diff --git a/i/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
--- a/i/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -2149,6 +2149,13 @@
     if (Size >= 16 && (!Subtarget.isUnalignedMem16Slow() ||
                        ((DstAlign == 0 || DstAlign >= 16) &&
                         (SrcAlign == 0 || SrcAlign >= 16)))) {
+      // FIXME: Check if unaligned 64-byte accesses are slow.
+      if (Size >= 64 && Subtarget.hasAVX512() &&
+          (Subtarget.getPreferVectorWidth() >= 512)) {
+        if (Subtarget.hasBWI() && Subtarget.hasVLX())
+          return MVT::v64i8;
+        return MVT::v16i32;
+      }
       // FIXME: Check if unaligned 32-byte accesses are slow.
       if (Size >= 32 && Subtarget.hasAVX() &&
           (Subtarget.getPreferVectorWidth() >= 256)) {
diff --git a/i/llvm/test/CodeGen/X86/memcpy.ll b/llvm/test/CodeGen/X86/memcpy.ll
--- a/i/llvm/test/CodeGen/X86/memcpy.ll
+++ b/llvm/test/CodeGen/X86/memcpy.ll
@@ -1,6 +1,9 @@
 ; 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
 
 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
@@ -48,6 +51,30 @@
 ; 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), %zmm0
+; LINUX-KNL-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
 ; DARWIN-LABEL: test3:
 ; DARWIN:       ## %bb.0: ## %entry
 ; DARWIN-NEXT:    movq 56(%rsi), %rax
@@ -79,6 +106,30 @@
 ; LINUX-NEXT:    popq %rdx
 ; LINUX-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), %zmm0
+; LINUX-KNL-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
 ; DARWIN-LABEL: test3_minsize:
 ; DARWIN:       ## %bb.0:
 ; DARWIN-NEXT:    pushq $64
@@ -95,6 +146,30 @@
 ; LINUX-NEXT:    popq %rdx
 ; LINUX-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), %zmm0
+; LINUX-KNL-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
 ; DARWIN-LABEL: test3_minsize_optsize:
 ; DARWIN:       ## %bb.0:
 ; DARWIN-NEXT:    pushq $64
@@ -126,6 +201,30 @@
 ; LINUX-NEXT:    movq %rax, (%rdi)
 ; LINUX-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), %zmm0
+; LINUX-KNL-NEXT:    vmovups %zmm0, (%rdi)
+; LINUX-KNL-NEXT:    retq
+;
 ; DARWIN-LABEL: test4:
 ; DARWIN:       ## %bb.0: ## %entry
 ; DARWIN-NEXT:    movq 56(%rsi), %rax