[llvm] 04b403d - [X86] combineConcatVectorOps - only concatenate single-use subops
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sun Oct 1 06:28:12 PDT 2023
Author: Simon Pilgrim
Date: 2023-10-01T14:27:55+01:00
New Revision: 04b403d8cc6990171ca0d9bd8d1571e56d231a4d
URL: https://github.com/llvm/llvm-project/commit/04b403d8cc6990171ca0d9bd8d1571e56d231a4d
DIFF: https://github.com/llvm/llvm-project/commit/04b403d8cc6990171ca0d9bd8d1571e56d231a4d.diff
LOG: [X86] combineConcatVectorOps - only concatenate single-use subops
We could maybe extend this by allowing the lowest subop to have multiple uses and extract the lowest subvector result of the concatenated op, but let's just get the fix in first.
Fixes #67333
Added:
llvm/test/CodeGen/X86/pr67333.ll
Modified:
llvm/lib/Target/X86/X86ISelLowering.cpp
llvm/test/CodeGen/X86/subvector-broadcast.ll
llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
llvm/test/CodeGen/X86/widen_bitcnt.ll
llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
Removed:
################################################################################
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 08b9098871b354b..0921f1a5c00fc7d 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -54448,7 +54448,7 @@ static SDValue combineConcatVectorOps(const SDLoc &DL, MVT VT,
// TODO - combineX86ShufflesRecursively should handle shuffle concatenation
// but it currently struggles with
diff erent vector widths.
if (llvm::all_of(Ops, [Op0](SDValue Op) {
- return Op.getOpcode() == Op0.getOpcode();
+ return Op.getOpcode() == Op0.getOpcode() && Op.hasOneUse();
})) {
auto ConcatSubOperand = [&](EVT VT, ArrayRef<SDValue> SubOps, unsigned I) {
SmallVector<SDValue> Subs;
diff --git a/llvm/test/CodeGen/X86/pr67333.ll b/llvm/test/CodeGen/X86/pr67333.ll
new file mode 100644
index 000000000000000..64c7f4fb143bfef
--- /dev/null
+++ b/llvm/test/CodeGen/X86/pr67333.ll
@@ -0,0 +1,190 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s
+
+declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #0
+declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #0
+
+define void @SHA256_Compress_Generic(ptr noundef %ctx) #1 {
+; CHECK-LABEL: SHA256_Compress_Generic:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: movbel 0, %eax
+; CHECK-NEXT: movbel 12(%rdi), %ecx
+; CHECK-NEXT: vmovd %eax, %xmm0
+; CHECK-NEXT: vmovdqa {{.*#+}} xmm1 = [128,128,128,128,0,1,2,3,128,128,128,128,128,128,128,128]
+; CHECK-NEXT: vpshufb %xmm1, %xmm0, %xmm2
+; CHECK-NEXT: vpsrld $17, %xmm2, %xmm0
+; CHECK-NEXT: vpslld $15, %xmm2, %xmm3
+; CHECK-NEXT: vpor %xmm0, %xmm3, %xmm0
+; CHECK-NEXT: vpsrld $19, %xmm2, %xmm3
+; CHECK-NEXT: vpslld $13, %xmm2, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpxor %xmm3, %xmm0, %xmm0
+; CHECK-NEXT: vpxor %xmm2, %xmm0, %xmm0
+; CHECK-NEXT: vmovd %ecx, %xmm3
+; CHECK-NEXT: vpshufb %xmm1, %xmm3, %xmm1
+; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm1
+; CHECK-NEXT: vpsrld $17, %xmm1, %xmm0
+; CHECK-NEXT: vpslld $15, %xmm1, %xmm3
+; CHECK-NEXT: vpor %xmm0, %xmm3, %xmm0
+; CHECK-NEXT: vpsrld $19, %xmm1, %xmm3
+; CHECK-NEXT: vpslld $13, %xmm1, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpxor %xmm3, %xmm0, %xmm0
+; CHECK-NEXT: vpxor %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
+; CHECK-NEXT: vpsrld $17, %xmm0, %xmm3
+; CHECK-NEXT: vpslld $15, %xmm0, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpsrld $19, %xmm0, %xmm4
+; CHECK-NEXT: vpslld $13, %xmm0, %xmm5
+; CHECK-NEXT: vpor %xmm4, %xmm5, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsrld $10, %xmm0, %xmm0
+; CHECK-NEXT: vpxor %xmm0, %xmm3, %xmm0
+; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
+; CHECK-NEXT: vpsrld $17, %xmm0, %xmm3
+; CHECK-NEXT: vpslld $15, %xmm0, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpsrld $19, %xmm0, %xmm4
+; CHECK-NEXT: vpslld $13, %xmm0, %xmm5
+; CHECK-NEXT: vpor %xmm4, %xmm5, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsrld $10, %xmm0, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpblendd {{.*#+}} xmm2 = xmm1[0],xmm2[1],xmm1[2,3]
+; CHECK-NEXT: vpshufd {{.*#+}} xmm2 = xmm2[1,0,2,3]
+; CHECK-NEXT: vpaddd %xmm3, %xmm2, %xmm2
+; CHECK-NEXT: vpsrld $17, %xmm2, %xmm3
+; CHECK-NEXT: vpslld $15, %xmm2, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpsrld $19, %xmm2, %xmm4
+; CHECK-NEXT: vpslld $13, %xmm2, %xmm5
+; CHECK-NEXT: vpor %xmm4, %xmm5, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsrld $10, %xmm2, %xmm2
+; CHECK-NEXT: vpxor %xmm2, %xmm3, %xmm2
+; CHECK-NEXT: vpsrlq $32, %xmm1, %xmm3
+; CHECK-NEXT: vpaddd %xmm2, %xmm3, %xmm1
+; CHECK-NEXT: vpsrld $17, %xmm1, %xmm2
+; CHECK-NEXT: vpslld $15, %xmm1, %xmm4
+; CHECK-NEXT: vpor %xmm2, %xmm4, %xmm2
+; CHECK-NEXT: vpsrld $19, %xmm1, %xmm4
+; CHECK-NEXT: vpslld $13, %xmm1, %xmm5
+; CHECK-NEXT: vpor %xmm4, %xmm5, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm2, %xmm2
+; CHECK-NEXT: vpsrld $10, %xmm1, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm2, %xmm2
+; CHECK-NEXT: vpaddd %xmm2, %xmm3, %xmm2
+; CHECK-NEXT: vpsrld $17, %xmm2, %xmm3
+; CHECK-NEXT: vpslld $15, %xmm2, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpsrld $19, %xmm2, %xmm4
+; CHECK-NEXT: vpslld $13, %xmm2, %xmm5
+; CHECK-NEXT: vpor %xmm4, %xmm5, %xmm4
+; CHECK-NEXT: vpxor %xmm4, %xmm3, %xmm3
+; CHECK-NEXT: vpsrld $10, %xmm2, %xmm2
+; CHECK-NEXT: vpxor %xmm2, %xmm3, %xmm2
+; CHECK-NEXT: vpaddd %xmm2, %xmm0, %xmm0
+; CHECK-NEXT: vpsrld $17, %xmm0, %xmm2
+; CHECK-NEXT: vpslld $15, %xmm0, %xmm3
+; CHECK-NEXT: vpor %xmm2, %xmm3, %xmm2
+; CHECK-NEXT: vpsrld $19, %xmm0, %xmm3
+; CHECK-NEXT: vpslld $13, %xmm0, %xmm4
+; CHECK-NEXT: vpor %xmm3, %xmm4, %xmm3
+; CHECK-NEXT: vpxor %xmm3, %xmm2, %xmm2
+; CHECK-NEXT: vpsrld $10, %xmm0, %xmm3
+; CHECK-NEXT: vpxor %xmm3, %xmm2, %xmm2
+; CHECK-NEXT: vpsllq $32, %xmm1, %xmm3
+; CHECK-NEXT: vpaddd %xmm2, %xmm3, %xmm2
+; CHECK-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; CHECK-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; CHECK-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,2,1,3]
+; CHECK-NEXT: vmovdqu %ymm0, 132(%rdi)
+; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: retq
+entry:
+ %0 = load i32, ptr null, align 4
+ %1 = tail call i32 asm "bswap $0", "=r,0,~{dirflag},~{fpsr},~{flags}"(i32 %0) #3
+ %arrayidx14 = getelementptr inbounds [64 x i32], ptr %ctx, i64 0, i64 3
+ %2 = load i32, ptr %arrayidx14, align 4
+ %3 = tail call i32 asm "bswap $0", "=r,0,~{dirflag},~{fpsr},~{flags}"(i32 %2) #3
+ %4 = insertelement <2 x i32> zeroinitializer, i32 %1, i64 1
+ %5 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %4, <2 x i32> %4, <2 x i32> <i32 15, i32 15>)
+ %6 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %4, <2 x i32> %4, <2 x i32> <i32 13, i32 13>)
+ %7 = xor <2 x i32> %5, %6
+ %8 = lshr <2 x i32> %4, zeroinitializer
+ %9 = xor <2 x i32> %7, %8
+ %10 = insertelement <2 x i32> zeroinitializer, i32 %3, i64 0
+ %11 = shufflevector <2 x i32> zeroinitializer, <2 x i32> %10, <2 x i32> <i32 1, i32 2>
+ %12 = add <2 x i32> %11, %9
+ %13 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %12, <2 x i32> %12, <2 x i32> <i32 15, i32 15>)
+ %14 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %12, <2 x i32> %12, <2 x i32> <i32 13, i32 13>)
+ %15 = xor <2 x i32> %13, %14
+ %16 = lshr <2 x i32> %12, zeroinitializer
+ %17 = xor <2 x i32> %15, %16
+ %18 = add <2 x i32> %4, %17
+ %19 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %18, <2 x i32> %18, <2 x i32> <i32 15, i32 15>)
+ %20 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %18, <2 x i32> %18, <2 x i32> <i32 13, i32 13>)
+ %21 = xor <2 x i32> %19, %20
+ %22 = lshr <2 x i32> %18, <i32 10, i32 10>
+ %23 = xor <2 x i32> %21, %22
+ %24 = add <2 x i32> %4, %23
+ %25 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %24, <2 x i32> %24, <2 x i32> <i32 15, i32 15>)
+ %26 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %24, <2 x i32> %24, <2 x i32> <i32 13, i32 13>)
+ %27 = xor <2 x i32> %25, %26
+ %28 = lshr <2 x i32> %24, <i32 10, i32 10>
+ %29 = xor <2 x i32> %27, %28
+ %30 = shufflevector <2 x i32> %4, <2 x i32> %12, <2 x i32> <i32 1, i32 2>
+ %31 = add <2 x i32> %30, %29
+ %32 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %31, <2 x i32> %31, <2 x i32> <i32 15, i32 15>)
+ %33 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %31, <2 x i32> %31, <2 x i32> <i32 13, i32 13>)
+ %34 = xor <2 x i32> %32, %33
+ %35 = lshr <2 x i32> %31, <i32 10, i32 10>
+ %36 = xor <2 x i32> %34, %35
+ %37 = shufflevector <2 x i32> %12, <2 x i32> zeroinitializer, <2 x i32> <i32 1, i32 2>
+ %38 = add <2 x i32> %37, %36
+ %arrayidx918 = getelementptr inbounds [64 x i32], ptr %ctx, i64 0, i64 33
+ store <2 x i32> %38, ptr %arrayidx918, align 4
+ %arrayidx1012 = getelementptr inbounds [64 x i32], ptr %ctx, i64 0, i64 35
+ %39 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %38, <2 x i32> %38, <2 x i32> <i32 15, i32 15>)
+ %40 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %38, <2 x i32> %38, <2 x i32> <i32 13, i32 13>)
+ %41 = xor <2 x i32> %39, %40
+ %42 = lshr <2 x i32> %38, <i32 10, i32 10>
+ %43 = xor <2 x i32> %41, %42
+ %44 = add <2 x i32> %37, %43
+ store <2 x i32> zeroinitializer, ptr %arrayidx1012, align 4
+ %arrayidx1106 = getelementptr inbounds [64 x i32], ptr %ctx, i64 0, i64 37
+ %45 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %44, <2 x i32> %44, <2 x i32> <i32 15, i32 15>)
+ %46 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %44, <2 x i32> %44, <2 x i32> <i32 13, i32 13>)
+ %47 = xor <2 x i32> %45, %46
+ %48 = lshr <2 x i32> %44, <i32 10, i32 10>
+ %49 = xor <2 x i32> %47, %48
+ %50 = lshr <2 x i32> %24, zeroinitializer
+ %51 = add <2 x i32> %50, %49
+ store <2 x i32> %51, ptr %arrayidx1106, align 4
+ %arrayidx1200 = getelementptr inbounds [64 x i32], ptr %ctx, i64 0, i64 39
+ %52 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %51, <2 x i32> %51, <2 x i32> <i32 15, i32 15>)
+ %53 = tail call <2 x i32> @llvm.fshl.v2i32(<2 x i32> %51, <2 x i32> %51, <2 x i32> <i32 13, i32 13>)
+ %54 = xor <2 x i32> %52, %53
+ %55 = lshr <2 x i32> %51, <i32 10, i32 10>
+ %56 = xor <2 x i32> %54, %55
+ %57 = shufflevector <2 x i32> %38, <2 x i32> zeroinitializer, <2 x i32> <i32 poison, i32 0>
+ %58 = insertelement <2 x i32> %57, i32 0, i64 0
+ %59 = add <2 x i32> %58, %56
+ store <2 x i32> %59, ptr %arrayidx1200, align 4
+ ret void
+
+; uselistorder directives
+ uselistorder <2 x i32> %4, { 7, 0, 1, 6, 5, 4, 3, 2 }
+ uselistorder <2 x i32> %38, { 6, 5, 4, 3, 2, 1, 0 }
+}
+
+declare <2 x i32> @llvm.fshl.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) #2
+
+; uselistorder directives
+uselistorder ptr @llvm.fshl.v2i32, { 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 }
+
+attributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: readwrite) }
+attributes #1 = { nounwind sspstrong memory(argmem: readwrite) uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "probe-stack"="inline-asm" "stack-protector-buffer-size"="8" "target-cpu"="skylake" "target-features"="+adx,+aes,+avx,+avx2,+bmi,+bmi2,+clflushopt,+cmov,+crc32,+cx16,+cx8,+f16c,+fma,+fsgsbase,+fxsr,+invpcid,+lzcnt,+mmx,+movbe,+pclmul,+popcnt,+prfchw,+rdrnd,+rdseed,+sahf,+sgx,+sse,+sse2,+sse3,+sse4.1,+sse4.2,+ssse3,+x87,+xsave,+xsavec,+xsaveopt,+xsaves" }
+attributes #2 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
+attributes #3 = { nounwind memory(none) }
diff --git a/llvm/test/CodeGen/X86/subvector-broadcast.ll b/llvm/test/CodeGen/X86/subvector-broadcast.ll
index b6a245a98634ff8..e104d9d2b4c1570 100644
--- a/llvm/test/CodeGen/X86/subvector-broadcast.ll
+++ b/llvm/test/CodeGen/X86/subvector-broadcast.ll
@@ -1756,8 +1756,8 @@ define void @PR51226() {
; X86-AVX2-LABEL: PR51226:
; X86-AVX2: # %bb.0:
; X86-AVX2-NEXT: vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X86-AVX2-NEXT: vpslld $16, %xmm0, %xmm0
; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X86-AVX2-NEXT: vpslld $16, %ymm0, %ymm0
; X86-AVX2-NEXT: vxorps %xmm1, %xmm1, %xmm1
; X86-AVX2-NEXT: vminps %ymm1, %ymm0, %ymm0
; X86-AVX2-NEXT: vmovups %ymm0, (%eax)
@@ -1767,8 +1767,8 @@ define void @PR51226() {
; X86-AVX512-LABEL: PR51226:
; X86-AVX512: # %bb.0:
; X86-AVX512-NEXT: vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X86-AVX512-NEXT: vpslld $16, %xmm0, %xmm0
; X86-AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X86-AVX512-NEXT: vpslld $16, %ymm0, %ymm0
; X86-AVX512-NEXT: vxorps %xmm1, %xmm1, %xmm1
; X86-AVX512-NEXT: vminps %ymm1, %ymm0, %ymm0
; X86-AVX512-NEXT: vmovups %ymm0, (%eax)
@@ -1789,8 +1789,8 @@ define void @PR51226() {
; X64-AVX2-LABEL: PR51226:
; X64-AVX2: # %bb.0:
; X64-AVX2-NEXT: vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X64-AVX2-NEXT: vpslld $16, %xmm0, %xmm0
; X64-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX2-NEXT: vpslld $16, %ymm0, %ymm0
; X64-AVX2-NEXT: vxorps %xmm1, %xmm1, %xmm1
; X64-AVX2-NEXT: vminps %ymm1, %ymm0, %ymm0
; X64-AVX2-NEXT: vmovups %ymm0, (%rax)
@@ -1800,8 +1800,8 @@ define void @PR51226() {
; X64-AVX512-LABEL: PR51226:
; X64-AVX512: # %bb.0:
; X64-AVX512-NEXT: vpmovzxwd {{.*#+}} xmm0 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero
+; X64-AVX512-NEXT: vpslld $16, %xmm0, %xmm0
; X64-AVX512-NEXT: vinserti128 $1, %xmm0, %ymm0, %ymm0
-; X64-AVX512-NEXT: vpslld $16, %ymm0, %ymm0
; X64-AVX512-NEXT: vxorps %xmm1, %xmm1, %xmm1
; X64-AVX512-NEXT: vminps %ymm1, %ymm0, %ymm0
; X64-AVX512-NEXT: vmovups %ymm0, (%rax)
diff --git a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
index 97be882ffd8e43f..7893a799c207ab3 100644
--- a/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
+++ b/llvm/test/CodeGen/X86/vector-interleaved-store-i8-stride-7.ll
@@ -725,17 +725,19 @@ define void @store_i8_stride7_vf8(ptr %in.vecptr0, ptr %in.vecptr1, ptr %in.vecp
; AVX512BW-SLOW-NEXT: vmovq {{.*#+}} xmm3 = mem[0],zero
; AVX512BW-SLOW-NEXT: vinserti128 $1, %xmm1, %ymm0, %ymm0
; AVX512BW-SLOW-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
-; AVX512BW-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
-; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} zmm1 = zmm0[0,2,1,3,4,6,5,7]
+; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} ymm1 = ymm0[0,2,1,3]
+; AVX512BW-SLOW-NEXT: vinserti64x4 $1, %ymm1, %zmm1, %zmm1
; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm1 = zero,zero,zero,zero,zmm1[0,8],zero,zero,zero,zero,zero,zmm1[1,9],zero,zero,zero,zero,zero,zero,zero,zmm1[18],zero,zero,zero,zero,zero,zero,zmm1[19],zero,zero,zero,zero,zmm1[36,44],zero,zero,zero,zero,zero,zmm1[37,45],zero,zero,zero,zero,zero,zmm1[38,46,54],zero,zero,zero,zero,zero,zero,zmm1[55],zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} zmm0 = zmm0[1,3,0,2,5,7,4,6]
+; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[1,3,0,2]
+; AVX512BW-SLOW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm0 = zero,zero,zero,zero,zero,zero,zmm0[0],zero,zero,zero,zero,zero,zero,zmm0[1],zero,zero,zero,zero,zmm0[18,26],zero,zero,zero,zero,zero,zmm0[19,27],zero,zero,zero,zero,zero,zero,zero,zmm0[36],zero,zero,zero,zero,zero,zero,zmm0[37],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm0[55,63],zero,zero,zero,zero,zero,zero,zero,zero,zero
; AVX512BW-SLOW-NEXT: vporq %zmm1, %zmm0, %zmm0
; AVX512BW-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm1
-; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm2 = zmm1[0,8],zero,zero,zero,zero,zero,zmm1[1,9],zero,zero,zero,zero,zero,zmm1[2,10,18,26],zero,zero,zero,zero,zero,zmm1[19,27],zero,zero,zero,zero,zero,zmm1[20,28],zero,zero,zero,zmm1[37,45],zero,zero,zero,zero,zero,zmm1[38,46],zero,zero,zero,zero,zero,zero,zero,zmm1[55,63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} zmm1 = zmm1[2,3,0,1,6,7,4,5]
-; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm1 = zero,zero,zmm1[0,8],zero,zero,zero,zero,zero,zmm1[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm1[19,27],zero,zero,zero,zero,zero,zmm1[20,28],zero,zero,zero,zero,zero,zero,zero,zmm1[37,45],zero,zero,zero,zero,zero,zmm1[38,46],zero,zero,zero,zmm1[55,63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
-; AVX512BW-SLOW-NEXT: vporq %zmm2, %zmm1, %zmm1
+; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm1 = zmm1[0,8],zero,zero,zero,zero,zero,zmm1[1,9],zero,zero,zero,zero,zero,zmm1[2,10,18,26],zero,zero,zero,zero,zero,zmm1[19,27],zero,zero,zero,zero,zero,zmm1[20,28],zero,zero,zero,zmm1[37,45],zero,zero,zero,zero,zero,zmm1[38,46],zero,zero,zero,zero,zero,zero,zero,zmm1[55,63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-SLOW-NEXT: vpermq {{.*#+}} ymm2 = ymm2[2,3,0,1]
+; AVX512BW-SLOW-NEXT: vinserti64x4 $1, %ymm2, %zmm2, %zmm2
+; AVX512BW-SLOW-NEXT: vpshufb {{.*#+}} zmm2 = zero,zero,zmm2[0,8],zero,zero,zero,zero,zero,zmm2[1,9],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zmm2[19,27],zero,zero,zero,zero,zero,zmm2[20,28],zero,zero,zero,zero,zero,zero,zero,zmm2[37,45],zero,zero,zero,zero,zero,zmm2[38,46],zero,zero,zero,zmm2[55,63],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; AVX512BW-SLOW-NEXT: vporq %zmm1, %zmm2, %zmm1
; AVX512BW-SLOW-NEXT: movabsq $63546854584629360, %rcx # imm = 0xE1C3870E1C3870
; AVX512BW-SLOW-NEXT: kmovq %rcx, %k1
; AVX512BW-SLOW-NEXT: vmovdqu8 %zmm0, %zmm1 {%k1}
diff --git a/llvm/test/CodeGen/X86/widen_bitcnt.ll b/llvm/test/CodeGen/X86/widen_bitcnt.ll
index 50f19d31183e5e3..da468b6d809e899 100644
--- a/llvm/test/CodeGen/X86/widen_bitcnt.ll
+++ b/llvm/test/CodeGen/X86/widen_bitcnt.ll
@@ -632,45 +632,71 @@ define <8 x i32> @widen_ctlz_v2i32_v8i32(<2 x i32> %a0, <2 x i32> %a1, <2 x i32>
; AVX2: # %bb.0:
; AVX2-NEXT: # kill: def $xmm1 killed $xmm1 def $ymm1
; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm5 = [4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0]
+; AVX2-NEXT: vpshufb %xmm0, %xmm5, %xmm4
+; AVX2-NEXT: vpsrlw $4, %xmm0, %xmm6
+; AVX2-NEXT: vpbroadcastb {{.*#+}} xmm7 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %xmm7, %xmm6, %xmm8
+; AVX2-NEXT: vpxor %xmm6, %xmm6, %xmm6
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm8, %xmm9
+; AVX2-NEXT: vpand %xmm4, %xmm9, %xmm4
+; AVX2-NEXT: vpshufb %xmm8, %xmm5, %xmm8
+; AVX2-NEXT: vpaddb %xmm4, %xmm8, %xmm4
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm0, %xmm8
+; AVX2-NEXT: vpsrlw $8, %xmm8, %xmm8
+; AVX2-NEXT: vpand %xmm4, %xmm8, %xmm8
+; AVX2-NEXT: vpsrlw $8, %xmm4, %xmm4
+; AVX2-NEXT: vpaddw %xmm4, %xmm8, %xmm4
+; AVX2-NEXT: vpshufb %xmm1, %xmm5, %xmm8
+; AVX2-NEXT: vpsrlw $4, %xmm1, %xmm9
+; AVX2-NEXT: vpand %xmm7, %xmm9, %xmm9
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm9, %xmm10
+; AVX2-NEXT: vpand %xmm10, %xmm8, %xmm8
+; AVX2-NEXT: vpshufb %xmm9, %xmm5, %xmm9
+; AVX2-NEXT: vpaddb %xmm9, %xmm8, %xmm8
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm1, %xmm9
+; AVX2-NEXT: vpsrlw $8, %xmm9, %xmm9
+; AVX2-NEXT: vpand %xmm9, %xmm8, %xmm9
+; AVX2-NEXT: vpsrlw $8, %xmm8, %xmm8
+; AVX2-NEXT: vpaddw %xmm9, %xmm8, %xmm8
+; AVX2-NEXT: vpshufb %xmm2, %xmm5, %xmm9
+; AVX2-NEXT: vpsrlw $4, %xmm2, %xmm10
+; AVX2-NEXT: vpand %xmm7, %xmm10, %xmm10
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm10, %xmm11
+; AVX2-NEXT: vpand %xmm11, %xmm9, %xmm9
+; AVX2-NEXT: vpshufb %xmm10, %xmm5, %xmm10
+; AVX2-NEXT: vpaddb %xmm10, %xmm9, %xmm9
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm2, %xmm10
+; AVX2-NEXT: vpsrlw $8, %xmm10, %xmm10
+; AVX2-NEXT: vpand %xmm10, %xmm9, %xmm10
+; AVX2-NEXT: vpsrlw $8, %xmm9, %xmm9
+; AVX2-NEXT: vpaddw %xmm10, %xmm9, %xmm9
+; AVX2-NEXT: vpshufb %xmm3, %xmm5, %xmm10
+; AVX2-NEXT: vpsrlw $4, %xmm3, %xmm11
+; AVX2-NEXT: vpand %xmm7, %xmm11, %xmm7
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm7, %xmm11
+; AVX2-NEXT: vpand %xmm11, %xmm10, %xmm10
+; AVX2-NEXT: vpshufb %xmm7, %xmm5, %xmm5
+; AVX2-NEXT: vpaddb %xmm5, %xmm10, %xmm5
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm3, %xmm6
+; AVX2-NEXT: vpsrlw $8, %xmm6, %xmm6
+; AVX2-NEXT: vpand %xmm6, %xmm5, %xmm6
+; AVX2-NEXT: vpsrlw $8, %xmm5, %xmm5
+; AVX2-NEXT: vpaddw %xmm6, %xmm5, %xmm5
+; AVX2-NEXT: vinserti128 $1, %xmm5, %ymm8, %ymm5
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm1, %ymm1
-; AVX2-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0,4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0]
-; AVX2-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX2-NEXT: vpshufb %ymm1, %ymm3, %ymm4
-; AVX2-NEXT: vpsrlw $4, %ymm1, %ymm5
-; AVX2-NEXT: vpbroadcastb {{.*#+}} ymm6 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX2-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX2-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm5, %ymm8
-; AVX2-NEXT: vpand %ymm4, %ymm8, %ymm4
-; AVX2-NEXT: vpshufb %ymm5, %ymm3, %ymm5
-; AVX2-NEXT: vpaddb %ymm5, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm1, %ymm5
-; AVX2-NEXT: vpsrlw $8, %ymm5, %ymm5
-; AVX2-NEXT: vpand %ymm5, %ymm4, %ymm5
-; AVX2-NEXT: vpsrlw $8, %ymm4, %ymm4
-; AVX2-NEXT: vpaddw %ymm5, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqw %ymm7, %ymm1, %ymm1
+; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX2-NEXT: vpcmpeqw %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpsrld $16, %ymm1, %ymm1
-; AVX2-NEXT: vpand %ymm1, %ymm4, %ymm1
-; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
-; AVX2-NEXT: vpaddd %ymm1, %ymm4, %ymm1
+; AVX2-NEXT: vpand %ymm1, %ymm5, %ymm1
+; AVX2-NEXT: vpsrld $16, %ymm5, %ymm5
+; AVX2-NEXT: vpaddd %ymm1, %ymm5, %ymm1
+; AVX2-NEXT: vinserti128 $1, %xmm9, %ymm4, %ymm4
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm2
-; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm4
-; AVX2-NEXT: vpand %ymm6, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm4, %ymm5
-; AVX2-NEXT: vpand %ymm5, %ymm2, %ymm2
-; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm0, %ymm3
-; AVX2-NEXT: vpsrlw $8, %ymm3, %ymm3
-; AVX2-NEXT: vpand %ymm3, %ymm2, %ymm3
-; AVX2-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX2-NEXT: vpaddw %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpcmpeqw %ymm7, %ymm0, %ymm0
+; AVX2-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0
-; AVX2-NEXT: vpand %ymm0, %ymm2, %ymm0
-; AVX2-NEXT: vpsrld $16, %ymm2, %ymm2
+; AVX2-NEXT: vpand %ymm0, %ymm4, %ymm0
+; AVX2-NEXT: vpsrld $16, %ymm4, %ymm2
; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT: retq
@@ -987,45 +1013,71 @@ define <8 x i32> @widen_ctlz_undef_v2i32_v8i32(<2 x i32> %a0, <2 x i32> %a1, <2
; AVX2: # %bb.0:
; AVX2-NEXT: # kill: def $xmm1 killed $xmm1 def $ymm1
; AVX2-NEXT: # kill: def $xmm0 killed $xmm0 def $ymm0
+; AVX2-NEXT: vmovdqa {{.*#+}} xmm5 = [4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0]
+; AVX2-NEXT: vpshufb %xmm0, %xmm5, %xmm4
+; AVX2-NEXT: vpsrlw $4, %xmm0, %xmm6
+; AVX2-NEXT: vpbroadcastb {{.*#+}} xmm7 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT: vpand %xmm7, %xmm6, %xmm8
+; AVX2-NEXT: vpxor %xmm6, %xmm6, %xmm6
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm8, %xmm9
+; AVX2-NEXT: vpand %xmm4, %xmm9, %xmm4
+; AVX2-NEXT: vpshufb %xmm8, %xmm5, %xmm8
+; AVX2-NEXT: vpaddb %xmm4, %xmm8, %xmm4
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm0, %xmm8
+; AVX2-NEXT: vpsrlw $8, %xmm8, %xmm8
+; AVX2-NEXT: vpand %xmm4, %xmm8, %xmm8
+; AVX2-NEXT: vpsrlw $8, %xmm4, %xmm4
+; AVX2-NEXT: vpaddw %xmm4, %xmm8, %xmm4
+; AVX2-NEXT: vpshufb %xmm1, %xmm5, %xmm8
+; AVX2-NEXT: vpsrlw $4, %xmm1, %xmm9
+; AVX2-NEXT: vpand %xmm7, %xmm9, %xmm9
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm9, %xmm10
+; AVX2-NEXT: vpand %xmm10, %xmm8, %xmm8
+; AVX2-NEXT: vpshufb %xmm9, %xmm5, %xmm9
+; AVX2-NEXT: vpaddb %xmm9, %xmm8, %xmm8
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm1, %xmm9
+; AVX2-NEXT: vpsrlw $8, %xmm9, %xmm9
+; AVX2-NEXT: vpand %xmm9, %xmm8, %xmm9
+; AVX2-NEXT: vpsrlw $8, %xmm8, %xmm8
+; AVX2-NEXT: vpaddw %xmm9, %xmm8, %xmm8
+; AVX2-NEXT: vpshufb %xmm2, %xmm5, %xmm9
+; AVX2-NEXT: vpsrlw $4, %xmm2, %xmm10
+; AVX2-NEXT: vpand %xmm7, %xmm10, %xmm10
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm10, %xmm11
+; AVX2-NEXT: vpand %xmm11, %xmm9, %xmm9
+; AVX2-NEXT: vpshufb %xmm10, %xmm5, %xmm10
+; AVX2-NEXT: vpaddb %xmm10, %xmm9, %xmm9
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm2, %xmm10
+; AVX2-NEXT: vpsrlw $8, %xmm10, %xmm10
+; AVX2-NEXT: vpand %xmm10, %xmm9, %xmm10
+; AVX2-NEXT: vpsrlw $8, %xmm9, %xmm9
+; AVX2-NEXT: vpaddw %xmm10, %xmm9, %xmm9
+; AVX2-NEXT: vpshufb %xmm3, %xmm5, %xmm10
+; AVX2-NEXT: vpsrlw $4, %xmm3, %xmm11
+; AVX2-NEXT: vpand %xmm7, %xmm11, %xmm7
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm7, %xmm11
+; AVX2-NEXT: vpand %xmm11, %xmm10, %xmm10
+; AVX2-NEXT: vpshufb %xmm7, %xmm5, %xmm5
+; AVX2-NEXT: vpaddb %xmm5, %xmm10, %xmm5
+; AVX2-NEXT: vpcmpeqb %xmm6, %xmm3, %xmm6
+; AVX2-NEXT: vpsrlw $8, %xmm6, %xmm6
+; AVX2-NEXT: vpand %xmm6, %xmm5, %xmm6
+; AVX2-NEXT: vpsrlw $8, %xmm5, %xmm5
+; AVX2-NEXT: vpaddw %xmm6, %xmm5, %xmm5
+; AVX2-NEXT: vinserti128 $1, %xmm5, %ymm8, %ymm5
; AVX2-NEXT: vinserti128 $1, %xmm3, %ymm1, %ymm1
-; AVX2-NEXT: vbroadcasti128 {{.*#+}} ymm3 = [4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0,4,3,2,2,1,1,1,1,0,0,0,0,0,0,0,0]
-; AVX2-NEXT: # ymm3 = mem[0,1,0,1]
-; AVX2-NEXT: vpshufb %ymm1, %ymm3, %ymm4
-; AVX2-NEXT: vpsrlw $4, %ymm1, %ymm5
-; AVX2-NEXT: vpbroadcastb {{.*#+}} ymm6 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
-; AVX2-NEXT: vpand %ymm6, %ymm5, %ymm5
-; AVX2-NEXT: vpxor %xmm7, %xmm7, %xmm7
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm5, %ymm8
-; AVX2-NEXT: vpand %ymm4, %ymm8, %ymm4
-; AVX2-NEXT: vpshufb %ymm5, %ymm3, %ymm5
-; AVX2-NEXT: vpaddb %ymm5, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm1, %ymm5
-; AVX2-NEXT: vpsrlw $8, %ymm5, %ymm5
-; AVX2-NEXT: vpand %ymm5, %ymm4, %ymm5
-; AVX2-NEXT: vpsrlw $8, %ymm4, %ymm4
-; AVX2-NEXT: vpaddw %ymm5, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqw %ymm7, %ymm1, %ymm1
+; AVX2-NEXT: vpxor %xmm3, %xmm3, %xmm3
+; AVX2-NEXT: vpcmpeqw %ymm3, %ymm1, %ymm1
; AVX2-NEXT: vpsrld $16, %ymm1, %ymm1
-; AVX2-NEXT: vpand %ymm1, %ymm4, %ymm1
-; AVX2-NEXT: vpsrld $16, %ymm4, %ymm4
-; AVX2-NEXT: vpaddd %ymm1, %ymm4, %ymm1
+; AVX2-NEXT: vpand %ymm1, %ymm5, %ymm1
+; AVX2-NEXT: vpsrld $16, %ymm5, %ymm5
+; AVX2-NEXT: vpaddd %ymm1, %ymm5, %ymm1
+; AVX2-NEXT: vinserti128 $1, %xmm9, %ymm4, %ymm4
; AVX2-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
-; AVX2-NEXT: vpshufb %ymm0, %ymm3, %ymm2
-; AVX2-NEXT: vpsrlw $4, %ymm0, %ymm4
-; AVX2-NEXT: vpand %ymm6, %ymm4, %ymm4
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm4, %ymm5
-; AVX2-NEXT: vpand %ymm5, %ymm2, %ymm2
-; AVX2-NEXT: vpshufb %ymm4, %ymm3, %ymm3
-; AVX2-NEXT: vpaddb %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpcmpeqb %ymm7, %ymm0, %ymm3
-; AVX2-NEXT: vpsrlw $8, %ymm3, %ymm3
-; AVX2-NEXT: vpand %ymm3, %ymm2, %ymm3
-; AVX2-NEXT: vpsrlw $8, %ymm2, %ymm2
-; AVX2-NEXT: vpaddw %ymm3, %ymm2, %ymm2
-; AVX2-NEXT: vpcmpeqw %ymm7, %ymm0, %ymm0
+; AVX2-NEXT: vpcmpeqw %ymm3, %ymm0, %ymm0
; AVX2-NEXT: vpsrld $16, %ymm0, %ymm0
-; AVX2-NEXT: vpand %ymm0, %ymm2, %ymm0
-; AVX2-NEXT: vpsrld $16, %ymm2, %ymm2
+; AVX2-NEXT: vpand %ymm0, %ymm4, %ymm0
+; AVX2-NEXT: vpsrld $16, %ymm4, %ymm2
; AVX2-NEXT: vpaddd %ymm0, %ymm2, %ymm0
; AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[2],ymm1[2]
; AVX2-NEXT: retq
diff --git a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
index 8f9d2169aa19b95..92973274f291909 100644
--- a/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
+++ b/llvm/test/CodeGen/X86/zero_extend_vector_inreg_of_broadcast.ll
@@ -5682,9 +5682,10 @@ define void @vec512_i8_widen_to_i16_factor2_broadcast_to_v32i16_factor32(ptr %in
;
; AVX512BW-LABEL: vec512_i8_widen_to_i16_factor2_broadcast_to_v32i16_factor32:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[0],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[16],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[32],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero,zmm0[48],zero
; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
@@ -5794,9 +5795,10 @@ define void @vec512_i8_widen_to_i32_factor4_broadcast_to_v16i32_factor16(ptr %in
;
; AVX512BW-LABEL: vec512_i8_widen_to_i32_factor4_broadcast_to_v16i32_factor16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0],zero,zero,zero,zmm0[0],zero,zero,zero,zmm0[0],zero,zero,zero,zmm0[0],zero,zero,zero,zmm0[16],zero,zero,zero,zmm0[16],zero,zero,zero,zmm0[16],zero,zero,zero,zmm0[16],zero,zero,zero,zmm0[32],zero,zero,zero,zmm0[32],zero,zero,zero,zmm0[32],zero,zero,zero,zmm0[32],zero,zero,zero,zmm0[48],zero,zero,zero,zmm0[48],zero,zero,zero,zmm0[48],zero,zero,zero,zmm0[48],zero,zero,zero
; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
@@ -5906,9 +5908,10 @@ define void @vec512_i8_widen_to_i64_factor8_broadcast_to_v8i64_factor8(ptr %in.v
;
; AVX512BW-LABEL: vec512_i8_widen_to_i64_factor8_broadcast_to_v8i64_factor8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0],zero,zero,zero,zero,zero,zero,zero,zmm0[0],zero,zero,zero,zero,zero,zero,zero,zmm0[16],zero,zero,zero,zero,zero,zero,zero,zmm0[16],zero,zero,zero,zero,zero,zero,zero,zmm0[32],zero,zero,zero,zero,zero,zero,zero,zmm0[32],zero,zero,zero,zero,zero,zero,zero,zmm0[48],zero,zero,zero,zero,zero,zero,zero,zmm0[48],zero,zero,zero,zero,zero,zero,zero
; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
@@ -5999,9 +6002,10 @@ define void @vec512_i8_widen_to_i128_factor16_broadcast_to_v4i128_factor4(ptr %i
;
; AVX512BW-LABEL: vec512_i8_widen_to_i128_factor16_broadcast_to_v4i128_factor4:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vshufi64x2 {{.*#+}} zmm0 = zmm0[0,1,0,1,0,1,0,1]
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
; AVX512BW-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
@@ -6205,12 +6209,12 @@ define void @vec512_i16_widen_to_i32_factor2_broadcast_to_v16i32_factor16(ptr %i
;
; AVX512BW-LABEL: vec512_i16_widen_to_i32_factor2_broadcast_to_v16i32_factor16:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,33,0,35,0,37,0,39,0,41,0,43,0,45,0,47,0,49,0,51,0,53,0,55,0,57,0,59,0,61,0,63]
-; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm2
-; AVX512BW-NEXT: vpaddb (%rdx), %zmm2, %zmm0
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0,1],zero,zero,zmm0[0,1],zero,zero,zmm0[0,1],zero,zero,zmm0[0,1],zero,zero,zmm0[16,17],zero,zero,zmm0[16,17],zero,zero,zmm0[16,17],zero,zero,zmm0[16,17],zero,zero,zmm0[32,33],zero,zero,zmm0[32,33],zero,zero,zmm0[32,33],zero,zero,zmm0[32,33],zero,zero,zmm0[48,49],zero,zero,zmm0[48,49],zero,zero,zmm0[48,49],zero,zero,zmm0[48,49],zero,zero
+; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
@@ -6324,12 +6328,12 @@ define void @vec512_i16_widen_to_i64_factor4_broadcast_to_v8i64_factor8(ptr %in.
;
; AVX512BW-LABEL: vec512_i16_widen_to_i64_factor4_broadcast_to_v8i64_factor8:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,33,34,35,0,37,38,39,0,41,42,43,0,45,46,47,0,49,50,51,0,53,54,55,0,57,58,59,0,61,62,63]
-; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm2
-; AVX512BW-NEXT: vpaddb (%rdx), %zmm2, %zmm0
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512BW-NEXT: vpshufb {{.*#+}} zmm0 = zmm0[0,1],zero,zero,zero,zero,zero,zero,zmm0[0,1],zero,zero,zero,zero,zero,zero,zmm0[16,17],zero,zero,zero,zero,zero,zero,zmm0[16,17],zero,zero,zero,zero,zero,zero,zmm0[32,33],zero,zero,zero,zero,zero,zero,zmm0[32,33],zero,zero,zero,zero,zero,zero,zmm0[48,49],zero,zero,zero,zero,zero,zero,zmm0[48,49],zero,zero,zero,zero,zero,zero
+; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
@@ -6443,12 +6447,12 @@ define void @vec512_i16_widen_to_i128_factor8_broadcast_to_v4i128_factor4(ptr %i
;
; AVX512BW-LABEL: vec512_i16_widen_to_i128_factor8_broadcast_to_v4i128_factor4:
; AVX512BW: # %bb.0:
-; AVX512BW-NEXT: vmovdqa64 (%rdi), %zmm0
-; AVX512BW-NEXT: vpaddb (%rsi), %zmm0, %zmm0
-; AVX512BW-NEXT: vpxor %xmm1, %xmm1, %xmm1
-; AVX512BW-NEXT: vmovdqa64 {{.*#+}} zmm2 = [0,33,34,35,36,37,38,39,0,41,42,43,44,45,46,47,0,49,50,51,52,53,54,55,0,57,58,59,60,61,62,63]
-; AVX512BW-NEXT: vpermi2w %zmm1, %zmm0, %zmm2
-; AVX512BW-NEXT: vpaddb (%rdx), %zmm2, %zmm0
+; AVX512BW-NEXT: vmovdqa (%rdi), %ymm0
+; AVX512BW-NEXT: vpaddb (%rsi), %ymm0, %ymm0
+; AVX512BW-NEXT: vpermq {{.*#+}} ymm0 = ymm0[0,1,0,1]
+; AVX512BW-NEXT: vinserti64x4 $1, %ymm0, %zmm0, %zmm0
+; AVX512BW-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %zmm0, %zmm0
+; AVX512BW-NEXT: vpaddb (%rdx), %zmm0, %zmm0
; AVX512BW-NEXT: vmovdqa64 %zmm0, (%rcx)
; AVX512BW-NEXT: vzeroupper
; AVX512BW-NEXT: retq
More information about the llvm-commits
mailing list