[llvm-branch-commits] [llvm] abc60e9 - [X86] vec_fabs.ll - add SSE test coverage

[Simon Pilgrim via llvm-branch-commits]

Author: Simon Pilgrim
Date: 2023-11-30T10:07:00Z
New Revision: abc60e9808820c3f6614e6815909d43ed085460e

URL: 
https://github.com/llvm/llvm-project/commit/abc60e9808820c3f6614e6815909d43ed085460e
DIFF: 
https://github.com/llvm/llvm-project/commit/abc60e9808820c3f6614e6815909d43ed085460e.diff

LOG: [X86] vec_fabs.ll - add SSE test coverage

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 ec02dfda30c8502..c17341c2c8b077e 100644
--- a/llvm/test/CodeGen/X86/vec_fabs.ll
+++ b/llvm/test/CodeGen/X86/vec_fabs.ll
@@ -1,24 +1,31 @@
 ; 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-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-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
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse2 | FileCheck %s 
--check-prefixes=X86,X86-SSE
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx | FileCheck %s 
--check-prefixes=X86,X86-AVX,X86-AVX1OR2,X86-AVX1
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx2 | FileCheck %s 
--check-prefixes=X86,X86-AVX,X86-AVX1OR2,X86-AVX2
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512vl | FileCheck %s 
--check-prefixes=X86,X86-AVX,X86-AVX512,X86-AVX512VL
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512fp16 | FileCheck 
%s --check-prefixes=X86,X86-AVX,X86-AVX512,X86-AVX512FP16
+; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+avx512dq,+avx512vl | 
FileCheck %s --check-prefixes=X86,X86-AVX,X86-AVX512,X86-AVX512VLDQ
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse2 | FileCheck %s 
--check-prefixes=X64,X64-SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s 
--check-prefixes=X64,X64-AVX,X64-AVX1OR2,X64-AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s 
--check-prefixes=X64,X64-AVX,X64-AVX1OR2,X64-AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vl | FileCheck 
%s --check-prefixes=X64,X64-AVX,X64-AVX512,X64-AVX512VL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512fp16 | FileCheck 
%s --check-prefixes=X64,X64-AVX,X64-AVX512,X64-AVX512FP16
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512dq,+avx512vl | 
FileCheck %s --check-prefixes=X64,X64-AVX,X64-AVX512,X64-AVX512VLDQ
 
 ;
 ; 128-bit Vectors
 ;
 
-define <2 x double> @fabs_v2f64(<2 x double> %p) {
-; X86-AVX-LABEL: fabs_v2f64:
-; X86-AVX:       # %bb.0:
-; X86-AVX-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
-; X86-AVX-NEXT:    retl
+define <2 x double> @fabs_v2f64(<2 x double> %p) nounwind {
+; X86-SSE-LABEL: fabs_v2f64:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    retl
+;
+; X86-AVX1OR2-LABEL: fabs_v2f64:
+; X86-AVX1OR2:       # %bb.0:
+; X86-AVX1OR2-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
+; X86-AVX1OR2-NEXT:    retl
 ;
 ; X86-AVX512VL-LABEL: fabs_v2f64:
 ; X86-AVX512VL:       # %bb.0:
@@ -35,10 +42,15 @@ define <2 x double> @fabs_v2f64(<2 x double> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to2}, %xmm0, %xmm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
-; X64-AVX-LABEL: fabs_v2f64:
-; X64-AVX:       # %bb.0:
-; X64-AVX-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
-; X64-AVX-NEXT:    retq
+; X64-SSE-LABEL: fabs_v2f64:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    retq
+;
+; X64-AVX1OR2-LABEL: fabs_v2f64:
+; X64-AVX1OR2:       # %bb.0:
+; X64-AVX1OR2-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
+; X64-AVX1OR2-NEXT:    retq
 ;
 ; X64-AVX512VL-LABEL: fabs_v2f64:
 ; X64-AVX512VL:       # %bb.0:
@@ -59,7 +71,12 @@ define <2 x double> @fabs_v2f64(<2 x double> %p) {
 }
 declare <2 x double> @llvm.fabs.v2f64(<2 x double> %p)
 
-define <4 x float> @fabs_v4f32(<4 x float> %p) {
+define <4 x float> @fabs_v4f32(<4 x float> %p) nounwind {
+; X86-SSE-LABEL: fabs_v4f32:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v4f32:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0, %xmm0
@@ -86,6 +103,11 @@ define <4 x float> @fabs_v4f32(<4 x float> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %xmm0, %xmm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v4f32:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v4f32:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0, %xmm0
@@ -116,7 +138,134 @@ define <4 x float> @fabs_v4f32(<4 x float> %p) {
 }
 declare <4 x float> @llvm.fabs.v4f32(<4 x float> %p)
 
-define <8 x half> @fabs_v8f16(ptr %p) {
+define <8 x half> @fabs_v8f16(ptr %p) nounwind {
+; X86-SSE-LABEL: fabs_v8f16:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    subl $148, %esp
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE-NEXT:    movdqa (%eax), %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm1, (%esp)
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-SSE-NEXT:    movhps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X86-SSE-NEXT:    # xmm1 = xmm1[0,1],mem[0,1]
+; X86-SSE-NEXT:    movaps %xmm1, %xmm0
+; X86-SSE-NEXT:    addl $148, %esp
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v8f16:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax
@@ -138,6 +287,74 @@ define <8 x half> @fabs_v8f16(ptr %p) {
 ; X86-AVX512-NEXT:    vpand (%eax), %xmm0, %xmm0
 ; X86-AVX512-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v8f16:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    subq $72, %rsp
+; X64-SSE-NEXT:    movdqa (%rdi), %xmm0
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movdqa %xmm1, %xmm0
+; X64-SSE-NEXT:    addq $72, %rsp
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v8f16:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vmovaps (%rdi), %xmm0
@@ -165,7 +382,14 @@ declare <8 x half> @llvm.fabs.v8f16(<8 x half> %p)
 ; 256-bit Vectors
 ;
 
-define <4 x double> @fabs_v4f64(<4 x double> %p) {
+define <4 x double> @fabs_v4f64(<4 x double> %p) nounwind {
+; X86-SSE-LABEL: fabs_v4f64:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN]
+; X86-SSE-NEXT:    andps %xmm2, %xmm0
+; X86-SSE-NEXT:    andps %xmm2, %xmm1
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v4f64:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %ymm0, %ymm0
@@ -192,6 +416,13 @@ define <4 x double> @fabs_v4f64(<4 x double> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to4}, %ymm0, %ymm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v4f64:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN]
+; X64-SSE-NEXT:    andps %xmm2, %xmm0
+; X64-SSE-NEXT:    andps %xmm2, %xmm1
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v4f64:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
@@ -222,7 +453,14 @@ define <4 x double> @fabs_v4f64(<4 x double> %p) {
 }
 declare <4 x double> @llvm.fabs.v4f64(<4 x double> %p)
 
-define <8 x float> @fabs_v8f32(<8 x float> %p) {
+define <8 x float> @fabs_v8f32(<8 x float> %p) nounwind {
+; X86-SSE-LABEL: fabs_v8f32:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN,NaN,NaN]
+; X86-SSE-NEXT:    andps %xmm2, %xmm0
+; X86-SSE-NEXT:    andps %xmm2, %xmm1
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v8f32:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}, %ymm0, %ymm0
@@ -249,6 +487,13 @@ define <8 x float> @fabs_v8f32(<8 x float> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %ymm0, %ymm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v8f32:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN,NaN,NaN]
+; X64-SSE-NEXT:    andps %xmm2, %xmm0
+; X64-SSE-NEXT:    andps %xmm2, %xmm1
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v8f32:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %ymm0, %ymm0
@@ -279,7 +524,256 @@ 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) {
+define <16 x half> @fabs_v16f16(ptr %p) nounwind {
+; X86-SSE-LABEL: fabs_v16f16:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    subl $324, %esp # imm = 0x144
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE-NEXT:    movaps (%eax), %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqa 16(%eax), %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm1, (%esp)
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqa %xmm0, %xmm3
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; X86-SSE-NEXT:    movdqu %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; X86-SSE-NEXT:    movhps {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X86-SSE-NEXT:    # xmm1 = xmm1[0,1],mem[0,1]
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    addl $324, %esp # imm = 0x144
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v16f16:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax
@@ -301,6 +795,138 @@ define <16 x half> @fabs_v16f16(ptr %p) {
 ; X86-AVX512-NEXT:    vpand (%eax), %ymm0, %ymm0
 ; X86-AVX512-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v16f16:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    subq $88, %rsp
+; X64-SSE-NEXT:    movdqa (%rdi), %xmm1
+; X64-SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movaps 16(%rdi), %xmm0
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa %xmm1, %xmm0
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd (%rsp), %xmm0 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq (%rsp), %xmm0 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq (%rsp), %xmm1 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    addq $88, %rsp
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v16f16:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vmovaps (%rdi), %ymm0
@@ -328,13 +954,28 @@ declare <16 x half> @llvm.fabs.v16f16(<16 x half> %p)
 ; 512-bit Vectors
 ;
 
-define <8 x double> @fabs_v8f64(<8 x double> %p) {
-; X86-AVX-LABEL: fabs_v8f64:
-; X86-AVX:       # %bb.0:
-; X86-AVX-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [NaN,NaN,NaN,NaN]
-; X86-AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; X86-AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-; X86-AVX-NEXT:    retl
+define <8 x double> @fabs_v8f64(<8 x double> %p) nounwind {
+; X86-SSE-LABEL: fabs_v8f64:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    pushl %ebp
+; X86-SSE-NEXT:    movl %esp, %ebp
+; X86-SSE-NEXT:    andl $-16, %esp
+; X86-SSE-NEXT:    subl $16, %esp
+; X86-SSE-NEXT:    movaps {{.*#+}} xmm3 = [NaN,NaN]
+; X86-SSE-NEXT:    andps %xmm3, %xmm0
+; X86-SSE-NEXT:    andps %xmm3, %xmm1
+; X86-SSE-NEXT:    andps %xmm3, %xmm2
+; X86-SSE-NEXT:    andps 8(%ebp), %xmm3
+; X86-SSE-NEXT:    movl %ebp, %esp
+; X86-SSE-NEXT:    popl %ebp
+; X86-SSE-NEXT:    retl
+;
+; X86-AVX1OR2-LABEL: fabs_v8f64:
+; X86-AVX1OR2:       # %bb.0:
+; X86-AVX1OR2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [NaN,NaN,NaN,NaN]
+; X86-AVX1OR2-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; X86-AVX1OR2-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; X86-AVX1OR2-NEXT:    retl
 ;
 ; X86-AVX512VL-LABEL: fabs_v8f64:
 ; X86-AVX512VL:       # %bb.0:
@@ -351,12 +992,21 @@ define <8 x double> @fabs_v8f64(<8 x double> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandpd {{\.?LCPI[0-9]+_[0-9]+}}{1to8}, %zmm0, %zmm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
-; X64-AVX-LABEL: fabs_v8f64:
-; X64-AVX:       # %bb.0:
-; X64-AVX-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [NaN,NaN,NaN,NaN]
-; X64-AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; X64-AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-; X64-AVX-NEXT:    retq
+; X64-SSE-LABEL: fabs_v8f64:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movaps {{.*#+}} xmm4 = [NaN,NaN]
+; X64-SSE-NEXT:    andps %xmm4, %xmm0
+; X64-SSE-NEXT:    andps %xmm4, %xmm1
+; X64-SSE-NEXT:    andps %xmm4, %xmm2
+; X64-SSE-NEXT:    andps %xmm4, %xmm3
+; X64-SSE-NEXT:    retq
+;
+; X64-AVX1OR2-LABEL: fabs_v8f64:
+; X64-AVX1OR2:       # %bb.0:
+; X64-AVX1OR2-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [NaN,NaN,NaN,NaN]
+; X64-AVX1OR2-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; X64-AVX1OR2-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; X64-AVX1OR2-NEXT:    retq
 ;
 ; X64-AVX512VL-LABEL: fabs_v8f64:
 ; X64-AVX512VL:       # %bb.0:
@@ -377,13 +1027,28 @@ define <8 x double> @fabs_v8f64(<8 x double> %p) {
 }
 declare <8 x double> @llvm.fabs.v8f64(<8 x double> %p)
 
-define <16 x float> @fabs_v16f32(<16 x float> %p) {
-; X86-AVX-LABEL: fabs_v16f32:
-; X86-AVX:       # %bb.0:
-; X86-AVX-NEXT:    vbroadcastss {{.*#+}} ymm2 = 
[NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
-; X86-AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; X86-AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-; X86-AVX-NEXT:    retl
+define <16 x float> @fabs_v16f32(<16 x float> %p) nounwind {
+; X86-SSE-LABEL: fabs_v16f32:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    pushl %ebp
+; X86-SSE-NEXT:    movl %esp, %ebp
+; X86-SSE-NEXT:    andl $-16, %esp
+; X86-SSE-NEXT:    subl $16, %esp
+; X86-SSE-NEXT:    movaps {{.*#+}} xmm3 = [NaN,NaN,NaN,NaN]
+; X86-SSE-NEXT:    andps %xmm3, %xmm0
+; X86-SSE-NEXT:    andps %xmm3, %xmm1
+; X86-SSE-NEXT:    andps %xmm3, %xmm2
+; X86-SSE-NEXT:    andps 8(%ebp), %xmm3
+; X86-SSE-NEXT:    movl %ebp, %esp
+; X86-SSE-NEXT:    popl %ebp
+; X86-SSE-NEXT:    retl
+;
+; X86-AVX1OR2-LABEL: fabs_v16f32:
+; X86-AVX1OR2:       # %bb.0:
+; X86-AVX1OR2-NEXT:    vbroadcastss {{.*#+}} ymm2 = 
[NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X86-AVX1OR2-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; X86-AVX1OR2-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; X86-AVX1OR2-NEXT:    retl
 ;
 ; X86-AVX512VL-LABEL: fabs_v16f32:
 ; X86-AVX512VL:       # %bb.0:
@@ -400,12 +1065,21 @@ define <16 x float> @fabs_v16f32(<16 x float> %p) {
 ; X86-AVX512VLDQ-NEXT:    vandps {{\.?LCPI[0-9]+_[0-9]+}}{1to16}, %zmm0, %zmm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
-; X64-AVX-LABEL: fabs_v16f32:
-; X64-AVX:       # %bb.0:
-; X64-AVX-NEXT:    vbroadcastss {{.*#+}} ymm2 = 
[NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
-; X64-AVX-NEXT:    vandps %ymm2, %ymm0, %ymm0
-; X64-AVX-NEXT:    vandps %ymm2, %ymm1, %ymm1
-; X64-AVX-NEXT:    retq
+; X64-SSE-LABEL: fabs_v16f32:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movaps {{.*#+}} xmm4 = [NaN,NaN,NaN,NaN]
+; X64-SSE-NEXT:    andps %xmm4, %xmm0
+; X64-SSE-NEXT:    andps %xmm4, %xmm1
+; X64-SSE-NEXT:    andps %xmm4, %xmm2
+; X64-SSE-NEXT:    andps %xmm4, %xmm3
+; X64-SSE-NEXT:    retq
+;
+; X64-AVX1OR2-LABEL: fabs_v16f32:
+; X64-AVX1OR2:       # %bb.0:
+; X64-AVX1OR2-NEXT:    vbroadcastss {{.*#+}} ymm2 = 
[NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
+; X64-AVX1OR2-NEXT:    vandps %ymm2, %ymm0, %ymm0
+; X64-AVX1OR2-NEXT:    vandps %ymm2, %ymm1, %ymm1
+; X64-AVX1OR2-NEXT:    retq
 ;
 ; X64-AVX512VL-LABEL: fabs_v16f32:
 ; X64-AVX512VL:       # %bb.0:
@@ -426,7 +1100,502 @@ 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) {
+define <32 x half> @fabs_v32f16(ptr %p) nounwind {
+; X86-SSE-LABEL: fabs_v32f16:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    subl $644, %esp # imm = 0x284
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE-NEXT:    movaps (%eax), %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movaps 16(%eax), %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movaps 32(%eax), %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqa 48(%eax), %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrlq $48, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    psrld $16, %xmm0
+; X86-SSE-NEXT:    pextrw $0, %xmm0, %eax
+; X86-SSE-NEXT:    movw %ax, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movups %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __extendhfsf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm0, (%esp)
+; X86-SSE-NEXT:    fstps {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; X86-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}, %xmm0
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    movss %xmm1, (%esp)
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqa %xmm0, %xmm3
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqa %xmm0, %xmm3
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    punpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqa %xmm0, %xmm3
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
+; X86-SSE-NEXT:    punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; X86-SSE-NEXT:    movdqu %xmm1, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm2 = 
xmm2[0],xmm0[0],xmm2[1],xmm0[1],xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm0 = 
xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X86-SSE-NEXT:    movdqu %xmm0, {{[-0-9]+}}(%e{{[sb]}}p) # 16-byte Spill
+; X86-SSE-NEXT:    calll __truncsfhf2
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Reload
+; X86-SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = 
xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; X86-SSE-NEXT:    movdqu {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; X86-SSE-NEXT:    movhps {{[-0-9]+}}(%e{{[sb]}}p), %xmm3 # 16-byte Folded 
Reload
+; X86-SSE-NEXT:    # xmm3 = xmm3[0,1],mem[0,1]
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm0 # 16-byte Reload
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm1 # 16-byte Reload
+; X86-SSE-NEXT:    movups {{[-0-9]+}}(%e{{[sb]}}p), %xmm2 # 16-byte Reload
+; X86-SSE-NEXT:    addl $644, %esp # imm = 0x284
+; X86-SSE-NEXT:    retl
+;
 ; X86-AVX1-LABEL: fabs_v32f16:
 ; X86-AVX1:       # %bb.0:
 ; X86-AVX1-NEXT:    movl {{[0-9]+}}(%esp), %eax
@@ -466,6 +1635,266 @@ define <32 x half> @fabs_v32f16(ptr %p) {
 ; X86-AVX512VLDQ-NEXT:    vpandq (%eax), %zmm0, %zmm0
 ; X86-AVX512VLDQ-NEXT:    retl
 ;
+; X64-SSE-LABEL: fabs_v32f16:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    subq $120, %rsp
+; X64-SSE-NEXT:    movdqa (%rdi), %xmm1
+; X64-SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps 16(%rdi), %xmm0
+; X64-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movaps 32(%rdi), %xmm0
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps 48(%rdi), %xmm0
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa %xmm1, %xmm0
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd (%rsp), %xmm0 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq (%rsp), %xmm0 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd (%rsp), %xmm0 # 16-byte Folded Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa (%rsp), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm1 = 
xmm1[0],xmm0[0],xmm1[1],xmm0[1],xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm1 = xmm1[0],mem[0]
+; X64-SSE-NEXT:    movdqa %xmm1, (%rsp) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrldq {{.*#+}} xmm0 = 
xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    punpckhqdq {{.*#+}} xmm0 = xmm0[1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrlq $48, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm0 = 
xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
+; X64-SSE-NEXT:    movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    andps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    psrld $16, %xmm0
+; X64-SSE-NEXT:    callq __extendhfsf2@PLT
+; X64-SSE-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-SSE-NEXT:    callq __truncsfhf2@PLT
+; X64-SSE-NEXT:    movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Reload
+; X64-SSE-NEXT:    punpcklwd {{.*#+}} xmm3 = 
xmm3[0],xmm0[0],xmm3[1],xmm0[1],xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; X64-SSE-NEXT:    punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm3 = xmm3[0],mem[0],xmm3[1],mem[1]
+; X64-SSE-NEXT:    punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded 
Reload
+; X64-SSE-NEXT:    # xmm3 = xmm3[0],mem[0]
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
+; X64-SSE-NEXT:    movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
+; X64-SSE-NEXT:    movaps (%rsp), %xmm2 # 16-byte Reload
+; X64-SSE-NEXT:    addq $120, %rsp
+; X64-SSE-NEXT:    retq
+;
 ; X64-AVX1-LABEL: fabs_v32f16:
 ; X64-AVX1:       # %bb.0:
 ; X64-AVX1-NEXT:    vbroadcastss {{.*#+}} ymm1 = 
[NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN,NaN]
@@ -520,7 +1949,7 @@ declare <32 x half> @llvm.fabs.v32f16(<32 x half> %p)
 ; We should generate:
 ;    mov     (put constant value in return register)
 
-define i64 @fabs_v2f32_1() {
+define i64 @fabs_v2f32_1() nounwind {
 ; X86-LABEL: fabs_v2f32_1:
 ; X86:       # %bb.0:
 ; X86-NEXT:    xorl %eax, %eax
@@ -537,7 +1966,7 @@ define i64 @fabs_v2f32_1() {
  ret i64 %ret
 }
 
-define i64 @fabs_v2f32_2() {
+define i64 @fabs_v2f32_2() nounwind {
 ; X86-LABEL: fabs_v2f32_2:
 ; X86:       # %bb.0:
 ; X86-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
@@ -557,28 +1986,52 @@ define i64 @fabs_v2f32_2() {
 declare <2 x float> @llvm.fabs.v2f32(<2 x float> %p)
 
 ; PR70947 - remove duplicate xmm/ymm constant loads
-define void @PR70947(ptr %src, ptr %dst) {
-; X86-LABEL: PR70947:
-; X86:       # %bb.0:
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
-; X86-NEXT:    movl {{[0-9]+}}(%esp), %ecx
-; X86-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X86-NEXT:    vandps (%ecx), %ymm0, %ymm1
-; X86-NEXT:    vandps 32(%ecx), %xmm0, %xmm0
-; X86-NEXT:    vmovups %ymm1, (%eax)
-; X86-NEXT:    vmovups %xmm0, 16(%eax)
-; X86-NEXT:    vzeroupper
-; X86-NEXT:    retl
+define void @PR70947(ptr %src, ptr %dst) nounwind {
+; X86-SSE-LABEL: PR70947:
+; X86-SSE:       # %bb.0:
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-SSE-NEXT:    movups (%ecx), %xmm0
+; X86-SSE-NEXT:    movups 32(%ecx), %xmm1
+; X86-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN]
+; X86-SSE-NEXT:    andps %xmm2, %xmm0
+; X86-SSE-NEXT:    andps %xmm2, %xmm1
+; X86-SSE-NEXT:    movups %xmm0, (%eax)
+; X86-SSE-NEXT:    movups %xmm1, 16(%eax)
+; X86-SSE-NEXT:    retl
+;
+; X86-AVX-LABEL: PR70947:
+; X86-AVX:       # %bb.0:
+; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-AVX-NEXT:    movl {{[0-9]+}}(%esp), %ecx
+; X86-AVX-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
+; X86-AVX-NEXT:    vandps (%ecx), %ymm0, %ymm1
+; X86-AVX-NEXT:    vandps 32(%ecx), %xmm0, %xmm0
+; X86-AVX-NEXT:    vmovups %ymm1, (%eax)
+; X86-AVX-NEXT:    vmovups %xmm0, 16(%eax)
+; X86-AVX-NEXT:    vzeroupper
+; X86-AVX-NEXT:    retl
 ;
-; X64-LABEL: PR70947:
-; X64:       # %bb.0:
-; X64-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
-; X64-NEXT:    vandps (%rdi), %ymm0, %ymm1
-; X64-NEXT:    vandps 32(%rdi), %xmm0, %xmm0
-; X64-NEXT:    vmovups %ymm1, (%rsi)
-; X64-NEXT:    vmovups %xmm0, 16(%rsi)
-; X64-NEXT:    vzeroupper
-; X64-NEXT:    retq
+; X64-SSE-LABEL: PR70947:
+; X64-SSE:       # %bb.0:
+; X64-SSE-NEXT:    movups (%rdi), %xmm0
+; X64-SSE-NEXT:    movups 32(%rdi), %xmm1
+; X64-SSE-NEXT:    movaps {{.*#+}} xmm2 = [NaN,NaN]
+; X64-SSE-NEXT:    andps %xmm2, %xmm0
+; X64-SSE-NEXT:    andps %xmm2, %xmm1
+; X64-SSE-NEXT:    movups %xmm0, (%rsi)
+; X64-SSE-NEXT:    movups %xmm1, 16(%rsi)
+; X64-SSE-NEXT:    retq
+;
+; X64-AVX-LABEL: PR70947:
+; X64-AVX:       # %bb.0:
+; X64-AVX-NEXT:    vbroadcastsd {{.*#+}} ymm0 = [NaN,NaN,NaN,NaN]
+; X64-AVX-NEXT:    vandps (%rdi), %ymm0, %ymm1
+; X64-AVX-NEXT:    vandps 32(%rdi), %xmm0, %xmm0
+; X64-AVX-NEXT:    vmovups %ymm1, (%rsi)
+; X64-AVX-NEXT:    vmovups %xmm0, 16(%rsi)
+; X64-AVX-NEXT:    vzeroupper
+; X64-AVX-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


        
_______________________________________________
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits