[llvm] 0b04406 - [X86] vec_fabs.ll - regenerate checks and add common AVX512 prefixes
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Fri Nov 17 02:31:36 PST 2023
Author: Simon Pilgrim
Date: 2023-11-17T10:31:19Z
New Revision: 0b0440613f94cced3b047f0105a7770b07662e20
URL: https://github.com/llvm/llvm-project/commit/0b0440613f94cced3b047f0105a7770b07662e20
DIFF: https://github.com/llvm/llvm-project/commit/0b0440613f94cced3b047f0105a7770b07662e20.diff
LOG: [X86] vec_fabs.ll - regenerate checks and add common AVX512 prefixes
Added:
Modified:
llvm/test/CodeGen/X86/vec_fabs.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/vec_fabs.ll b/llvm/test/CodeGen/X86/vec_fabs.ll
index cc106890eeeb95d..f691cb76bc684e6 100644
--- a/llvm/test/CodeGen/X86/vec_fabs.ll
+++ b/llvm/test/CodeGen/X86/vec_fabs.ll
@@ -1,14 +1,14 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=X86,X86-AVX,X86-AVX1
; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=X86,X86-AVX,X86-AVX2
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefixes=X86,X86-AVX512VL
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512fp16 | FileCheck %s --check-prefixes=X86,X86-AVX512FP16
-; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=X86,X86-AVX512VLDQ
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefixes=X86,X86-AVX512,X86-AVX512VL
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512fp16 | FileCheck %s --check-prefixes=X86,X86-AVX512,X86-AVX512FP16
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=X86,X86-AVX512,X86-AVX512VLDQ
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefixes=X64,X64-AVX,X64-AVX1
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefixes=X64,X64-AVX,X64-AVX2
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefixes=X64,X64-AVX512VL
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512fp16 | FileCheck %s --check-prefixes=X64,X64-AVX512FP16
-; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=X64,X64-AVX512VLDQ
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck %s --check-prefixes=X64,X64-AVX512,X64-AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512fp16 | FileCheck %s --check-prefixes=X64,X64-AVX512,X64-AVX512FP16
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | FileCheck %s --check-prefixes=X64,X64-AVX512,X64-AVX512VLDQ
;
; 128-bit Vectors
@@ -25,6 +25,11 @@ define <2 x double> @fabs_v2f64(<2 x double> %p) {
; X86-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}{1to2}, %xmm0, %xmm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v2f64:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to2}, %xmm0, %xmm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v2f64:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to2}, %xmm0, %xmm0
@@ -40,6 +45,11 @@ define <2 x double> @fabs_v2f64(<2 x double> %p) {
; X64-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v2f64:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v2f64:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to2}, %xmm0, %xmm0
@@ -66,6 +76,11 @@ define <4 x float> @fabs_v4f32(<4 x float> %p) {
; X86-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %xmm0, %xmm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v4f32:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %xmm0, %xmm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v4f32:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %xmm0, %xmm0
@@ -87,6 +102,11 @@ define <4 x float> @fabs_v4f32(<4 x float> %p) {
; X64-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v4f32:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v4f32:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %xmm0, %xmm0
@@ -99,36 +119,42 @@ declare <4 x float> @llvm.fabs.v4f32(<4 x float> %p)
define <8 x half> @fabs_v8f16(ptr %p) {
; X86-AVX1-LABEL: fabs_v8f16:
; X86-AVX1: # %bb.0:
-; X86-AVX1-NEXT: movl 4(%esp), [[ADDRREG:%.*]]
-; X86-AVX1-NEXT: vmovaps ([[ADDRREG]]), %xmm0
+; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX1-NEXT: vmovaps (%eax), %xmm0
; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
; X86-AVX1-NEXT: retl
-
+;
; X86-AVX2-LABEL: fabs_v8f16:
; X86-AVX2: # %bb.0:
-; X86-AVX2-NEXT: movl 4(%esp), [[REG:%.*]]
-; X86-AVX2-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
-; X86-AVX2-NEXT: vpand ([[REG]]), %xmm0, %xmm0
+; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX2-NEXT: vpbroadcastw {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX2-NEXT: vpand (%eax), %xmm0, %xmm0
; X86-AVX2-NEXT: retl
-
-; X64-AVX512VL-LABEL: fabs_v8f16:
-; X64-AVX512VL: # %bb.0:
-; X64-AVX512VL-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; X64-AVX512VL-NEXT: vpand (%rdi), %xmm0, %xmm0
-; X64-AVX512VL-NEXT: retq
-
+;
+; X86-AVX512-LABEL: fabs_v8f16:
+; X86-AVX512: # %bb.0:
+; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512-NEXT: vpbroadcastw {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX512-NEXT: vpand (%eax), %xmm0, %xmm0
+; X86-AVX512-NEXT: retl
+;
; X64-AVX1-LABEL: fabs_v8f16:
; X64-AVX1: # %bb.0:
; X64-AVX1-NEXT: vmovaps (%rdi), %xmm0
; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
; X64-AVX1-NEXT: retq
-
+;
; X64-AVX2-LABEL: fabs_v8f16:
; X64-AVX2: # %bb.0:
-; X64-AVX2-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-AVX2-NEXT: vpbroadcastw {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
; X64-AVX2-NEXT: vpand (%rdi), %xmm0, %xmm0
; X64-AVX2-NEXT: retq
-
+;
+; X64-AVX512-LABEL: fabs_v8f16:
+; X64-AVX512: # %bb.0:
+; X64-AVX512-NEXT: vpbroadcastw {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X64-AVX512-NEXT: vpand (%rdi), %xmm0, %xmm0
+; X64-AVX512-NEXT: retq
%v = load <8 x half>, ptr %p, align 16
%nnv = call <8 x half> @llvm.fabs.v8f16(<8 x half> %v)
ret <8 x half> %nnv
@@ -156,6 +182,11 @@ define <4 x double> @fabs_v4f64(<4 x double> %p) {
; X86-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm0, %ymm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v4f64:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm0, %ymm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v4f64:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm0, %ymm0
@@ -177,6 +208,11 @@ define <4 x double> @fabs_v4f64(<4 x double> %p) {
; X64-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %ymm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v4f64:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %ymm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v4f64:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to4}, %ymm0, %ymm0
@@ -203,6 +239,11 @@ define <8 x float> @fabs_v8f32(<8 x float> %p) {
; X86-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v8f32:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v8f32:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm0
@@ -224,6 +265,11 @@ define <8 x float> @fabs_v8f32(<8 x float> %p) {
; X64-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v8f32:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v8f32:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %ymm0, %ymm0
@@ -234,19 +280,1377 @@ define <8 x float> @fabs_v8f32(<8 x float> %p) {
declare <8 x float> @llvm.fabs.v8f32(<8 x float> %p)
define <16 x half> @fabs_v16f16(ptr %p) {
+; X86-AVX1-LABEL: fabs_v16f16:
+; X86-AVX1: # %bb.0:
+; X86-AVX1-NEXT: pushl %esi
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX1-NEXT: subl $308, %esp # imm = 0x134
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 316
+; X86-AVX1-NEXT: .cfi_offset %esi, -8
+; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-AVX1-NEXT: vmovdqa (%esi), %xmm0
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovaps 16(%esi), %xmm1
+; X86-AVX1-NEXT: vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 4(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 8(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 12(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 20(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 24(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 28(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm1, (%esp)
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X86-AVX1-NEXT: vinsertf128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: addl $308, %esp # imm = 0x134
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX1-NEXT: popl %esi
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX1-NEXT: retl
+;
+; X86-AVX2-LABEL: fabs_v16f16:
+; X86-AVX2: # %bb.0:
+; X86-AVX2-NEXT: pushl %esi
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX2-NEXT: subl $372, %esp # imm = 0x174
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 380
+; X86-AVX2-NEXT: .cfi_offset %esi, -8
+; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-AVX2-NEXT: vmovdqa (%esi), %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovaps 16(%esi), %xmm1
+; X86-AVX2-NEXT: vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; X86-AVX2-NEXT: vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 4(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 20(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 8(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 24(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 12(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 28(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vpand {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm1, (%esp)
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm1
+; X86-AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; X86-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vzeroupper
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X86-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X86-AVX2-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X86-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; X86-AVX2-NEXT: addl $372, %esp # imm = 0x174
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX2-NEXT: popl %esi
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX2-NEXT: retl
+;
+; X86-AVX512VL-LABEL: fabs_v16f16:
+; X86-AVX512VL: # %bb.0:
+; X86-AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512VL-NEXT: movzwl 28(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm0
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm3
+; X86-AVX512VL-NEXT: vmovdqa (%eax), %xmm1
+; X86-AVX512VL-NEXT: vmovdqa 16(%eax), %xmm2
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm4 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm4, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; X86-AVX512VL-NEXT: movzwl 12(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm3, %ymm4, %ymm3
+; X86-AVX512VL-NEXT: movzwl 24(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VL-NEXT: movzwl 8(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X86-AVX512VL-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[4],ymm3[4],ymm4[5],ymm3[5]
+; X86-AVX512VL-NEXT: movzwl 20(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VL-NEXT: vpsrlq $48, %xmm2, %xmm5
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VL-NEXT: movzwl 4(%eax), %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpsrlq $48, %xmm1, %xmm6
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm6, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm2
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vmovd %xmm2, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X86-AVX512VL-NEXT: vmovd %xmm0, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X86-AVX512VL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
+; X86-AVX512VL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; X86-AVX512VL-NEXT: retl
+;
; X86-AVX512FP16-LABEL: fabs_v16f16:
; X86-AVX512FP16: # %bb.0:
-; X86-AVX512FP16-NEXT: movl 4(%esp), [[REG:%.*]]
-; X86-AVX512FP16-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}, [[YMM:%ymm[0-9]+]]
-; X86-AVX512FP16-NEXT: vpand ([[REG]]), [[YMM]], [[YMM]]
-; X86-AVX512FP16-NEXT: retl
-
+; X86-AVX512FP16-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512FP16-NEXT: vpbroadcastw {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX512FP16-NEXT: vpand (%eax), %ymm0, %ymm0
+; X86-AVX512FP16-NEXT: retl
+;
+; X86-AVX512VLDQ-LABEL: fabs_v16f16:
+; X86-AVX512VLDQ: # %bb.0:
+; X86-AVX512VLDQ-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512VLDQ-NEXT: movzwl 28(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm0
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm3
+; X86-AVX512VLDQ-NEXT: vmovdqa (%eax), %xmm1
+; X86-AVX512VLDQ-NEXT: vmovdqa 16(%eax), %xmm2
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm4 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; X86-AVX512VLDQ-NEXT: movzwl 12(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm5 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm3, %ymm4, %ymm3
+; X86-AVX512VLDQ-NEXT: movzwl 24(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm5 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: movzwl 8(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X86-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[4],ymm3[4],ymm4[5],ymm3[5]
+; X86-AVX512VLDQ-NEXT: movzwl 20(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovd %xmm4, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm4
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, %xmm2, %xmm5
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: movzwl 4(%eax), %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, %xmm1, %xmm6
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vmovd %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X86-AVX512VLDQ-NEXT: vmovd %xmm0, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X86-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
+; X86-AVX512VLDQ-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; X86-AVX512VLDQ-NEXT: retl
+;
+; X64-AVX1-LABEL: fabs_v16f16:
+; X64-AVX1: # %bb.0:
+; X64-AVX1-NEXT: pushq %rbx
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX1-NEXT: subq $80, %rsp
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 96
+; X64-AVX1-NEXT: .cfi_offset %rbx, -16
+; X64-AVX1-NEXT: movq %rdi, %rbx
+; X64-AVX1-NEXT: vbroadcastss 28(%rdi), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps (%rbx), %xmm0
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa 16(%rbx), %xmm0
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 24(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 20(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 12(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 8(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 4(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: addq $80, %rsp
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX1-NEXT: popq %rbx
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X64-AVX1-NEXT: retq
+;
+; X64-AVX2-LABEL: fabs_v16f16:
+; X64-AVX2: # %bb.0:
+; X64-AVX2-NEXT: pushq %rbx
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX2-NEXT: subq $128, %rsp
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 144
+; X64-AVX2-NEXT: .cfi_offset %rbx, -16
+; X64-AVX2-NEXT: movq %rdi, %rbx
+; X64-AVX2-NEXT: vpinsrw $0, 28(%rdi), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; X64-AVX2-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
+; X64-AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps (%rbx), %xmm0
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa 16(%rbx), %xmm0
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 12(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 24(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 8(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 20(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 4(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; X64-AVX2-NEXT: addq $128, %rsp
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX2-NEXT: popq %rbx
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X64-AVX2-NEXT: retq
+;
+; X64-AVX512VL-LABEL: fabs_v16f16:
+; X64-AVX512VL: # %bb.0:
+; X64-AVX512VL-NEXT: movzwl 28(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm0
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm0, %xmm1
+; X64-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X64-AVX512VL-NEXT: vmovdqa (%rdi), %xmm1
+; X64-AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm2
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm4 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; X64-AVX512VL-NEXT: movzwl 12(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm3, %ymm4, %ymm3
+; X64-AVX512VL-NEXT: movzwl 24(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VL-NEXT: movzwl 8(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X64-AVX512VL-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[4],ymm3[4],ymm4[5],ymm3[5]
+; X64-AVX512VL-NEXT: movzwl 20(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm2, %xmm5
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VL-NEXT: movzwl 4(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm1, %xmm6
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm2
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vmovd %xmm2, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VL-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X64-AVX512VL-NEXT: vmovd %xmm0, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
+; X64-AVX512VL-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; X64-AVX512VL-NEXT: retq
+;
; X64-AVX512FP16-LABEL: fabs_v16f16:
; X64-AVX512FP16: # %bb.0:
-; X64-AVX512FP16-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), [[YMM:%ymm[0-9]+]]
-; X64-AVX512FP16-NEXT: vpand (%rdi), [[YMM]], [[YMM]]
-; X64-AVX512FP16-NEXT: retq
-;
+; X64-AVX512FP16-NEXT: vpbroadcastw {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X64-AVX512FP16-NEXT: vpand (%rdi), %ymm0, %ymm0
+; X64-AVX512FP16-NEXT: retq
+;
+; X64-AVX512VLDQ-LABEL: fabs_v16f16:
+; X64-AVX512VLDQ: # %bb.0:
+; X64-AVX512VLDQ-NEXT: movzwl 28(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm0
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm0, %xmm1
+; X64-AVX512VLDQ-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X64-AVX512VLDQ-NEXT: vmovdqa (%rdi), %xmm1
+; X64-AVX512VLDQ-NEXT: vmovdqa 16(%rdi), %xmm2
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm4 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm3[0],xmm4[0],xmm3[1],xmm4[1],xmm3[2],xmm4[2],xmm3[3],xmm4[3]
+; X64-AVX512VLDQ-NEXT: movzwl 12(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm5 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm3, %ymm4, %ymm3
+; X64-AVX512VLDQ-NEXT: movzwl 24(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm5 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VLDQ-NEXT: movzwl 8(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm6 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X64-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} ymm3 = ymm4[0],ymm3[0],ymm4[1],ymm3[1],ymm4[4],ymm3[4],ymm4[5],ymm3[5]
+; X64-AVX512VLDQ-NEXT: movzwl 20(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm2, %xmm5
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm4[0],xmm5[0],xmm4[1],xmm5[1],xmm4[2],xmm5[2],xmm4[3],xmm5[3]
+; X64-AVX512VLDQ-NEXT: movzwl 4(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm1, %xmm6
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm5, %ymm4
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm2
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vmovd %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm5[0],xmm2[0],xmm5[1],xmm2[1],xmm5[2],xmm2[2],xmm5[3],xmm2[3]
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovd %xmm5, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X64-AVX512VLDQ-NEXT: vmovd %xmm0, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm5[0],xmm0[0],xmm5[1],xmm0[1],xmm5[2],xmm0[2],xmm5[3],xmm0[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X64-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm4[0],ymm0[1],ymm4[1],ymm0[4],ymm4[4],ymm0[5],ymm4[5]
+; X64-AVX512VLDQ-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[2],ymm3[2]
+; X64-AVX512VLDQ-NEXT: retq
%v = load <16 x half>, ptr %p, align 32
%nnv = call <16 x half> @llvm.fabs.v16f16(<16 x half> %v)
ret <16 x half> %nnv
@@ -270,6 +1674,11 @@ define <8 x double> @fabs_v8f64(<8 x double> %p) {
; X86-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm0, %zmm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v8f64:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm0, %zmm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v8f64:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm0, %zmm0
@@ -287,6 +1696,11 @@ define <8 x double> @fabs_v8f64(<8 x double> %p) {
; X64-AVX512VL-NEXT: vpandq {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v8f64:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v8f64:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandpd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to8}, %zmm0, %zmm0
@@ -309,6 +1723,11 @@ define <16 x float> @fabs_v16f32(<16 x float> %p) {
; X86-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm0
; X86-AVX512VL-NEXT: retl
;
+; X86-AVX512FP16-LABEL: fabs_v16f32:
+; X86-AVX512FP16: # %bb.0:
+; X86-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm0
+; X86-AVX512FP16-NEXT: retl
+;
; X86-AVX512VLDQ-LABEL: fabs_v16f32:
; X86-AVX512VLDQ: # %bb.0:
; X86-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm0
@@ -326,6 +1745,11 @@ define <16 x float> @fabs_v16f32(<16 x float> %p) {
; X64-AVX512VL-NEXT: vpandd {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
; X64-AVX512VL-NEXT: retq
;
+; X64-AVX512FP16-LABEL: fabs_v16f32:
+; X64-AVX512FP16: # %bb.0:
+; X64-AVX512FP16-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
+; X64-AVX512FP16-NEXT: retq
+;
; X64-AVX512VLDQ-LABEL: fabs_v16f32:
; X64-AVX512VLDQ: # %bb.0:
; X64-AVX512VLDQ-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip){1to16}, %zmm0, %zmm0
@@ -336,19 +1760,2704 @@ define <16 x float> @fabs_v16f32(<16 x float> %p) {
declare <16 x float> @llvm.fabs.v16f32(<16 x float> %p)
define <32 x half> @fabs_v32f16(ptr %p) {
+; X86-AVX1-LABEL: fabs_v32f16:
+; X86-AVX1: # %bb.0:
+; X86-AVX1-NEXT: pushl %esi
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX1-NEXT: subl $644, %esp # imm = 0x284
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 652
+; X86-AVX1-NEXT: .cfi_offset %esi, -8
+; X86-AVX1-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-AVX1-NEXT: vmovdqa 32(%esi), %xmm0
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 36(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 40(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 44(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqa 48(%esi), %xmm0
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 52(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 56(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 60(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqa (%esi), %xmm1
+; X86-AVX1-NEXT: vmovdqu %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovaps 16(%esi), %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vpsrld $16, %xmm1, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 4(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 8(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 12(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 20(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 24(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vbroadcastss 28(%esi), %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX1-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __extendhfsf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX1-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX1-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vmovss %xmm1, (%esp)
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm2 = xmm3[0],xmm2[0],xmm3[1],xmm2[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm1 = xmm2[0],xmm1[0]
+; X86-AVX1-NEXT: vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{.*#+}} xmm0 = xmm1[0],xmm0[0]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX1-NEXT: vpunpckldq {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX1-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX1-NEXT: vzeroupper
+; X86-AVX1-NEXT: calll __truncsfhf2
+; X86-AVX1-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X86-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X86-AVX1-NEXT: vinsertf128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; X86-AVX1-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload
+; X86-AVX1-NEXT: addl $644, %esp # imm = 0x284
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX1-NEXT: popl %esi
+; X86-AVX1-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX1-NEXT: retl
+;
+; X86-AVX2-LABEL: fabs_v32f16:
+; X86-AVX2: # %bb.0:
+; X86-AVX2-NEXT: pushl %esi
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX2-NEXT: subl $708, %esp # imm = 0x2C4
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 716
+; X86-AVX2-NEXT: .cfi_offset %esi, -8
+; X86-AVX2-NEXT: movl {{[0-9]+}}(%esp), %esi
+; X86-AVX2-NEXT: vmovdqa 32(%esi), %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqa 48(%esi), %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vbroadcastss {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; X86-AVX2-NEXT: vmovups %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps %xmm1, %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 36(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 52(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 40(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 56(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 44(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 60(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqa (%esi), %xmm1
+; X86-AVX2-NEXT: vmovdqu %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovaps 16(%esi), %xmm0
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vpsrld $16, %xmm1, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 4(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 20(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 8(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 24(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 12(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovd %xmm0, (%esp)
+; X86-AVX2-NEXT: vpinsrw $0, 28(%esi), %xmm0, %xmm0
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X86-AVX2-NEXT: vpextrw $0, %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vandps {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __extendhfsf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm0, (%esp)
+; X86-AVX2-NEXT: fstps {{[0-9]+}}(%esp)
+; X86-AVX2-NEXT: vmovd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-AVX2-NEXT: vpand {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vmovss %xmm1, (%esp)
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm1
+; X86-AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm1
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm3, %xmm3 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1],xmm3[2],mem[2],xmm3[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm2, %ymm3, %ymm2
+; X86-AVX2-NEXT: vpunpckldq {{.*#+}} ymm1 = ymm2[0],ymm1[0],ymm2[1],ymm1[1],ymm2[4],ymm1[4],ymm2[5],ymm1[5]
+; X86-AVX2-NEXT: vpunpcklqdq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[2],ymm0[2]
+; X86-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm2, %xmm2 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm2 = xmm2[0],mem[0],xmm2[1],mem[1],xmm2[2],mem[2],xmm2[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm1
+; X86-AVX2-NEXT: vpunpckldq {{.*#+}} ymm0 = ymm1[0],ymm0[0],ymm1[1],ymm0[1],ymm1[4],ymm0[4],ymm1[5],ymm0[5]
+; X86-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm1, %xmm1 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1],xmm1[2],mem[2],xmm1[3],mem[3]
+; X86-AVX2-NEXT: vinserti128 $1, %xmm0, %ymm1, %ymm0
+; X86-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%e{{[sb]}}p) # 32-byte Spill
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{[-0-9]+}}(%e{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X86-AVX2-NEXT: vmovdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX2-NEXT: vzeroupper
+; X86-AVX2-NEXT: calll __truncsfhf2
+; X86-AVX2-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X86-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X86-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X86-AVX2-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; X86-AVX2-NEXT: # ymm1 = ymm0[0],mem[0],ymm0[2],mem[2]
+; X86-AVX2-NEXT: vmovups {{[-0-9]+}}(%e{{[sb]}}p), %ymm0 # 32-byte Reload
+; X86-AVX2-NEXT: addl $708, %esp # imm = 0x2C4
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X86-AVX2-NEXT: popl %esi
+; X86-AVX2-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX2-NEXT: retl
+;
+; X86-AVX512VL-LABEL: fabs_v32f16:
+; X86-AVX512VL: # %bb.0:
+; X86-AVX512VL-NEXT: subl $128, %esp
+; X86-AVX512VL-NEXT: .cfi_def_cfa_offset 132
+; X86-AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512VL-NEXT: movzwl 60(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm0
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vmovdqa 48(%eax), %xmm3
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm2 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vmovdqa %xmm3, %xmm4
+; X86-AVX512VL-NEXT: vmovdqu %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm2
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vmovd %xmm2, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm2
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X86-AVX512VL-NEXT: movzwl 44(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm2
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vmovd %xmm2, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm2
+; X86-AVX512VL-NEXT: vmovdqa 32(%eax), %xmm3
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm5 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VL-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm5
+; X86-AVX512VL-NEXT: movzwl 28(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VL-NEXT: vmovdqa (%eax), %xmm1
+; X86-AVX512VL-NEXT: vmovdqu %xmm1, (%esp) # 16-byte Spill
+; X86-AVX512VL-NEXT: vmovdqa 16(%eax), %xmm2
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm7 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vmovdqu %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm7, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm7
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vmovd %xmm7, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X86-AVX512VL-NEXT: movzwl 12(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm7
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vmovd %xmm7, %ecx
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm7 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm7, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm7
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VL-NEXT: vmovd %xmm7, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm1
+; X86-AVX512VL-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; X86-AVX512VL-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill
+; X86-AVX512VL-NEXT: movzwl 56(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm6 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; X86-AVX512VL-NEXT: movzwl 40(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vmovdqa %xmm3, %xmm4
+; X86-AVX512VL-NEXT: vmovdqu %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm6
+; X86-AVX512VL-NEXT: movzwl 24(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm1 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VL-NEXT: movzwl 8(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vmovdqu (%esp), %xmm2 # 16-byte Reload
+; X86-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm1 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm7, %ymm1, %ymm1
+; X86-AVX512VL-NEXT: vinserti64x4 $1, %ymm6, %zmm1, %zmm1
+; X86-AVX512VL-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %zmm1, %zmm1 # 64-byte Folded Reload
+; X86-AVX512VL-NEXT: # zmm1 = zmm1[0],mem[0],zmm1[1],mem[1],zmm1[4],mem[4],zmm1[5],mem[5],zmm1[8],mem[8],zmm1[9],mem[9],zmm1[12],mem[12],zmm1[13],mem[13]
+; X86-AVX512VL-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill
+; X86-AVX512VL-NEXT: movzwl 52(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Reload
+; X86-AVX512VL-NEXT: vpsrlq $48, %xmm3, %xmm6
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VL-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; X86-AVX512VL-NEXT: movzwl 36(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpsrlq $48, %xmm4, %xmm1
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm6
+; X86-AVX512VL-NEXT: movzwl 20(%eax), %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpsrlq $48, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VL-NEXT: movzwl %dx, %edx
+; X86-AVX512VL-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VL-NEXT: movzwl 4(%eax), %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VL-NEXT: vpsrlq $48, %xmm2, %xmm1
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm7, %ymm1, %ymm1
+; X86-AVX512VL-NEXT: vinserti64x4 $1, %ymm6, %zmm1, %zmm6
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm3, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm3, %xmm4
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm4, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; X86-AVX512VL-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm2, %xmm3
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm3, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm3
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm3, %xmm3
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm3, %xmm3
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm3, %xmm3
+; X86-AVX512VL-NEXT: vmovd %xmm3, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm1, %ymm3
+; X86-AVX512VL-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm2
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VL-NEXT: vmovd %xmm2, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X86-AVX512VL-NEXT: vmovdqu (%esp), %xmm4 # 16-byte Reload
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm4, %eax
+; X86-AVX512VL-NEXT: movzwl %ax, %eax
+; X86-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VL-NEXT: vpsrld $16, %xmm4, %xmm1
+; X86-AVX512VL-NEXT: vpextrw $0, %xmm1, %ecx
+; X86-AVX512VL-NEXT: movzwl %cx, %ecx
+; X86-AVX512VL-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VL-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X86-AVX512VL-NEXT: vmovd %xmm0, %eax
+; X86-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX512VL-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X86-AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; X86-AVX512VL-NEXT: vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm6[0],zmm0[1],zmm6[1],zmm0[4],zmm6[4],zmm0[5],zmm6[5],zmm0[8],zmm6[8],zmm0[9],zmm6[9],zmm0[12],zmm6[12],zmm0[13],zmm6[13]
+; X86-AVX512VL-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; X86-AVX512VL-NEXT: # zmm0 = zmm0[0],mem[0],zmm0[2],mem[2],zmm0[4],mem[4],zmm0[6],mem[6]
+; X86-AVX512VL-NEXT: addl $128, %esp
+; X86-AVX512VL-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX512VL-NEXT: retl
+;
; X86-AVX512FP16-LABEL: fabs_v32f16:
; X86-AVX512FP16: # %bb.0:
-; X86-AVX512FP16-NEXT: movl 4(%esp), [[REG:%.*]]
-; X86-AVX512FP16-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}, [[ZMM:%zmm[0-9]+]]
-; X86-AVX512FP16-NEXT: vpandq ([[REG]]), [[ZMM]], [[ZMM]]
-; X86-AVX512FP16-NEXT: retl
-
+; X86-AVX512FP16-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512FP16-NEXT: vpbroadcastw {{.*#+}} zmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX512FP16-NEXT: vpandq (%eax), %zmm0, %zmm0
+; X86-AVX512FP16-NEXT: retl
+;
+; X86-AVX512VLDQ-LABEL: fabs_v32f16:
+; X86-AVX512VLDQ: # %bb.0:
+; X86-AVX512VLDQ-NEXT: subl $128, %esp
+; X86-AVX512VLDQ-NEXT: .cfi_def_cfa_offset 132
+; X86-AVX512VLDQ-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512VLDQ-NEXT: movzwl 60(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm0
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovdqa 48(%eax), %xmm3
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm2 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vmovdqa %xmm3, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovdqu %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vmovd %xmm2, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm2
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X86-AVX512VLDQ-NEXT: movzwl 44(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vmovd %xmm2, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm2
+; X86-AVX512VLDQ-NEXT: vmovdqa 32(%eax), %xmm3
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm5 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm5, %xmm5
+; X86-AVX512VLDQ-NEXT: vmovd %xmm5, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm2[0],xmm5[0],xmm2[1],xmm5[1],xmm2[2],xmm5[2],xmm2[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm1, %ymm2, %ymm5
+; X86-AVX512VLDQ-NEXT: movzwl 28(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VLDQ-NEXT: vmovdqa (%eax), %xmm1
+; X86-AVX512VLDQ-NEXT: vmovdqu %xmm1, (%esp) # 16-byte Spill
+; X86-AVX512VLDQ-NEXT: vmovdqa 16(%eax), %xmm2
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm7 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vmovdqu %xmm2, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vmovd %xmm7, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X86-AVX512VLDQ-NEXT: movzwl 12(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vmovd %xmm7, %ecx
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm7 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X86-AVX512VLDQ-NEXT: vmovd %xmm7, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm7[0],xmm1[0],xmm7[1],xmm1[1],xmm7[2],xmm1[2],xmm7[3],xmm1[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm1
+; X86-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm5, %zmm1, %zmm1
+; X86-AVX512VLDQ-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill
+; X86-AVX512VLDQ-NEXT: movzwl 56(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm6 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; X86-AVX512VLDQ-NEXT: movzwl 40(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm1 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vmovdqa %xmm3, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovdqu %xmm3, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm6
+; X86-AVX512VLDQ-NEXT: movzwl 24(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm1 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VLDQ-NEXT: movzwl 8(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vmovdqu (%esp), %xmm2 # 16-byte Reload
+; X86-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm1 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm7, %ymm1, %ymm1
+; X86-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm6, %zmm1, %zmm1
+; X86-AVX512VLDQ-NEXT: vpunpckldq {{[-0-9]+}}(%e{{[sb]}}p), %zmm1, %zmm1 # 64-byte Folded Reload
+; X86-AVX512VLDQ-NEXT: # zmm1 = zmm1[0],mem[0],zmm1[1],mem[1],zmm1[4],mem[4],zmm1[5],mem[5],zmm1[8],mem[8],zmm1[9],mem[9],zmm1[12],mem[12],zmm1[13],mem[13]
+; X86-AVX512VLDQ-NEXT: vmovdqu64 %zmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 64-byte Spill
+; X86-AVX512VLDQ-NEXT: movzwl 52(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Reload
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, %xmm3, %xmm6
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X86-AVX512VLDQ-NEXT: vmovd %xmm6, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm6
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm1[0],xmm6[0],xmm1[1],xmm6[1],xmm1[2],xmm6[2],xmm1[3],xmm6[3]
+; X86-AVX512VLDQ-NEXT: movzwl 36(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, %xmm4, %xmm1
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm1, %ymm6
+; X86-AVX512VLDQ-NEXT: movzwl 20(%eax), %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Folded Reload
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: movzwl %dx, %edx
+; X86-AVX512VLDQ-NEXT: vmovd %edx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %edx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %edx, %xmm0, %xmm7
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm1[0],xmm7[0],xmm1[1],xmm7[1],xmm1[2],xmm7[2],xmm1[3],xmm7[3]
+; X86-AVX512VLDQ-NEXT: movzwl 4(%eax), %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: vpsrlq $48, %xmm2, %xmm1
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %ecx, %xmm0, %xmm5
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm5[0],xmm1[1],xmm5[1],xmm1[2],xmm5[2],xmm1[3],xmm5[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm7, %ymm1, %ymm1
+; X86-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm6, %zmm1, %zmm6
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm3, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm3, %xmm4
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X86-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm1[0],xmm4[0],xmm1[1],xmm4[1],xmm1[2],xmm4[2],xmm1[3],xmm4[3]
+; X86-AVX512VLDQ-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm2, %xmm3
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm3, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm3
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm3, %xmm3
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm3, %xmm3
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm3, %xmm3
+; X86-AVX512VLDQ-NEXT: vmovd %xmm3, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm3[0],xmm1[1],xmm3[1],xmm1[2],xmm3[2],xmm1[3],xmm3[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm1, %ymm3
+; X86-AVX512VLDQ-NEXT: vmovdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X86-AVX512VLDQ-NEXT: vmovd %xmm2, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
+; X86-AVX512VLDQ-NEXT: vmovdqu (%esp), %xmm4 # 16-byte Reload
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %eax
+; X86-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X86-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X86-AVX512VLDQ-NEXT: vpsrld $16, %xmm4, %xmm1
+; X86-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %ecx
+; X86-AVX512VLDQ-NEXT: movzwl %cx, %ecx
+; X86-AVX512VLDQ-NEXT: vmovd %ecx, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X86-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm1
+; X86-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X86-AVX512VLDQ-NEXT: vmovd %xmm0, %eax
+; X86-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X86-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-AVX512VLDQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X86-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; X86-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm6[0],zmm0[1],zmm6[1],zmm0[4],zmm6[4],zmm0[5],zmm6[5],zmm0[8],zmm6[8],zmm0[9],zmm6[9],zmm0[12],zmm6[12],zmm0[13],zmm6[13]
+; X86-AVX512VLDQ-NEXT: vpunpcklqdq {{[-0-9]+}}(%e{{[sb]}}p), %zmm0, %zmm0 # 64-byte Folded Reload
+; X86-AVX512VLDQ-NEXT: # zmm0 = zmm0[0],mem[0],zmm0[2],mem[2],zmm0[4],mem[4],zmm0[6],mem[6]
+; X86-AVX512VLDQ-NEXT: addl $128, %esp
+; X86-AVX512VLDQ-NEXT: .cfi_def_cfa_offset 4
+; X86-AVX512VLDQ-NEXT: retl
+;
+; X64-AVX1-LABEL: fabs_v32f16:
+; X64-AVX1: # %bb.0:
+; X64-AVX1-NEXT: pushq %rbx
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX1-NEXT: subq $128, %rsp
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 144
+; X64-AVX1-NEXT: .cfi_offset %rbx, -16
+; X64-AVX1-NEXT: movq %rdi, %rbx
+; X64-AVX1-NEXT: vbroadcastss 28(%rdi), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps (%rbx), %xmm0
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa 16(%rbx), %xmm1
+; X64-AVX1-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps 32(%rbx), %xmm0
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps 48(%rbx), %xmm0
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 24(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 20(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 12(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 8(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 4(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: vmovups %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 60(%rbx), %xmm0
+; X64-AVX1-NEXT: vzeroupper
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 56(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 52(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 44(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 40(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vbroadcastss 36(%rbx), %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vmovdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-AVX1-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX1-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX1-NEXT: vpand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX1-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX1-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX1-NEXT: vpunpckldq (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-AVX1-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX1-NEXT: # xmm0 = xmm0[0],mem[0]
+; X64-AVX1-NEXT: vinsertf128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 16-byte Folded Reload
+; X64-AVX1-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; X64-AVX1-NEXT: addq $128, %rsp
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX1-NEXT: popq %rbx
+; X64-AVX1-NEXT: .cfi_def_cfa_offset 8
+; X64-AVX1-NEXT: retq
+;
+; X64-AVX2-LABEL: fabs_v32f16:
+; X64-AVX2: # %bb.0:
+; X64-AVX2-NEXT: pushq %rbx
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX2-NEXT: subq $192, %rsp
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 208
+; X64-AVX2-NEXT: .cfi_offset %rbx, -16
+; X64-AVX2-NEXT: movq %rdi, %rbx
+; X64-AVX2-NEXT: vpinsrw $0, 28(%rdi), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpbroadcastd {{.*#+}} xmm1 = [NaN,NaN,NaN,NaN]
+; X64-AVX2-NEXT: vmovdqa %xmm1, (%rsp) # 16-byte Spill
+; X64-AVX2-NEXT: vpand %xmm1, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps (%rbx), %xmm0
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa 16(%rbx), %xmm1
+; X64-AVX2-NEXT: vmovdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps 32(%rbx), %xmm0
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps 48(%rbx), %xmm0
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 12(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 24(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 8(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 20(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 4(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[2],mem[2]
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 60(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 44(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 56(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 40(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 52(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vpinsrw $0, 36(%rbx), %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrlq $48, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vmovdqu %ymm0, {{[-0-9]+}}(%r{{[sb]}}p) # 32-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vzeroupper
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vmovdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vandps (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-AVX2-NEXT: vpsrld $16, %xmm0, %xmm0
+; X64-AVX2-NEXT: callq __extendhfsf2 at PLT
+; X64-AVX2-NEXT: vpand (%rsp), %xmm0, %xmm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: callq __truncsfhf2 at PLT
+; X64-AVX2-NEXT: vmovdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-AVX2-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-AVX2-NEXT: vinserti128 $1, {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 16-byte Folded Reload
+; X64-AVX2-NEXT: vpunpckldq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm0 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm0 = ymm0[0],mem[0],ymm0[1],mem[1],ymm0[4],mem[4],ymm0[5],mem[5]
+; X64-AVX2-NEXT: vpunpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %ymm0, %ymm1 # 32-byte Folded Reload
+; X64-AVX2-NEXT: # ymm1 = ymm0[0],mem[0],ymm0[2],mem[2]
+; X64-AVX2-NEXT: vmovups {{[-0-9]+}}(%r{{[sb]}}p), %ymm0 # 32-byte Reload
+; X64-AVX2-NEXT: addq $192, %rsp
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 16
+; X64-AVX2-NEXT: popq %rbx
+; X64-AVX2-NEXT: .cfi_def_cfa_offset 8
+; X64-AVX2-NEXT: retq
+;
+; X64-AVX512VL-LABEL: fabs_v32f16:
+; X64-AVX512VL: # %bb.0:
+; X64-AVX512VL-NEXT: movzwl 60(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm0
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm0, %xmm1
+; X64-AVX512VL-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vmovd %xmm1, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VL-NEXT: vmovdqa (%rdi), %xmm1
+; X64-AVX512VL-NEXT: vmovdqa 16(%rdi), %xmm2
+; X64-AVX512VL-NEXT: vmovdqa 32(%rdi), %xmm3
+; X64-AVX512VL-NEXT: vmovdqa 48(%rdi), %xmm4
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm6 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VL-NEXT: movzwl 44(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm7 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm5, %ymm6, %ymm5
+; X64-AVX512VL-NEXT: movzwl 28(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm7 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VL-NEXT: movzwl 12(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm8 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm5, %zmm6, %zmm5
+; X64-AVX512VL-NEXT: movzwl 56(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm7 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VL-NEXT: movzwl 40(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm8 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VL-NEXT: movzwl 24(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm8 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VL-NEXT: movzwl 8(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpsrldq {{.*#+}} xmm9 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm9, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm9
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vmovd %xmm9, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm9
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm6, %zmm7, %zmm6
+; X64-AVX512VL-NEXT: vpunpckldq {{.*#+}} zmm5 = zmm6[0],zmm5[0],zmm6[1],zmm5[1],zmm6[4],zmm5[4],zmm6[5],zmm5[5],zmm6[8],zmm5[8],zmm6[9],zmm5[9],zmm6[12],zmm5[12],zmm6[13],zmm5[13]
+; X64-AVX512VL-NEXT: movzwl 52(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VL-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm4, %xmm7
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VL-NEXT: movzwl 36(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm3, %xmm8
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VL-NEXT: movzwl 20(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm2, %xmm8
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VL-NEXT: movzwl 4(%rdi), %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VL-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VL-NEXT: vpsrlq $48, %xmm1, %xmm9
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm9, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm9
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm9, %xmm9
+; X64-AVX512VL-NEXT: vmovd %xmm9, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm9
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm6, %zmm7, %zmm6
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm3, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VL-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm3, %xmm3
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm3, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm3
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm3, %xmm3
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm3, %xmm3
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm3, %xmm3
+; X64-AVX512VL-NEXT: vmovd %xmm3, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm3[0],xmm7[1],xmm3[1],xmm7[2],xmm3[2],xmm7[3],xmm3[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm4, %ymm3, %ymm3
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm2
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X64-AVX512VL-NEXT: vmovd %xmm2, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VL-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VL-NEXT: vpsrld $16, %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VL-NEXT: movzwl %ax, %eax
+; X64-AVX512VL-NEXT: vmovd %eax, %xmm1
+; X64-AVX512VL-NEXT: vcvtph2ps %xmm1, %xmm1
+; X64-AVX512VL-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X64-AVX512VL-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X64-AVX512VL-NEXT: vmovd %xmm0, %eax
+; X64-AVX512VL-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X64-AVX512VL-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; X64-AVX512VL-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X64-AVX512VL-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; X64-AVX512VL-NEXT: vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm6[0],zmm0[1],zmm6[1],zmm0[4],zmm6[4],zmm0[5],zmm6[5],zmm0[8],zmm6[8],zmm0[9],zmm6[9],zmm0[12],zmm6[12],zmm0[13],zmm6[13]
+; X64-AVX512VL-NEXT: vpunpcklqdq {{.*#+}} zmm0 = zmm0[0],zmm5[0],zmm0[2],zmm5[2],zmm0[4],zmm5[4],zmm0[6],zmm5[6]
+; X64-AVX512VL-NEXT: retq
+;
; X64-AVX512FP16-LABEL: fabs_v32f16:
; X64-AVX512FP16: # %bb.0:
-; X64-AVX512FP16-NEXT: vpbroadcastw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), [[ZMM:%zmm[0-9]+]]
-; X64-AVX512FP16-NEXT: vpandq (%rdi), [[ZMM]], [[ZMM]]
-; X64-AVX512FP16-NEXT: retq
-
+; X64-AVX512FP16-NEXT: vpbroadcastw {{.*#+}} zmm0 = [NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X64-AVX512FP16-NEXT: vpandq (%rdi), %zmm0, %zmm0
+; X64-AVX512FP16-NEXT: retq
+;
+; X64-AVX512VLDQ-LABEL: fabs_v32f16:
+; X64-AVX512VLDQ: # %bb.0:
+; X64-AVX512VLDQ-NEXT: movzwl 60(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm0
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm0, %xmm1
+; X64-AVX512VLDQ-NEXT: vpbroadcastd {{.*#+}} xmm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vmovd %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm5
+; X64-AVX512VLDQ-NEXT: vmovdqa (%rdi), %xmm1
+; X64-AVX512VLDQ-NEXT: vmovdqa 16(%rdi), %xmm2
+; X64-AVX512VLDQ-NEXT: vmovdqa 32(%rdi), %xmm3
+; X64-AVX512VLDQ-NEXT: vmovdqa 48(%rdi), %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm6 = xmm4[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm5 = xmm5[0],xmm6[0],xmm5[1],xmm6[1],xmm5[2],xmm6[2],xmm5[3],xmm6[3]
+; X64-AVX512VLDQ-NEXT: movzwl 44(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm7 = xmm3[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm5, %ymm6, %ymm5
+; X64-AVX512VLDQ-NEXT: movzwl 28(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm7 = xmm2[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VLDQ-NEXT: movzwl 12(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm8 = xmm1[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm5, %zmm6, %zmm5
+; X64-AVX512VLDQ-NEXT: movzwl 56(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm7 = xmm4[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VLDQ-NEXT: movzwl 40(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm8 = xmm3[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VLDQ-NEXT: movzwl 24(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm8 = xmm2[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VLDQ-NEXT: movzwl 8(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpsrldq {{.*#+}} xmm9 = xmm1[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm9, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm9
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vmovd %xmm9, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm9
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; X64-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm6, %zmm7, %zmm6
+; X64-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} zmm5 = zmm6[0],zmm5[0],zmm6[1],zmm5[1],zmm6[4],zmm5[4],zmm6[5],zmm5[5],zmm6[8],zmm5[8],zmm6[9],zmm5[9],zmm6[12],zmm5[12],zmm6[13],zmm5[13]
+; X64-AVX512VLDQ-NEXT: movzwl 52(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm6, %xmm6
+; X64-AVX512VLDQ-NEXT: vmovd %xmm6, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm6
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm4, %xmm7
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm6 = xmm6[0],xmm7[0],xmm6[1],xmm7[1],xmm6[2],xmm7[2],xmm6[3],xmm7[3]
+; X64-AVX512VLDQ-NEXT: movzwl 36(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm3, %xmm8
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm6, %ymm7, %ymm6
+; X64-AVX512VLDQ-NEXT: movzwl 20(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm2, %xmm8
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm7 = xmm7[0],xmm8[0],xmm7[1],xmm8[1],xmm7[2],xmm8[2],xmm7[3],xmm8[3]
+; X64-AVX512VLDQ-NEXT: movzwl 4(%rdi), %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm8, %xmm8
+; X64-AVX512VLDQ-NEXT: vmovd %xmm8, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm8
+; X64-AVX512VLDQ-NEXT: vpsrlq $48, %xmm1, %xmm9
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm9, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm9
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm9, %xmm9
+; X64-AVX512VLDQ-NEXT: vmovd %xmm9, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm9
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm8 = xmm8[0],xmm9[0],xmm8[1],xmm9[1],xmm8[2],xmm9[2],xmm8[3],xmm9[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm7, %ymm8, %ymm7
+; X64-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm6, %zmm7, %zmm6
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm4 = xmm7[0],xmm4[0],xmm7[1],xmm4[1],xmm7[2],xmm4[2],xmm7[3],xmm4[3]
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm3, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm7, %xmm7
+; X64-AVX512VLDQ-NEXT: vmovd %xmm7, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm7
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm3, %xmm3
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm3, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm3
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm3, %xmm3
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm3, %xmm3
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm3, %xmm3
+; X64-AVX512VLDQ-NEXT: vmovd %xmm3, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm3
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm3 = xmm7[0],xmm3[0],xmm7[1],xmm3[1],xmm7[2],xmm3[2],xmm7[3],xmm3[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm4, %ymm3, %ymm3
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm2
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm2, %xmm2
+; X64-AVX512VLDQ-NEXT: vmovd %xmm2, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm2
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm2 = xmm4[0],xmm2[0],xmm4[1],xmm2[1],xmm4[2],xmm2[2],xmm4[3],xmm2[3]
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm4, %xmm4
+; X64-AVX512VLDQ-NEXT: vmovd %xmm4, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm4
+; X64-AVX512VLDQ-NEXT: vpsrld $16, %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vpextrw $0, %xmm1, %eax
+; X64-AVX512VLDQ-NEXT: movzwl %ax, %eax
+; X64-AVX512VLDQ-NEXT: vmovd %eax, %xmm1
+; X64-AVX512VLDQ-NEXT: vcvtph2ps %xmm1, %xmm1
+; X64-AVX512VLDQ-NEXT: vpand %xmm0, %xmm1, %xmm0
+; X64-AVX512VLDQ-NEXT: vcvtps2ph $4, %xmm0, %xmm0
+; X64-AVX512VLDQ-NEXT: vmovd %xmm0, %eax
+; X64-AVX512VLDQ-NEXT: vpinsrw $0, %eax, %xmm0, %xmm0
+; X64-AVX512VLDQ-NEXT: vpunpcklwd {{.*#+}} xmm0 = xmm4[0],xmm0[0],xmm4[1],xmm0[1],xmm4[2],xmm0[2],xmm4[3],xmm0[3]
+; X64-AVX512VLDQ-NEXT: vinserti128 $1, %xmm2, %ymm0, %ymm0
+; X64-AVX512VLDQ-NEXT: vinserti64x4 $1, %ymm3, %zmm0, %zmm0
+; X64-AVX512VLDQ-NEXT: vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm6[0],zmm0[1],zmm6[1],zmm0[4],zmm6[4],zmm0[5],zmm6[5],zmm0[8],zmm6[8],zmm0[9],zmm6[9],zmm0[12],zmm6[12],zmm0[13],zmm6[13]
+; X64-AVX512VLDQ-NEXT: vpunpcklqdq {{.*#+}} zmm0 = zmm0[0],zmm5[0],zmm0[2],zmm5[2],zmm0[4],zmm5[4],zmm0[6],zmm5[6]
+; X64-AVX512VLDQ-NEXT: retq
%v = load <32 x half>, ptr %p, align 64
%nnv = call <32 x half> @llvm.fabs.v32f16(<32 x half> %v)
ret <32 x half> %nnv
@@ -433,29 +4542,17 @@ define void @PR70947(ptr %src, ptr %dst) {
; X86-AVX2-NEXT: vzeroupper
; X86-AVX2-NEXT: retl
;
-; X86-AVX512VL-LABEL: PR70947:
-; X86-AVX512VL: # %bb.0:
-; X86-AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-AVX512VL-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-AVX512VL-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X86-AVX512VL-NEXT: vandps (%ecx), %ymm0, %ymm1
-; X86-AVX512VL-NEXT: vandps 32(%ecx), %xmm0, %xmm0
-; X86-AVX512VL-NEXT: vmovups %ymm1, (%eax)
-; X86-AVX512VL-NEXT: vmovups %xmm0, 16(%eax)
-; X86-AVX512VL-NEXT: vzeroupper
-; X86-AVX512VL-NEXT: retl
-;
-; X86-AVX512VLDQ-LABEL: PR70947:
-; X86-AVX512VLDQ: # %bb.0:
-; X86-AVX512VLDQ-NEXT: movl {{[0-9]+}}(%esp), %eax
-; X86-AVX512VLDQ-NEXT: movl {{[0-9]+}}(%esp), %ecx
-; X86-AVX512VLDQ-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X86-AVX512VLDQ-NEXT: vandps (%ecx), %ymm0, %ymm1
-; X86-AVX512VLDQ-NEXT: vandps 32(%ecx), %xmm0, %xmm0
-; X86-AVX512VLDQ-NEXT: vmovups %ymm1, (%eax)
-; X86-AVX512VLDQ-NEXT: vmovups %xmm0, 16(%eax)
-; X86-AVX512VLDQ-NEXT: vzeroupper
-; X86-AVX512VLDQ-NEXT: retl
+; X86-AVX512-LABEL: PR70947:
+; X86-AVX512: # %bb.0:
+; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %eax
+; X86-AVX512-NEXT: movl {{[0-9]+}}(%esp), %ecx
+; X86-AVX512-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX512-NEXT: vandps (%ecx), %ymm0, %ymm1
+; X86-AVX512-NEXT: vandps 32(%ecx), %xmm0, %xmm0
+; X86-AVX512-NEXT: vmovups %ymm1, (%eax)
+; X86-AVX512-NEXT: vmovups %xmm0, 16(%eax)
+; X86-AVX512-NEXT: vzeroupper
+; X86-AVX512-NEXT: retl
;
; X64-AVX1-LABEL: PR70947:
; X64-AVX1: # %bb.0:
@@ -478,25 +4575,15 @@ define void @PR70947(ptr %src, ptr %dst) {
; X64-AVX2-NEXT: vzeroupper
; X64-AVX2-NEXT: retq
;
-; X64-AVX512VL-LABEL: PR70947:
-; X64-AVX512VL: # %bb.0:
-; X64-AVX512VL-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X64-AVX512VL-NEXT: vandps (%rdi), %ymm0, %ymm1
-; X64-AVX512VL-NEXT: vandps 32(%rdi), %xmm0, %xmm0
-; X64-AVX512VL-NEXT: vmovups %ymm1, (%rsi)
-; X64-AVX512VL-NEXT: vmovups %xmm0, 16(%rsi)
-; X64-AVX512VL-NEXT: vzeroupper
-; X64-AVX512VL-NEXT: retq
-;
-; X64-AVX512VLDQ-LABEL: PR70947:
-; X64-AVX512VLDQ: # %bb.0:
-; X64-AVX512VLDQ-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X64-AVX512VLDQ-NEXT: vandps (%rdi), %ymm0, %ymm1
-; X64-AVX512VLDQ-NEXT: vandps 32(%rdi), %xmm0, %xmm0
-; X64-AVX512VLDQ-NEXT: vmovups %ymm1, (%rsi)
-; X64-AVX512VLDQ-NEXT: vmovups %xmm0, 16(%rsi)
-; X64-AVX512VLDQ-NEXT: vzeroupper
-; X64-AVX512VLDQ-NEXT: retq
+; X64-AVX512-LABEL: PR70947:
+; X64-AVX512: # %bb.0:
+; X64-AVX512-NEXT: vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX512-NEXT: vandps (%rdi), %ymm0, %ymm1
+; X64-AVX512-NEXT: vandps 32(%rdi), %xmm0, %xmm0
+; X64-AVX512-NEXT: vmovups %ymm1, (%rsi)
+; X64-AVX512-NEXT: vmovups %xmm0, 16(%rsi)
+; X64-AVX512-NEXT: vzeroupper
+; X64-AVX512-NEXT: retq
%src4 = getelementptr inbounds double, ptr %src, i64 4
%dst4 = getelementptr inbounds i32, ptr %dst, i64 4
%ld0 = load <4 x double>, ptr %src, align 8
More information about the llvm-commits
mailing list