Re: x86: Don't check PREFETCHWT1 in tst-cpu-features-cpuinfo.c

[Noah Goldstein <goldstein.w.n@gmail.com>]

On Fri, Mar 10, 2023 at 1:04 AM Noah Goldstein <goldstein.w.n@gmail.com> wrote:
>
> On Thu, Mar 9, 2023 at 9:43 PM DJ Delorie via Libc-alpha
> <libc-alpha@sourceware.org> wrote:
> >
> >
> > As not noted in the kernel sources:
> >
> > > /* Intel-defined CPU features, CPUID level 0x00000007:0 (ECX), word 16 */
> > > #define X86_FEATURE_AVX512VBMI          (16*32+ 1) /* AVX512 Vector Bit Manipulation instructions*/
> > > #define X86_FEATURE_UMIP                (16*32+ 2) /* User Mode Instruction Protection */
> >
> > vs glibc:
> >
> > > /* ECX.  */
> > > #define bit_cpu_PREFETCHWT1   (1u << 0)
> > > #define bit_cpu_AVX512_VBMI   (1u << 1)
> > > #define bit_cpu_UMIP          (1u << 2)
> >
> > Tested on the one machine I could find in our inventory that set that flag:
> >
> > - FAIL: elf/tst-cpu-features-cpuinfo
> > - FAIL: elf/tst-cpu-features-cpuinfo-static
> > -     27 FAIL
> > +     23 FAIL
> >
> > (there were two unrelated timeouts in the "before" results)
> >
> > From c4a62abbeac4ced531ced3999a2cd2d4fab6bdc6 Mon Sep 17 00:00:00 2001
> > From: DJ Delorie <dj@redhat.com>
> > Date: Thu, 9 Mar 2023 22:32:54 -0500
> > Subject: x86: Don't check PREFETCHWT1 in tst-cpu-features-cpuinfo.c
> >
> > Don't check PREFETCHWT1 against /proc/cpuinfo since kernel doesn't report
> > PREFETCHWT1 in /proc/cpuinfo.
> >
> > diff --git a/sysdeps/x86/tst-cpu-features-cpuinfo.c b/sysdeps/x86/tst-cpu-features-cpuinfo.c
> > index c25240774e..e963592c4b 100644
> > --- a/sysdeps/x86/tst-cpu-features-cpuinfo.c
> > +++ b/sysdeps/x86/tst-cpu-features-cpuinfo.c
> > @@ -217,7 +217,10 @@ do_test (int argc, char **argv)
> >    fails += CHECK_PROC (pku, PKU);
> >    fails += CHECK_PROC (popcnt, POPCNT);
> >    fails += CHECK_PROC (3dnowprefetch, PREFETCHW);
> > +#if 0
> > +  /* NB: /proc/cpuinfo doesn't report this feature.  */
> >    fails += CHECK_PROC (prefetchwt1, PREFETCHWT1);
> > +#endif
> >  #if 0
> >    /* NB: /proc/cpuinfo doesn't report this feature.  */
> >    fails += CHECK_PROC (ptwrite, PTWRITE);
> >
>
> LGTM.
Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>