From: Craig Topper <craig.topper@intel.com>
Date: Sun, 22 Sep 2019 06:52:25 +0000 (+0000)
Subject: [X86] Add test memset and memcpy testcases for D67874. NFC
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=38014c553f0d8aa8b5db4fe7b0a2d98c044580f3;p=platform%2Fupstream%2Fllvm.git

[X86] Add test memset and memcpy testcases for D67874. NFC

llvm-svn: 372494
---

diff --git a/llvm/test/CodeGen/X86/memcpy.ll b/llvm/test/CodeGen/X86/memcpy.ll
index 597b776..d2454d5 100644
--- a/llvm/test/CodeGen/X86/memcpy.ll
+++ b/llvm/test/CodeGen/X86/memcpy.ll
@@ -1,6 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=core2 | FileCheck %s -check-prefix=LINUX
-; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core2 | FileCheck %s -check-prefix=DARWIN
+; RUN: llc < %s -mtriple=x86_64-apple-darwin      -mcpu=core2   | FileCheck %s -check-prefix=DARWIN
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=core2   | FileCheck %s -check-prefix=LINUX
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=skylake | FileCheck %s -check-prefix=LINUX-SKL
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=skx     | FileCheck %s -check-prefix=LINUX-SKX
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=knl     | FileCheck %s -check-prefix=LINUX-KNL
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=avx512bw | FileCheck %s -check-prefix=LINUX-AVX512BW
 
 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i1) nounwind
 declare void @llvm.memcpy.p256i8.p256i8.i64(i8 addrspace(256)* nocapture, i8 addrspace(256)* nocapture, i64, i1) nounwind
@@ -8,13 +12,29 @@ declare void @llvm.memcpy.p256i8.p256i8.i64(i8 addrspace(256)* nocapture, i8 add
 
 ; Variable memcpy's should lower to calls.
 define i8* @test1(i8* %a, i8* %b, i64 %n) nounwind {
+; DARWIN-LABEL: test1:
+; DARWIN:       ## %bb.0: ## %entry
+; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+;
 ; LINUX-LABEL: test1:
 ; LINUX:       # %bb.0: # %entry
 ; LINUX-NEXT:    jmp memcpy # TAILCALL
 ;
-; DARWIN-LABEL: test1:
-; DARWIN:       ## %bb.0: ## %entry
-; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+; LINUX-SKL-LABEL: test1:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-SKX-LABEL: test1:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-KNL-LABEL: test1:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-AVX512BW-LABEL: test1:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    jmp memcpy # TAILCALL
 entry:
 	tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %b, i64 %n, i1 0 )
 	ret i8* %a
@@ -22,13 +42,29 @@ entry:
 
 ; Variable memcpy's should lower to calls.
 define i8* @test2(i64* %a, i64* %b, i64 %n) nounwind {
+; DARWIN-LABEL: test2:
+; DARWIN:       ## %bb.0: ## %entry
+; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+;
 ; LINUX-LABEL: test2:
 ; LINUX:       # %bb.0: # %entry
 ; LINUX-NEXT:    jmp memcpy # TAILCALL
 ;
-; DARWIN-LABEL: test2:
-; DARWIN:       ## %bb.0: ## %entry
-; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+; LINUX-SKL-LABEL: test2:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-SKX-LABEL: test2:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-KNL-LABEL: test2:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-AVX512BW-LABEL: test2:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    jmp memcpy # TAILCALL
 entry:
 	%tmp14 = bitcast i64* %a to i8*
 	%tmp25 = bitcast i64* %b to i8*
@@ -43,11 +79,6 @@ entry:
 ; hurting performance so it should just ignore optsize when expanding memcpy.
 ; rdar://8821501
 define void @test3(i8* nocapture %A, i8* nocapture %B) nounwind optsize noredzone {
-; LINUX-LABEL: test3:
-; LINUX:       # %bb.0: # %entry
-; LINUX-NEXT:    movl $64, %edx
-; LINUX-NEXT:    jmp memcpy # TAILCALL
-;
 ; DARWIN-LABEL: test3:
 ; DARWIN:       ## %bb.0: ## %entry
 ; DARWIN-NEXT:    movq 56(%rsi), %rax
@@ -67,45 +98,169 @@ define void @test3(i8* nocapture %A, i8* nocapture %B) nounwind optsize noredzon
 ; DARWIN-NEXT:    movq %rcx, 8(%rdi)
 ; DARWIN-NEXT:    movq %rax, (%rdi)
 ; DARWIN-NEXT:    retq
+;
+; LINUX-LABEL: test3:
+; LINUX:       # %bb.0: # %entry
+; LINUX-NEXT:    movl $64, %edx
+; LINUX-NEXT:    jmp memcpy # TAILCALL
+;
+; LINUX-SKL-LABEL: test3:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKL-NEXT:    vzeroupper
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test3:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKX-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKX-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKX-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKX-NEXT:    vzeroupper
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test3:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-KNL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-KNL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-KNL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test3:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    vmovups (%rsi), %zmm0
+; LINUX-AVX512BW-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    vzeroupper
+; LINUX-AVX512BW-NEXT:    retq
 entry:
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i1 false)
   ret void
 }
 
 define void @test3_minsize(i8* nocapture %A, i8* nocapture %B) nounwind minsize noredzone {
+; DARWIN-LABEL: test3_minsize:
+; DARWIN:       ## %bb.0:
+; DARWIN-NEXT:    pushq $64
+; DARWIN-NEXT:    popq %rdx
+; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+;
 ; LINUX-LABEL: test3_minsize:
 ; LINUX:       # %bb.0:
 ; LINUX-NEXT:    pushq $64
 ; LINUX-NEXT:    popq %rdx
 ; LINUX-NEXT:    jmp memcpy # TAILCALL
 ;
-; DARWIN-LABEL: test3_minsize:
-; DARWIN:       ## %bb.0:
-; DARWIN-NEXT:    pushq $64
-; DARWIN-NEXT:    popq %rdx
-; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+; LINUX-SKL-LABEL: test3_minsize:
+; LINUX-SKL:       # %bb.0:
+; LINUX-SKL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKL-NEXT:    vzeroupper
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test3_minsize:
+; LINUX-SKX:       # %bb.0:
+; LINUX-SKX-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKX-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKX-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKX-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKX-NEXT:    vzeroupper
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test3_minsize:
+; LINUX-KNL:       # %bb.0:
+; LINUX-KNL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-KNL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-KNL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-KNL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test3_minsize:
+; LINUX-AVX512BW:       # %bb.0:
+; LINUX-AVX512BW-NEXT:    vmovups (%rsi), %zmm0
+; LINUX-AVX512BW-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    vzeroupper
+; LINUX-AVX512BW-NEXT:    retq
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i1 false)
   ret void
 }
 
 define void @test3_minsize_optsize(i8* nocapture %A, i8* nocapture %B) nounwind optsize minsize noredzone {
+; DARWIN-LABEL: test3_minsize_optsize:
+; DARWIN:       ## %bb.0:
+; DARWIN-NEXT:    pushq $64
+; DARWIN-NEXT:    popq %rdx
+; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+;
 ; LINUX-LABEL: test3_minsize_optsize:
 ; LINUX:       # %bb.0:
 ; LINUX-NEXT:    pushq $64
 ; LINUX-NEXT:    popq %rdx
 ; LINUX-NEXT:    jmp memcpy # TAILCALL
 ;
-; DARWIN-LABEL: test3_minsize_optsize:
-; DARWIN:       ## %bb.0:
-; DARWIN-NEXT:    pushq $64
-; DARWIN-NEXT:    popq %rdx
-; DARWIN-NEXT:    jmp _memcpy ## TAILCALL
+; LINUX-SKL-LABEL: test3_minsize_optsize:
+; LINUX-SKL:       # %bb.0:
+; LINUX-SKL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKL-NEXT:    vzeroupper
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test3_minsize_optsize:
+; LINUX-SKX:       # %bb.0:
+; LINUX-SKX-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKX-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKX-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKX-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKX-NEXT:    vzeroupper
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test3_minsize_optsize:
+; LINUX-KNL:       # %bb.0:
+; LINUX-KNL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-KNL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-KNL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-KNL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test3_minsize_optsize:
+; LINUX-AVX512BW:       # %bb.0:
+; LINUX-AVX512BW-NEXT:    vmovups (%rsi), %zmm0
+; LINUX-AVX512BW-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    vzeroupper
+; LINUX-AVX512BW-NEXT:    retq
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i1 false)
   ret void
 }
 
 ; Large constant memcpy's should be inlined when not optimizing for size.
 define void @test4(i8* nocapture %A, i8* nocapture %B) nounwind noredzone {
+; DARWIN-LABEL: test4:
+; DARWIN:       ## %bb.0: ## %entry
+; DARWIN-NEXT:    movq 56(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 56(%rdi)
+; DARWIN-NEXT:    movq 48(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 48(%rdi)
+; DARWIN-NEXT:    movq 40(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 40(%rdi)
+; DARWIN-NEXT:    movq 32(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 32(%rdi)
+; DARWIN-NEXT:    movq 24(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 24(%rdi)
+; DARWIN-NEXT:    movq 16(%rsi), %rax
+; DARWIN-NEXT:    movq %rax, 16(%rdi)
+; DARWIN-NEXT:    movq (%rsi), %rax
+; DARWIN-NEXT:    movq 8(%rsi), %rcx
+; DARWIN-NEXT:    movq %rcx, 8(%rdi)
+; DARWIN-NEXT:    movq %rax, (%rdi)
+; DARWIN-NEXT:    retq
+;
 ; LINUX-LABEL: test4:
 ; LINUX:       # %bb.0: # %entry
 ; LINUX-NEXT:    movq 56(%rsi), %rax
@@ -126,25 +281,38 @@ define void @test4(i8* nocapture %A, i8* nocapture %B) nounwind noredzone {
 ; LINUX-NEXT:    movq %rax, (%rdi)
 ; LINUX-NEXT:    retq
 ;
-; DARWIN-LABEL: test4:
-; DARWIN:       ## %bb.0: ## %entry
-; DARWIN-NEXT:    movq 56(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 56(%rdi)
-; DARWIN-NEXT:    movq 48(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 48(%rdi)
-; DARWIN-NEXT:    movq 40(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 40(%rdi)
-; DARWIN-NEXT:    movq 32(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 32(%rdi)
-; DARWIN-NEXT:    movq 24(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 24(%rdi)
-; DARWIN-NEXT:    movq 16(%rsi), %rax
-; DARWIN-NEXT:    movq %rax, 16(%rdi)
-; DARWIN-NEXT:    movq (%rsi), %rax
-; DARWIN-NEXT:    movq 8(%rsi), %rcx
-; DARWIN-NEXT:    movq %rcx, 8(%rdi)
-; DARWIN-NEXT:    movq %rax, (%rdi)
-; DARWIN-NEXT:    retq
+; LINUX-SKL-LABEL: test4:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKL-NEXT:    vzeroupper
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test4:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-SKX-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-SKX-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-SKX-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-SKX-NEXT:    vzeroupper
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test4:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    vmovups (%rsi), %ymm0
+; LINUX-KNL-NEXT:    vmovups 32(%rsi), %ymm1
+; LINUX-KNL-NEXT:    vmovups %ymm1, 32(%rdi)
+; LINUX-KNL-NEXT:    vmovups %ymm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test4:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    vmovups (%rsi), %zmm0
+; LINUX-AVX512BW-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    vzeroupper
+; LINUX-AVX512BW-NEXT:    retq
 entry:
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %A, i8* %B, i64 64, i1 false)
   ret void
@@ -154,6 +322,14 @@ entry:
 @.str = private unnamed_addr constant [30 x i8] c"\00aaaaaaaaaaaaaaaaaaaaaaaaaaaa\00", align 1
 
 define void @test5(i8* nocapture %C) nounwind uwtable ssp {
+; DARWIN-LABEL: test5:
+; DARWIN:       ## %bb.0: ## %entry
+; DARWIN-NEXT:    movabsq $7016996765293437281, %rax ## imm = 0x6161616161616161
+; DARWIN-NEXT:    movq %rax, 8(%rdi)
+; DARWIN-NEXT:    movabsq $7016996765293437184, %rax ## imm = 0x6161616161616100
+; DARWIN-NEXT:    movq %rax, (%rdi)
+; DARWIN-NEXT:    retq
+;
 ; LINUX-LABEL: test5:
 ; LINUX:       # %bb.0: # %entry
 ; LINUX-NEXT:    movabsq $7016996765293437281, %rax # imm = 0x6161616161616161
@@ -162,13 +338,29 @@ define void @test5(i8* nocapture %C) nounwind uwtable ssp {
 ; LINUX-NEXT:    movq %rax, (%rdi)
 ; LINUX-NEXT:    retq
 ;
-; DARWIN-LABEL: test5:
-; DARWIN:       ## %bb.0: ## %entry
-; DARWIN-NEXT:    movabsq $7016996765293437281, %rax ## imm = 0x6161616161616161
-; DARWIN-NEXT:    movq %rax, 8(%rdi)
-; DARWIN-NEXT:    movabsq $7016996765293437184, %rax ## imm = 0x6161616161616100
-; DARWIN-NEXT:    movq %rax, (%rdi)
-; DARWIN-NEXT:    retq
+; LINUX-SKL-LABEL: test5:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    vmovups {{.*}}(%rip), %xmm0
+; LINUX-SKL-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test5:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    vmovups {{.*}}(%rip), %xmm0
+; LINUX-SKX-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test5:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    vmovups {{.*}}(%rip), %xmm0
+; LINUX-KNL-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test5:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    vmovups {{.*}}(%rip), %xmm0
+; LINUX-AVX512BW-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    retq
 entry:
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %C, i8* getelementptr inbounds ([30 x i8], [30 x i8]* @.str, i64 0, i64 0), i64 16, i1 false)
   ret void
@@ -179,17 +371,41 @@ entry:
 @.str2 = private unnamed_addr constant [2 x i8] c"x\00", align 1
 
 define void @test6() nounwind uwtable {
+; DARWIN-LABEL: test6:
+; DARWIN:       ## %bb.0: ## %entry
+; DARWIN-NEXT:    movw $0, 8
+; DARWIN-NEXT:    movq $120, 0
+; DARWIN-NEXT:    retq
+;
 ; LINUX-LABEL: test6:
 ; LINUX:       # %bb.0: # %entry
 ; LINUX-NEXT:    movw $0, 8
 ; LINUX-NEXT:    movq $120, 0
 ; LINUX-NEXT:    retq
 ;
-; DARWIN-LABEL: test6:
-; DARWIN:       ## %bb.0: ## %entry
-; DARWIN-NEXT:    movw $0, 8
-; DARWIN-NEXT:    movq $120, 0
-; DARWIN-NEXT:    retq
+; LINUX-SKL-LABEL: test6:
+; LINUX-SKL:       # %bb.0: # %entry
+; LINUX-SKL-NEXT:    movw $0, 8
+; LINUX-SKL-NEXT:    movq $120, 0
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: test6:
+; LINUX-SKX:       # %bb.0: # %entry
+; LINUX-SKX-NEXT:    movw $0, 8
+; LINUX-SKX-NEXT:    movq $120, 0
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: test6:
+; LINUX-KNL:       # %bb.0: # %entry
+; LINUX-KNL-NEXT:    movw $0, 8
+; LINUX-KNL-NEXT:    movq $120, 0
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: test6:
+; LINUX-AVX512BW:       # %bb.0: # %entry
+; LINUX-AVX512BW-NEXT:    movw $0, 8
+; LINUX-AVX512BW-NEXT:    movq $120, 0
+; LINUX-AVX512BW-NEXT:    retq
 entry:
   tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* null, i8* getelementptr inbounds ([2 x i8], [2 x i8]* @.str2, i64 0, i64 0), i64 10, i1 false)
   ret void
@@ -198,6 +414,16 @@ entry:
 define void @PR15348(i8* %a, i8* %b) {
 ; Ensure that alignment of '0' in an @llvm.memcpy intrinsic results in
 ; unaligned loads and stores.
+; DARWIN-LABEL: PR15348:
+; DARWIN:       ## %bb.0:
+; DARWIN-NEXT:    movb 16(%rsi), %al
+; DARWIN-NEXT:    movb %al, 16(%rdi)
+; DARWIN-NEXT:    movq (%rsi), %rax
+; DARWIN-NEXT:    movq 8(%rsi), %rcx
+; DARWIN-NEXT:    movq %rcx, 8(%rdi)
+; DARWIN-NEXT:    movq %rax, (%rdi)
+; DARWIN-NEXT:    retq
+;
 ; LINUX-LABEL: PR15348:
 ; LINUX:       # %bb.0:
 ; LINUX-NEXT:    movb 16(%rsi), %al
@@ -208,15 +434,37 @@ define void @PR15348(i8* %a, i8* %b) {
 ; LINUX-NEXT:    movq %rax, (%rdi)
 ; LINUX-NEXT:    retq
 ;
-; DARWIN-LABEL: PR15348:
-; DARWIN:       ## %bb.0:
-; DARWIN-NEXT:    movb 16(%rsi), %al
-; DARWIN-NEXT:    movb %al, 16(%rdi)
-; DARWIN-NEXT:    movq (%rsi), %rax
-; DARWIN-NEXT:    movq 8(%rsi), %rcx
-; DARWIN-NEXT:    movq %rcx, 8(%rdi)
-; DARWIN-NEXT:    movq %rax, (%rdi)
-; DARWIN-NEXT:    retq
+; LINUX-SKL-LABEL: PR15348:
+; LINUX-SKL:       # %bb.0:
+; LINUX-SKL-NEXT:    movb 16(%rsi), %al
+; LINUX-SKL-NEXT:    movb %al, 16(%rdi)
+; LINUX-SKL-NEXT:    vmovups (%rsi), %xmm0
+; LINUX-SKL-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: PR15348:
+; LINUX-SKX:       # %bb.0:
+; LINUX-SKX-NEXT:    movb 16(%rsi), %al
+; LINUX-SKX-NEXT:    movb %al, 16(%rdi)
+; LINUX-SKX-NEXT:    vmovups (%rsi), %xmm0
+; LINUX-SKX-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: PR15348:
+; LINUX-KNL:       # %bb.0:
+; LINUX-KNL-NEXT:    movb 16(%rsi), %al
+; LINUX-KNL-NEXT:    movb %al, 16(%rdi)
+; LINUX-KNL-NEXT:    vmovups (%rsi), %xmm0
+; LINUX-KNL-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: PR15348:
+; LINUX-AVX512BW:       # %bb.0:
+; LINUX-AVX512BW-NEXT:    movb 16(%rsi), %al
+; LINUX-AVX512BW-NEXT:    movb %al, 16(%rdi)
+; LINUX-AVX512BW-NEXT:    vmovups (%rsi), %xmm0
+; LINUX-AVX512BW-NEXT:    vmovups %xmm0, (%rdi)
+; LINUX-AVX512BW-NEXT:    retq
   call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* %b, i64 17, i1 false)
   ret void
 }
@@ -224,6 +472,14 @@ define void @PR15348(i8* %a, i8* %b) {
 ; Memcpys from / to address space 256 should be lowered to appropriate loads /
 ; stores if small enough.
 define void @addrspace256(i8 addrspace(256)* %a, i8 addrspace(256)* %b) nounwind {
+; DARWIN-LABEL: addrspace256:
+; DARWIN:       ## %bb.0:
+; DARWIN-NEXT:    movq %gs:(%rsi), %rax
+; DARWIN-NEXT:    movq %gs:8(%rsi), %rcx
+; DARWIN-NEXT:    movq %rcx, %gs:8(%rdi)
+; DARWIN-NEXT:    movq %rax, %gs:(%rdi)
+; DARWIN-NEXT:    retq
+;
 ; LINUX-LABEL: addrspace256:
 ; LINUX:       # %bb.0:
 ; LINUX-NEXT:    movq %gs:(%rsi), %rax
@@ -232,13 +488,29 @@ define void @addrspace256(i8 addrspace(256)* %a, i8 addrspace(256)* %b) nounwind
 ; LINUX-NEXT:    movq %rax, %gs:(%rdi)
 ; LINUX-NEXT:    retq
 ;
-; DARWIN-LABEL: addrspace256:
-; DARWIN:       ## %bb.0:
-; DARWIN-NEXT:    movq %gs:(%rsi), %rax
-; DARWIN-NEXT:    movq %gs:8(%rsi), %rcx
-; DARWIN-NEXT:    movq %rcx, %gs:8(%rdi)
-; DARWIN-NEXT:    movq %rax, %gs:(%rdi)
-; DARWIN-NEXT:    retq
+; LINUX-SKL-LABEL: addrspace256:
+; LINUX-SKL:       # %bb.0:
+; LINUX-SKL-NEXT:    vmovups %gs:(%rsi), %xmm0
+; LINUX-SKL-NEXT:    vmovups %xmm0, %gs:(%rdi)
+; LINUX-SKL-NEXT:    retq
+;
+; LINUX-SKX-LABEL: addrspace256:
+; LINUX-SKX:       # %bb.0:
+; LINUX-SKX-NEXT:    vmovups %gs:(%rsi), %xmm0
+; LINUX-SKX-NEXT:    vmovups %xmm0, %gs:(%rdi)
+; LINUX-SKX-NEXT:    retq
+;
+; LINUX-KNL-LABEL: addrspace256:
+; LINUX-KNL:       # %bb.0:
+; LINUX-KNL-NEXT:    vmovups %gs:(%rsi), %xmm0
+; LINUX-KNL-NEXT:    vmovups %xmm0, %gs:(%rdi)
+; LINUX-KNL-NEXT:    retq
+;
+; LINUX-AVX512BW-LABEL: addrspace256:
+; LINUX-AVX512BW:       # %bb.0:
+; LINUX-AVX512BW-NEXT:    vmovups %gs:(%rsi), %xmm0
+; LINUX-AVX512BW-NEXT:    vmovups %xmm0, %gs:(%rdi)
+; LINUX-AVX512BW-NEXT:    retq
   tail call void @llvm.memcpy.p256i8.p256i8.i64(i8 addrspace(256)* align 8 %a, i8 addrspace(256)* align 8 %b, i64 16, i1 false)
   ret void
 }
diff --git a/llvm/test/CodeGen/X86/memset-nonzero.ll b/llvm/test/CodeGen/X86/memset-nonzero.ll
index 8166d00..8d1eeba 100644
--- a/llvm/test/CodeGen/X86/memset-nonzero.ll
+++ b/llvm/test/CodeGen/X86/memset-nonzero.ll
@@ -4,6 +4,8 @@
 ; RUN: llc -mtriple=x86_64-unknown-unknown < %s -mattr=sse2,-slow-unaligned-mem-16 | FileCheck %s --check-prefix=SSE2FAST
 ; RUN: llc -mtriple=x86_64-unknown-unknown < %s -mattr=avx | FileCheck %s --check-prefix=AVX --check-prefix=AVX1
 ; RUN: llc -mtriple=x86_64-unknown-unknown < %s -mattr=avx2 | FileCheck %s --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc -mtriple=x86_64-unknown-unknown < %s -mattr=avx512f | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: llc -mtriple=x86_64-unknown-unknown < %s -mattr=avx512bw | FileCheck %s --check-prefix=AVX --check-prefix=AVX512 --check-prefix=AVX512BW
 
 ; https://llvm.org/bugs/show_bug.cgi?id=27100
 
@@ -227,6 +229,13 @@ define void @memset_16_nonconst_bytes(i8* %x, i8 %c) {
 ; AVX2-NEXT:    vpbroadcastb %xmm0, %xmm0
 ; AVX2-NEXT:    vmovdqu %xmm0, (%rdi)
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: memset_16_nonconst_bytes:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovd %esi, %xmm0
+; AVX512-NEXT:    vpbroadcastb %xmm0, %xmm0
+; AVX512-NEXT:    vmovdqu %xmm0, (%rdi)
+; AVX512-NEXT:    retq
   tail call void @llvm.memset.p0i8.i64(i8* %x, i8 %c, i64 16, i1 false)
   ret void
 }
@@ -270,6 +279,14 @@ define void @memset_32_nonconst_bytes(i8* %x, i8 %c) {
 ; AVX2-NEXT:    vmovdqu %ymm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: memset_32_nonconst_bytes:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovd %esi, %xmm0
+; AVX512-NEXT:    vpbroadcastb %xmm0, %ymm0
+; AVX512-NEXT:    vmovdqu %ymm0, (%rdi)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   tail call void @llvm.memset.p0i8.i64(i8* %x, i8 %c, i64 32, i1 false)
   ret void
 }
@@ -322,6 +339,15 @@ define void @memset_64_nonconst_bytes(i8* %x, i8 %c) {
 ; AVX2-NEXT:    vmovdqu %ymm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: memset_64_nonconst_bytes:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovd %esi, %xmm0
+; AVX512-NEXT:    vpbroadcastb %xmm0, %ymm0
+; AVX512-NEXT:    vmovdqu %ymm0, 32(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, (%rdi)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   tail call void @llvm.memset.p0i8.i64(i8* %x, i8 %c, i64 64, i1 false)
   ret void
 }
@@ -390,6 +416,17 @@ define void @memset_128_nonconst_bytes(i8* %x, i8 %c) {
 ; AVX2-NEXT:    vmovdqu %ymm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: memset_128_nonconst_bytes:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovd %esi, %xmm0
+; AVX512-NEXT:    vpbroadcastb %xmm0, %ymm0
+; AVX512-NEXT:    vmovdqu %ymm0, 96(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 64(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 32(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, (%rdi)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   tail call void @llvm.memset.p0i8.i64(i8* %x, i8 %c, i64 128, i1 false)
   ret void
 }
@@ -455,6 +492,21 @@ define void @memset_256_nonconst_bytes(i8* %x, i8 %c) {
 ; AVX2-NEXT:    vmovdqu %ymm0, (%rdi)
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: memset_256_nonconst_bytes:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vmovd %esi, %xmm0
+; AVX512-NEXT:    vpbroadcastb %xmm0, %ymm0
+; AVX512-NEXT:    vmovdqu %ymm0, 224(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 192(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 160(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 128(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 96(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 64(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, 32(%rdi)
+; AVX512-NEXT:    vmovdqu %ymm0, (%rdi)
+; AVX512-NEXT:    vzeroupper
+; AVX512-NEXT:    retq
   tail call void @llvm.memset.p0i8.i64(i8* %x, i8 %c, i64 256, i1 false)
   ret void
 }
diff --git a/llvm/test/CodeGen/X86/memset-zero.ll b/llvm/test/CodeGen/X86/memset-zero.ll
index fb28742..14d7472 100644
--- a/llvm/test/CodeGen/X86/memset-zero.ll
+++ b/llvm/test/CodeGen/X86/memset-zero.ll
@@ -4,6 +4,7 @@
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux -mcpu=nehalem | FileCheck %s --check-prefix=NEHALEM
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux -mcpu=sandybridge | FileCheck %s --check-prefix=SANDYBRIDGE
 ; RUN: llc < %s -mtriple=x86_64-unknown-linux -mcpu=skylake | FileCheck %s --check-prefix=SKYLAKE
+; RUN: llc < %s -mtriple=x86_64-unknown-linux -mcpu=knl | FileCheck %s --check-prefix=KNL
 
 
 declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1) nounwind
@@ -28,6 +29,10 @@ define void @memset_0(i8* %a) nounwind  {
 ; SKYLAKE-LABEL: memset_0:
 ; SKYLAKE:       # %bb.0: # %entry
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_0:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 0, i1 false)
 	ret void
@@ -59,6 +64,11 @@ define void @memset_4(i8* %a) nounwind  {
 ; SKYLAKE:       # %bb.0: # %entry
 ; SKYLAKE-NEXT:    movl $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_4:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movl $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 4, i1 false)
 	ret void
@@ -95,6 +105,12 @@ define void @memset_5(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movb $0, 4(%rdi)
 ; SKYLAKE-NEXT:    movl $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_5:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movb $0, 4(%rdi)
+; KNL-NEXT:    movl $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 5, i1 false)
 	ret void
@@ -131,6 +147,12 @@ define void @memset_7(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movl $0, 3(%rdi)
 ; SKYLAKE-NEXT:    movl $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_7:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movl $0, 3(%rdi)
+; KNL-NEXT:    movl $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 7, i1 false)
 	ret void
@@ -163,6 +185,11 @@ define void @memset_8(i8* %a) nounwind  {
 ; SKYLAKE:       # %bb.0: # %entry
 ; SKYLAKE-NEXT:    movq $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_8:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movq $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 8, i1 false)
 	ret void
@@ -200,6 +227,12 @@ define void @memset_11(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movl $0, 7(%rdi)
 ; SKYLAKE-NEXT:    movq $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_11:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movl $0, 7(%rdi)
+; KNL-NEXT:    movq $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 11, i1 false)
 	ret void
@@ -238,6 +271,12 @@ define void @memset_13(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movq $0, 5(%rdi)
 ; SKYLAKE-NEXT:    movq $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_13:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movq $0, 5(%rdi)
+; KNL-NEXT:    movq $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 13, i1 false)
 	ret void
@@ -276,6 +315,12 @@ define void @memset_15(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movq $0, 7(%rdi)
 ; SKYLAKE-NEXT:    movq $0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_15:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    movq $0, 7(%rdi)
+; KNL-NEXT:    movq $0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 15, i1 false)
 	ret void
@@ -314,6 +359,12 @@ define void @memset_16(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
 ; SKYLAKE-NEXT:    vmovups %xmm0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_16:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %xmm0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 16, i1 false)
 	ret void
@@ -357,6 +408,13 @@ define void @memset_17(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vmovups %xmm0, (%rdi)
 ; SKYLAKE-NEXT:    movb $0, 16(%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_17:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %xmm0, (%rdi)
+; KNL-NEXT:    movb $0, 16(%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 17, i1 false)
 	ret void
@@ -400,6 +458,13 @@ define void @memset_19(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vmovups %xmm0, (%rdi)
 ; SKYLAKE-NEXT:    movl $0, 15(%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_19:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %xmm0, (%rdi)
+; KNL-NEXT:    movl $0, 15(%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 19, i1 false)
 	ret void
@@ -447,6 +512,13 @@ define void @memset_31(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vmovups %xmm0, 15(%rdi)
 ; SKYLAKE-NEXT:    vmovups %xmm0, (%rdi)
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_31:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %xmm0, 15(%rdi)
+; KNL-NEXT:    vmovups %xmm0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 31, i1 false)
 	ret void
@@ -494,6 +566,12 @@ define void @memset_32(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vmovups %ymm0, (%rdi)
 ; SKYLAKE-NEXT:    vzeroupper
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_32:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %ymm0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 32, i1 false)
 	ret void
@@ -540,6 +618,12 @@ define void @memset_32_align32(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    vmovaps %ymm0, (%rdi)
 ; SKYLAKE-NEXT:    vzeroupper
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_32_align32:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovaps %ymm0, (%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* align 32 %a, i8 0, i64 32, i1 false)
 	ret void
@@ -593,7 +677,152 @@ define void @memset_35(i8* %a) nounwind  {
 ; SKYLAKE-NEXT:    movl $0, 31(%rdi)
 ; SKYLAKE-NEXT:    vzeroupper
 ; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_35:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %ymm0, (%rdi)
+; KNL-NEXT:    movl $0, 31(%rdi)
+; KNL-NEXT:    retq
 entry:
 	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 35, i1 false)
 	ret void
 }
+
+define void @memset_64(i8* %a) nounwind  {
+; X86-LABEL: memset_64:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl $0, 60(%eax)
+; X86-NEXT:    movl $0, 56(%eax)
+; X86-NEXT:    movl $0, 52(%eax)
+; X86-NEXT:    movl $0, 48(%eax)
+; X86-NEXT:    movl $0, 44(%eax)
+; X86-NEXT:    movl $0, 40(%eax)
+; X86-NEXT:    movl $0, 36(%eax)
+; X86-NEXT:    movl $0, 32(%eax)
+; X86-NEXT:    movl $0, 28(%eax)
+; X86-NEXT:    movl $0, 24(%eax)
+; X86-NEXT:    movl $0, 20(%eax)
+; X86-NEXT:    movl $0, 16(%eax)
+; X86-NEXT:    movl $0, 12(%eax)
+; X86-NEXT:    movl $0, 8(%eax)
+; X86-NEXT:    movl $0, 4(%eax)
+; X86-NEXT:    movl $0, (%eax)
+; X86-NEXT:    retl
+;
+; CORE2-LABEL: memset_64:
+; CORE2:       # %bb.0: # %entry
+; CORE2-NEXT:    movq $0, 56(%rdi)
+; CORE2-NEXT:    movq $0, 48(%rdi)
+; CORE2-NEXT:    movq $0, 40(%rdi)
+; CORE2-NEXT:    movq $0, 32(%rdi)
+; CORE2-NEXT:    movq $0, 24(%rdi)
+; CORE2-NEXT:    movq $0, 16(%rdi)
+; CORE2-NEXT:    movq $0, 8(%rdi)
+; CORE2-NEXT:    movq $0, (%rdi)
+; CORE2-NEXT:    retq
+;
+; NEHALEM-LABEL: memset_64:
+; NEHALEM:       # %bb.0: # %entry
+; NEHALEM-NEXT:    xorps %xmm0, %xmm0
+; NEHALEM-NEXT:    movups %xmm0, 48(%rdi)
+; NEHALEM-NEXT:    movups %xmm0, 32(%rdi)
+; NEHALEM-NEXT:    movups %xmm0, 16(%rdi)
+; NEHALEM-NEXT:    movups %xmm0, (%rdi)
+; NEHALEM-NEXT:    retq
+;
+; SANDYBRIDGE-LABEL: memset_64:
+; SANDYBRIDGE:       # %bb.0: # %entry
+; SANDYBRIDGE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; SANDYBRIDGE-NEXT:    vmovups %xmm0, 16(%rdi)
+; SANDYBRIDGE-NEXT:    vmovups %xmm0, (%rdi)
+; SANDYBRIDGE-NEXT:    vmovups %xmm0, 48(%rdi)
+; SANDYBRIDGE-NEXT:    vmovups %xmm0, 32(%rdi)
+; SANDYBRIDGE-NEXT:    retq
+;
+; SKYLAKE-LABEL: memset_64:
+; SKYLAKE:       # %bb.0: # %entry
+; SKYLAKE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; SKYLAKE-NEXT:    vmovups %ymm0, 32(%rdi)
+; SKYLAKE-NEXT:    vmovups %ymm0, (%rdi)
+; SKYLAKE-NEXT:    vzeroupper
+; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_64:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovups %ymm0, 32(%rdi)
+; KNL-NEXT:    vmovups %ymm0, (%rdi)
+; KNL-NEXT:    retq
+entry:
+	call void @llvm.memset.p0i8.i64(i8* %a, i8 0, i64 64, i1 false)
+	ret void
+}
+
+define void @memset_64_align64(i8* %a) nounwind  {
+; X86-LABEL: memset_64_align64:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    movl $0, 60(%eax)
+; X86-NEXT:    movl $0, 56(%eax)
+; X86-NEXT:    movl $0, 52(%eax)
+; X86-NEXT:    movl $0, 48(%eax)
+; X86-NEXT:    movl $0, 44(%eax)
+; X86-NEXT:    movl $0, 40(%eax)
+; X86-NEXT:    movl $0, 36(%eax)
+; X86-NEXT:    movl $0, 32(%eax)
+; X86-NEXT:    movl $0, 28(%eax)
+; X86-NEXT:    movl $0, 24(%eax)
+; X86-NEXT:    movl $0, 20(%eax)
+; X86-NEXT:    movl $0, 16(%eax)
+; X86-NEXT:    movl $0, 12(%eax)
+; X86-NEXT:    movl $0, 8(%eax)
+; X86-NEXT:    movl $0, 4(%eax)
+; X86-NEXT:    movl $0, (%eax)
+; X86-NEXT:    retl
+;
+; CORE2-LABEL: memset_64_align64:
+; CORE2:       # %bb.0: # %entry
+; CORE2-NEXT:    xorps %xmm0, %xmm0
+; CORE2-NEXT:    movaps %xmm0, 48(%rdi)
+; CORE2-NEXT:    movaps %xmm0, 32(%rdi)
+; CORE2-NEXT:    movaps %xmm0, 16(%rdi)
+; CORE2-NEXT:    movaps %xmm0, (%rdi)
+; CORE2-NEXT:    retq
+;
+; NEHALEM-LABEL: memset_64_align64:
+; NEHALEM:       # %bb.0: # %entry
+; NEHALEM-NEXT:    xorps %xmm0, %xmm0
+; NEHALEM-NEXT:    movaps %xmm0, 48(%rdi)
+; NEHALEM-NEXT:    movaps %xmm0, 32(%rdi)
+; NEHALEM-NEXT:    movaps %xmm0, 16(%rdi)
+; NEHALEM-NEXT:    movaps %xmm0, (%rdi)
+; NEHALEM-NEXT:    retq
+;
+; SANDYBRIDGE-LABEL: memset_64_align64:
+; SANDYBRIDGE:       # %bb.0: # %entry
+; SANDYBRIDGE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; SANDYBRIDGE-NEXT:    vmovaps %ymm0, 32(%rdi)
+; SANDYBRIDGE-NEXT:    vmovaps %ymm0, (%rdi)
+; SANDYBRIDGE-NEXT:    vzeroupper
+; SANDYBRIDGE-NEXT:    retq
+;
+; SKYLAKE-LABEL: memset_64_align64:
+; SKYLAKE:       # %bb.0: # %entry
+; SKYLAKE-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; SKYLAKE-NEXT:    vmovaps %ymm0, 32(%rdi)
+; SKYLAKE-NEXT:    vmovaps %ymm0, (%rdi)
+; SKYLAKE-NEXT:    vzeroupper
+; SKYLAKE-NEXT:    retq
+;
+; KNL-LABEL: memset_64_align64:
+; KNL:       # %bb.0: # %entry
+; KNL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; KNL-NEXT:    vmovaps %ymm0, 32(%rdi)
+; KNL-NEXT:    vmovaps %ymm0, (%rdi)
+; KNL-NEXT:    retq
+entry:
+	call void @llvm.memset.p0i8.i64(i8* align 64 %a, i8 0, i64 64, i1 false)
+	ret void
+}