[igt-dev] [PATCH i-g-t v3 26/31] lib/i915/perf: Enable multi-tile support for perf library

[Dixit, Ashutosh]

On Thu, 16 Mar 2023 16:33:18 -0700, Umesh Nerlige Ramappa wrote:
>

Hi Umesh,

> @@ -584,20 +584,25 @@ static const char *intel_sysfs_attr_name[2][RPS_MAX_ATTR] =
>		"gt/gt0/rps_min_freq_mhz",
>		"gt/gt0/rps_max_freq_mhz",
>	},
> +	{
> +		"gt/gt1/rps_min_freq_mhz",
> +		"gt/gt1/rps_max_freq_mhz",
> +	},
>  };
>
>  static const char *
> -intel_sysfs_attr_id_to_name(int sysfs_dirfd, intel_sysfs_attr_id id)
> +intel_sysfs_attr_id_to_name(int sysfs_dirfd, intel_sysfs_attr_id id, int gt)
>  {
>	assert(id < RPS_MAX_ATTR);
> +	assert(gt < sizeof(intel_sysfs_attr_name) - 1);

Isn't this sizeof pretty large? ;-)

> diff --git a/tests/i915/perf.c b/tests/i915/perf.c
> index 5f516aef..f3434b99 100644
> --- a/tests/i915/perf.c
> +++ b/tests/i915/perf.c
> @@ -1123,7 +1123,7 @@ init_sys_info(void)
>  {
>	igt_assert_neq(devid, 0);
>
> -	intel_perf = intel_perf_for_fd(drm_fd);
> +	intel_perf = intel_perf_for_fd(drm_fd, 0);

Here the question is why only gt0 when on MTL we have tests running on gt1
too? Because I suspect the only thing gt dependent are the freq's and we
are not using those on gt1? Still probably better to have an array of
intel_perf, one per gt? Is that a later patch?

Otherwise lgtm so:

Reviewed-by: Ashutosh Dixit <ashutosh.dixit at intel.com>