qiurui
/
glibc
mirror of git://sourceware.org/git/glibc.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

memcpy_falkor: Fix code style in comments

This commit is contained in:
Siddhesh Poyarekar 2017-08-09 12:57:17 +05:30
parent 36ada5f681
commit 0e02b5107e
2 changed files with 7 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
ChangeLog
View File

@ -1,5 +1,8 @@
2017-08-09 Siddhesh Poyarekar <siddhesh@sourceware.org> 2017-08-09 Siddhesh Poyarekar <siddhesh@sourceware.org>
* sysdeps/aarch64/multiarch/memcpy_falkor.S: Fix code style in
comments.
* manual/tunables.texi (Tunable glibc.tune.cpu): Add falkor. * manual/tunables.texi (Tunable glibc.tune.cpu): Add falkor.
* sysdeps/aarch64/multiarch/Makefile (sysdep_routines): Add * sysdeps/aarch64/multiarch/Makefile (sysdep_routines): Add
memcpy_falkor. memcpy_falkor.

11
sysdeps/aarch64/multiarch/memcpy_falkor.S
View File

@ -20,10 +20,8 @@
#include <sysdep.h> #include <sysdep.h>
/* Assumptions: /* Assumptions:
*
* ARMv8-a, AArch64, falkor, unaligned accesses. ARMv8-a, AArch64, falkor, unaligned accesses. */
*
*/
#define dstin x0 #define dstin x0
#define src x1 #define src x1
@ -53,14 +51,13 @@
conditionals, since the former would unnecessarily break across multiple conditionals, since the former would unnecessarily break across multiple
issue groups. The medium copy group has been enlarged to 128 bytes since issue groups. The medium copy group has been enlarged to 128 bytes since
bumping up the small copies up to 32 bytes allows us to do that without bumping up the small copies up to 32 bytes allows us to do that without
cost and also allows us the reduce the size of the prep code before loop64. cost and also allows us to reduce the size of the prep code before loop64.
All copies are done only via two registers r6 and r7. This is to ensure All copies are done only via two registers r6 and r7. This is to ensure
that all loads hit a single hardware prefetcher which can get correctly that all loads hit a single hardware prefetcher which can get correctly
trained to prefetch a single stream. trained to prefetch a single stream.
The non-temporal stores help optimize cache utilization. The non-temporal stores help optimize cache utilization. */
*/
#if IS_IN (libc) #if IS_IN (libc)
ENTRY_ALIGN (__memcpy_falkor, 6) ENTRY_ALIGN (__memcpy_falkor, 6)