x86: Add evex optimized functions for the wchar_t strcpy family
commit64b8b6516b3cba19dba4c8f4f9b97daa0556fd98
authorNoah Goldstein <goldstein.w.n@gmail.com>
Wed, 9 Nov 2022 01:38:40 +0000 (8 17:38 -0800)
committerNoah Goldstein <goldstein.w.n@gmail.com>
Wed, 9 Nov 2022 03:22:33 +0000 (8 19:22 -0800)
tree93e9e1c8bd54c3a10eb9c00db604bd74696adb10
parent642933158e7cf072d873231b1a9bb03291f2b989
x86: Add evex optimized functions for the wchar_t strcpy family

Implemented:
    wcscat-evex  (+ 905 bytes)
    wcscpy-evex  (+ 674 bytes)
    wcpcpy-evex  (+ 709 bytes)
    wcsncpy-evex (+1358 bytes)
    wcpncpy-evex (+1467 bytes)
    wcsncat-evex (+1213 bytes)

Performance Changes:
    Times are from N = 10 runs of the benchmark suite and are reported
    as geometric mean of all ratios of New Implementation / Best Old
    Implementation. Best Old Implementation was determined with the
    highest ISA implementation.

    wcscat-evex     -> 0.991
    wcscpy-evex     -> 0.587
    wcpcpy-evex     -> 0.695
    wcsncpy-evex    -> 0.719
    wcpncpy-evex    -> 0.694
    wcsncat-evex    -> 0.979

Code Size Changes:
    This change  increase the size of libc.so by ~6.3kb bytes. For
    reference the patch optimizing the normal strcpy family functions
    decreases libc.so by ~5.7kb.

Full check passes on x86-64 and build succeeds for all ISA levels w/
and w/o multiarch.
33 files changed:
sysdeps/x86_64/Makefile
sysdeps/x86_64/multiarch/Makefile
sysdeps/x86_64/multiarch/ifunc-impl-list.c
sysdeps/x86_64/multiarch/ifunc-wcs.h [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcpcpy-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcpcpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 80% similarity]
sysdeps/x86_64/multiarch/wcpcpy.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 58% similarity]
sysdeps/x86_64/multiarch/wcpncpy-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcpncpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 79% similarity]
sysdeps/x86_64/multiarch/wcpncpy.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 57% similarity]
sysdeps/x86_64/multiarch/wcscat-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcscat-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 80% similarity]
sysdeps/x86_64/multiarch/wcscat.c [copied from sysdeps/x86_64/multiarch/wcscpy.c with 59% similarity]
sysdeps/x86_64/multiarch/wcscpy-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcscpy-generic.c
sysdeps/x86_64/multiarch/wcscpy.c
sysdeps/x86_64/multiarch/wcsncat-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcsncat-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 79% similarity]
sysdeps/x86_64/multiarch/wcsncat.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 60% similarity]
sysdeps/x86_64/multiarch/wcsncpy-evex.S [new file with mode: 0644]
sysdeps/x86_64/multiarch/wcsncpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 79% similarity]
sysdeps/x86_64/multiarch/wcsncpy.c [copied from sysdeps/x86_64/multiarch/wcscpy.c with 58% similarity]
sysdeps/x86_64/wcpcpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 67% similarity]
sysdeps/x86_64/wcpcpy.S [copied from sysdeps/x86_64/wcscpy.S with 65% similarity]
sysdeps/x86_64/wcpncpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 66% similarity]
sysdeps/x86_64/wcpncpy.S [copied from sysdeps/x86_64/wcscpy.S with 65% similarity]
sysdeps/x86_64/wcscat-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 67% similarity]
sysdeps/x86_64/wcscat.S [copied from sysdeps/x86_64/wcscpy.S with 65% similarity]
sysdeps/x86_64/wcscpy.S
sysdeps/x86_64/wcsncat-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 66% similarity]
sysdeps/x86_64/wcsncat.S [copied from sysdeps/x86_64/wcscpy.S with 65% similarity]
sysdeps/x86_64/wcsncpy-generic.c [copied from sysdeps/x86_64/multiarch/wcscpy-generic.c with 66% similarity]
sysdeps/x86_64/wcsncpy.S [copied from sysdeps/x86_64/wcscpy.S with 65% similarity]