[PATCH 11/11] powerpc64le: Add glibc-hwcaps support
Adhemerval Zanella
adhemerval.zanella@linaro.org
Wed Dec 2 13:46:24 GMT 2020
On 09/11/2020 15:41, Florian Weimer via Libc-alpha wrote:
> The "power10" and "power9" subdirectories are selected in a way
> that matches the -mcpu=power10 and -mcpu=power9 options of GCC.
LGTM, a comment about testing below.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
> ---
> elf/Makefile | 2 +-
> elf/tst-glibc-hwcaps-cache.script | 6 +++
> sysdeps/powerpc/powerpc64/le/Makefile | 28 ++++++++++
> .../powerpc/powerpc64/le/dl-hwcaps-subdirs.c | 46 ++++++++++++++++
> .../powerpc/powerpc64/le/tst-glibc-hwcaps.c | 54 +++++++++++++++++++
> 5 files changed, 135 insertions(+), 1 deletion(-)
> create mode 100644 sysdeps/powerpc/powerpc64/le/dl-hwcaps-subdirs.c
> create mode 100644 sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c
>
> diff --git a/elf/Makefile b/elf/Makefile
> index 0a44b19a0e..60e8662e5d 100644
> --- a/elf/Makefile
> +++ b/elf/Makefile
> @@ -1826,7 +1826,7 @@ $(objpfx)argv0test.out: tst-rtld-argv0.sh $(objpfx)ld.so \
>
> # Most likely search subdirectories, for each supported architecture.
> # Used to obtain test coverage wide test coverage.
> -glibc-hwcaps-first-subdirs-for-tests = x86-64-v2
> +glibc-hwcaps-first-subdirs-for-tests = power9 x86-64-v2
>
> # The test modules are parameterized by preprocessor macros.
> LDFLAGS-libmarkermod1-1.so += -Wl,-soname,libmarkermod1.so
Ok.
> diff --git a/elf/tst-glibc-hwcaps-cache.script b/elf/tst-glibc-hwcaps-cache.script
> index 2aef9fb177..a3e3317bbf 100644
> --- a/elf/tst-glibc-hwcaps-cache.script
> +++ b/elf/tst-glibc-hwcaps-cache.script
> @@ -10,3 +10,9 @@ mkdirp 0770 $L/glibc-hwcaps/x86-64-v4
> cp $B/elf/libmarkermod4-2.so $L/glibc-hwcaps/x86-64-v2/libmarkermod4.so
> cp $B/elf/libmarkermod4-3.so $L/glibc-hwcaps/x86-64-v3/libmarkermod4.so
> cp $B/elf/libmarkermod4-4.so $L/glibc-hwcaps/x86-64-v4/libmarkermod4.so
> +
> +mkdirp 0770 $L/glibc-hwcaps/power9
> +cp $B/elf/libmarkermod2-2.so $L/glibc-hwcaps/power9/libmarkermod2.so
> +mkdirp 0770 $L/glibc-hwcaps/power10
> +cp $B/elf/libmarkermod3-2.so $L/glibc-hwcaps/power9/libmarkermod3.so
> +cp $B/elf/libmarkermod3-3.so $L/glibc-hwcaps/power10/libmarkermod3.so
Ok.
> diff --git a/sysdeps/powerpc/powerpc64/le/Makefile b/sysdeps/powerpc/powerpc64/le/Makefile
> index 033dc77b01..7c036b45fc 100644
> --- a/sysdeps/powerpc/powerpc64/le/Makefile
> +++ b/sysdeps/powerpc/powerpc64/le/Makefile
> @@ -188,3 +188,31 @@ ifeq ($(subdir),nptl)
> CFLAGS-tst-thread_local1.cc += -mno-float128
> CFLAGS-tst-minstack-throw.cc += -mno-float128
> endif
> +
> +ifeq ($(subdir),elf)
> +$(objpfx)tst-glibc-hwcaps: \
> + $(objpfx)libmarkermod2-1.so $(objpfx)libmarkermod3-1.so
> +$(objpfx)tst-glibc-hwcaps.out: \
> + $(objpfx)libmarkermod2.so \
> + $(objpfx)glibc-hwcaps/power9/libmarkermod2.so \
> + $(objpfx)libmarkermod3.so \
> + $(objpfx)glibc-hwcaps/power9/libmarkermod3.so \
> + $(objpfx)glibc-hwcaps/power10/libmarkermod3.so \
> +
> +$(objpfx)glibc-hwcaps/power9/libmarkermod2.so: $(objpfx)libmarkermod2-2.so
> + $(make-target-directory)
> + cp $< $@
> +$(objpfx)glibc-hwcaps/power9/libmarkermod3.so: $(objpfx)libmarkermod3-2.so
> + $(make-target-directory)
> + cp $< $@
> +$(objpfx)glibc-hwcaps/power10/libmarkermod3.so: $(objpfx)libmarkermod3-3.so
> + $(make-target-directory)
> + cp $< $@
> +
> +ifeq (no,$(build-hardcoded-path-in-tests))
> +# This is an ld.so.cache test, and RPATH/RUNPATH in the executable
> +# interferes with its test objectives.
> +tests-container += tst-glibc-hwcaps-cache
> +endif
> +
> +endif # $(subdir) == elf
Ok.
> diff --git a/sysdeps/powerpc/powerpc64/le/dl-hwcaps-subdirs.c b/sysdeps/powerpc/powerpc64/le/dl-hwcaps-subdirs.c
> new file mode 100644
> index 0000000000..0ce76c2fe5
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/le/dl-hwcaps-subdirs.c
> @@ -0,0 +1,46 @@
> +/* Architecture-specific glibc-hwcaps subdirectories. powerpc64le version.
> + Copyright (C) 2020 Free Software Foundation, Inc.
> + This file is part of the GNU C Library.
> +
> + The GNU C Library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + The GNU C Library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with the GNU C Library; if not, see
> + <https://www.gnu.org/licenses/>. */
> +
> +#include <dl-hwcaps.h>
> +#include <ldsodefs.h>
> +
> +const char _dl_hwcaps_subdirs[] = "power10:power9";
> +enum { subdirs_count = 2 }; /* Number of components in _dl_hwcaps_subdirs. */
> +
> +uint32_t
> +_dl_hwcaps_subdirs_active (void)
> +{
> + int active = 0;
> +
> + /* Test in reverse preference order. Altivec and VSX are implied by
> + the powerpc64le ABI definition. */
> +
> + /* POWER9. GCC enables float128 hardware support for -mcpu=power9. */
> + if ((GLRO (dl_hwcap2) & PPC_FEATURE2_ARCH_3_00) == 0
> + || (GLRO (dl_hwcap2) & PPC_FEATURE2_HAS_IEEE128) == 0)
> + return _dl_hwcaps_subdirs_build_bitmask (subdirs_count, active);
> + ++active;
Should we test PPC_FEATURE2_DARN as well? I think cryptographic and related
libraries might use it as source of entropy.
> +
> + /* POWER10. GCC defines __MMA__ for -mcpu=power10. */
Double space after period.
> + if ((GLRO (dl_hwcap2) & PPC_FEATURE2_ARCH_3_1) == 0
> + || (GLRO (dl_hwcap2) & PPC_FEATURE2_MMA) == 0)
> + return _dl_hwcaps_subdirs_build_bitmask (subdirs_count, active);
> + ++active;
> +
> + return _dl_hwcaps_subdirs_build_bitmask (subdirs_count, active);
> +}
Ok.
> diff --git a/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c b/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c
> new file mode 100644
> index 0000000000..e510fca80a
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/le/tst-glibc-hwcaps.c
> @@ -0,0 +1,54 @@
> +/* glibc-hwcaps subdirectory test. powerpc64le version.
> + Copyright (C) 2020 Free Software Foundation, Inc.
> + This file is part of the GNU C Library.
> +
> + The GNU C Library is free software; you can redistribute it and/or
> + modify it under the terms of the GNU Lesser General Public
> + License as published by the Free Software Foundation; either
> + version 2.1 of the License, or (at your option) any later version.
> +
> + The GNU C Library is distributed in the hope that it will be useful,
> + but WITHOUT ANY WARRANTY; without even the implied warranty of
> + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + Lesser General Public License for more details.
> +
> + You should have received a copy of the GNU Lesser General Public
> + License along with the GNU C Library; if not, see
> + <https://www.gnu.org/licenses/>. */
> +
> +#include <stdio.h>
> +#include <string.h>
> +#include <support/check.h>
> +#include <sys/auxv.h>
> +#include <sys/param.h>
> +
> +extern int marker2 (void);
> +extern int marker3 (void);
> +
> +/* Return the POWER level, 8 for the baseline. */
> +static int
> +compute_level (void)
> +{
> + const char *platform = (const char *) getauxval (AT_PLATFORM);
> + if (strcmp (platform, "power8") == 0)
> + return 8;
> + if (strcmp (platform, "power9") == 0)
> + return 9;
> + if (strcmp (platform, "power10") == 0)
> + return 10;
Why not use HWCAP? Is it to decouple from the loader code that uses it?
I am asking because by using hwcap we can check if this tests is working
on different chips by using qemu-ppc64le with -mcpu.
> + printf ("warning: unrecognized AT_PLATFORM value: %s\n", platform);
> + /* Assume that the new platform supports POWER10. */
> + return 10;
> +}
> +
> +static int
> +do_test (void)
> +{
> + int level = compute_level ();
> + printf ("info: detected POWER level: %d\n", level);
> + TEST_COMPARE (marker2 (), MIN (level - 7, 2));
> + TEST_COMPARE (marker3 (), MIN (level - 7, 3));
> + return 0;
> +}
> +
> +#include <support/test-driver.c>
>
Ok.
More information about the Libc-alpha
mailing list