about summary refs log tree commit diff
path: root/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
diff options
context:
space:
mode:
authorH.J. Lu <hjl.tools@gmail.com>2016-01-15 12:14:18 -0800
committerH.J. Lu <hjl.tools@gmail.com>2016-01-15 12:39:49 -0800
commita806d37903f0db0081a49a440b1f3a82f5e65916 (patch)
tree3897e0431012c2117ff1f5186c9af6dd55a98845 /sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
parente8ae91992f4eca858f344e10e75783e66c228e35 (diff)
downloadglibc-a806d37903f0db0081a49a440b1f3a82f5e65916.tar.gz
glibc-a806d37903f0db0081a49a440b1f3a82f5e65916.tar.xz
glibc-a806d37903f0db0081a49a440b1f3a82f5e65916.zip
Use vmovdqu64 load with vmovntdq store
Diffstat (limited to 'sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S')
-rw-r--r--sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S16
1 files changed, 8 insertions, 8 deletions
diff --git a/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S b/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
index 75bc836d0c..56e727bfc9 100644
--- a/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
+++ b/sysdeps/x86_64/multiarch/memcpy-avx512-no-vzeroupper.S
@@ -353,10 +353,10 @@ L(gobble_256bytes_nt_loop):
 	prefetchnta 0x340(%rsi)
 	prefetchnta 0x380(%rsi)
 	prefetchnta 0x3C0(%rsi)
-	vmovups	(%rsi), %zmm0
-	vmovups	0x40(%rsi), %zmm1
-	vmovups	0x80(%rsi), %zmm2
-	vmovups	0xC0(%rsi), %zmm3
+	vmovdqu64	(%rsi), %zmm0
+	vmovdqu64	0x40(%rsi), %zmm1
+	vmovdqu64	0x80(%rsi), %zmm2
+	vmovdqu64	0xC0(%rsi), %zmm3
 	vmovntdq %zmm0, (%rdi)
 	vmovntdq %zmm1, 0x40(%rdi)
 	vmovntdq %zmm2, 0x80(%rdi)
@@ -391,10 +391,10 @@ L(gobble_256bytes_nt_loop_bkw):
 	prefetchnta -0x3C0(%rcx)
 	prefetchnta -0x380(%rcx)
 	prefetchnta -0x340(%rcx)
-	vmovups	-0x100(%rcx), %zmm0
-	vmovups	-0xC0(%rcx), %zmm1
-	vmovups	-0x80(%rcx), %zmm2
-	vmovups	-0x40(%rcx), %zmm3
+	vmovdqu64	-0x100(%rcx), %zmm0
+	vmovdqu64	-0xC0(%rcx), %zmm1
+	vmovdqu64	-0x80(%rcx), %zmm2
+	vmovdqu64	-0x40(%rcx), %zmm3
 	vmovntdq %zmm0,	-0x100(%r9)
 	vmovntdq %zmm1,	-0xC0(%r9)
 	vmovntdq %zmm2,	-0x80(%r9)