X86 Tests: Add more sdiv combine cases. NFC

Add cases with vector non-splat pow2 contant divider.

llvm-svn: 323329

1 files changed, 3160 insertions, 9 deletions

diff --git a/llvm/test/CodeGen/X86/combine-sdiv.ll b/llvm/test/CodeGen/X86/combine-sdiv.ll
index b32a58c0ec6..9662958a9a9 100644
--- a/llvm/test/CodeGen/X86/combine-sdiv.ll
+++ b/llvm/test/CodeGen/X86/combine-sdiv.ll
@@ -1,7 +1,9 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
 ; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx  | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=CHECK,AVX,AVX2ORLATER,AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefixes=CHECK,AVX,AVX2ORLATER,AVX512,AVX512F
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f,+avx512bw,+avx512vl | FileCheck %s --check-prefixes=CHECK,AVX,AVX2ORLATER,AVX512,AVX512BW
 
 ; fold (sdiv undef, x) -> 0
 define i32 @combine_sdiv_undef0(i32 %x) {
@@ -195,11 +197,11 @@ define <4 x i32> @combine_vec_sdiv_by_pos1(<4 x i32> %x) {
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: combine_vec_sdiv_by_pos1:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm0
-; AVX2-NEXT:    retq
+; AVX2ORLATER-LABEL: combine_vec_sdiv_by_pos1:
+; AVX2ORLATER:       # %bb.0:
+; AVX2ORLATER-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX2ORLATER-NEXT:    vpsrlvd {{.*}}(%rip), %xmm0, %xmm0
+; AVX2ORLATER-NEXT:    retq
   %1 = and <4 x i32> %x, <i32 255, i32 255, i32 255, i32 255>
   %2 = sdiv <4 x i32> %1, <i32 1, i32 4, i32 8, i32 16>
   ret <4 x i32> %2
@@ -228,8 +230,2224 @@ define <4 x i32> @combine_vec_sdiv_by_pow2a(<4 x i32> %x) {
   ret <4 x i32> %1
 }
 
-define <4 x i32> @combine_vec_sdiv_by_pow2b(<4 x i32> %x) {
-; SSE-LABEL: combine_vec_sdiv_by_pow2b:
+define <16 x i8> @combine_vec_sdiv_by_pow2b_v16i8(<16 x i8> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrb $1, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $6, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $2, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pextrb $0, %xmm0, %ecx
+; SSE-NEXT:    movd %ecx, %xmm1
+; SSE-NEXT:    pinsrb $1, %eax, %xmm1
+; SSE-NEXT:    pextrb $2, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrb $7, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $2, %eax, %xmm1
+; SSE-NEXT:    pextrb $3, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $4, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $4, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $3, %eax, %xmm1
+; SSE-NEXT:    pextrb $4, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $5, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $3, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $4, %eax, %xmm1
+; SSE-NEXT:    pextrb $5, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $3, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $5, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $5, %eax, %xmm1
+; SSE-NEXT:    pextrb $6, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $2, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $6, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $6, %eax, %xmm1
+; SSE-NEXT:    pextrb $7, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrb $7, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $7, %eax, %xmm1
+; SSE-NEXT:    pextrb $8, %xmm0, %eax
+; SSE-NEXT:    pinsrb $8, %eax, %xmm1
+; SSE-NEXT:    pextrb $9, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $6, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $2, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $9, %eax, %xmm1
+; SSE-NEXT:    pextrb $10, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrb $7, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $10, %eax, %xmm1
+; SSE-NEXT:    pextrb $11, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $4, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $4, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $11, %eax, %xmm1
+; SSE-NEXT:    pextrb $12, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $5, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $3, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $12, %eax, %xmm1
+; SSE-NEXT:    pextrb $13, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $3, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $5, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $13, %eax, %xmm1
+; SSE-NEXT:    pextrb $14, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarb $7, %cl
+; SSE-NEXT:    shrb $2, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb $6, %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $14, %eax, %xmm1
+; SSE-NEXT:    pextrb $15, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrb $7, %cl
+; SSE-NEXT:    addb %al, %cl
+; SSE-NEXT:    sarb %cl
+; SSE-NEXT:    movzbl %cl, %eax
+; SSE-NEXT:    pinsrb $15, %eax, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_sdiv_by_pow2b_v16i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpextrb $1, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $6, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $2, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpextrb $0, %xmm0, %ecx
+; AVX-NEXT:    vmovd %ecx, %xmm1
+; AVX-NEXT:    vpinsrb $1, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $2, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    shrb $7, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $2, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $3, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $4, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $4, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $3, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $4, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $5, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $3, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $4, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $5, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $3, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $5, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $5, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $6, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $2, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $6, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $6, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $7, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    shrb $7, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $7, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $8, %xmm0, %eax
+; AVX-NEXT:    vpinsrb $8, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $9, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $6, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $2, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $9, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $10, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    shrb $7, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $10, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $11, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $4, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $4, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $11, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $12, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $5, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $3, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $12, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $13, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $3, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $5, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $13, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $14, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarb $7, %cl
+; AVX-NEXT:    shrb $2, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb $6, %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $14, %eax, %xmm1, %xmm1
+; AVX-NEXT:    vpextrb $15, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    shrb $7, %cl
+; AVX-NEXT:    addb %al, %cl
+; AVX-NEXT:    sarb %cl
+; AVX-NEXT:    movzbl %cl, %eax
+; AVX-NEXT:    vpinsrb $15, %eax, %xmm1, %xmm0
+; AVX-NEXT:    retq
+  %1 = sdiv <16 x i8> %x, <i8 1, i8 4, i8 2, i8 16, i8 8, i8 32, i8 64, i8 2, i8 1, i8 4, i8 2, i8 16, i8 8, i8 32, i8 64, i8 2>
+  ret <16 x i8> %1
+}
+
+define <8 x i16> @combine_vec_sdiv_by_pow2b_v8i16(<8 x i16> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm0
+; SSE-NEXT:    pextrw $2, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm0
+; SSE-NEXT:    pextrw $3, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm0
+; SSE-NEXT:    pextrw $4, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm0
+; SSE-NEXT:    pextrw $5, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm0
+; SSE-NEXT:    pextrw $6, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm0
+; SSE-NEXT:    pextrw $7, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX1-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vxorps %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX512F-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: combine_vec_sdiv_by_pow2b_v8i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX512BW-NEXT:    retq
+  %1 = sdiv <8 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
+  ret <8 x i16> %1
+}
+
+define <16 x i16> @combine_vec_sdiv_by_pow2b_v16i16(<16 x i16> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movdqa %xmm1, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm3
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm3[0,1],xmm0[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm0
+; SSE-NEXT:    pextrw $2, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm0
+; SSE-NEXT:    pextrw $3, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm0
+; SSE-NEXT:    pextrw $4, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm0
+; SSE-NEXT:    pextrw $5, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm0
+; SSE-NEXT:    pextrw $6, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm0
+; SSE-NEXT:    pextrw $7, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm2[0,1],xmm1[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm1
+; SSE-NEXT:    pextrw $2, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm1
+; SSE-NEXT:    pextrw $3, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm1
+; SSE-NEXT:    pextrw $4, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm1
+; SSE-NEXT:    pextrw $5, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm1
+; SSE-NEXT:    pextrw $6, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm1
+; SSE-NEXT:    pextrw $7, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm3, %xmm1
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm3 = xmm1[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm3, %xmm1
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm2 = xmm0[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX512F-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm3 = xmm1[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm3, %xmm3
+; AVX512F-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm3, %xmm1
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX512F-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: combine_vec_sdiv_by_pow2b_v16i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX512BW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm3 = xmm1[0],xmm2[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm3, %xmm1
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm2 = xmm0[0],xmm2[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX512BW-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm0
+; AVX512BW-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX512BW-NEXT:    retq
+  %1 = sdiv <16 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
+  ret <16 x i16> %1
+}
+
+define <32 x i16> @combine_vec_sdiv_by_pow2b_v32i16(<32 x i16> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    movdqa %xmm3, %xmm4
+; SSE-NEXT:    movdqa %xmm1, %xmm5
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm3, %xmm3
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm1[0,1],xmm0[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm0
+; SSE-NEXT:    pextrw $2, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm0
+; SSE-NEXT:    pextrw $3, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm0
+; SSE-NEXT:    pextrw $4, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm0
+; SSE-NEXT:    pextrw $5, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm0
+; SSE-NEXT:    pextrw $6, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm0
+; SSE-NEXT:    pextrw $7, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pblendw {{.*#+}} xmm1 = xmm5[0,1],xmm1[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm1
+; SSE-NEXT:    pextrw $2, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm1
+; SSE-NEXT:    pextrw $3, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm1
+; SSE-NEXT:    pextrw $4, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm1
+; SSE-NEXT:    pextrw $5, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm1
+; SSE-NEXT:    pextrw $6, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm1
+; SSE-NEXT:    pextrw $7, %xmm5, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm1
+; SSE-NEXT:    pxor %xmm5, %xmm5
+; SSE-NEXT:    pblendw {{.*#+}} xmm5 = xmm2[0,1],xmm5[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm5
+; SSE-NEXT:    pextrw $2, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm5
+; SSE-NEXT:    pextrw $3, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm5
+; SSE-NEXT:    pextrw $4, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm5
+; SSE-NEXT:    pextrw $5, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm5
+; SSE-NEXT:    pextrw $6, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm5
+; SSE-NEXT:    pextrw $7, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm5
+; SSE-NEXT:    pblendw {{.*#+}} xmm3 = xmm4[0,1],xmm3[2,3,4,5,6,7]
+; SSE-NEXT:    pextrw $1, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $14, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $2, %cx
+; SSE-NEXT:    pinsrw $1, %ecx, %xmm3
+; SSE-NEXT:    pextrw $2, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $2, %ecx, %xmm3
+; SSE-NEXT:    pextrw $3, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $12, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $4, %cx
+; SSE-NEXT:    pinsrw $3, %ecx, %xmm3
+; SSE-NEXT:    pextrw $4, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $13, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $3, %cx
+; SSE-NEXT:    pinsrw $4, %ecx, %xmm3
+; SSE-NEXT:    pextrw $5, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $11, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $5, %cx
+; SSE-NEXT:    pinsrw $5, %ecx, %xmm3
+; SSE-NEXT:    pextrw $6, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarw $15, %cx
+; SSE-NEXT:    movzwl %cx, %ecx
+; SSE-NEXT:    shrl $10, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw $6, %cx
+; SSE-NEXT:    pinsrw $6, %ecx, %xmm3
+; SSE-NEXT:    pextrw $7, %xmm4, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    shrl $15, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarw %cx
+; SSE-NEXT:    pinsrw $7, %ecx, %xmm3
+; SSE-NEXT:    movdqa %xmm5, %xmm2
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm3[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm3[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX1-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3,4,5,6,7]
+; AVX1-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $14, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $2, %cx
+; AVX1-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $12, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $4, %cx
+; AVX1-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $13, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $3, %cx
+; AVX1-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $11, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $5, %cx
+; AVX1-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarw $15, %cx
+; AVX1-NEXT:    movzwl %cx, %ecx
+; AVX1-NEXT:    shrl $10, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw $6, %cx
+; AVX1-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    shrl $15, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarw %cx
+; AVX1-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm3, %ymm1, %ymm1
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm3
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm4 = xmm3[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm4 = xmm0[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm0
+; AVX2-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm3
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm4 = xmm3[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX2-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX2-NEXT:    vpblendw {{.*#+}} xmm2 = xmm1[0,1],xmm2[2,3,4,5,6,7]
+; AVX2-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $14, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $2, %cx
+; AVX2-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $12, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $4, %cx
+; AVX2-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $13, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $3, %cx
+; AVX2-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $11, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $5, %cx
+; AVX2-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarw $15, %cx
+; AVX2-NEXT:    movzwl %cx, %ecx
+; AVX2-NEXT:    shrl $10, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw $6, %cx
+; AVX2-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX2-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    shrl $15, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarw %cx
+; AVX2-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX2-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
+; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX512F-NEXT:    vxorps %xmm2, %xmm2, %xmm2
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm4 = xmm3[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm4 = xmm0[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm0
+; AVX512F-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vextractf128 $1, %ymm1, %xmm3
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm4 = xmm3[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX512F-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX512F-NEXT:    vmovss {{.*#+}} xmm2 = xmm1[0],xmm2[1,2,3]
+; AVX512F-NEXT:    vpextrw $1, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $14, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $2, %cx
+; AVX512F-NEXT:    vpinsrw $1, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $2, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $2, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $3, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $12, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $4, %cx
+; AVX512F-NEXT:    vpinsrw $3, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $4, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $13, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $3, %cx
+; AVX512F-NEXT:    vpinsrw $4, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $5, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $11, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $5, %cx
+; AVX512F-NEXT:    vpinsrw $5, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $6, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    sarw $15, %cx
+; AVX512F-NEXT:    movzwl %cx, %ecx
+; AVX512F-NEXT:    shrl $10, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw $6, %cx
+; AVX512F-NEXT:    vpinsrw $6, %ecx, %xmm2, %xmm2
+; AVX512F-NEXT:    vpextrw $7, %xmm1, %eax
+; AVX512F-NEXT:    movl %eax, %ecx
+; AVX512F-NEXT:    shrl $15, %ecx
+; AVX512F-NEXT:    addl %eax, %ecx
+; AVX512F-NEXT:    sarw %cx
+; AVX512F-NEXT:    vpinsrw $7, %ecx, %xmm2, %xmm1
+; AVX512F-NEXT:    vinserti128 $1, %xmm3, %ymm1, %ymm1
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: combine_vec_sdiv_by_pow2b_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vextractf32x4 $3, %zmm0, %xmm2
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm3 = xmm2[0],xmm1[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $2, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $3, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $4, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $5, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $6, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm3, %xmm3
+; AVX512BW-NEXT:    vpextrw $7, %xmm2, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm3, %xmm2
+; AVX512BW-NEXT:    vextractf32x4 $2, %zmm0, %xmm3
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm4 = xmm3[0],xmm1[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX512BW-NEXT:    vinserti128 $1, %xmm2, %ymm3, %ymm2
+; AVX512BW-NEXT:    vextractf128 $1, %ymm0, %xmm3
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm4 = xmm3[0],xmm1[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $2, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $3, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $4, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $5, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $6, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm4, %xmm4
+; AVX512BW-NEXT:    vpextrw $7, %xmm3, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm4, %xmm3
+; AVX512BW-NEXT:    vmovss {{.*#+}} xmm1 = xmm0[0],xmm1[1,2,3]
+; AVX512BW-NEXT:    vpextrw $1, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $14, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $2, %cx
+; AVX512BW-NEXT:    vpinsrw $1, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $2, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $2, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $3, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $12, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $4, %cx
+; AVX512BW-NEXT:    vpinsrw $3, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $4, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $13, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $3, %cx
+; AVX512BW-NEXT:    vpinsrw $4, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $5, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $11, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $5, %cx
+; AVX512BW-NEXT:    vpinsrw $5, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $6, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    sarw $15, %cx
+; AVX512BW-NEXT:    movzwl %cx, %ecx
+; AVX512BW-NEXT:    shrl $10, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw $6, %cx
+; AVX512BW-NEXT:    vpinsrw $6, %ecx, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpextrw $7, %xmm0, %eax
+; AVX512BW-NEXT:    movl %eax, %ecx
+; AVX512BW-NEXT:    shrl $15, %ecx
+; AVX512BW-NEXT:    addl %eax, %ecx
+; AVX512BW-NEXT:    sarw %cx
+; AVX512BW-NEXT:    vpinsrw $7, %ecx, %xmm1, %xmm0
+; AVX512BW-NEXT:    vinserti128 $1, %xmm3, %ymm0, %ymm0
+; AVX512BW-NEXT:    vinserti64x4 $1, %ymm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    retq
+  %1 = sdiv <32 x i16> %x, <i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2, i16 1, i16 4, i16 2, i16 16, i16 8, i16 32, i16 64, i16 2>
+  ret <32 x i16> %1
+}
+
+define <4 x i32> @combine_vec_sdiv_by_pow2b_v4i32(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v4i32:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    pextrd $1, %xmm0, %eax
 ; SSE-NEXT:    movl %eax, %ecx
@@ -254,7 +2472,7 @@ define <4 x i32> @combine_vec_sdiv_by_pow2b(<4 x i32> %x) {
 ; SSE-NEXT:    pinsrd $3, %eax, %xmm0
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: combine_vec_sdiv_by_pow2b:
+; AVX-LABEL: combine_vec_sdiv_by_pow2b_v4i32:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vpextrd $1, %xmm0, %eax
 ; AVX-NEXT:    movl %eax, %ecx
@@ -281,3 +2499,936 @@ define <4 x i32> @combine_vec_sdiv_by_pow2b(<4 x i32> %x) {
   %1 = sdiv <4 x i32> %x, <i32 1, i32 4, i32 8, i32 16>
   ret <4 x i32> %1
 }
+
+define <8 x i32> @combine_vec_sdiv_by_pow2b_v8i32(<8 x i32> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v8i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrd $1, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm0, %eax
+; SSE-NEXT:    pextrd $3, %xmm0, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm0
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm0
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm0
+; SSE-NEXT:    pextrd $1, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm1, %eax
+; SSE-NEXT:    pextrd $3, %xmm1, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm1
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm1
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm1
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpextrd $1, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm2
+; AVX1-NEXT:    vpextrd $2, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrd $3, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX1-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm2
+; AVX1-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2ORLATER-LABEL: combine_vec_sdiv_by_pow2b_v8i32:
+; AVX2ORLATER:       # %bb.0:
+; AVX2ORLATER-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2ORLATER-NEXT:    vpextrd $1, %xmm1, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $30, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $2, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm2
+; AVX2ORLATER-NEXT:    vpextrd $2, %xmm1, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $29, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $3, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2ORLATER-NEXT:    vpextrd $3, %xmm1, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $28, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $4, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX2ORLATER-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $30, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $2, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm2
+; AVX2ORLATER-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $29, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $3, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX2ORLATER-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX2ORLATER-NEXT:    movl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $31, %ecx
+; AVX2ORLATER-NEXT:    shrl $28, %ecx
+; AVX2ORLATER-NEXT:    addl %eax, %ecx
+; AVX2ORLATER-NEXT:    sarl $4, %ecx
+; AVX2ORLATER-NEXT:    vpinsrd $3, %ecx, %xmm2, %xmm0
+; AVX2ORLATER-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2ORLATER-NEXT:    retq
+  %1 = sdiv <8 x i32> %x, <i32 1, i32 4, i32 8, i32 16, i32 1, i32 4, i32 8, i32 16>
+  ret <8 x i32> %1
+}
+
+define <16 x i32> @combine_vec_sdiv_by_pow2b_v16i32(<16 x i32> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v16i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrd $1, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm0, %eax
+; SSE-NEXT:    pextrd $3, %xmm0, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm0
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm0
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm0
+; SSE-NEXT:    pextrd $1, %xmm1, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm1, %eax
+; SSE-NEXT:    pextrd $3, %xmm1, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm1
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm1
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm1
+; SSE-NEXT:    pextrd $1, %xmm2, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm2, %eax
+; SSE-NEXT:    pextrd $3, %xmm2, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm2
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm2
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm2
+; SSE-NEXT:    pextrd $1, %xmm3, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    pextrd $2, %xmm3, %eax
+; SSE-NEXT:    pextrd $3, %xmm3, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm3
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm3
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm3
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v16i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX1-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX1-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm3
+; AVX1-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX1-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX1-NEXT:    vpextrd $1, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $30, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $2, %ecx
+; AVX1-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm3
+; AVX1-NEXT:    vpextrd $2, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $29, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $3, %ecx
+; AVX1-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX1-NEXT:    vpextrd $3, %xmm1, %eax
+; AVX1-NEXT:    movl %eax, %ecx
+; AVX1-NEXT:    sarl $31, %ecx
+; AVX1-NEXT:    shrl $28, %ecx
+; AVX1-NEXT:    addl %eax, %ecx
+; AVX1-NEXT:    sarl $4, %ecx
+; AVX1-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_vec_sdiv_by_pow2b_v16i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $30, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $2, %ecx
+; AVX2-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX2-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $29, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $3, %ecx
+; AVX2-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $28, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $4, %ecx
+; AVX2-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX2-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $30, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $2, %ecx
+; AVX2-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm3
+; AVX2-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $29, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $3, %ecx
+; AVX2-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $28, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $4, %ecx
+; AVX2-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm0
+; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $30, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $2, %ecx
+; AVX2-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX2-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $29, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $3, %ecx
+; AVX2-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $28, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $4, %ecx
+; AVX2-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX2-NEXT:    vpextrd $1, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $30, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $2, %ecx
+; AVX2-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm3
+; AVX2-NEXT:    vpextrd $2, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $29, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $3, %ecx
+; AVX2-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX2-NEXT:    vpextrd $3, %xmm1, %eax
+; AVX2-NEXT:    movl %eax, %ecx
+; AVX2-NEXT:    sarl $31, %ecx
+; AVX2-NEXT:    shrl $28, %ecx
+; AVX2-NEXT:    addl %eax, %ecx
+; AVX2-NEXT:    sarl $4, %ecx
+; AVX2-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm1
+; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: combine_vec_sdiv_by_pow2b_v16i32:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vextracti32x4 $3, %zmm0, %xmm1
+; AVX512-NEXT:    vpextrd $1, %xmm1, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $30, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $2, %ecx
+; AVX512-NEXT:    vpinsrd $1, %ecx, %xmm1, %xmm2
+; AVX512-NEXT:    vpextrd $2, %xmm1, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $29, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $3, %ecx
+; AVX512-NEXT:    vpinsrd $2, %ecx, %xmm2, %xmm2
+; AVX512-NEXT:    vpextrd $3, %xmm1, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $28, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $4, %ecx
+; AVX512-NEXT:    vpinsrd $3, %ecx, %xmm2, %xmm1
+; AVX512-NEXT:    vextracti32x4 $2, %zmm0, %xmm2
+; AVX512-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $30, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $2, %ecx
+; AVX512-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX512-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $29, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $3, %ecx
+; AVX512-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX512-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $28, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $4, %ecx
+; AVX512-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX512-NEXT:    vpextrd $1, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $30, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $2, %ecx
+; AVX512-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm3
+; AVX512-NEXT:    vpextrd $2, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $29, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $3, %ecx
+; AVX512-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX512-NEXT:    vpextrd $3, %xmm2, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $28, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $4, %ecx
+; AVX512-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm2
+; AVX512-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $30, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $2, %ecx
+; AVX512-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm3
+; AVX512-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $29, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $3, %ecx
+; AVX512-NEXT:    vpinsrd $2, %ecx, %xmm3, %xmm3
+; AVX512-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX512-NEXT:    movl %eax, %ecx
+; AVX512-NEXT:    sarl $31, %ecx
+; AVX512-NEXT:    shrl $28, %ecx
+; AVX512-NEXT:    addl %eax, %ecx
+; AVX512-NEXT:    sarl $4, %ecx
+; AVX512-NEXT:    vpinsrd $3, %ecx, %xmm3, %xmm0
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT:    retq
+  %1 = sdiv <16 x i32> %x, <i32 1, i32 4, i32 8, i32 16, i32 1, i32 4, i32 8, i32 16, i32 1, i32 4, i32 8, i32 16, i32 1, i32 4, i32 8, i32 16>
+  ret <16 x i32> %1
+}
+
+define <2 x i64> @combine_vec_sdiv_by_pow2b_v2i64(<2 x i64> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v2i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrq $1, %xmm0, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $62, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $2, %rcx
+; SSE-NEXT:    movq %rcx, %xmm1
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_sdiv_by_pow2b_v2i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX-NEXT:    movq %rax, %rcx
+; AVX-NEXT:    sarq $63, %rcx
+; AVX-NEXT:    shrq $62, %rcx
+; AVX-NEXT:    addq %rax, %rcx
+; AVX-NEXT:    sarq $2, %rcx
+; AVX-NEXT:    vmovq %rcx, %xmm1
+; AVX-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX-NEXT:    retq
+  %1 = sdiv <2 x i64> %x, <i64 1, i64 4>
+  ret <2 x i64> %1
+}
+
+define <4 x i64> @combine_vec_sdiv_by_pow2b_v4i64(<4 x i64> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v4i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrq $1, %xmm1, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $60, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $4, %rcx
+; SSE-NEXT:    movq %rcx, %xmm2
+; SSE-NEXT:    movq %xmm1, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $61, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $3, %rcx
+; SSE-NEXT:    movq %rcx, %xmm1
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; SSE-NEXT:    pextrq $1, %xmm0, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $62, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $2, %rcx
+; SSE-NEXT:    movq %rcx, %xmm2
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $60, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $4, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm2
+; AVX1-NEXT:    vmovq %xmm1, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $61, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $3, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm1
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $62, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $2, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm2
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2ORLATER-LABEL: combine_vec_sdiv_by_pow2b_v4i64:
+; AVX2ORLATER:       # %bb.0:
+; AVX2ORLATER-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2ORLATER-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2ORLATER-NEXT:    movq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $63, %rcx
+; AVX2ORLATER-NEXT:    shrq $60, %rcx
+; AVX2ORLATER-NEXT:    addq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $4, %rcx
+; AVX2ORLATER-NEXT:    vmovq %rcx, %xmm2
+; AVX2ORLATER-NEXT:    vmovq %xmm1, %rax
+; AVX2ORLATER-NEXT:    movq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $63, %rcx
+; AVX2ORLATER-NEXT:    shrq $61, %rcx
+; AVX2ORLATER-NEXT:    addq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $3, %rcx
+; AVX2ORLATER-NEXT:    vmovq %rcx, %xmm1
+; AVX2ORLATER-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX2ORLATER-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX2ORLATER-NEXT:    movq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $63, %rcx
+; AVX2ORLATER-NEXT:    shrq $62, %rcx
+; AVX2ORLATER-NEXT:    addq %rax, %rcx
+; AVX2ORLATER-NEXT:    sarq $2, %rcx
+; AVX2ORLATER-NEXT:    vmovq %rcx, %xmm2
+; AVX2ORLATER-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX2ORLATER-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
+; AVX2ORLATER-NEXT:    retq
+  %1 = sdiv <4 x i64> %x, <i64 1, i64 4, i64 8, i64 16>
+  ret <4 x i64> %1
+}
+
+define <8 x i64> @combine_vec_sdiv_by_pow2b_v8i64(<8 x i64> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_v8i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrq $1, %xmm1, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $60, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $4, %rcx
+; SSE-NEXT:    movq %rcx, %xmm4
+; SSE-NEXT:    movq %xmm1, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $61, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $3, %rcx
+; SSE-NEXT:    movq %rcx, %xmm1
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
+; SSE-NEXT:    pextrq $1, %xmm3, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $60, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $4, %rcx
+; SSE-NEXT:    movq %rcx, %xmm4
+; SSE-NEXT:    movq %xmm3, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $61, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $3, %rcx
+; SSE-NEXT:    movq %rcx, %xmm3
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm3 = xmm3[0],xmm4[0]
+; SSE-NEXT:    pextrq $1, %xmm0, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $62, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $2, %rcx
+; SSE-NEXT:    movq %rcx, %xmm4
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm4[0]
+; SSE-NEXT:    pextrq $1, %xmm2, %rax
+; SSE-NEXT:    movq %rax, %rcx
+; SSE-NEXT:    sarq $63, %rcx
+; SSE-NEXT:    shrq $62, %rcx
+; SSE-NEXT:    addq %rax, %rcx
+; SSE-NEXT:    sarq $2, %rcx
+; SSE-NEXT:    movq %rcx, %xmm4
+; SSE-NEXT:    punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm4[0]
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: combine_vec_sdiv_by_pow2b_v8i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $60, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $4, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm3
+; AVX1-NEXT:    vmovq %xmm2, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $61, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $3, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm2
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $62, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $2, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm3
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $60, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $4, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm3
+; AVX1-NEXT:    vmovq %xmm2, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $61, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $3, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm2
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    sarq $63, %rcx
+; AVX1-NEXT:    shrq $62, %rcx
+; AVX1-NEXT:    addq %rax, %rcx
+; AVX1-NEXT:    sarq $2, %rcx
+; AVX1-NEXT:    vmovq %rcx, %xmm3
+; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: combine_vec_sdiv_by_pow2b_v8i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $60, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $4, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm3
+; AVX2-NEXT:    vmovq %xmm2, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $61, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $3, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm2
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX2-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $62, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $2, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm3
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $60, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $4, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm3
+; AVX2-NEXT:    vmovq %xmm2, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $61, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $3, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm2
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX2-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-NEXT:    movq %rax, %rcx
+; AVX2-NEXT:    sarq $63, %rcx
+; AVX2-NEXT:    shrq $62, %rcx
+; AVX2-NEXT:    addq %rax, %rcx
+; AVX2-NEXT:    sarq $2, %rcx
+; AVX2-NEXT:    vmovq %rcx, %xmm3
+; AVX2-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX2-NEXT:    vinserti128 $1, %xmm2, %ymm1, %ymm1
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: combine_vec_sdiv_by_pow2b_v8i64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vextracti32x4 $3, %zmm0, %xmm1
+; AVX512-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $60, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $4, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm2
+; AVX512-NEXT:    vmovq %xmm1, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $61, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $3, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm1
+; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX512-NEXT:    vextracti32x4 $2, %zmm0, %xmm2
+; AVX512-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $62, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $2, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm3
+; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX512-NEXT:    vinserti128 $1, %xmm1, %ymm2, %ymm1
+; AVX512-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX512-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $60, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $4, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm3
+; AVX512-NEXT:    vmovq %xmm2, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $61, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $3, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm2
+; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX512-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512-NEXT:    movq %rax, %rcx
+; AVX512-NEXT:    sarq $63, %rcx
+; AVX512-NEXT:    shrq $62, %rcx
+; AVX512-NEXT:    addq %rax, %rcx
+; AVX512-NEXT:    sarq $2, %rcx
+; AVX512-NEXT:    vmovq %rcx, %xmm3
+; AVX512-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX512-NEXT:    vinserti128 $1, %xmm2, %ymm0, %ymm0
+; AVX512-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512-NEXT:    retq
+  %1 = sdiv <8 x i64> %x, <i64 1, i64 4, i64 8, i64 16, i64 1, i64 4, i64 8, i64 16>
+  ret <8 x i64> %1
+}
+
+define <4 x i32> @combine_vec_sdiv_by_pow2b_PosAndNeg(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_sdiv_by_pow2b_PosAndNeg:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pextrd $1, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $2, %ecx
+; SSE-NEXT:    negl %ecx
+; SSE-NEXT:    pextrd $2, %xmm0, %eax
+; SSE-NEXT:    pextrd $3, %xmm0, %edx
+; SSE-NEXT:    pinsrd $1, %ecx, %xmm0
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    sarl $3, %ecx
+; SSE-NEXT:    pinsrd $2, %ecx, %xmm0
+; SSE-NEXT:    movl %edx, %eax
+; SSE-NEXT:    sarl $31, %eax
+; SSE-NEXT:    shrl $28, %eax
+; SSE-NEXT:    addl %edx, %eax
+; SSE-NEXT:    sarl $4, %eax
+; SSE-NEXT:    negl %eax
+; SSE-NEXT:    pinsrd $3, %eax, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_sdiv_by_pow2b_PosAndNeg:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $30, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    sarl $2, %ecx
+; AVX-NEXT:    negl %ecx
+; AVX-NEXT:    vpinsrd $1, %ecx, %xmm0, %xmm1
+; AVX-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $29, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    sarl $3, %ecx
+; AVX-NEXT:    vpinsrd $2, %ecx, %xmm1, %xmm1
+; AVX-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $28, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    sarl $4, %ecx
+; AVX-NEXT:    negl %ecx
+; AVX-NEXT:    vpinsrd $3, %ecx, %xmm1, %xmm0
+; AVX-NEXT:    retq
+  %1 = sdiv <4 x i32> %x, <i32 1, i32 -4, i32 8, i32 -16>
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @combine_vec_sdiv_by_pow2b_undef1(<4 x i32> %x) {
+; CHECK-LABEL: combine_vec_sdiv_by_pow2b_undef1:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    retq
+  %1 = sdiv <4 x i32> %x, <i32 undef, i32 -4, i32 undef, i32 -16>
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @combine_vec_sdiv_by_pow2b_undef2(<4 x i32> %x) {
+; CHECK-LABEL: combine_vec_sdiv_by_pow2b_undef2:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    retq
+  %1 = sdiv <4 x i32> %x, <i32 undef, i32 4, i32 undef, i32 16>
+  ret <4 x i32> %1
+}
+
+define <4 x i32> @combine_vec_sdiv_by_pow2b_undef3(<4 x i32> %x) {
+; CHECK-LABEL: combine_vec_sdiv_by_pow2b_undef3:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    retq
+  %1 = sdiv <4 x i32> %x, <i32 undef, i32 -4, i32 undef, i32 16>
+  ret <4 x i32> %1
+}