Message ID | 20210708081308.640036-1-goldstein.w.n@gmail.com |
---|---|
State | New |
Headers | show |
Series | [v1] x86: Remove wcsnlen-sse4_1 from wcslen ifunc-impl-list [BZ #28064] | expand |
On Thu, Jul 8, 2021 at 1:22 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote: > > The following commit > > commit 6f573a27b6c8b4236445810a44660612323f5a73 > Author: Noah Goldstein <goldstein.w.n@gmail.com> > Date: Wed Jun 23 01:19:34 2021 -0400 > > x86-64: Add wcslen optimize for sse4.1 > > Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did > not add wcslen-sse4.1 to wcslen ifunc implementation list. This commit > fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc > implementation list and adding wcslen-sse4.1 to the ifunc > implementation list. > > Testing: > test-wcslen.c is passing as well as all other tests in wcsmbs and > string. > > Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com> > --- > I was unable to actually reproduce the bug on my machine, even after > reordering ifunc-wcslen.h to prefer the sse4.1 > implementation. Although this most definetly is a bug. > > > I ran the following command. Is there an issue with the command? Or > does rdx happen to always have a len >= actual length of the string? That must be the case. We can do something similar to sysdeps/x86_64/x32/tst-size_t-strnlen.c to write a testcase. > $> rm -rf build; mkdir -p build/glibc; (cd build/glibc/; unset LD_LIBRARY_PATH; /absolute/path/to/src/glibc/configure --prefix=/usr; make --silent; make xcheck; make -r -C /absolute/path/to/src/glibc/string/ objdir=`pwd` check; make -r -C /absolute/path/to/src/glibc/wcsmbs/ objdir=`pwd` check); > > > > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > index dbd1ebf298..39ab10613b 100644 > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > @@ -684,9 +684,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > && CPU_FEATURE_USABLE (AVX512BW) > && CPU_FEATURE_USABLE (BMI2)), > __wcslen_evex) > - IFUNC_IMPL_ADD (array, i, wcsnlen, > + IFUNC_IMPL_ADD (array, i, wcslen, > CPU_FEATURE_USABLE (SSE4_1), > - __wcsnlen_sse4_1) > + __wcslen_sse4_1) > IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) > > /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */ > -- > 2.25.1 > LGTM. Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
On Thu, Jul 8, 2021 at 5:21 AM H.J. Lu <hjl.tools@gmail.com> wrote: > > On Thu, Jul 8, 2021 at 1:22 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote: > > > > The following commit > > > > commit 6f573a27b6c8b4236445810a44660612323f5a73 > > Author: Noah Goldstein <goldstein.w.n@gmail.com> > > Date: Wed Jun 23 01:19:34 2021 -0400 > > > > x86-64: Add wcslen optimize for sse4.1 > > > > Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did > > not add wcslen-sse4.1 to wcslen ifunc implementation list. This commit > > fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc > > implementation list and adding wcslen-sse4.1 to the ifunc > > implementation list. > > > > Testing: > > test-wcslen.c is passing as well as all other tests in wcsmbs and > > string. > > > > Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com> > > --- > > I was unable to actually reproduce the bug on my machine, even after > > reordering ifunc-wcslen.h to prefer the sse4.1 > > implementation. Although this most definetly is a bug. > > > > > > I ran the following command. Is there an issue with the command? Or > > does rdx happen to always have a len >= actual length of the string? > > That must be the case. We can do something similar to > > sysdeps/x86_64/x32/tst-size_t-strnlen.c > > to write a testcase. Here is the patch with a testcase. Please include it in your fix and mention BZ #28064 in your commit log. > > $> rm -rf build; mkdir -p build/glibc; (cd build/glibc/; unset LD_LIBRARY_PATH; /absolute/path/to/src/glibc/configure --prefix=/usr; make --silent; make xcheck; make -r -C /absolute/path/to/src/glibc/string/ objdir=`pwd` check; make -r -C /absolute/path/to/src/glibc/wcsmbs/ objdir=`pwd` check); > > > > > > > > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > index dbd1ebf298..39ab10613b 100644 > > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > @@ -684,9 +684,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > > && CPU_FEATURE_USABLE (AVX512BW) > > && CPU_FEATURE_USABLE (BMI2)), > > __wcslen_evex) > > - IFUNC_IMPL_ADD (array, i, wcsnlen, > > + IFUNC_IMPL_ADD (array, i, wcslen, > > CPU_FEATURE_USABLE (SSE4_1), > > - __wcsnlen_sse4_1) > > + __wcslen_sse4_1) > > IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) > > > > /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */ > > -- > > 2.25.1 > > > > LGTM. > > Reviewed-by: H.J. Lu <hjl.tools@gmail.com> > > -- > H.J.
On Thu, Jul 8, 2021 at 6:57 AM H.J. Lu <hjl.tools@gmail.com> wrote: > > On Thu, Jul 8, 2021 at 5:21 AM H.J. Lu <hjl.tools@gmail.com> wrote: > > > > On Thu, Jul 8, 2021 at 1:22 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote: > > > > > > The following commit > > > > > > commit 6f573a27b6c8b4236445810a44660612323f5a73 > > > Author: Noah Goldstein <goldstein.w.n@gmail.com> > > > Date: Wed Jun 23 01:19:34 2021 -0400 > > > > > > x86-64: Add wcslen optimize for sse4.1 > > > > > > Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did > > > not add wcslen-sse4.1 to wcslen ifunc implementation list. This commit > > > fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc > > > implementation list and adding wcslen-sse4.1 to the ifunc > > > implementation list. > > > > > > Testing: > > > test-wcslen.c is passing as well as all other tests in wcsmbs and > > > string. > > > > > > Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com> > > > --- > > > I was unable to actually reproduce the bug on my machine, even after > > > reordering ifunc-wcslen.h to prefer the sse4.1 > > > implementation. Although this most definetly is a bug. > > > > > > > > > I ran the following command. Is there an issue with the command? Or > > > does rdx happen to always have a len >= actual length of the string? > > > > That must be the case. We can do something similar to > > > > sysdeps/x86_64/x32/tst-size_t-strnlen.c > > > > to write a testcase. > > Here is the patch with a testcase. Please include it in your fix and > mention BZ #28064 in your commit log. Please use this patch passing a pointer to do_strlen for x32. > > > $> rm -rf build; mkdir -p build/glibc; (cd build/glibc/; unset LD_LIBRARY_PATH; /absolute/path/to/src/glibc/configure --prefix=/usr; make --silent; make xcheck; make -r -C /absolute/path/to/src/glibc/string/ objdir=`pwd` check; make -r -C /absolute/path/to/src/glibc/wcsmbs/ objdir=`pwd` check); > > > > > > > > > > > > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++-- > > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > > > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > index dbd1ebf298..39ab10613b 100644 > > > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > @@ -684,9 +684,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, > > > && CPU_FEATURE_USABLE (AVX512BW) > > > && CPU_FEATURE_USABLE (BMI2)), > > > __wcslen_evex) > > > - IFUNC_IMPL_ADD (array, i, wcsnlen, > > > + IFUNC_IMPL_ADD (array, i, wcslen, > > > CPU_FEATURE_USABLE (SSE4_1), > > > - __wcsnlen_sse4_1) > > > + __wcslen_sse4_1) > > > IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) > > > > > > /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */ > > > -- > > > 2.25.1 > > > > > > > LGTM. > > > > Reviewed-by: H.J. Lu <hjl.tools@gmail.com> > > > > -- > > H.J. > > > > -- > H.J.
On Thu, Jul 8, 2021 at 11:08 AM H.J. Lu <hjl.tools@gmail.com> wrote: > On Thu, Jul 8, 2021 at 6:57 AM H.J. Lu <hjl.tools@gmail.com> wrote: > > > > On Thu, Jul 8, 2021 at 5:21 AM H.J. Lu <hjl.tools@gmail.com> wrote: > > > > > > On Thu, Jul 8, 2021 at 1:22 AM Noah Goldstein <goldstein.w.n@gmail.com> > wrote: > > > > > > > > The following commit > > > > > > > > commit 6f573a27b6c8b4236445810a44660612323f5a73 > > > > Author: Noah Goldstein <goldstein.w.n@gmail.com> > > > > Date: Wed Jun 23 01:19:34 2021 -0400 > > > > > > > > x86-64: Add wcslen optimize for sse4.1 > > > > > > > > Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did > > > > not add wcslen-sse4.1 to wcslen ifunc implementation list. This > commit > > > > fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc > > > > implementation list and adding wcslen-sse4.1 to the ifunc > > > > implementation list. > > > > > > > > Testing: > > > > test-wcslen.c is passing as well as all other tests in wcsmbs and > > > > string. > > > > > > > > Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com> > > > > --- > > > > I was unable to actually reproduce the bug on my machine, even after > > > > reordering ifunc-wcslen.h to prefer the sse4.1 > > > > implementation. Although this most definetly is a bug. > > > > > > > > > > > > I ran the following command. Is there an issue with the command? Or > > > > does rdx happen to always have a len >= actual length of the string? > > > > > > That must be the case. We can do something similar to > > > > > > sysdeps/x86_64/x32/tst-size_t-strnlen.c > > > > > > to write a testcase. > > > > Here is the patch with a testcase. Please include it in your fix and > > mention BZ #28064 in your commit log. > Done. > > Please use this patch passing a pointer to do_strlen for x32. > Done. Made it first patch in patchset. > > > > > $> rm -rf build; mkdir -p build/glibc; (cd build/glibc/; unset > LD_LIBRARY_PATH; /absolute/path/to/src/glibc/configure --prefix=/usr; make > --silent; make xcheck; make -r -C /absolute/path/to/src/glibc/string/ > objdir=`pwd` check; make -r -C /absolute/path/to/src/glibc/wcsmbs/ > objdir=`pwd` check); > > > > > > > > > > > > > > > > sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++-- > > > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > > > > > diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > > index dbd1ebf298..39ab10613b 100644 > > > > --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > > +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c > > > > @@ -684,9 +684,9 @@ __libc_ifunc_impl_list (const char *name, struct > libc_ifunc_impl *array, > > > > && CPU_FEATURE_USABLE (AVX512BW) > > > > && CPU_FEATURE_USABLE (BMI2)), > > > > __wcslen_evex) > > > > - IFUNC_IMPL_ADD (array, i, wcsnlen, > > > > + IFUNC_IMPL_ADD (array, i, wcslen, > > > > CPU_FEATURE_USABLE (SSE4_1), > > > > - __wcsnlen_sse4_1) > > > > + __wcslen_sse4_1) > > > > IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) > > > > > > > > /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */ > > > > -- > > > > 2.25.1 > > > > > > > > > > LGTM. > > > > > > Reviewed-by: H.J. Lu <hjl.tools@gmail.com> > > > > > > -- > > > H.J. > > > > > > > > -- > > H.J. > > > > -- > H.J. >
diff --git a/sysdeps/x86_64/multiarch/ifunc-impl-list.c b/sysdeps/x86_64/multiarch/ifunc-impl-list.c index dbd1ebf298..39ab10613b 100644 --- a/sysdeps/x86_64/multiarch/ifunc-impl-list.c +++ b/sysdeps/x86_64/multiarch/ifunc-impl-list.c @@ -684,9 +684,9 @@ __libc_ifunc_impl_list (const char *name, struct libc_ifunc_impl *array, && CPU_FEATURE_USABLE (AVX512BW) && CPU_FEATURE_USABLE (BMI2)), __wcslen_evex) - IFUNC_IMPL_ADD (array, i, wcsnlen, + IFUNC_IMPL_ADD (array, i, wcslen, CPU_FEATURE_USABLE (SSE4_1), - __wcsnlen_sse4_1) + __wcslen_sse4_1) IFUNC_IMPL_ADD (array, i, wcslen, 1, __wcslen_sse2)) /* Support sysdeps/x86_64/multiarch/wcsnlen.c. */
The following commit commit 6f573a27b6c8b4236445810a44660612323f5a73 Author: Noah Goldstein <goldstein.w.n@gmail.com> Date: Wed Jun 23 01:19:34 2021 -0400 x86-64: Add wcslen optimize for sse4.1 Added wcsnlen-sse4.1 to the wcslen ifunc implementation list and did not add wcslen-sse4.1 to wcslen ifunc implementation list. This commit fixes that by removing wcsnlen-sse4.1 from the wcslen ifunc implementation list and adding wcslen-sse4.1 to the ifunc implementation list. Testing: test-wcslen.c is passing as well as all other tests in wcsmbs and string. Signed-off-by: Noah Goldstein <goldstein.w.n@gmail.com> --- I was unable to actually reproduce the bug on my machine, even after reordering ifunc-wcslen.h to prefer the sse4.1 implementation. Although this most definetly is a bug. I ran the following command. Is there an issue with the command? Or does rdx happen to always have a len >= actual length of the string? $> rm -rf build; mkdir -p build/glibc; (cd build/glibc/; unset LD_LIBRARY_PATH; /absolute/path/to/src/glibc/configure --prefix=/usr; make --silent; make xcheck; make -r -C /absolute/path/to/src/glibc/string/ objdir=`pwd` check; make -r -C /absolute/path/to/src/glibc/wcsmbs/ objdir=`pwd` check); sysdeps/x86_64/multiarch/ifunc-impl-list.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)