From: Riana Tauro <riana.tauro@intel.com>
To: Vinay Belgaumkar <vinay.belgaumkar@intel.com>,
<intel-gfx@lists.freedesktop.org>,
<igt-dev@lists.freedesktop.org>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>,
Rodrigo Vivi <rodrigo.vivi@intel.com>
Subject: Re: [PATCH i-g-t 2/2] tests/xe_pmu: Add frequency test
Date: Thu, 3 Apr 2025 16:09:32 +0530 [thread overview]
Message-ID: <5894cf23-7110-4dfd-bbb8-a758f167b468@intel.com> (raw)
In-Reply-To: <20250402013912.2981094-3-vinay.belgaumkar@intel.com>
Hi Vinay
VF igt patches were merged so this will need a rebase
On 4/2/2025 7:09 AM, Vinay Belgaumkar wrote:
> Add a basic test that uses PMU to read GT actual and requested
> frequencies while running a workload.
>
> Cc: Lucas De Marchi <lucas.demarchi@intel.com>
> Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
> ---
> tests/intel/xe_pmu.c | 93 ++++++++++++++++++++++++++++++++++++++++++++
> 1 file changed, 93 insertions(+)
>
> diff --git a/tests/intel/xe_pmu.c b/tests/intel/xe_pmu.c
> index 66edf24ad..43085b4d0 100644
> --- a/tests/intel/xe_pmu.c
> +++ b/tests/intel/xe_pmu.c
> @@ -226,6 +226,95 @@ static void test_gt_c6_idle(int xe, unsigned int gt)
> close(pmu_fd);
> }
>
> +/**
> + * SUBTEST: gt-frequency
> + * Description: Validate we can collect accurate frequency PMU stats
> + * while running a workload.
> + */
> +static void test_gt_frequency(int fd, struct drm_xe_engine_class_instance *eci)
> +{
> + struct xe_cork *cork = NULL;
> + uint64_t end[2], start[2];
> + unsigned long config_rq_freq, config_act_freq;
> + double min[2], max[2];
> + uint32_t gt = eci->gt_id;
> + uint32_t orig_min = xe_gt_get_freq(fd, eci->gt_id, "min");
> + uint32_t orig_max = xe_gt_get_freq(fd, eci->gt_id, "max");
> + uint32_t vm;
> + int pmu_fd[2];
> +
> + config_rq_freq = get_event_config(fd, gt, NULL, "gt-requested-frequency");
> + pmu_fd[0] = open_group(fd, config_rq_freq, -1);
> +
> + config_act_freq = get_event_config(fd, gt, NULL, "gt-actual-frequency");
> + pmu_fd[1] = open_group(fd, config_act_freq, pmu_fd[0]);
> +
> + vm = xe_vm_create(fd, 0, 0);
> +
> + cork = xe_cork_create_opts(fd, eci, vm, 1, 1);
> + xe_cork_sync_start(fd, cork);
> +
> + /*
> + * Set GPU to min frequency and read PMU counters.
> + */
> + igt_assert(xe_gt_set_freq(fd, gt, "max", orig_min) > 0);
> + igt_assert(xe_gt_get_freq(fd, gt, "max") == orig_min);
> +
> + pmu_read_multi(pmu_fd[0], 2, start);
> + usleep(SLEEP_DURATION * USEC_PER_SEC);
> + pmu_read_multi(pmu_fd[0], 2, end);
> +
> + min[0] = (end[0] - start[0]);
> + min[1] = (end[1] - start[1]);
> +
> + /*
> + * Set GPU to max frequency and read PMU counters.
> + */
> + igt_assert(xe_gt_set_freq(fd, gt, "max", orig_max) > 0);
> + igt_assert(xe_gt_get_freq(fd, gt, "max") == orig_max);
> + igt_assert(xe_gt_set_freq(fd, gt, "min", orig_max) > 0);
> + igt_assert(xe_gt_get_freq(fd, gt, "min") == orig_max);
Should a exit handler be added to restore frequencies in case of assert
failure?> +
> + pmu_read_multi(pmu_fd[0], 2, start);
> + usleep(SLEEP_DURATION * USEC_PER_SEC);
> + pmu_read_multi(pmu_fd[0], 2, end);
> +
> + max[0] = (end[0] - start[0]);
> + max[1] = (end[1] - start[1]);
> +
> + if (!cork->ended)This can be removed. The engine activity had a trailing idle case which
ended the cork before. The lib already has this check> +
xe_cork_sync_end(fd, cork);
> +
> + /*
> + * Restore min/max.
> + */
> + igt_assert(xe_gt_set_freq(fd, gt, "min", orig_min) > 0);
> + igt_assert(xe_gt_get_freq(fd, gt, "min") == orig_min);
> +
> + igt_info("Minimum frequency: requested %.1f, actual %.1f\n",
> + min[0], min[1]);
> + igt_info("Maximum frequency: requested %.1f, actual %.1f\n",
> + max[0], max[1]);
> +
> + close(pmu_fd[0]);
> + close(pmu_fd[1]);
> +
> + if (cork)
> + xe_cork_destroy(fd, cork);
> +
> + xe_vm_destroy(fd, vm);
> +
> + close(pmu_fd[0]);
> + close(pmu_fd[1]);
> +
> + assert_within_epsilon(min[0], orig_min, tolerance);
> + /*
> + * On thermally throttled devices we cannot be sure maximum frequency
> + * can be reached so use larger tolerance downwards.
> + */
> + assert_within_epsilon_up_down(max[0], orig_max, tolerance, 0.15f);
> +}
> +
> igt_main
> {
> int fd, gt;
> @@ -254,6 +343,10 @@ igt_main
> test_each_engine("engine-activity-load", fd, eci)
> engine_activity(fd, eci, TEST_LOAD);
>
> + igt_describe("Validate PMU GT freq measured is within the tolerance");Can't we have this per-gt instead of engine?
Thanks
Riana> + test_each_engine("gt-frequency", fd, eci)
> + test_gt_frequency(fd, eci);
> +
> igt_fixture {
> close(fd);
> }
next prev parent reply other threads:[~2025-04-03 10:40 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2025-04-02 1:39 [PATCH i-g-t 0/2] Add PMU tests for GT frequency Vinay Belgaumkar
2025-04-02 1:39 ` [PATCH i-g-t 1/2] lib/xe_gt: Move get/set GT freq utils to lib Vinay Belgaumkar
2025-04-03 10:50 ` Riana Tauro
2025-04-02 1:39 ` [PATCH i-g-t 2/2] tests/xe_pmu: Add frequency test Vinay Belgaumkar
2025-04-03 10:39 ` Riana Tauro [this message]
2025-04-02 2:19 ` ✓ Xe.CI.BAT: success for Add PMU tests for GT frequency Patchwork
2025-04-02 2:38 ` ✓ i915.CI.BAT: " Patchwork
2025-04-02 5:24 ` ✗ i915.CI.Full: failure " Patchwork
2025-04-02 5:47 ` ✗ Xe.CI.Full: " Patchwork
2025-04-06 23:48 ` Patchwork
-- strict thread matches above, loose matches on Subject: below --
2025-04-07 23:44 [PATCH i-g-t 0/2] Add PMU test " Vinay Belgaumkar
2025-04-07 23:44 ` [PATCH i-g-t 2/2] tests/xe_pmu: Add frequency test Vinay Belgaumkar
2025-04-09 9:58 ` Riana Tauro
2025-04-09 11:06 ` Kamil Konieczny
2025-04-10 1:26 ` Belgaumkar, Vinay
2025-04-10 1:31 ` Belgaumkar, Vinay
2025-04-10 1:33 [PATCH i-g-t 0/2] Add PMU test for GT frequency Vinay Belgaumkar
2025-04-10 1:33 ` [PATCH i-g-t 2/2] tests/xe_pmu: Add frequency test Vinay Belgaumkar
2025-04-11 7:33 ` Riana Tauro
2025-04-11 22:36 ` Belgaumkar, Vinay
2025-04-11 22:45 [PATCH i-g-t 0/2] Add PMU test for GT frequency Vinay Belgaumkar
2025-04-11 22:46 ` [PATCH i-g-t 2/2] tests/xe_pmu: Add frequency test Vinay Belgaumkar
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=5894cf23-7110-4dfd-bbb8-a758f167b468@intel.com \
--to=riana.tauro@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=intel-gfx@lists.freedesktop.org \
--cc=lucas.demarchi@intel.com \
--cc=rodrigo.vivi@intel.com \
--cc=vinay.belgaumkar@intel.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox