11-NEXT:    s_lshr_b32 s9, s4, 8
 ; GFX11-NEXT:    s_waitcnt_depctr 0xfff
 ; GFX11-NEXT:    v_dual_mul_f32 v0, 0x4f7ffffe, v0 :: v_dual_mul_f32 v1, 0x4f7ffffe, v1
-; GFX11-NEXT:    s_lshr_b32 s10, s4, 8
-; GFX11-NEXT:    s_lshr_b32 s11, s4, 16
-; GFX11-NEXT:    s_and_b32 s10, s10, 0xff
+; GFX11-NEXT:    s_lshr_b32 s10, s4, 16
+; GFX11-NEXT:    s_and_b32 s9, s9, 0xff
+; GFX11-NEXT:    s_and_b32 s11, s4, 0xff
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; GFX11-NEXT:    v_cvt_u32_f32_e32 v1, v1
-; GFX11-NEXT:    s_and_b32 s12, s4, 0xff
-; GFX11-NEXT:    s_lshl_b32 s10, s10, s9
-; GFX11-NEXT:    s_and_b32 s11, s11, 0xff
+; GFX11-NEXT:    s_lshl_b32 s9, s9, 8
+; GFX11-NEXT:    s_and_b32 s10, s10, 0xff
+; GFX11-NEXT:    s_or_b32 s9, s11, s9
 ; GFX11-NEXT:    v_mul_lo_u32 v2, 0xffffffe8, v0
-; GFX11-NEXT:    s_or_b32 s10, s12, s10
 ; GFX11-NEXT:    v_mul_lo_u32 v3, 0xffffffe8, v1
-; GFX11-NEXT:    s_bfe_u32 s11, s11, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s10, s10, 0x100000
-; GFX11-NEXT:    s_lshl_b32 s11, s11, 16
-; GFX11-NEXT:    s_lshr_b32 s12, s5, 8
-; GFX11-NEXT:    s_or_b32 s10, s10, s11
-; GFX11-NEXT:    v_mul_hi_u32 v2, v0, v2
+; GFX11-NEXT:    s_and_b32 s10, 0xffff, s10
+; GFX11-NEXT:    s_and_b32 s9, 0xffff, s9
+; GFX11-NEXT:    s_lshl_b32 s10, s10, 16
+; GFX11-NEXT:    s_lshr_b32 s11, s5, 8
+; GFX11-NEXT:    s_or_b32 s9, s9, s10
 ; GFX11-NEXT:    s_and_b32 s5, s5, 0xff
+; GFX11-NEXT:    v_mul_hi_u32 v2, v0, v2
 ; GFX11-NEXT:    s_lshr_b32 s4, s4, 24
-; GFX11-NEXT:    s_lshl_b32 s5, s5, s9
-; GFX11-NEXT:    s_and_b32 s11, s12, 0xff
+; GFX11-NEXT:    s_lshl_b32 s5, s5, 8
+; GFX11-NEXT:    s_and_b32 s10, s11, 0xff
 ; GFX11-NEXT:    s_or_b32 s4, s4, s5
-; GFX11-NEXT:    s_bfe_u32 s5, s11, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX11-NEXT:    s_and_b32 s5, 0xffff, s10
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX11-NEXT:    s_lshl_b32 s5, s5, 16
 ; GFX11-NEXT:    v_add_nc_u32_e32 v0, v0, v2
 ; GFX11-NEXT:    v_mul_hi_u32 v2, v1, v3
-; GFX11-NEXT:    s_lshl_b32 s5, s5, 16
-; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
 ; GFX11-NEXT:    s_or_b32 s4, s4, s5
-; GFX11-NEXT:    v_mul_hi_u32 v0, s10, v0
-; GFX11-NEXT:    s_lshl_b32 s1, s1, s9
-; GFX11-NEXT:    s_lshr_b32 s11, s2, 16
+; GFX11-NEXT:    s_and_b32 s1, s1, 0xff
+; GFX11-NEXT:    s_lshr_b32 s10, s2, 16
+; GFX11-NEXT:    v_mul_hi_u32 v0, s9, v0
+; GFX11-NEXT:    s_lshl_b32 s1, s1, 8
+; GFX11-NEXT:    s_lshr_b32 s5, s2, 24
 ; GFX11-NEXT:    s_or_b32 s1, s8, s1
 ; GFX11-NEXT:    v_add_nc_u32_e32 v1, v1, v2
 ; GFX11-NEXT:    s_lshr_b32 s8, s2, 8
-; GFX11-NEXT:    s_lshr_b32 s5, s2, 24
+; GFX11-NEXT:    s_and_b32 s2, s2, 0xff
 ; GFX11-NEXT:    s_and_b32 s8, s8, 0xff
 ; GFX11-NEXT:    v_mul_lo_u32 v0, v0, 24
 ; GFX11-NEXT:    v_mul_hi_u32 v1, s4, v1
-; GFX11-NEXT:    s_and_b32 s2, s2, 0xff
-; GFX11-NEXT:    s_lshl_b32 s8, s8, s9
-; GFX11-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX11-NEXT:    s_lshl_b32 s8, s8, 8
+; GFX11-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX11-NEXT:    s_or_b32 s2, s2, s8
-; GFX11-NEXT:    s_and_b32 s8, s11, 0xff
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX11-NEXT:    v_sub_nc_u32_e32 v0, s10, v0
+; GFX11-NEXT:    s_and_b32 s8, s10, 0xff
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX11-NEXT:    s_and_b32 s8, 0xffff, s8
+; GFX11-NEXT:    v_sub_nc_u32_e32 v0, s9, v0
 ; GFX11-NEXT:    v_mul_lo_u32 v1, v1, 24
-; GFX11-NEXT:    s_bfe_u32 s8, s8, 0x100000
-; GFX11-NEXT:    s_lshr_b32 s10, s3, 8
+; GFX11-NEXT:    s_lshr_b32 s9, s3, 8
 ; GFX11-NEXT:    s_and_b32 s3, s3, 0xff
+; GFX11-NEXT:    s_lshl_b32 s8, s8, 16
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX11-NEXT:    s_lshl_b32 s8, s8, 16
-; GFX11-NEXT:    s_lshl_b32 s3, s3, s9
+; GFX11-NEXT:    s_lshl_b32 s3, s3, 8
+; GFX11-NEXT:    s_or_b32 s2, s2, s8
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v1, s4, v1
-; GFX11-NEXT:    s_and_b32 s4, s10, 0xff
+; GFX11-NEXT:    s_and_b32 s4, s9, 0xff
 ; GFX11-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
-; GFX11-NEXT:    s_or_b32 s2, s2, s8
-; GFX11-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX11-NEXT:    v_subrev_nc_u32_e32 v3, 24, v1
+; GFX11-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX11-NEXT:    s_or_b32 s3, s5, s3
+; GFX11-NEXT:    v_subrev_nc_u32_e32 v3, 24, v1
+; GFX11-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v2, 24, v0
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v0
-; GFX11-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX11-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX11-NEXT:    s_lshl_b32 s4, s4, 16
 ; GFX11-NEXT:    s_lshl_b32 s5, s6, 17
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX11-NEXT:    v_cndmask_b32_e32 v0, v0, v2, vcc_lo
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
-; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
 ; GFX11-NEXT:    s_or_b32 s0, s5, s0
-; GFX11-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc_lo
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX11-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc_lo
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_subrev_nc_u32_e32 v3, 24, v1
 ; GFX11-NEXT:    v_cmp_le_u32_e32 vcc_lo, 24, v1
 ; GFX11-NEXT:    v_cndmask_b32_e32 v1, v1, v3, vcc_lo
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_3) | instid1(VALU_DEP_2)
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v3, 23, v1
 ; GFX11-NEXT:    v_and_b32_e32 v1, 0xffffff, v1
 ; GFX11-NEXT:    v_sub_nc_u32_e32 v2, 23, v0

 }
 
 define amdgpu_ps i16 @s_fshr_i16(i16 inreg %lhs, i16 inreg %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: s_fshr_i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s3, s2, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s2
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX6-NEXT:    s_bfe_u32 s2, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s3
 ; GFX6-NEXT:    s_and_b32 s1, s1, 0xffff
 ; GFX6-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s3, s2, 15
 ; GFX8-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX8-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s3
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s3, s2, 15
 ; GFX9-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX9-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX9-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, s3, 0x100000
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_and_b32 s2, 0xffff, s3
 ; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i16:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s3, s2, 15
-; GFX10-NEXT:    s_bfe_u32 s4, 1, 0x100000
 ; GFX10-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX10-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX10-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i16:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s3, s2, 15
-; GFX11-NEXT:    s_bfe_u32 s4, 1, 0x100000
 ; GFX11-NEXT:    s_and_not1_b32 s2, 15, s2
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX11-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX11-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
 ; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
   ret i16 %result
 }
 
 define amdgpu_ps i16 @s_fshr_i16_4(i16 inreg %lhs, i16 inreg %rhs) {
 ; GFX6-LABEL: s_fshr_i16_4:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 12
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0xc0004
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i16_4:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 12
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i16_4:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, 4, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 12
+; GFX9-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i16_4:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, 4, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 12
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i16_4:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_bfe_u32 s2, 12, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, 4, 0x100000
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 12
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 4
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 4)
   ret i16 %result
 }
 
 define amdgpu_ps i16 @s_fshr_i16_5(i16 inreg %lhs, i16 inreg %rhs) {
 ; GFX6-LABEL: s_fshr_i16_5:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 11
 ; GFX6-NEXT:    s_bfe_u32 s1, s1, 0xb0005
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_i16_5:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 11
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_i16_5:
 ; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s2, 5, 0x100000
-; GFX9-NEXT:    s_lshr_b32 s1, s1, s2
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 11
+; GFX9-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX9-NEXT:    s_or_b32 s0, s0, s1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: s_fshr_i16_5:
 ; GFX10:       ; %bb.0:
-; GFX10-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s3, 5, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX10-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 11
+; GFX10-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX10-NEXT:    s_or_b32 s0, s0, s1
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: s_fshr_i16_5:
 ; GFX11:       ; %bb.0:
-; GFX11-NEXT:    s_bfe_u32 s2, 11, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s3, 5, 0x100000
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX11-NEXT:    s_lshr_b32 s1, s1, s3
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 11
+; GFX11-NEXT:    s_lshr_b32 s1, s1, 5
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_or_b32 s0, s0, s1
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 5)
   ret i16 %result
 }
 
 define i16 @v_fshr_i16(i16 %lhs, i16 %rhs, i16 %amt) {
 ; GFX6-LABEL: v_fshr_i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v3, 15, v2
 ; GFX6-NEXT:    v_xor_b32_e32 v2, -1, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v2, v0
-; GFX6-NEXT:    v_bfe_u32 v2, v3, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v3
 ; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 
 define amdgpu_ps half @v_fshr_i16_ssv(i16 inreg %lhs, i16 inreg %rhs, i16 %amt) {
 ; GFX6-LABEL: v_fshr_i16_ssv:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX6-NEXT:    v_and_b32_e32 v0, 15, v0
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshl_b32_e32 v0, s0, v0
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    s_and_b32 s0, s1, 0xffff
 ; GFX6-NEXT:    v_lshr_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_i16_ssv:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
-; GFX8-NEXT:    s_bfe_u32 s2, 1, 0x100000
 ; GFX8-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s2
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v0, v0, s0
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v1, v1, s1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_i16_ssv:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    v_and_b32_e32 v1, 15, v0
 ; GFX9-NEXT:    v_xor_b32_e32 v0, -1, v0
-; GFX9-NEXT:    s_bfe_u32 s2, 1, 0x100000
 ; GFX9-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s2
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX9-NEXT:    v_lshlrev_b16_e64 v0, v0, s0
 ; GFX9-NEXT:    v_lshrrev_b16_e64 v1, v1, s1
 ; GFX9-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshr_i16_ssv:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_xor_b32_e32 v1, -1, v0
 ; GFX10-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX10-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s2
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX10-NEXT:    v_and_b32_e32 v1, 15, v1
 ; GFX10-NEXT:    v_lshrrev_b16 v0, v0, s1
 ; GFX10-NEXT:    v_lshlrev_b16 v1, v1, s0
 ; GFX10-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshr_i16_ssv:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_xor_b32_e32 v1, -1, v0
 ; GFX11-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX11-NEXT:    s_bfe_u32 s2, 1, 0x100000
-; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s2
-; GFX11-NEXT:    v_and_b32_e32 v1, 15, v1
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GFX11-NEXT:    v_and_b32_e32 v1, 15, v1
 ; GFX11-NEXT:    v_lshrrev_b16 v0, v0, s1
+; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_lshlrev_b16 v1, v1, s0
-; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, v1, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps half @v_fshr_i16_svs(i16 inreg %lhs, i16 %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: v_fshr_i16_svs:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s2, s1, 15
 ; GFX6-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s1
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, s1, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_i16_svs:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s2, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX8-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, s2, v0
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_i16_svs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 15
 ; GFX9-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX9-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX9-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX9-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX9-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX9-NEXT:    v_lshrrev_b16_e32 v0, s2, v0
 ; GFX9-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshr_i16_svs:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    s_and_b32 s2, s1, 15
-; GFX10-NEXT:    s_bfe_u32 s3, 1, 0x100000
 ; GFX10-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX10-NEXT:    v_lshrrev_b16 v0, s2, v0
-; GFX10-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX10-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshr_i16_svs:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    s_and_b32 s2, s1, 15
-; GFX11-NEXT:    s_bfe_u32 s3, 1, 0x100000
 ; GFX11-NEXT:    s_and_not1_b32 s1, 15, s1
 ; GFX11-NEXT:    v_lshrrev_b16 v0, s2, v0
-; GFX11-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX11-NEXT:    s_lshl_b32 s0, s0, s1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps half @v_fshr_i16_vss(i16 %lhs, i16 inreg %rhs, i16 inreg %amt) {
 ; GFX6-LABEL: v_fshr_i16_vss:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s2, s1, 15
 ; GFX6-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s1, v0
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    s_and_b32 s0, s0, 0xffff
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_i16_vss:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_and_b32 s2, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, s1, v0
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_i16_vss:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 15
 ; GFX9-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, 1, v0
 ; GFX9-NEXT:    v_lshlrev_b16_e32 v0, s1, v0
-; GFX9-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX9-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX9-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX9-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX9-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX9-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
 ; GFX10-LABEL: v_fshr_i16_vss:
 ; GFX10:       ; %bb.0:
 ; GFX10-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX10-NEXT:    s_andn2_b32 s2, 15, s1
 ; GFX10-NEXT:    s_and_b32 s1, s1, 15
-; GFX10-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX10-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX10-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX10-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX10-NEXT:    v_lshlrev_b16 v0, s2, v0
 ; GFX10-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX10-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX11-LABEL: v_fshr_i16_vss:
 ; GFX11:       ; %bb.0:
 ; GFX11-NEXT:    v_lshlrev_b16 v0, 1, v0
 ; GFX11-NEXT:    s_and_not1_b32 s2, 15, s1
 ; GFX11-NEXT:    s_and_b32 s1, s1, 15
-; GFX11-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX11-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX11-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX11-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX11-NEXT:    v_lshlrev_b16 v0, s2, v0
 ; GFX11-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instid1(SALU_CYCLE_1)
 ; GFX11-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX11-NEXT:    ; return to shader part epilog
   %result = call i16 @llvm.fshr.i16(i16 %lhs, i16 %rhs, i16 %amt)
   %cast.result = bitcast i16 %result to half
   ret half %cast.result
 }
 
 define amdgpu_ps i32 @s_fshr_v2i16(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: s_fshr_v2i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s5, s5, 16
 ; GFX6-NEXT:    s_and_b32 s4, s4, 0xffff
 ; GFX6-NEXT:    s_or_b32 s4, s5, s4
-; GFX6-NEXT:    s_bfe_u32 s5, 1, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s5
-; GFX6-NEXT:    s_bfe_u32 s6, s2, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s7, 14, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s1, s1, s5
+; GFX6-NEXT:    s_bfe_u32 s5, s2, 0xf0001
+; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX6-NEXT:    s_lshr_b32 s5, s5, 14
+; GFX6-NEXT:    s_or_b32 s0, s0, s5
 ; GFX6-NEXT:    s_bfe_u32 s5, s3, 0xf0001
-; GFX6-NEXT:    s_lshr_b32 s6, s6, s7
-; GFX6-NEXT:    s_lshr_b32 s5, s5, s7
+; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX6-NEXT:    s_lshr_b32 s5, s5, 14
 ; GFX6-NEXT:    s_xor_b32 s4, s4, -1
-; GFX6-NEXT:    s_or_b32 s0, s0, s6
 ; GFX6-NEXT:    s_or_b32 s1, s1, s5
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX6-NEXT:    s_lshr_b32 s5, s4, 16
 ; GFX6-NEXT:    s_and_b32 s6, s4, 15
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
 ; GFX6-NEXT:    s_bfe_u32 s2, s2, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX6-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX6-NEXT:    s_or_b32 s0, s0, s2
 ; GFX6-NEXT:    s_and_b32 s2, s5, 15
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s5
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s2
 ; GFX6-NEXT:    s_bfe_u32 s2, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s3, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s4
 ; GFX6-NEXT:    s_lshr_b32 s2, s2, s3
 ; GFX6-NEXT:    s_or_b32 s1, s1, s2
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_v2i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s5, 1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s6, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s7, 15, 0x100000
+; GFX8-NEXT:    s_and_b32 s5, 0xffff, s1
 ; GFX8-NEXT:    s_lshr_b32 s3, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s4, s1, 16
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s5
-; GFX8-NEXT:    s_lshr_b32 s6, s6, s7
-; GFX8-NEXT:    s_or_b32 s0, s0, s6
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s5
-; GFX8-NEXT:    s_lshr_b32 s6, s4, s7
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s5
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_lshr_b32 s5, s5, 15
+; GFX8-NEXT:    s_or_b32 s0, s0, s5
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
+; GFX8-NEXT:    s_lshr_b32 s5, s4, 15
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX8-NEXT:    s_xor_b32 s2, s2, -1
-; GFX8-NEXT:    s_or_b32 s3, s3, s6
-; GFX8-NEXT:    s_lshr_b32 s6, s2, 16
-; GFX8-NEXT:    s_and_b32 s7, s2, 15
+; GFX8-NEXT:    s_or_b32 s3, s3, s5
+; GFX8-NEXT:    s_lshr_b32 s5, s2, 16
+; GFX8-NEXT:    s_and_b32 s6, s2, 15
 ; GFX8-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s1, s1, s5
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s7
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX8-NEXT:    s_lshr_b32 s1, s1, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_lshl_b32 s0, s0, s6
 ; GFX8-NEXT:    s_lshr_b32 s1, s1, s2
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
-; GFX8-NEXT:    s_and_b32 s1, s6, 15
-; GFX8-NEXT:    s_lshl_b32 s4, s4, s5
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_andn2_b32 s2, 15, s6
+; GFX8-NEXT:    s_and_b32 s1, s5, 15
+; GFX8-NEXT:    s_lshl_b32 s4, s4, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_andn2_b32 s2, 15, s5
 ; GFX8-NEXT:    s_lshl_b32 s1, s3, s1
-; GFX8-NEXT:    s_bfe_u32 s3, s4, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s5
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s4
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_lshr_b32 s2, s3, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX8-NEXT:    s_or_b32 s0, s0, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_v2i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s4, s0, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 0x10001

 
 define <2 x i16> @v_fshr_v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt) {
 ; GFX6-LABEL: v_fshr_v2i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v5, 16, v5
 ; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_or_b32_e32 v4, v5, v4
-; GFX6-NEXT:    s_bfe_u32 s4, 1, 0x100000
 ; GFX6-NEXT:    v_bfe_u32 v5, v2, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s5, 14, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
-; GFX6-NEXT:    v_lshrrev_b32_e32 v5, s5, v5
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX6-NEXT:    v_lshrrev_b32_e32 v5, 14, v5
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v5
 ; GFX6-NEXT:    v_bfe_u32 v5, v3, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s4, v1
-; GFX6-NEXT:    v_lshrrev_b32_e32 v5, s5, v5
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v5, 14, v5
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v5
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v5, 16, v4
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v4
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 1, v2
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v6, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_bfe_u32 v2, v2, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v6, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v4, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v5
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v5
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v3
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v2, v1
 ; GFX6-NEXT:    v_bfe_u32 v2, v3, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v3, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v4
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v2, v3, v2
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_v2i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, 1, v0

 ; GFX8-NEXT:    v_or_b32_e32 v2, v3, v2
 ; GFX8-NEXT:    v_and_b32_e32 v3, 15, v4
 ; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX8-NEXT:    v_and_b32_e32 v4, 15, v4
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, 1, v1
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v3, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, v4, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v1, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v2i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v3, 0xf000f, v2
 ; GFX9-NEXT:    v_xor_b32_e32 v2, -1, v2
; GFX11-NEXT:    s_setpc_b64 s[30:31]
   %result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
   ret <2 x i16> %result
 }
 
 define <2 x i16> @v_fshr_v2i16_4_8(<2 x i16> %lhs, <2 x i16> %rhs) {
 ; GFX6-LABEL: v_fshr_v2i16_4_8:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX6-NEXT:    s_bfe_u32 s4, 12, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
 ; GFX6-NEXT:    v_bfe_u32 v2, v2, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s4, 3, 0x100000
-; GFX6-NEXT:    v_lshrrev_b32_e32 v2, s4, v2
-; GFX6-NEXT:    s_bfe_u32 s4, 8, 0x100000
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 12, v0
+; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 3, v2
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v2
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s4, v1
 ; GFX6-NEXT:    v_bfe_u32 v2, v3, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s4, 7, 0x100000
-; GFX6-NEXT:    v_lshrrev_b32_e32 v2, s4, v2
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 8, v1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 7, v2
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v2
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_v2i16_4_8:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v2, 16, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, 12, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, 4, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v3
 ; GFX8-NEXT:    v_mov_b32_e32 v3, 8
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v2, 8, v2
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v1, v3, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v2i16_4_8:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_mov_b32_e32 v2, 0x8000c
 ; GFX9-NEXT:    v_pk_lshlrev_b16 v0, v2, v0

 }
 
 define amdgpu_ps float @v_fshr_v2i16_ssv(<2 x i16> inreg %lhs, <2 x i16> inreg %rhs, <2 x i16> %amt) {
 ; GFX6-LABEL: v_fshr_v2i16_ssv:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_or_b32_e32 v0, v1, v0
-; GFX6-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s5, s2, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s6, 14, 0x100000
+; GFX6-NEXT:    s_bfe_u32 s4, s2, 0xf0001
 ; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX6-NEXT:    s_lshr_b32 s5, s5, s6
+; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX6-NEXT:    s_lshr_b32 s4, s4, 14
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v0
 ; GFX6-NEXT:    v_xor_b32_e32 v0, -1, v0
-; GFX6-NEXT:    s_or_b32 s0, s0, s5
+; GFX6-NEXT:    s_or_b32 s0, s0, s4
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX6-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s0, v2
 ; GFX6-NEXT:    s_bfe_u32 s0, s2, 0xf0001
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshr_b32_e32 v0, s0, v0
-; GFX6-NEXT:    s_lshl_b32 s1, s1, s4
 ; GFX6-NEXT:    s_bfe_u32 s4, s3, 0xf0001
 ; GFX6-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_and_b32_e32 v2, 15, v1
 ; GFX6-NEXT:    v_xor_b32_e32 v1, -1, v1
-; GFX6-NEXT:    s_lshr_b32 s4, s4, s6
+; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX6-NEXT:    s_lshr_b32 s4, s4, 14
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX6-NEXT:    v_and_b32_e32 v1, 15, v1
 ; GFX6-NEXT:    s_or_b32 s1, s1, s4
-; GFX6-NEXT:    v_bfe_u32 v2, v2, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v2, 0xffff, v2
 ; GFX6-NEXT:    s_bfe_u32 s0, s3, 0xf0001
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX6-NEXT:    v_lshl_b32_e32 v2, s1, v2
 ; GFX6-NEXT:    v_lshr_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_v2i16_ssv:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s4, 1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s5, s1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s6, 15, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s1
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s4
-; GFX8-NEXT:    s_lshr_b32 s5, s5, s6
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_lshr_b32 s4, s4, 15
 ; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
-; GFX8-NEXT:    s_or_b32 s0, s0, s5
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s4
+; GFX8-NEXT:    s_or_b32 s0, s0, s4
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v0
 ; GFX8-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
 ; GFX8-NEXT:    v_xor_b32_e32 v0, -1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v2, v2, s0
-; GFX8-NEXT:    s_bfe_u32 s0, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s1
 ; GFX8-NEXT:    v_and_b32_e32 v0, 15, v0
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s4
-; GFX8-NEXT:    s_lshr_b32 s5, s3, s6
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s4
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX8-NEXT:    s_lshr_b32 s4, s3, 15
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v0, v0, s0
-; GFX8-NEXT:    s_lshl_b32 s2, s2, s4
+; GFX8-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX8-NEXT:    v_and_b32_e32 v2, 15, v1
 ; GFX8-NEXT:    v_xor_b32_e32 v1, -1, v1
-; GFX8-NEXT:    s_bfe_u32 s0, s3, 0x100000
-; GFX8-NEXT:    s_or_b32 s2, s2, s5
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s3
+; GFX8-NEXT:    s_or_b32 s2, s2, s4
 ; GFX8-NEXT:    v_and_b32_e32 v1, 15, v1
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s4
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
 ; GFX8-NEXT:    v_lshlrev_b16_e64 v2, v2, s2
 ; GFX8-NEXT:    v_lshrrev_b16_e64 v1, v1, s0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v0, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_v2i16_ssv:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s2, s0, 16
 ; GFX9-NEXT:    v_and_b32_e32 v1, 0xf000f, v0
 ; GFX9-NEXT:    v_xor_b32_e32 v0, -1, v0
; GFX11-NEXT:    ; return to shader part epilog
   %result = call <2 x i16> @llvm.fshr.v2i16(<2 x i16> %lhs, <2 x i16> %rhs, <2 x i16> %amt)
   %cast = bitcast <2 x i16> %result to float
   ret float %cast
 }
 
 define amdgpu_ps float @v_fshr_v2i16_svs(<2 x i16> inreg %lhs, <2 x i16> %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: v_fshr_v2i16_svs:
 ; GFX6:       ; %bb.0:
+; GFX6-NEXT:    v_bfe_u32 v2, v0, 1, 15
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 16
 ; GFX6-NEXT:    s_and_b32 s2, s2, 0xffff
-; GFX6-NEXT:    s_or_b32 s2, s3, s2
-; GFX6-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX6-NEXT:    v_bfe_u32 v2, v0, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s4, 14, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s3
-; GFX6-NEXT:    v_lshrrev_b32_e32 v2, s4, v2
+; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v2, 14, v2
 ; GFX6-NEXT:    v_bfe_u32 v3, v1, 1, 15
+; GFX6-NEXT:    s_or_b32 s2, s3, s2
 ; GFX6-NEXT:    v_or_b32_e32 v2, s0, v2
-; GFX6-NEXT:    s_lshl_b32 s0, s1, s3
-; GFX6-NEXT:    v_lshrrev_b32_e32 v3, s4, v3
+; GFX6-NEXT:    s_lshl_b32 s0, s1, 1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 14, v3
 ; GFX6-NEXT:    v_or_b32_e32 v3, s0, v3
 ; GFX6-NEXT:    s_xor_b32 s0, s2, -1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
 ; GFX6-NEXT:    s_lshr_b32 s1, s0, 16
 ; GFX6-NEXT:    s_and_b32 s2, s0, 15
 ; GFX6-NEXT:    s_andn2_b32 s0, 15, s0
 ; GFX6-NEXT:    v_bfe_u32 v0, v0, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    s_and_b32 s0, s1, 15
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v1
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, s2, v2
 ; GFX6-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    v_or_b32_e32 v0, v2, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, s0, v3
 ; GFX6-NEXT:    v_bfe_u32 v1, v1, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s0, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s1
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    v_or_b32_e32 v1, v2, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_v2i16_svs:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s3, 1, 0x100000
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s3
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v1, 15, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 15
 ; GFX8-NEXT:    v_or_b32_e32 v1, s0, v1
-; GFX8-NEXT:    s_lshl_b32 s0, s2, s3
+; GFX8-NEXT:    s_lshl_b32 s0, s2, 1
 ; GFX8-NEXT:    v_lshrrev_b16_sdwa v2, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    v_or_b32_e32 v2, s0, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, 1, v0
 ; GFX8-NEXT:    v_mov_b32_e32 v4, 1
 ; GFX8-NEXT:    s_xor_b32 s0, s1, -1
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
 ; GFX8-NEXT:    s_lshr_b32 s1, s0, 16
 ; GFX8-NEXT:    s_and_b32 s2, s0, 15
 ; GFX8-NEXT:    s_andn2_b32 s0, 15, s0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, 1, v3
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, s0, v3
 ; GFX8-NEXT:    s_and_b32 s0, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, 1, v0
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v2, s0, v2
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v0, s1, v0
-; GFX8-NEXT:    v_lshlrev_b16_e32 v1, s2, v1
 ; GFX8-NEXT:    v_or_b32_e32 v0, v2, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
+; GFX8-NEXT:    v_lshlrev_b16_e32 v1, s2, v1
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v3
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_v2i16_svs:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s3, s0, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 0x10001
 ; GFX9-NEXT:    s_lshl_b32 s3, s3, 1

 }
 
 define amdgpu_ps float @v_fshr_v2i16_vss(<2 x i16> %lhs, <2 x i16> inreg %rhs, <2 x i16> inreg %amt) {
 ; GFX6-LABEL: v_fshr_v2i16_vss:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 16
 ; GFX6-NEXT:    s_and_b32 s2, s2, 0xffff
 ; GFX6-NEXT:    s_or_b32 s2, s3, s2
-; GFX6-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s3, v0
-; GFX6-NEXT:    s_bfe_u32 s4, s0, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, 14, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s3, v1
+; GFX6-NEXT:    s_bfe_u32 s3, s0, 0xf0001
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX6-NEXT:    s_lshr_b32 s3, s3, 14
+; GFX6-NEXT:    v_or_b32_e32 v0, s3, v0
 ; GFX6-NEXT:    s_bfe_u32 s3, s1, 0xf0001
-; GFX6-NEXT:    s_lshr_b32 s4, s4, s5
-; GFX6-NEXT:    s_lshr_b32 s3, s3, s5
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v1
+; GFX6-NEXT:    s_lshr_b32 s3, s3, 14
 ; GFX6-NEXT:    s_xor_b32 s2, s2, -1
-; GFX6-NEXT:    v_or_b32_e32 v0, s4, v0
 ; GFX6-NEXT:    v_or_b32_e32 v1, s3, v1
 ; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX6-NEXT:    s_lshr_b32 s3, s2, 16
 ; GFX6-NEXT:    s_and_b32 s4, s2, 15
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s2
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_bfe_u32 s0, s0, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s2
 ; GFX6-NEXT:    v_or_b32_e32 v0, s0, v0
 ; GFX6-NEXT:    s_and_b32 s0, s3, 15
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
 ; GFX6-NEXT:    s_andn2_b32 s2, 15, s3
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s0, v1
 ; GFX6-NEXT:    s_bfe_u32 s0, s1, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX6-NEXT:    v_or_b32_e32 v1, s0, v1
-; GFX6-NEXT:    v_bfe_u32 v1, v1, 0, 16
-; GFX6-NEXT:    v_bfe_u32 v0, v0, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX6-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v1
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: v_fshr_v2i16_vss:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s3, s0, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s4, 15, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s0
 ; GFX8-NEXT:    s_lshr_b32 s2, s0, 16
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, 1, v0
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s4
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 15
 ; GFX8-NEXT:    v_mov_b32_e32 v2, 1
 ; GFX8-NEXT:    v_or_b32_e32 v1, s3, v1
 ; GFX8-NEXT:    v_lshlrev_b16_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT:    s_lshr_b32 s3, s2, s4
-; GFX8-NEXT:    v_or_b32_e32 v0, s3, v0
-; GFX8-NEXT:    s_bfe_u32 s3, 1, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s3
+; GFX8-NEXT:    s_lshr_b32 s3, s2, 15
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
 ; GFX8-NEXT:    s_xor_b32 s1, s1, -1
-; GFX8-NEXT:    s_lshr_b32 s4, s1, 16
-; GFX8-NEXT:    s_and_b32 s5, s1, 15
+; GFX8-NEXT:    v_or_b32_e32 v0, s3, v0
+; GFX8-NEXT:    s_lshr_b32 s3, s1, 16
+; GFX8-NEXT:    s_and_b32 s4, s1, 15
 ; GFX8-NEXT:    s_andn2_b32 s1, 15, s1
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s3
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX8-NEXT:    v_lshlrev_b16_e32 v1, s5, v1
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX8-NEXT:    v_lshlrev_b16_e32 v1, s4, v1
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
-; GFX8-NEXT:    s_lshl_b32 s2, s2, s3
+; GFX8-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX8-NEXT:    v_or_b32_e32 v1, s0, v1
-; GFX8-NEXT:    s_and_b32 s0, s4, 15
-; GFX8-NEXT:    s_andn2_b32 s1, 15, s4
+; GFX8-NEXT:    s_and_b32 s0, s3, 15
+; GFX8-NEXT:    s_andn2_b32 s1, 15, s3
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, s0, v0
-; GFX8-NEXT:    s_bfe_u32 s0, s2, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s0, s0, s3
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s2
+; GFX8-NEXT:    s_lshr_b32 s0, s0, 1
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshr_b32 s0, s0, s1
 ; GFX8-NEXT:    v_or_b32_e32 v0, s0, v0
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v1, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: v_fshr_v2i16_vss:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_and_b32 s2, s1, 0xf000f
 ; GFX9-NEXT:    s_andn2_b32 s1, 0xf000f, s1
 ; GFX9-NEXT:    v_pk_lshlrev_b16 v0, 1, v0 op_sel_hi:[0,1]

 define amdgpu_ps i48 @s_fshr_v3i16(<3 x i16> inreg %lhs, <3 x i16> inreg %rhs, <3 x i16> inreg %amt) {
 ; GFX6-LABEL: s_fshr_v3i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_and_b32 s7, s7, 0xffff
 ; GFX6-NEXT:    s_and_b32 s6, s6, 0xffff
 ; GFX6-NEXT:    s_lshl_b32 s7, s7, 16
 ; GFX6-NEXT:    s_or_b32 s6, s6, s7
 ; GFX6-NEXT:    s_and_b32 s7, s8, 0xffff
-; GFX6-NEXT:    s_bfe_u32 s8, 1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s9, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s10, 14, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s8
-; GFX6-NEXT:    s_lshr_b32 s9, s9, s10
-; GFX6-NEXT:    s_or_b32 s0, s0, s9
-; GFX6-NEXT:    s_bfe_u32 s9, s4, 0xf0001
-; GFX6-NEXT:    s_lshl_b32 s1, s1, s8
-; GFX6-NEXT:    s_lshr_b32 s9, s9, s10
+; GFX6-NEXT:    s_bfe_u32 s8, s3, 0xf0001
+; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX6-NEXT:    s_lshr_b32 s8, s8, 14
+; GFX6-NEXT:    s_or_b32 s0, s0, s8
+; GFX6-NEXT:    s_bfe_u32 s8, s4, 0xf0001
+; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX6-NEXT:    s_lshr_b32 s8, s8, 14
 ; GFX6-NEXT:    s_xor_b32 s6, s6, -1
-; GFX6-NEXT:    s_or_b32 s1, s1, s9
+; GFX6-NEXT:    s_or_b32 s1, s1, s8
 ; GFX6-NEXT:    s_lshl_b32 s3, s3, 1
-; GFX6-NEXT:    s_lshr_b32 s9, s6, 16
-; GFX6-NEXT:    s_and_b32 s11, s6, 15
+; GFX6-NEXT:    s_lshr_b32 s8, s6, 16
+; GFX6-NEXT:    s_and_b32 s9, s6, 15
 ; GFX6-NEXT:    s_andn2_b32 s6, 15, s6
-; GFX6-NEXT:    s_bfe_u32 s11, s11, 0x100000
+; GFX6-NEXT:    s_and_b32 s9, 0xffff, s9
 ; GFX6-NEXT:    s_bfe_u32 s3, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s6, s6, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s11
+; GFX6-NEXT:    s_and_b32 s6, 0xffff, s6
+; GFX6-NEXT:    s_lshl_b32 s0, s0, s9
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s6
 ; GFX6-NEXT:    s_or_b32 s0, s0, s3
-; GFX6-NEXT:    s_and_b32 s3, s9, 15
+; GFX6-NEXT:    s_and_b32 s3, s8, 15
 ; GFX6-NEXT:    s_lshl_b32 s4, s4, 1
-; GFX6-NEXT:    s_andn2_b32 s6, 15, s9
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_andn2_b32 s6, 15, s8
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s3
 ; GFX6-NEXT:    s_bfe_u32 s3, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s6, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s6
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s4
 ; GFX6-NEXT:    s_or_b32 s1, s1, s3
 ; GFX6-NEXT:    s_bfe_u32 s3, s5, 0xf0001
-; GFX6-NEXT:    s_lshl_b32 s2, s2, s8
-; GFX6-NEXT:    s_lshr_b32 s3, s3, s10
+; GFX6-NEXT:    s_lshl_b32 s2, s2, 1
+; GFX6-NEXT:    s_lshr_b32 s3, s3, 14
 ; GFX6-NEXT:    s_xor_b32 s4, s7, -1
 ; GFX6-NEXT:    s_or_b32 s2, s2, s3
 ; GFX6-NEXT:    s_lshl_b32 s3, s5, 1
 ; GFX6-NEXT:    s_and_b32 s5, s4, 15
 ; GFX6-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_bfe_u32 s3, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, s5
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s4
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_or_b32 s2, s2, s3
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
-; GFX6-NEXT:    s_bfe_u32 s1, s2, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s2
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_v3i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s8, 1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s9, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s10, 15, 0x100000
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s2
 ; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s7, s2, 16
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s8
-; GFX8-NEXT:    s_lshr_b32 s9, s9, s10
-; GFX8-NEXT:    s_or_b32 s0, s0, s9
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s8
-; GFX8-NEXT:    s_lshr_b32 s9, s7, s10
-; GFX8-NEXT:    s_lshl_b32 s2, s2, s8
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_lshr_b32 s8, s8, 15
+; GFX8-NEXT:    s_or_b32 s0, s0, s8
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 1
+; GFX8-NEXT:    s_lshr_b32 s8, s7, 15
+; GFX8-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX8-NEXT:    s_xor_b32 s4, s4, -1
-; GFX8-NEXT:    s_or_b32 s6, s6, s9
-; GFX8-NEXT:    s_lshr_b32 s9, s4, 16
-; GFX8-NEXT:    s_and_b32 s11, s4, 15
+; GFX8-NEXT:    s_or_b32 s6, s6, s8
+; GFX8-NEXT:    s_lshr_b32 s8, s4, 16
+; GFX8-NEXT:    s_and_b32 s9, s4, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s11, s11, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s2, s2, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s11
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s9, 0xffff, s9
+; GFX8-NEXT:    s_lshr_b32 s2, s2, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_lshl_b32 s0, s0, s9
 ; GFX8-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
-; GFX8-NEXT:    s_and_b32 s2, s9, 15
-; GFX8-NEXT:    s_lshl_b32 s7, s7, s8
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_andn2_b32 s4, 15, s9
+; GFX8-NEXT:    s_and_b32 s2, s8, 15
+; GFX8-NEXT:    s_lshl_b32 s7, s7, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_andn2_b32 s4, 15, s8
 ; GFX8-NEXT:    s_lshl_b32 s2, s6, s2
-; GFX8-NEXT:    s_bfe_u32 s6, s7, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s6, s6, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s7
+; GFX8-NEXT:    s_lshr_b32 s6, s6, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshr_b32 s4, s6, s4
 ; GFX8-NEXT:    s_or_b32 s2, s2, s4
-; GFX8-NEXT:    s_bfe_u32 s4, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s3
 ; GFX8-NEXT:    s_and_b32 s5, s5, 0xffff
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s8
-; GFX8-NEXT:    s_lshr_b32 s4, s4, s10
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX8-NEXT:    s_lshr_b32 s4, s4, 15
 ; GFX8-NEXT:    s_or_b32 s1, s1, s4
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s8
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX8-NEXT:    s_xor_b32 s4, s5, -1
 ; GFX8-NEXT:    s_and_b32 s5, s4, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s5, s5, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX8-NEXT:    s_and_b32 s5, 0xffff, s5
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, s5
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
 ; GFX8-NEXT:    s_or_b32 s1, s1, s3
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_v3i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 0x10001
 ; GFX9-NEXT:    s_lshl_b32 s7, s7, 1
 ; GFX9-NEXT:    s_and_b32 s6, s4, 0xf000f

 ; GFX6-LABEL: v_fshr_v3i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_and_b32_e32 v7, 0xffff, v7
 ; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v7, 16, v7
 ; GFX6-NEXT:    v_or_b32_e32 v6, v6, v7
 ; GFX6-NEXT:    v_and_b32_e32 v7, 0xffff, v8
-; GFX6-NEXT:    s_bfe_u32 s4, 1, 0x100000
 ; GFX6-NEXT:    v_bfe_u32 v8, v3, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s5, 14, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
-; GFX6-NEXT:    v_lshrrev_b32_e32 v8, s5, v8
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX6-NEXT:    v_lshrrev_b32_e32 v8, 14, v8
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v8
 ; GFX6-NEXT:    v_bfe_u32 v8, v4, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s4, v1
-; GFX6-NEXT:    v_lshrrev_b32_e32 v8, s5, v8
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v8, 14, v8
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v8
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v8, 16, v6
 ; GFX6-NEXT:    v_and_b32_e32 v9, 15, v6
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v3
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v9, v9, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v9, 0xffff, v9
 ; GFX6-NEXT:    v_bfe_u32 v3, v3, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v6, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v9, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v6, v3
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v3
 ; GFX6-NEXT:    v_and_b32_e32 v3, 15, v8
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, 1, v4
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v3, v3, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v3, 0xffff, v3
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v3, v1
 ; GFX6-NEXT:    v_bfe_u32 v3, v4, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v3
 ; GFX6-NEXT:    v_bfe_u32 v3, v5, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v2, s4, v2
-; GFX6-NEXT:    v_lshrrev_b32_e32 v3, s5, v3
+; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 1, v2
+; GFX6-NEXT:    v_lshrrev_b32_e32 v3, 14, v3
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v7
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v5
 ; GFX6-NEXT:    v_and_b32_e32 v5, 15, v4
 ; GFX6-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v4
-; GFX6-NEXT:    v_bfe_u32 v5, v5, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v5
 ; GFX6-NEXT:    v_bfe_u32 v3, v3, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v5, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v3
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_v3i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)

 ; GFX8-NEXT:    v_xor_b32_sdwa v4, v5, v4 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v3, 1, v3
 ; GFX8-NEXT:    v_and_b32_e32 v5, 15, v4
 ; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX8-NEXT:    v_and_b32_e32 v4, 15, v4
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, 1, v3
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v1, v5, v1
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, v4, v3
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v3
-; GFX8-NEXT:    v_mov_b32_e32 v3, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v3, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX8-NEXT:    v_bfe_u32 v1, v1, 0, 16
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v3i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v6, 0xf000f, v4
 ; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX9-NEXT:    v_and_b32_e32 v4, 0xf000f, v4

 ; GFX6-LABEL: s_fshr_v4i16:
 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_lshl_b32 s9, s9, 16
 ; GFX6-NEXT:    s_and_b32 s8, s8, 0xffff
 ; GFX6-NEXT:    s_or_b32 s8, s9, s8
 ; GFX6-NEXT:    s_lshl_b32 s9, s11, 16
 ; GFX6-NEXT:    s_and_b32 s10, s10, 0xffff
 ; GFX6-NEXT:    s_or_b32 s9, s9, s10
-; GFX6-NEXT:    s_bfe_u32 s10, 1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s11, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s12, 14, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s10
-; GFX6-NEXT:    s_lshr_b32 s11, s11, s12
-; GFX6-NEXT:    s_or_b32 s0, s0, s11
-; GFX6-NEXT:    s_bfe_u32 s11, s5, 0xf0001
-; GFX6-NEXT:    s_lshl_b32 s1, s1, s10
-; GFX6-NEXT:    s_lshr_b32 s11, s11, s12
+; GFX6-NEXT:    s_bfe_u32 s10, s4, 0xf0001
+; GFX6-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX6-NEXT:    s_lshr_b32 s10, s10, 14
+; GFX6-NEXT:    s_or_b32 s0, s0, s10
+; GFX6-NEXT:    s_bfe_u32 s10, s5, 0xf0001
+; GFX6-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX6-NEXT:    s_lshr_b32 s10, s10, 14
 ; GFX6-NEXT:    s_xor_b32 s8, s8, -1
-; GFX6-NEXT:    s_or_b32 s1, s1, s11
+; GFX6-NEXT:    s_or_b32 s1, s1, s10
 ; GFX6-NEXT:    s_lshl_b32 s4, s4, 1
-; GFX6-NEXT:    s_lshr_b32 s11, s8, 16
-; GFX6-NEXT:    s_and_b32 s13, s8, 15
+; GFX6-NEXT:    s_lshr_b32 s10, s8, 16
+; GFX6-NEXT:    s_and_b32 s11, s8, 15
 ; GFX6-NEXT:    s_andn2_b32 s8, 15, s8
-; GFX6-NEXT:    s_bfe_u32 s13, s13, 0x100000
+; GFX6-NEXT:    s_and_b32 s11, 0xffff, s11
 ; GFX6-NEXT:    s_bfe_u32 s4, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s8, s8, 0x100000
-; GFX6-NEXT:    s_lshl_b32 s0, s0, s13
+; GFX6-NEXT:    s_and_b32 s8, 0xffff, s8
+; GFX6-NEXT:    s_lshl_b32 s0, s0, s11
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s8
 ; GFX6-NEXT:    s_or_b32 s0, s0, s4
-; GFX6-NEXT:    s_and_b32 s4, s11, 15
+; GFX6-NEXT:    s_and_b32 s4, s10, 15
 ; GFX6-NEXT:    s_lshl_b32 s5, s5, 1
-; GFX6-NEXT:    s_andn2_b32 s8, 15, s11
-; GFX6-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX6-NEXT:    s_andn2_b32 s8, 15, s10
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s4
 ; GFX6-NEXT:    s_bfe_u32 s4, s5, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, s8, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s8
 ; GFX6-NEXT:    s_lshr_b32 s4, s4, s5
 ; GFX6-NEXT:    s_or_b32 s1, s1, s4
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
+; GFX6-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, 16
 ; GFX6-NEXT:    s_or_b32 s0, s0, s1
-; GFX6-NEXT:    s_lshl_b32 s1, s2, s10
+; GFX6-NEXT:    s_lshl_b32 s1, s2, 1
 ; GFX6-NEXT:    s_bfe_u32 s2, s6, 0xf0001
-; GFX6-NEXT:    s_lshr_b32 s2, s2, s12
+; GFX6-NEXT:    s_lshr_b32 s2, s2, 14
 ; GFX6-NEXT:    s_or_b32 s1, s1, s2
-; GFX6-NEXT:    s_lshl_b32 s2, s3, s10
+; GFX6-NEXT:    s_lshl_b32 s2, s3, 1
 ; GFX6-NEXT:    s_bfe_u32 s3, s7, 0xf0001
-; GFX6-NEXT:    s_lshr_b32 s3, s3, s12
+; GFX6-NEXT:    s_lshr_b32 s3, s3, 14
 ; GFX6-NEXT:    s_xor_b32 s5, s9, -1
 ; GFX6-NEXT:    s_or_b32 s2, s2, s3
 ; GFX6-NEXT:    s_lshl_b32 s3, s6, 1
 ; GFX6-NEXT:    s_lshl_b32 s4, s7, 1
 ; GFX6-NEXT:    s_lshr_b32 s6, s5, 16
 ; GFX6-NEXT:    s_and_b32 s7, s5, 15
 ; GFX6-NEXT:    s_andn2_b32 s5, 15, s5
-; GFX6-NEXT:    s_bfe_u32 s7, s7, 0x100000
+; GFX6-NEXT:    s_and_b32 s7, 0xffff, s7
 ; GFX6-NEXT:    s_bfe_u32 s3, s3, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX6-NEXT:    s_lshl_b32 s1, s1, s7
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s5
 ; GFX6-NEXT:    s_or_b32 s1, s1, s3
 ; GFX6-NEXT:    s_and_b32 s3, s6, 15
 ; GFX6-NEXT:    s_andn2_b32 s5, 15, s6
-; GFX6-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX6-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, s3
 ; GFX6-NEXT:    s_bfe_u32 s3, s4, 0xf0001
-; GFX6-NEXT:    s_bfe_u32 s4, s5, 0x100000
+; GFX6-NEXT:    s_and_b32 s4, 0xffff, s5
 ; GFX6-NEXT:    s_lshr_b32 s3, s3, s4
 ; GFX6-NEXT:    s_or_b32 s2, s2, s3
-; GFX6-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX6-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX6-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX6-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX6-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX6-NEXT:    s_or_b32 s1, s1, s2
 ; GFX6-NEXT:    ; return to shader part epilog
 ;
 ; GFX8-LABEL: s_fshr_v4i16:
 ; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_bfe_u32 s8, 1, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s9, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s10, 15, 0x100000
+; GFX8-NEXT:    s_and_b32 s8, 0xffff, s2
 ; GFX8-NEXT:    s_lshr_b32 s6, s0, 16
 ; GFX8-NEXT:    s_lshr_b32 s7, s2, 16
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s8
-; GFX8-NEXT:    s_lshr_b32 s9, s9, s10
-; GFX8-NEXT:    s_or_b32 s0, s0, s9
-; GFX8-NEXT:    s_lshl_b32 s6, s6, s8
-; GFX8-NEXT:    s_lshr_b32 s9, s7, s10
-; GFX8-NEXT:    s_lshl_b32 s2, s2, s8
+; GFX8-NEXT:    s_lshl_b32 s0, s0, 1
+; GFX8-NEXT:    s_lshr_b32 s8, s8, 15
+; GFX8-NEXT:    s_or_b32 s0, s0, s8
+; GFX8-NEXT:    s_lshl_b32 s6, s6, 1
+; GFX8-NEXT:    s_lshr_b32 s8, s7, 15
+; GFX8-NEXT:    s_lshl_b32 s2, s2, 1
 ; GFX8-NEXT:    s_xor_b32 s4, s4, -1
-; GFX8-NEXT:    s_or_b32 s6, s6, s9
-; GFX8-NEXT:    s_lshr_b32 s9, s4, 16
-; GFX8-NEXT:    s_and_b32 s11, s4, 15
+; GFX8-NEXT:    s_or_b32 s6, s6, s8
+; GFX8-NEXT:    s_lshr_b32 s8, s4, 16
+; GFX8-NEXT:    s_and_b32 s9, s4, 15
 ; GFX8-NEXT:    s_andn2_b32 s4, 15, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s11, s11, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s2, s2, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
-; GFX8-NEXT:    s_lshl_b32 s0, s0, s11
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s9, 0xffff, s9
+; GFX8-NEXT:    s_lshr_b32 s2, s2, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
+; GFX8-NEXT:    s_lshl_b32 s0, s0, s9
 ; GFX8-NEXT:    s_lshr_b32 s2, s2, s4
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
-; GFX8-NEXT:    s_and_b32 s2, s9, 15
-; GFX8-NEXT:    s_lshl_b32 s7, s7, s8
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_andn2_b32 s4, 15, s9
+; GFX8-NEXT:    s_and_b32 s2, s8, 15
+; GFX8-NEXT:    s_lshl_b32 s7, s7, 1
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_andn2_b32 s4, 15, s8
 ; GFX8-NEXT:    s_lshl_b32 s2, s6, s2
-; GFX8-NEXT:    s_bfe_u32 s6, s7, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s6, s6, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s4, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s7
+; GFX8-NEXT:    s_lshr_b32 s6, s6, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s4
 ; GFX8-NEXT:    s_lshr_b32 s4, s6, s4
 ; GFX8-NEXT:    s_or_b32 s2, s2, s4
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s0, s0, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s0, 0xffff, s0
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
-; GFX8-NEXT:    s_bfe_u32 s6, s3, 0x100000
+; GFX8-NEXT:    s_and_b32 s6, 0xffff, s3
 ; GFX8-NEXT:    s_or_b32 s0, s0, s2
 ; GFX8-NEXT:    s_lshr_b32 s2, s1, 16
 ; GFX8-NEXT:    s_lshr_b32 s4, s3, 16
-; GFX8-NEXT:    s_lshl_b32 s1, s1, s8
-; GFX8-NEXT:    s_lshr_b32 s6, s6, s10
+; GFX8-NEXT:    s_lshl_b32 s1, s1, 1
+; GFX8-NEXT:    s_lshr_b32 s6, s6, 15
 ; GFX8-NEXT:    s_or_b32 s1, s1, s6
-; GFX8-NEXT:    s_lshl_b32 s2, s2, s8
-; GFX8-NEXT:    s_lshr_b32 s6, s4, s10
-; GFX8-NEXT:    s_lshl_b32 s3, s3, s8
+; GFX8-NEXT:    s_lshl_b32 s2, s2, 1
+; GFX8-NEXT:    s_lshr_b32 s6, s4, 15
+; GFX8-NEXT:    s_lshl_b32 s3, s3, 1
 ; GFX8-NEXT:    s_xor_b32 s5, s5, -1
 ; GFX8-NEXT:    s_or_b32 s2, s2, s6
 ; GFX8-NEXT:    s_lshr_b32 s6, s5, 16
 ; GFX8-NEXT:    s_and_b32 s7, s5, 15
 ; GFX8-NEXT:    s_andn2_b32 s5, 15, s5
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s7, s7, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s8
-; GFX8-NEXT:    s_bfe_u32 s5, s5, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
+; GFX8-NEXT:    s_and_b32 s7, 0xffff, s7
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s5, 0xffff, s5
 ; GFX8-NEXT:    s_lshl_b32 s1, s1, s7
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, s5
 ; GFX8-NEXT:    s_or_b32 s1, s1, s3
 ; GFX8-NEXT:    s_and_b32 s3, s6, 15
-; GFX8-NEXT:    s_lshl_b32 s4, s4, s8
-; GFX8-NEXT:    s_bfe_u32 s3, s3, 0x100000
+; GFX8-NEXT:    s_lshl_b32 s4, s4, 1
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s3
 ; GFX8-NEXT:    s_andn2_b32 s5, 15, s6
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, s3
-; GFX8-NEXT:    s_bfe_u32 s3, s4, 0x100000
-; GFX8-NEXT:    s_lshr_b32 s3, s3, s8
-; GFX8-NEXT:    s_bfe_u32 s4, s5, 0x100000
+; GFX8-NEXT:    s_and_b32 s3, 0xffff, s4
+; GFX8-NEXT:    s_lshr_b32 s3, s3, 1
+; GFX8-NEXT:    s_and_b32 s4, 0xffff, s5
 ; GFX8-NEXT:    s_lshr_b32 s3, s3, s4
 ; GFX8-NEXT:    s_or_b32 s2, s2, s3
-; GFX8-NEXT:    s_bfe_u32 s2, s2, 0x100000
-; GFX8-NEXT:    s_bfe_u32 s1, s1, 0x100000
+; GFX8-NEXT:    s_and_b32 s2, 0xffff, s2
+; GFX8-NEXT:    s_and_b32 s1, 0xffff, s1
 ; GFX8-NEXT:    s_lshl_b32 s2, s2, 16
 ; GFX8-NEXT:    s_or_b32 s1, s1, s2
 ; GFX8-NEXT:    ; return to shader part epilog
 ;
 ; GFX9-LABEL: s_fshr_v4i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_lshr_b32 s7, s0, 16
 ; GFX9-NEXT:    s_lshl_b32 s0, s0, 0x10001

 ; GFX6:       ; %bb.0:
 ; GFX6-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v9, 16, v9
 ; GFX6-NEXT:    v_and_b32_e32 v8, 0xffff, v8
 ; GFX6-NEXT:    v_or_b32_e32 v8, v9, v8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v9, 16, v11
 ; GFX6-NEXT:    v_and_b32_e32 v10, 0xffff, v10
 ; GFX6-NEXT:    v_or_b32_e32 v9, v9, v10
-; GFX6-NEXT:    s_bfe_u32 s4, 1, 0x100000
 ; GFX6-NEXT:    v_bfe_u32 v10, v4, 1, 15
-; GFX6-NEXT:    s_bfe_u32 s5, 14, 0x100000
-; GFX6-NEXT:    v_lshlrev_b32_e32 v0, s4, v0
-; GFX6-NEXT:    v_lshrrev_b32_e32 v10, s5, v10
+; GFX6-NEXT:    v_lshlrev_b32_e32 v0, 1, v0
+; GFX6-NEXT:    v_lshrrev_b32_e32 v10, 14, v10
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v10
 ; GFX6-NEXT:    v_bfe_u32 v10, v5, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v1, s4, v1
-; GFX6-NEXT:    v_lshrrev_b32_e32 v10, s5, v10
+; GFX6-NEXT:    v_lshlrev_b32_e32 v1, 1, v1
+; GFX6-NEXT:    v_lshrrev_b32_e32 v10, 14, v10
 ; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v8
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v10
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v10, 16, v8
 ; GFX6-NEXT:    v_and_b32_e32 v11, 15, v8
 ; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, 1, v4
 ; GFX6-NEXT:    v_and_b32_e32 v8, 15, v8
-; GFX6-NEXT:    v_bfe_u32 v11, v11, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v11, 0xffff, v11
 ; GFX6-NEXT:    v_bfe_u32 v4, v4, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v8, v8, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v8, 0xffff, v8
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v0, v11, v0
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v8, v4
 ; GFX6-NEXT:    v_or_b32_e32 v0, v0, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v10
 ; GFX6-NEXT:    v_xor_b32_e32 v8, -1, v10
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v5, 1, v5
 ; GFX6-NEXT:    v_and_b32_e32 v8, 15, v8
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v1, v4, v1
 ; GFX6-NEXT:    v_bfe_u32 v4, v5, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v5, v8, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v8
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v5, v4
 ; GFX6-NEXT:    v_or_b32_e32 v1, v1, v4
 ; GFX6-NEXT:    v_bfe_u32 v4, v6, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v2, s4, v2
-; GFX6-NEXT:    v_lshrrev_b32_e32 v4, s5, v4
+; GFX6-NEXT:    v_lshlrev_b32_e32 v2, 1, v2
+; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 14, v4
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v4
 ; GFX6-NEXT:    v_bfe_u32 v4, v7, 1, 15
-; GFX6-NEXT:    v_lshlrev_b32_e32 v3, s4, v3
-; GFX6-NEXT:    v_lshrrev_b32_e32 v4, s5, v4
+; GFX6-NEXT:    v_lshlrev_b32_e32 v3, 1, v3
+; GFX6-NEXT:    v_lshrrev_b32_e32 v4, 14, v4
 ; GFX6-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v4, 1, v6
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v9
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v5, 1, v7
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v7, 16, v6
 ; GFX6-NEXT:    v_and_b32_e32 v8, 15, v6
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v6
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v8, v8, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v8, 0xffff, v8
 ; GFX6-NEXT:    v_bfe_u32 v4, v4, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v6, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v2, v8, v2
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v6, v4
 ; GFX6-NEXT:    v_or_b32_e32 v2, v2, v4
 ; GFX6-NEXT:    v_and_b32_e32 v4, 15, v7
 ; GFX6-NEXT:    v_xor_b32_e32 v6, -1, v7
 ; GFX6-NEXT:    v_and_b32_e32 v6, 15, v6
-; GFX6-NEXT:    v_bfe_u32 v4, v4, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX6-NEXT:    v_lshlrev_b32_e32 v3, v4, v3
 ; GFX6-NEXT:    v_bfe_u32 v4, v5, 1, 15
-; GFX6-NEXT:    v_bfe_u32 v5, v6, 0, 16
+; GFX6-NEXT:    v_and_b32_e32 v5, 0xffff, v6
 ; GFX6-NEXT:    v_lshrrev_b32_e32 v4, v5, v4
 ; GFX6-NEXT:    v_or_b32_e32 v3, v3, v4
 ; GFX6-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX8-LABEL: v_fshr_v4i16:
 ; GFX8:       ; %bb.0:
 ; GFX8-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v6, 1, v0

 ; GFX8-NEXT:    v_or_b32_e32 v4, v6, v4
 ; GFX8-NEXT:    v_and_b32_e32 v6, 15, v7
 ; GFX8-NEXT:    v_xor_b32_e32 v7, -1, v7
 ; GFX8-NEXT:    v_and_b32_e32 v7, 15, v7
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, 1, v2
 ; GFX8-NEXT:    v_lshlrev_b16_e32 v0, v6, v0
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v2, v7, v2
 ; GFX8-NEXT:    v_or_b32_e32 v0, v0, v2
-; GFX8-NEXT:    v_mov_b32_e32 v2, 16
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v0, v2, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
+; GFX8-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX8-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
 ; GFX8-NEXT:    v_or_b32_sdwa v0, v4, v0 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
-; GFX8-NEXT:    v_lshlrev_b16_e32 v4, 1, v1
-; GFX8-NEXT:    v_lshrrev_b16_e32 v6, 15, v3
-; GFX8-NEXT:    v_or_b32_e32 v4, v4, v6
-; GFX8-NEXT:    v_mov_b32_e32 v6, 1
-; GFX8-NEXT:    v_lshlrev_b16_sdwa v1, v6, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT:    v_lshrrev_b16_sdwa v7, v8, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v5
-; GFX8-NEXT:    v_or_b32_e32 v1, v1, v7
-; GFX8-NEXT:    v_lshlrev_b16_e32 v7, 1, v3
-; GFX8-NEXT:    v_lshlrev_b16_sdwa v3, v6, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
-; GFX8-NEXT:    v_lshrrev_b32_e32 v6, 16, v5
-; GFX8-NEXT:    v_and_b32_e32 v8, 15, v5
+; GFX8-NEXT:    v_lshlrev_b16_e32 v2, 1, v1
+; GFX8-NEXT:    v_lshrrev_b16_e32 v4, 15, v3
+; GFX8-NEXT:    v_or_b32_e32 v2, v2, v4
+; GFX8-NEXT:    v_mov_b32_e32 v4, 1
+; GFX8-NEXT:    v_lshlrev_b16_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT:    v_lshrrev_b16_sdwa v6, v8, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT:    v_or_b32_e32 v1, v1, v6
+; GFX8-NEXT:    v_lshlrev_b16_e32 v6, 1, v3
+; GFX8-NEXT:    v_lshlrev_b16_sdwa v3, v4, v3 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_1
+; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v5
+; GFX8-NEXT:    v_lshrrev_b32_e32 v5, 16, v4
+; GFX8-NEXT:    v_and_b32_e32 v7, 15, v4
+; GFX8-NEXT:    v_xor_b32_e32 v4, -1, v4
+; GFX8-NEXT:    v_and_b32_e32 v4, 15, v4
+; GFX8-NEXT:    v_lshrrev_b16_e32 v6, 1, v6
+; GFX8-NEXT:    v_lshlrev_b16_e32 v2, v7, v2
+; GFX8-NEXT:    v_lshrrev_b16_e32 v4, v4, v6
+; GFX8-NEXT:    v_or_b32_e32 v2, v2, v4
+; GFX8-NEXT:    v_and_b32_e32 v4, 15, v5
 ; GFX8-NEXT:    v_xor_b32_e32 v5, -1, v5
 ; GFX8-NEXT:    v_and_b32_e32 v5, 15, v5
-; GFX8-NEXT:    v_lshrrev_b16_e32 v7, 1, v7
-; GFX8-NEXT:    v_lshlrev_b16_e32 v4, v8, v4
-; GFX8-NEXT:    v_lshrrev_b16_e32 v5, v5, v7
-; GFX8-NEXT:    v_or_b32_e32 v4, v4, v5
-; GFX8-NEXT:    v_and_b32_e32 v5, 15, v6
-; GFX8-NEXT:    v_xor_b32_e32 v6, -1, v6
-; GFX8-NEXT:    v_and_b32_e32 v6, 15, v6
 ; GFX8-NEXT:    v_lshrrev_b16_e32 v3, 1, v3
-; GFX8-NEXT:    v_lshlrev_b16_e32 v1, v5, v1
-; GFX8-NEXT:    v_lshrrev_b16_e32 v3, v6, v3
+; GFX8-NEXT:    v_lshlrev_b16_e32 v1, v4, v1
+; GFX8-NEXT:    v_lshrrev_b16_e32 v3, v5, v3
 ; GFX8-NEXT:    v_or_b32_e32 v1, v1, v3
-; GFX8-NEXT:    v_lshlrev_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:DWORD src1_sel:WORD_0
-; GFX8-NEXT:    v_or_b32_sdwa v1, v4, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
+; GFX8-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX8-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
+; GFX8-NEXT:    v_or_b32_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:DWORD
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX9-LABEL: v_fshr_v4i16:
 ; GFX9:       ; %bb.0:
 ; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
 ; GFX9-NEXT:    v_and_b32_e32 v6, 0xf000f, v4
 ; GFX9-NEXT:    v_xor_b32_e32 v4, -1, v4
 ; GFX9-NEXT:    v_and_b32_e32 v4, 0xf000f, v4

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.sample.g16.a16.dim.ll

 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v4, v5, v4, 0x5040100
 ; GFX10-NEXT:    image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_d_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v5, v4, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v4, 0xffff, v4
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v5, 16, v4
 ; GFX10GISEL-NEXT:    image_sample_d v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v14, v7, v6, 0x5040100
 ; GFX10-NEXT:    image_sample_d v[0:3], v[8:15], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_d_3d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v7
+; GFX10GISEL-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v7, v8
-; GFX10GISEL-NEXT:    v_perm_b32 v6, v9, v6, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v6, v9, 16, v6
 ; GFX10GISEL-NEXT:    image_sample_d v[0:3], v[0:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %drdh, float %dsdv, float %dtdv, float %drdv, half %s, half %t, half %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v5, v6, v5, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_d_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v6, v5, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v5, 0xffff, v5
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v6, 16, v5
 ; GFX10GISEL-NEXT:    image_sample_c_d v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dsdv, half %s, half %clamp) {
 ; GFX10-LABEL: sample_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    image_sample_d_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_d_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v2, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_d_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.1d.v4f32.f32.f16(i32 15, float %dsdh, float %dsdv, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_d_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp) {
 ; GFX10-LABEL: sample_d_cl_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v11, v6
 ; GFX10-NEXT:    v_mov_b32_e32 v9, v3
 ; GFX10-NEXT:    v_mov_b32_e32 v8, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v7, v1
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v0
 ; GFX10-NEXT:    v_perm_b32 v10, v5, v4, 0x5040100
 ; GFX10-NEXT:    image_sample_d_cl v[0:3], v[6:11], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_d_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v7, v5
+; GFX10GISEL-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v5, v6
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v7, v4, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v7, 16, v4
 ; GFX10GISEL-NEXT:    image_sample_d_cl v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_d_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, half %s, half %clamp) {
 ; GFX10-LABEL: sample_c_d_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_d_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v3, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v3, v4, 16, v3
 ; GFX10GISEL-NEXT:    image_sample_c_d_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.1d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dsdv, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v12, v6, v5, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_cl v[0:3], v[7:13], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_d_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v8, v6
+; GFX10GISEL-NEXT:    v_and_b32_e32 v5, 0xffff, v5
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v6, v7
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v8, v5, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v8, 16, v5
 ; GFX10GISEL-NEXT:    image_sample_c_d_cl v[0:3], v[0:6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v4, v5, v4, 0x5040100
 ; GFX10-NEXT:    image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_cd_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v5, v4, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v4, 0xffff, v4
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v5, 16, v4
 ; GFX10GISEL-NEXT:    image_sample_cd v[0:3], v[0:4], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v5, v6, v5, 0x5040100
 ; GFX10-NEXT:    image_sample_c_cd v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_cd_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v6, v5, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v5, 0xffff, v5
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v6, 16, v5
 ; GFX10GISEL-NEXT:    image_sample_c_cd v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dsdv, half %s, half %clamp) {
 ; GFX10-LABEL: sample_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    image_sample_cd_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_cd_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v2, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_cd_cl v[0:3], v[0:2], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.1d.v4f32.f32.f16(i32 15, float %dsdh, float %dsdv, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_cd_cl_2d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp) {
 ; GFX10-LABEL: sample_cd_cl_2d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v11, v6
 ; GFX10-NEXT:    v_mov_b32_e32 v9, v3
 ; GFX10-NEXT:    v_mov_b32_e32 v8, v2
 ; GFX10-NEXT:    v_mov_b32_e32 v7, v1
 ; GFX10-NEXT:    v_mov_b32_e32 v6, v0
 ; GFX10-NEXT:    v_perm_b32 v10, v5, v4, 0x5040100
 ; GFX10-NEXT:    image_sample_cd_cl v[0:3], v[6:11], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_cd_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v7, v5
+; GFX10GISEL-NEXT:    v_and_b32_e32 v4, 0xffff, v4
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v5, v6
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v7, v4, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v7, 16, v4
 ; GFX10GISEL-NEXT:    image_sample_cd_cl v[0:3], v[0:5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f32.f16(i32 15, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_c_cd_cl_1d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, float %zcompare, float %dsdh, float %dsdv, half %s, half %clamp) {
 ; GFX10-LABEL: sample_c_cd_cl_1d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
 ; GFX10-NEXT:    image_sample_c_cd_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_cd_cl_1d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v3, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v3, v4, 16, v3
 ; GFX10GISEL-NEXT:    image_sample_c_cd_cl v[0:3], v[0:3], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_1D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.1d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dsdv, half %s, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v12, v6, v5, 0x5040100
 ; GFX10-NEXT:    image_sample_c_cd_cl v[0:3], v[7:13], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_cd_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v8, v6
+; GFX10GISEL-NEXT:    v_and_b32_e32 v5, 0xffff, v5
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v6, v7
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v8, v5, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v8, 16, v5
 ; GFX10GISEL-NEXT:    image_sample_c_cd_cl v[0:3], v[0:6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f32.f16(i32 15, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v14, v7, v6, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_o v0, v[8:15], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_d_o_2darray_V1:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v7
+; GFX10GISEL-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v7, v8
-; GFX10GISEL-NEXT:    v_perm_b32 v6, v9, v6, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v6, v9, 16, v6
 ; GFX10GISEL-NEXT:    image_sample_c_d_o v0, v[0:7], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f32.f16(i32 4, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret float %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v14, v7, v6, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_o v[0:1], v[8:15], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY a16
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_c_d_o_2darray_V2:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v7
+; GFX10GISEL-NEXT:    v_and_b32_e32 v6, 0xffff, v6
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v7, v8
-; GFX10GISEL-NEXT:    v_perm_b32 v6, v9, v6, 0x5040100
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v6, v9, 16, v6
 ; GFX10GISEL-NEXT:    image_sample_c_d_o v[0:1], v[0:7], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY a16
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f32.f16(i32 6, i32 %offset, float %zcompare, float %dsdh, float %dtdh, float %dsdv, float %dtdv, half %s, half %t, half %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <2 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
 ; GFX10-NEXT:    image_sample_d_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_d_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_d_g16 v[0:3], [v0, v1, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 
 define amdgpu_ps <4 x float> @sample_g16_noa16_d_3d(<8 x i32> inreg %rsrc, <4 x i32> inreg %samp, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r) {
 ; GFX10-LABEL: sample_g16_noa16_d_3d:
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    v_mov_b32_e32 v9, v3
 ; GFX10-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX10-NEXT:    v_perm_b32 v2, v1, v0, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v4, v4, v9, 0x5040100
 ; GFX10-NEXT:    image_sample_d_g16 v[0:3], v[2:8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_d_3d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v3
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v0
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v3, v2
-; GFX10GISEL-NEXT:    v_perm_b32 v2, v1, v0, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v4, v9, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v9, 0xffff, v9
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v2, v1, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v4, 16, v9
 ; GFX10GISEL-NEXT:    image_sample_d_g16 v[0:3], v[2:8], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_3D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.3d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %drdh, half %dsdv, half %dtdv, half %drdv, float %s, float %t, float %r, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_d_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v2, v4, v3, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10GISEL-NEXT:    v_and_b32_e32 v3, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v2, 16, v1
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v2, v4, 16, v3
 ; GFX10GISEL-NEXT:    image_sample_c_d_g16 v[0:3], [v0, v1, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
 ; GFX10-NEXT:    image_sample_d_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_d_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_d_cl_g16 v[0:3], [v0, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.d.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v3, v8, v1, 0x5040100
 ; GFX10-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[2:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_d_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v8, v2
-; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v3
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v2, v0
-; GFX10GISEL-NEXT:    v_perm_b32 v3, v8, v1, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v4, v9, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v1
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v3, v8, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v4, 16, v1
 ; GFX10GISEL-NEXT:    image_sample_c_d_cl_g16 v[0:3], v[2:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.d.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
 ; GFX10-NEXT:    image_sample_cd_g16 v[0:3], [v0, v2, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_cd_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_cd_g16 v[0:3], [v0, v1, v4, v5], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v3, v4, v3, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
 ; GFX10-NEXT:    image_sample_c_cd_g16 v[0:3], [v0, v1, v3, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_cd_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v2, v1, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v2, v4, v3, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v1
+; GFX10GISEL-NEXT:    v_and_b32_e32 v3, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v2, 16, v1
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v2, v4, 16, v3
 ; GFX10GISEL-NEXT:    image_sample_c_cd_g16 v[0:3], [v0, v1, v2, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v2, v3, v2, 0x5040100
 ; GFX10-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
 ; GFX10-NEXT:    image_sample_cd_cl_g16 v[0:3], [v0, v2, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_cd_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
-; GFX10GISEL-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v1, v3, v2, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v0
+; GFX10GISEL-NEXT:    v_and_b32_e32 v2, 0xffff, v2
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v0, v1, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v1, v3, 16, v2
 ; GFX10GISEL-NEXT:    image_sample_cd_cl_g16 v[0:3], [v0, v1, v4, v5, v6], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.cd.cl.2d.v4f32.f16.f32(i32 15, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    v_perm_b32 v3, v8, v1, 0x5040100
 ; GFX10-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[2:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_cd_cl_2d:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v8, v2
-; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v3
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v2, v0
-; GFX10GISEL-NEXT:    v_perm_b32 v3, v8, v1, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v4, v9, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v1
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v3
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v3, v8, 16, v0
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v4, 16, v1
 ; GFX10GISEL-NEXT:    image_sample_c_cd_cl_g16 v[0:3], v[2:7], s[0:7], s[8:11] dmask:0xf dim:SQ_RSRC_IMG_2D
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <4 x float> @llvm.amdgcn.image.sample.c.cd.cl.2d.v4f32.f16.f32(i32 15, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %clamp, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <4 x float> %v
 }
 

 ; GFX10-NEXT:    image_sample_c_d_o_g16 v0, v[2:8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_d_o_2darray_V1:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v2
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v10, v3
-; GFX10GISEL-NEXT:    v_mov_b32_e32 v11, v4
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v3, v1
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v10, v9, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v5, v11, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v4
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v9
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v5, 16, v1
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v10, 16, v0
 ; GFX10GISEL-NEXT:    image_sample_c_d_o_g16 v0, v[2:8], s[0:7], s[8:11] dmask:0x4 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call float @llvm.amdgcn.image.sample.c.d.o.2darray.f32.f16.f32(i32 4, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret float %v
 }
 

 ; GFX10-NEXT:    image_sample_c_d_o_g16 v[0:1], v[2:8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    ; return to shader part epilog
 ;
 ; GFX10GISEL-LABEL: sample_g16_noa16_c_d_o_2darray_V2:
 ; GFX10GISEL:       ; %bb.0: ; %main_body
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v9, v2
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v10, v3
-; GFX10GISEL-NEXT:    v_mov_b32_e32 v11, v4
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v2, v0
 ; GFX10GISEL-NEXT:    v_mov_b32_e32 v3, v1
-; GFX10GISEL-NEXT:    v_perm_b32 v4, v10, v9, 0x5040100
-; GFX10GISEL-NEXT:    v_perm_b32 v5, v5, v11, 0x5040100
+; GFX10GISEL-NEXT:    v_and_b32_e32 v1, 0xffff, v4
+; GFX10GISEL-NEXT:    v_and_b32_e32 v0, 0xffff, v9
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v5, v5, 16, v1
+; GFX10GISEL-NEXT:    v_lshl_or_b32 v4, v10, 16, v0
 ; GFX10GISEL-NEXT:    image_sample_c_d_o_g16 v[0:1], v[2:8], s[0:7], s[8:11] dmask:0x6 dim:SQ_RSRC_IMG_2D_ARRAY
 ; GFX10GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10GISEL-NEXT:    ; return to shader part epilog
 main_body:
   %v = call <2 x float> @llvm.amdgcn.image.sample.c.d.o.2darray.v2f32.f16.f32(i32 6, i32 %offset, float %zcompare, half %dsdh, half %dtdh, half %dsdv, half %dtdv, float %s, float %t, float %slice, <8 x i32> %rsrc, <4 x i32> %samp, i1 0, i32 0, i32 0)
   ret <2 x float> %v
 }
 

llvm/test/TableGen/GlobalISelEmitter-notype-output-pattern.td

 // RUN: llvm-tblgen -gen-global-isel -I %p/../../include -I %p/Common %s | FileCheck %s
 
 include "llvm/Target/Target.td"
 include "GlobalISelEmitterCommon.td"
 
-// No rule will be added to the match table.
 // CHECK:      constexpr static int64_t MatchTable0[] = {
-// CHECK-NEXT:   GIM_Reject
-// CHECK-NEXT: };
+// CHECK-NEXT:   GIM_Try,
+// CHECK-NEXT:     GIM_CheckOpcode{{.*}}TargetOpcode::G_ANYEXT,
+// CHECK-NEXT:     GIM_CheckType{{.*}}/*Type*/GILLT_s32,
+// CHECK-NEXT:     GIM_CheckType{{.*}}/*Type*/GILLT_s8,
+// CHECK-NEXT:     GIM_CheckRegBankForClass{{.*}}/*RC*/MyTarget::GPR32RegClassID,
+// CHECK-NEXT:     // (anyext:{{.*}}=>{{.*}}(SELECT_I4:
+// CHECK:          GIR_Done,
+// CHECK-NEXT:   // Label 0:
+// CHECK-NEXT:   GIM_Reject,
+// CHECK-NEXT:   };
 
 def SELECT_I4 : I<(outs GPR32:$dst), (ins GPR8:$cond, GPR32:$T, GPR32:$F), []>;
 def LI : I<(outs GPR32:$dst), (ins i32imm:$src), []>;
 
 def : Pat<(i32 (anyext i8:$in)), (SELECT_I4 $in, (LI 1), (LI 0))>;

llvm/utils/TableGen/GlobalISelEmitter.cpp

Error importChildMatcher(RuleMatcher &Rule, InstructionMatcher &InsnMatcher,
                            bool OperandIsAPointer, bool OperandIsImmArg,
                            unsigned OpIdx, unsigned &TempOpIdx);
 
   Expected<BuildMIAction &> createAndImportInstructionRenderer(
       RuleMatcher &M, InstructionMatcher &InsnMatcher,
       const TreePatternNode *Src, const TreePatternNode *Dst);
   Expected<action_iterator> createAndImportSubInstructionRenderer(
       action_iterator InsertPt, RuleMatcher &M, const TreePatternNode *Dst,
-      unsigned TempReg);
+      const TreePatternNode *Src, unsigned TempReg);
   Expected<action_iterator>
   createInstructionRenderer(action_iterator InsertPt, RuleMatcher &M,
                             const TreePatternNode *Dst);
 
   Expected<action_iterator> importExplicitDefRenderers(
       action_iterator InsertPt, RuleMatcher &M, BuildMIAction &DstMIBuilder,
       const TreePatternNode *Src, const TreePatternNode *Dst);
 
-  Expected<action_iterator>
-  importExplicitUseRenderers(action_iterator InsertPt, RuleMatcher &M,
-                             BuildMIAction &DstMIBuilder,
-                             const llvm::TreePatternNode *Dst);
-  Expected<action_iterator>
-  importExplicitUseRenderer(action_iterator InsertPt, RuleMatcher &Rule,
-                            BuildMIAction &DstMIBuilder,
-                            TreePatternNode *DstChild);
+  Expected<action_iterator> importExplicitUseRenderers(
+      action_iterator InsertPt, RuleMatcher &M, BuildMIAction &DstMIBuilder,
+      const llvm::TreePatternNode *Dst, const TreePatternNode *Src);
+  Expected<action_iterator> importExplicitUseRenderer(
+      action_iterator InsertPt, RuleMatcher &Rule, BuildMIAction &DstMIBuilder,
+      TreePatternNode *DstChild, const TreePatternNode *Src);
   Error importDefaultOperandRenderers(action_iterator InsertPt, RuleMatcher &M,
                                       BuildMIAction &DstMIBuilder,
                                       DagInit *DefaultOps) const;
   Error
   importImplicitDefRenderers(BuildMIAction &DstMIBuilder,
                              const std::vector<Record *> &ImplicitDefs) const;
 
   void emitCxxPredicateFns(raw_ostream &OS, StringRef CodeFieldName,
return failedImport(
         "Src pattern child def is an unsupported tablegen class");
   }
 
   return failedImport("Src pattern child is an unsupported kind");
 }
 
 Expected<action_iterator> GlobalISelEmitter::importExplicitUseRenderer(
     action_iterator InsertPt, RuleMatcher &Rule, BuildMIAction &DstMIBuilder,
-    TreePatternNode *DstChild) {
+    TreePatternNode *DstChild, const TreePatternNode *Src) {
 
   const auto &SubOperand = Rule.getComplexSubOperand(DstChild->getName());
   if (SubOperand) {
     DstMIBuilder.addRenderer<RenderComplexPatternOperand>(
         *std::get<0>(*SubOperand), DstChild->getName(),
         std::get<1>(*SubOperand), std::get<2>(*SubOperand));
     return InsertPt;
   }
if (DstChild->getOperator()->isSubClassOf("Instruction")) {
         return OpTy.takeError();
 
       unsigned TempRegID = Rule.allocateTempRegID();
       InsertPt = Rule.insertAction<MakeTempRegisterAction>(
           InsertPt, *OpTy, TempRegID);
       DstMIBuilder.addRenderer<TempRegRenderer>(TempRegID);
 
       auto InsertPtOrError = createAndImportSubInstructionRenderer(
-          ++InsertPt, Rule, DstChild, TempRegID);
+          ++InsertPt, Rule, DstChild, Src, TempRegID);
       if (auto Error = InsertPtOrError.takeError())
         return std::move(Error);
       return InsertPtOrError.get();
     }
 
     return failedImport("Dst pattern child isn't a leaf node or an MBB" + llvm::to_string(*DstChild));
   }
 
if (ChildRec->isSubClassOf("ComplexPattern")) {
           *ComplexPattern->second, DstChild->getName(),
           OM.getAllocatedTemporariesBaseID());
       return InsertPt;
     }
 
     return failedImport(
         "Dst pattern child def is an unsupported tablegen class");
   }
+
+  // Handle the case where the MVT/register class is omitted in the dest pattern
+  // but MVT exists in the source pattern.
+  if (isa<UnsetInit>(DstChild->getLeafValue())) {
+    for (unsigned NumOp = 0; NumOp < Src->getNumChildren(); NumOp++)
+      if (Src->getChild(NumOp)->getName() == DstChild->getName()) {
+        DstMIBuilder.addRenderer<CopyRenderer>(Src->getChild(NumOp)->getName());
+        return InsertPt;
+      }
+  }
   return failedImport("Dst pattern child is an unsupported kind");
 }
 
 Expected<BuildMIAction &> GlobalISelEmitter::createAndImportInstructionRenderer(
     RuleMatcher &M, InstructionMatcher &InsnMatcher, const TreePatternNode *Src,
     const TreePatternNode *Dst) {
   auto InsertPtOrError = createInstructionRenderer(M.actions_end(), M, Dst);
   if (auto Error = InsertPtOrError.takeError())
for (auto PhysInput : InsnMatcher.getPhysRegInputs()) {
     CopyToPhysRegMIBuilder.addRenderer<CopyPhysRegRenderer>(PhysInput.first);
   }
 
   if (auto Error =
           importExplicitDefRenderers(InsertPt, M, DstMIBuilder, Src, Dst)
               .takeError())
     return std::move(Error);
 
-  if (auto Error = importExplicitUseRenderers(InsertPt, M, DstMIBuilder, Dst)
-                       .takeError())
+  if (auto Error =
+          importExplicitUseRenderers(InsertPt, M, DstMIBuilder, Dst, Src)
     return std::move(Error);
 
   return DstMIBuilder;
 }
 
 Expected<action_iterator>
 GlobalISelEmitter::createAndImportSubInstructionRenderer(
     const action_iterator InsertPt, RuleMatcher &M, const TreePatternNode *Dst,
-    unsigned TempRegID) {
+    const TreePatternNode *Src, unsigned TempRegID) {
   auto InsertPtOrError = createInstructionRenderer(InsertPt, M, Dst);
 
   // TODO: Assert there's exactly one result.
 
   if (auto Error = InsertPtOrError.takeError())
     return std::move(Error);
 
   BuildMIAction &DstMIBuilder =
       *static_cast<BuildMIAction *>(InsertPtOrError.get()->get());
 
   // Assign the result to TempReg.
   DstMIBuilder.addRenderer<TempRegRenderer>(TempRegID, true);
 
-  InsertPtOrError =
-      importExplicitUseRenderers(InsertPtOrError.get(), M, DstMIBuilder, Dst);
+  InsertPtOrError = importExplicitUseRenderers(InsertPtOrError.get(), M,
+                                               DstMIBuilder, Dst, Src);
   if (auto Error = InsertPtOrError.takeError())
     return std::move(Error);
 
   // We need to make sure that when we import an INSERT_SUBREG as a
   // subinstruction that it ends up being constrained to the correct super
   // register and subregister classes.
   auto OpName = Target.getInstruction(Dst->getOperator()).TheDef->getName();
   if (OpName == "INSERT_SUBREG") {
for (unsigned I = SrcNumDefs; I < DstNumDefs; ++I) {
     DstMIBuilder.addRenderer<TempRegRenderer>(TempRegID, true, nullptr, true);
   }
 
   return InsertPt;
 }
 
 Expected<action_iterator> GlobalISelEmitter::importExplicitUseRenderers(
     action_iterator InsertPt, RuleMatcher &M, BuildMIAction &DstMIBuilder,
-    const llvm::TreePatternNode *Dst) {
+    const llvm::TreePatternNode *Dst, const llvm::TreePatternNode *Src) {
   const CodeGenInstruction *DstI = DstMIBuilder.getCGI();
   CodeGenInstruction *OrigDstI = &Target.getInstruction(Dst->getOperator());
 
   StringRef Name = OrigDstI->TheDef->getName();
   unsigned ExpectedDstINumUses = Dst->getNumChildren();
 
   // EXTRACT_SUBREG needs to use a subregister COPY.
   if (Name == "EXTRACT_SUBREG") {
if (!ValChild->isLeaf()) {
       if (!ExtractSrcTy)
         return ExtractSrcTy.takeError();
 
       unsigned TempRegID = M.allocateTempRegID();
       InsertPt = M.insertAction<MakeTempRegisterAction>(
         InsertPt, *ExtractSrcTy, TempRegID);
 
       auto InsertPtOrError = createAndImportSubInstructionRenderer(
-        ++InsertPt, M, ValChild, TempRegID);
+        ++InsertPt, M, ValChild, Src, TempRegID);
       if (auto Error = InsertPtOrError.takeError())
         return std::move(Error);
 
       DstMIBuilder.addRenderer<TempRegRenderer>(TempRegID, false, SubIdx);
       return InsertPt;
     }
 
     // If this is a source operand, this is just a subregister copy.
for (unsigned I = 1; I != ExpectedDstINumUses; I += 2) {
       TreePatternNode *ValChild = Dst->getChild(I);
       TreePatternNode *SubRegChild = Dst->getChild(I + 1);
 
       if (DefInit *SubRegInit =
               dyn_cast<DefInit>(SubRegChild->getLeafValue())) {
         CodeGenSubRegIndex *SubIdx = CGRegs.getSubRegIdx(SubRegInit->getDef());
 
         auto InsertPtOrError =
-            importExplicitUseRenderer(InsertPt, M, DstMIBuilder, ValChild);
+            importExplicitUseRenderer(InsertPt, M, DstMIBuilder, ValChild, Src);
         if (auto Error = InsertPtOrError.takeError())
           return std::move(Error);
         InsertPt = InsertPtOrError.get();
         DstMIBuilder.addRenderer<SubRegIndexRenderer>(SubIdx);
       }
     }
 
     return InsertPt;
if (CGP.operandHasDefault(OperandNode) &&
       if (auto Error = importDefaultOperandRenderers(
             InsertPt, M, DstMIBuilder, DefaultOps))
         return std::move(Error);
       ++NumDefaultOps;
       continue;
     }
 
     auto InsertPtOrError = importExplicitUseRenderer(InsertPt, M, DstMIBuilder,
-                                                     Dst->getChild(Child));
+                                                     Dst->getChild(Child), Src);
     if (auto Error = InsertPtOrError.takeError())
       return std::move(Error);
     InsertPt = InsertPtOrError.get();
     ++Child;
   }
 
   if (NumDefaultOps + ExpectedDstINumUses != DstINumUses)
     return failedImport("Expected " + llvm::to_string(DstINumUses) +