turn the default value of x86_rep_stosb_threshold from 2k to 1M

Signed-off-by: Yang Yanchao <yangyanchao6@huawei.com>

glibc.spec

@@ -65,7 +65,7 @@
 ##############################################################################
 Name: 	 	glibc
 Version: 	2.34
-Release: 	29
+Release: 	30
 Summary: 	The GNU libc libraries
 License:	%{all_license}
 URL: 		http://www.gnu.org/software/glibc/
@@ -138,7 +138,7 @@ Patch51: linux-Simplify-get_nprocs.patch
 Patch52: misc-Add-__get_nprocs_sched.patch
 Patch53: linux-Revert-the-use-of-sched_getaffinity-on-get_npr.patch
 
-#Patch9000: turn-REP_STOSB_THRESHOLD-from-2k-to-1M.patch
+Patch9000: turn-default-value-of-x86_rep_stosb_threshold_form_2K_to_1M.patch
 Patch9001: delete-no-hard-link-to-avoid-all_language-package-to.patch 
 Patch9002: 0001-add-base-files-for-libphtread-condition-family.patch
 Patch9003: 0002-add-header-files-for-libphtread_2_17_so.patch
@@ -1334,6 +1334,9 @@ fi
 %endif
 
 %changelog
+* Thu Dec 3 2021 Yang Yanchao <yangyanchao6@huawei.com> - 2.34-30
+- turn the default value of x86_rep_stosb_threshold from 2k to 1M
+
 * Thu Dec 2 2021 Qingqing Li <liqingqing3@huawei.com> - 2.34-29
 - revert the use of sched_getaffinity [BZ #28310]
 

turn-REP_STOSB_THRESHOLD-from-2k-to-1M.patch

@@ -1,45 +0,0 @@
-From dc8c5d3ba8ec3c2de8ca0898d682d89492d275b3 Mon Sep 17 00:00:00 2001
-From: Shuo Wang <wangshuo_1994@foxmail.com>
-Date: Tue, 2 Mar 2021 10:41:09 +0800
-Subject: [PATCH] turn REP_STOSB_THRESHOLD from 2k to 1M
-
-REP_STOSB_THRESHOLD is designed to choose vec mov or stosb.
-The default threshold (2k) will lead to performance degradation if the
-memcpy size is between 2k and 1M.
----
- sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S | 12 +++++++++++-
- 1 file changed, 11 insertions(+), 1 deletion(-)
-
-diff --git a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
-index faa40856..76f84748 100644
---- a/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
-+++ b/sysdeps/x86_64/multiarch/memset-vec-unaligned-erms.S
-@@ -58,6 +58,16 @@
- # endif
- #endif
- 
-+/* Threshold to use Enhanced REP STOSB.  Since there is overhead to set
-+   up REP STOSB operation, REP STOSB isn't faster on short data.  The
-+   memset micro benchmark in glibc shows that 2KB is the approximate
-+   value above which REP STOSB becomes faster on processors with
-+   Enhanced REP STOSB.  Since the stored value is fixed, larger register
-+   size has minimal impact on threshold.  */
-+#ifndef REP_STOSB_THRESHOLD
-+# define REP_STOSB_THRESHOLD           1048576
-+#endif
-+
- #ifndef SECTION
- # error SECTION is not defined!
- #endif
-@@ -171,7 +181,7 @@ ENTRY (MEMSET_SYMBOL (__memset, unaligned_erms))
- 	ret
- 
- L(stosb_more_2x_vec):
--	cmp	__x86_rep_stosb_threshold(%rip), %RDX_LP
-+	cmp	$REP_STOSB_THRESHOLD, %RDX_LP
- 	ja	L(stosb)
- #endif
- L(more_2x_vec):
--- 
-2.23.0
-

turn-default-value-of-x86_rep_stosb_threshold_form_2K_to_1M.patch

@@ -0,0 +1,29 @@
+From 4dee2794b8c78ccd540e3f72bc07585828e0143b Mon Sep 17 00:00:00 2001
+From: Yang Yanchao <yangyanchao6@huawei.com>
+Date: Thu, 2 Dec 2021 19:56:20 +0800
+Subject: [PATCH] turn the default value of x86_rep_stosb_threshold from 2k to 1M
+
+x86_rep_stosb_threshold is designed to choose vec mov or stosb.
+For the libMicro, after set this x86_rep_stosb_threshold to 1 MB. 
+The performance of memset_256_u, memset_4k_uc, and memset_1m is improved.
+The performance deteriorates in the memset_4k and memset_10k scenarios.
+---
+ sysdeps/x86/dl-tunables.list | 2 +-
+ 1 file changed, 1 insertion(+), 1 deletion(-)
+
+diff --git a/sysdeps/x86/dl-tunables.list b/sysdeps/x86/dl-tunables.list
+index dd6e1d65..a4c3af69 100644
+--- a/sysdeps/x86/dl-tunables.list
++++ b/sysdeps/x86/dl-tunables.list
+@@ -54,7 +54,7 @@ glibc {
+       # stored value is fixed, larger register size has minimal impact
+       # on threshold.
+       minval: 1
+-      default: 2048
++      default: 1048576
+     }
+     x86_data_cache_size {
+       type: SIZE_T
+-- 
+2.30.0
+