packages/glibc
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

aarch64: update a64fx memset not to degrade at 16KB

Browse Source
This commit is contained in:
liqingqing_1229 2021-10-29 11:07:01 +08:00
parent 97485454ba
commit 653f3153b9
2 changed files with 44 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
AArch64-Update-A64FX-memset-not-to-degrade-at-16KB.patch Normal file
View File

@ -0,0 +1,39 @@
From 23777232c23f80809613bdfa329f63aadf992922 Mon Sep 17 00:00:00 2001
From: Naohiro Tamura via Libc-alpha <libc-alpha@sourceware.org>
Date: Fri, 27 Aug 2021 05:03:04 +0000
Subject: [PATCH] AArch64: Update A64FX memset not to degrade at 16KB
This patch updates unroll8 code so as not to degrade at the peak
performance 16KB for both FX1000 and FX700.
Inserted 2 instructions at the beginning of the unroll8 loop,
cmp and branch, are a workaround that is found heuristically.
Reviewed-by: Wilco Dijkstra <Wilco.Dijkstra@arm.com>
---
sysdeps/aarch64/multiarch/memset_a64fx.S | 9 ++++++++-
1 file changed, 8 insertions(+), 1 deletion(-)
diff --git a/sysdeps/aarch64/multiarch/memset_a64fx.S b/sysdeps/aarch64/multiarch/memset_a64fx.S
index 7bf759b..f7dfdaa 100644
--- a/sysdeps/aarch64/multiarch/memset_a64fx.S
+++ b/sysdeps/aarch64/multiarch/memset_a64fx.S
@@ -96,7 +96,14 @@ L(vl_agnostic): // VL Agnostic
L(unroll8):
sub count, count, tmp1
.p2align 4
-1: st1b_unroll 0, 7
+ // The 2 instructions at the beginning of the following loop,
+ // cmp and branch, are a workaround so as not to degrade at
+ // the peak performance 16KB.
+ // It is found heuristically and the branch condition, b.ne,
+ // is chosen intentionally never to jump.
+1: cmp xzr, xzr
+ b.ne 1b
+ st1b_unroll 0, 7
add dst, dst, tmp1
subs count, count, tmp1
b.hi 1b
--
1.8.3.1

6
glibc.spec
View File

@ -65,7 +65,7 @@
##############################################################################
Name: glibc
Version: 2.34
Release: 18
Release: 19
Summary: The GNU libc libraries
License: %{all_license}
URL: http://www.gnu.org/software/glibc/
@ -118,6 +118,7 @@ Patch31: mtrace-Fix-output-with-PIE-and-ASLR-BZ-22716.patch
Patch32: rtld-copy-terminating-null-in-tunables_strdup-bug-28.patch
Patch33: Use-__executable_start-as-the-lowest-address-for-pro.patch
Patch34: x86-64-Use-testl-to-check-__x86_string_control.patch
Patch35: AArch64-Update-A64FX-memset-not-to-degrade-at-16KB.patch
#Patch9000: turn-REP_STOSB_THRESHOLD-from-2k-to-1M.patch
Patch9001: delete-no-hard-link-to-avoid-all_language-package-to.patch
@ -1307,6 +1308,9 @@ fi
%endif
%changelog
* Fri Oct 29 2021 Qingqing Li<liqingqing3@huawei.com> - 2.34-19
- aarch64: update a64fx memset not to degrade at 16KB
* Thu Oct 28 2021 Qingqing Li<liqingqing3@huawei.com> - 2.34-18
- use testl instead of andl to check __x86_string_control to
avoid updating __x86_string_control