Re: [PATCH] tools/perf/tests/expr: Make the system_tsc_freq test only for intel

[Namhyung Kim <namhyung@kernel.org>]

On Tue, Dec 03, 2024 at 10:42:45AM -0800, Namhyung Kim wrote:
> On Tue, Dec 03, 2024 at 10:16:06AM -0800, Namhyung Kim wrote:
> > Hello,
> > 
> > On Fri, Nov 08, 2024 at 10:50:10AM +0530, Athira Rajeev wrote:
> > > 
> > > 
> > > > On 7 Nov 2024, at 7:26 PM, Leo Yan <leo.yan@arm.com> wrote:
> > > > 
> > > > Hi Athira,
> > > > 
> > > > On Wed, Nov 06, 2024 at 03:04:57PM +0530, Athira Rajeev wrote:
> > > > 
> > > > [...]
> > > > 
> > > >>> Hi Athira,
> > > >>> 
> > > >>> sorry for the breakage and thank you for the detailed explanation. As
> > > >>> the code will run on AMD I think your change will break that - . It is
> > > >>> probably safest to keep the ".. else { .." for this case but guard it
> > > >>> in the ifdef.
> > > >>> 
> > > >> 
> > > >> Hi Ian
> > > >> 
> > > >> Thanks for your comments. Does the below change looks good ?
> > > >> 
> > > >> diff --git a/tools/perf/tests/expr.c b/tools/perf/tests/expr.c
> > > >> index e3aa9d4fcf3a..f5b2d96bb59b 100644
> > > >> --- a/tools/perf/tests/expr.c
> > > >> +++ b/tools/perf/tests/expr.c
> > > >> @@ -74,14 +74,12 @@ static int test__expr(struct test_suite *t __maybe_unused, int subtest __maybe_u
> > > >>    double val, num_cpus_online, num_cpus, num_cores, num_dies, num_packages;
> > > >>    int ret;
> > > >>    struct expr_parse_ctx *ctx;
> > > >> -    bool is_intel = false;
> > > >>    char strcmp_cpuid_buf[256];
> > > >>    struct perf_pmu *pmu = perf_pmus__find_core_pmu();
> > > >>    char *cpuid = perf_pmu__getcpuid(pmu);
> > > >>    char *escaped_cpuid1, *escaped_cpuid2;
> > > >> 
> > > >>    TEST_ASSERT_VAL("get_cpuid", cpuid);
> > > >> -    is_intel = strstr(cpuid, "Intel") != NULL;
> > > >> 
> > > >>    TEST_ASSERT_EQUAL("ids_union", test_ids_union(), 0);
> > > >> 
> > > >> @@ -244,11 +242,13 @@ static int test__expr(struct test_suite *t __maybe_unused, int subtest __maybe_u
> > > >>    if (num_dies) // Some platforms do not have CPU die support, for example s390
> > > >>        TEST_ASSERT_VAL("#num_dies >= #num_packages", num_dies >= num_packages);
> > > >> 
> > > >> +#if defined(__i386__) && defined(__x86_64__)
> > > >>    TEST_ASSERT_VAL("#system_tsc_freq", expr__parse(&val, ctx, "#system_tsc_freq") == 0);
> > > >> -    if (is_intel)
> > > >> +    if (strstr(cpuid, "Intel") != NULL)
> > > >>        TEST_ASSERT_VAL("#system_tsc_freq > 0", val > 0);
> > > >>    else
> > > >>        TEST_ASSERT_VAL("#system_tsc_freq == 0", fpclassify(val) == FP_ZERO);
> > > >> +#endif
> > > >> 
> > > >>    /*
> > > >>     * Source count returns the number of events aggregating in a leader
> > > > 
> > > > I confirmed the change above fixes the failure on Arm64.
> > > > 
> > > > Tested-by: Leo Yan <leo.yan@arm.com>
> > > Thanks Leo Yan for testing.
> > > 
> > > Hi Ian,
> > > 
> > > If the change above looks good, I will post a V2 . Please share your review comments
> > 
> > Sorry for the delay, it looks good to me.  Can you please send the v2?
> 
> After looking at another report, I think we need to check the value of
> TSC freq, not just the vendor.  Can you please test this?

Oops, nevermind.  I've realized we have two different issues at the same
time.  So !x86 archs should not use #system_tsc_freq at all, and only
*some* of (real) Intel machines have the value actually.  Hmm...

I think we need the original v2 here, and check the value even on Intel
separately.

Thanks,
Namhyung