| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This patch removes the power6 wcscpy optimization and uses generic
implementation instead. Currently, both power6 and power7 IFUNC variant
resulting binary are essentially the same and the generic implementation
with unrolling loop set to 8 also results in similar performance.
Checked on powerpc64-linux-gnu.
* sysdeps/powerpc/Makefile [$(subdir) == wcsmbs] (CFLAGS-wcscpy.c):
New rule.
* sysdeps/powerpc/power6/wcscpy.c: Remove file.
* sysdeps/powerpc/powerpc32/power4/multiarch/wcscpy-power6.c:
Likewise.
* sysdeps/powerpc/powerpc32/power4/multiarch/wcscpy-power7.c:
Likewise.
* sysdeps/powerpc/powerpc32/power4/multiarch/wcscpy-ppc32.c:
Likewise.
* sysdeps/powerpc/powerpc32/power4/multiarch/wcscpy.c: Likewise.
* sysdeps/powerpc/powerpc64/multiarch/wcscpy-power6.c: Likewise.
* sysdeps/powerpc/powerpc64/multiarch/wcscpy-power7.c: Likewise.
* sysdeps/powerpc/powerpc64/multiarch/wcscpy-ppc64.c: Likewise.
* sysdeps/powerpc/powerpc64/multiarch/wcscpy.c: Likewise.
* sysdeps/powerpc/powerpc64/power6/wcscpy.c: Likewise.
* sysdeps/powerpc/powerpc32/power4/multiarch/Makefile
[$(subdir) == wcsmbs] (sysdeps_routines): Remove wcscpy-power6 and
wcscpy-power7.
(CFLAGS-wcscpy-power7.c, CFLAGS-wcscpy-power6.c): Remove rule.
* sysdeps/powerpc/powerpc64/multiarch/Makefile: Likewise.
* sysdeps/powerpc/powerpc32/power4/multiarch/ifunc-impl-list.c:
Remove wcscpy optimizations.
* sysdeps/powerpc/powerpc64/multiarch/ifunc-impl-list.c: Likewise.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since the commit
commit 81a14439417552324ec6ca71f65ddf8e7cdd51c7
Author: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Date: Tue Feb 5 17:35:12 2019 -0200
wcsmbs: optimize wcscat
powerpc64 and powerpc64le builds fail when configured with
--disable-multi-arch and --with-cpu=power6 (or newer), due to an
undefined reference to __GI___wcscpy. This patch fixes this on
sysdeps/powerpc/powerpc64/power6/wcscpy.c, which is only used when
multi-arch is disabled.
This patch does nothing for the failures on 32-bits powerpc builds,
because the file is under the powerpc64 subdirectory, however, powerpc
builds were already failing with --disable-multi-arch, with multiple
error messages, even before the aforementioned commit.
Tested for powerpc, powerpc64, and powerpc64le with multi-arch enabled
(all pass) and disabled (powerpc still fails as explained above).
|
| | |
|
|
|
Initially based on the versions found in wcsmbs/* ; these files have
been changed by hand unrolling, and adding some additional variables
to allow some read-ahead to occur, which then relieves some of the
wait-for-increment/wait-for-load/wait-for-compare-results pressure
that was slowing down every iteration through the while-loop.
For 64-bit Power7, These changes give an approx 20% throughput boost
for the wcschr and wcsrchr functions; and approx 40% boost for the
wcscpy function. 32-bit improvements appear to be slightly better
with ~ %30 and ~ %45 respectively. Results for Power6 closely match
those for power7.
|
