From: "Sundaresan, Sujaritha" <sujaritha.sundaresan@intel.com>
To: Riana Tauro <riana.tauro@intel.com>, <igt-dev@lists.freedesktop.org>
Cc: badal.nilawar@intel.com
Subject: Re: [igt-dev] [PATCH i-g-t 2/3] tests/xe: Add a test to toggle GT C states and validate power
Date: Wed, 23 Aug 2023 14:10:47 +0530 [thread overview]
Message-ID: <596f4301-2088-6976-db5d-91dbe43fd86e@intel.com> (raw)
In-Reply-To: <20230817055301.855960-3-riana.tauro@intel.com>
On 8/17/2023 11:23 AM, Riana Tauro wrote:
> Add a test that toggles GT C states by checking if GT is in C0 when
> forcewake is acquired and in C6 once released.
> This test also validates power consumed by GPU in GT C6 is lesser
> than GT C0.
>
> v2: change test name (Suja, Anshuman)
> v3: add exit handler for closing forcewake handle
> define a macro for number of repetitions (Vinay)
> change forcewake open to readonly (Anshuman)
> modify test to validate power
>
> Signed-off-by: Riana Tauro <riana.tauro@intel.com>
> ---
> tests/xe/xe_pm_residency.c | 67 ++++++++++++++++++++++++++++++++++++++
> 1 file changed, 67 insertions(+)
>
> diff --git a/tests/xe/xe_pm_residency.c b/tests/xe/xe_pm_residency.c
> index fd4876eaf..3a6611fac 100644
> --- a/tests/xe/xe_pm_residency.c
> +++ b/tests/xe/xe_pm_residency.c
> @@ -14,14 +14,17 @@
>
> #include "igt.h"
> #include "igt_device.h"
> +#include "igt_power.h"
> #include "igt_sysfs.h"
>
> #include "xe/xe_query.h"
> #include "xe/xe_util.h"
>
> +#define NUM_REPS 16 /* No of Repetitions */
> #define SLEEP_DURATION 3000 /* in milliseconds */
>
> const double tolerance = 0.1;
> +int forcewake_handle = -1;
>
> #define assert_within_epsilon(x, ref, tol) \
> igt_assert_f((double)(x) <= (1.0 + (tol)) * (double)(ref) && \
> @@ -50,9 +53,19 @@ enum test_type {
> * Description: Validate idle residency measured over suspend(s2idle)
> * is greater than suspend time or within tolerance
> * Run type: FULL
> + *
> + * SUBTEST: toggle-gt-c6
> + * Description: toggles GT C states by acquiring/releasing forcewake,
> + * also validates power consumed by GPU in GT C6 is lesser than that of GT C0.
> + * Run type: FULL
> */
> IGT_TEST_DESCRIPTION("Tests for gtidle properties");
>
> +static void restore(int sig)
> +{
> + close(forcewake_handle);
> +}
> +
> static unsigned int measured_usleep(unsigned int usec)
> {
> struct timespec ts = { };
> @@ -117,6 +130,54 @@ static void test_idle_residency(int fd, int gt, enum test_type flag)
> assert_within_epsilon(residency_end - residency_start, elapsed_ms, tolerance);
> }
>
> +static void measure_power(struct igt_power *gpu, double *power)
> +{
> + struct power_sample power_sample[2];
> +
> + igt_power_get_energy(gpu, &power_sample[0]);
> + measured_usleep(SLEEP_DURATION * 1000);
> + igt_power_get_energy(gpu, &power_sample[1]);
> + *power = igt_power_get_mW(gpu, &power_sample[0], &power_sample[1]);
> +}
> +
> +static void toggle_gt_c6(int fd, int n)
> +{
> + double gt_c0_power, gt_c6_power;
> + int gt;
> + struct igt_power gpu;
> +
> + igt_power_open(fd, &gpu, "gpu");
> +
> + do {
> + forcewake_handle = igt_debugfs_open(fd, "forcewake_all", O_RDONLY);
> + igt_assert(forcewake_handle >= 0);
> + /* check if all gts are in C0 after forcewake is acquired */
> + xe_for_each_gt(fd, gt)
> + igt_assert_f(!xe_is_gt_in_c6(fd, gt),
> + "Forcewake acquired, GT should be in C0\n");
> +
> + if (n == NUM_REPS)
> + measure_power(&gpu, >_c0_power);
> +
> + close(forcewake_handle);
> + /* check if all gts are in C6 after forcewake is released */
> + xe_for_each_gt(fd, gt)
> + igt_assert_f(igt_wait(xe_is_gt_in_c6(fd, gt), 1000, 1),
> + "Forcewake released, GT should be in C6\n");
> +
> + if (n == NUM_REPS)
> + measure_power(&gpu, >_c6_power);
> + } while (n--);
> +
> + igt_power_close(&gpu);
> + igt_info("GPU consumed %fmW in GT C6 and %fmW in GT C0\n", gt_c6_power, gt_c0_power);
> +
> + /* FIXME: Remove dgfx check after hwmon is added */
> + if (!xe_has_vram(fd))
> + igt_assert_f(gt_c6_power < gt_c0_power,
> + "Power consumed in GT C6 should be lower than GT C0\n");
> +}
> +
> igt_main
> {
> uint32_t d3cold_allowed;
> @@ -152,6 +213,12 @@ igt_main
> xe_for_each_gt(fd, gt)
> test_idle_residency(fd, gt, TEST_IDLE);
>
> + igt_describe("Toggle GT C states by acquiring/releasing forcewake and validate power measured");
> + igt_subtest("toggle-gt-c6") {
> + igt_install_exit_handler(restore);
> + toggle_gt_c6(fd, NUM_REPS);
> + }
> +
> igt_fixture {
> close(fd);
> }
lgtm,
Reviewed-by: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
next prev parent reply other threads:[~2023-08-23 8:41 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-08-17 5:52 [igt-dev] [PATCH i-g-t 0/3] Toggle GT C States and validate power Riana Tauro
2023-08-17 5:52 ` [igt-dev] [PATCH i-g-t 1/3] lib/igt_power: Modify igt_power library for xe Riana Tauro
2023-08-18 8:00 ` Kamil Konieczny
2023-08-18 8:36 ` Riana Tauro
2023-08-17 5:53 ` [igt-dev] [PATCH i-g-t 2/3] tests/xe: Add a test to toggle GT C states and validate power Riana Tauro
2023-08-23 8:40 ` Sundaresan, Sujaritha [this message]
2023-08-17 5:53 ` [igt-dev] [PATCH i-g-t 3/3] HAX: Add toggle-gt-c6 to xe-fast-feedback.testlist Riana Tauro
2023-08-17 6:35 ` [igt-dev] ✗ GitLab.Pipeline: warning for Toggle GT C States and validate power Patchwork
2023-08-17 7:01 ` [igt-dev] ○ CI.xeBAT: info " Patchwork
2023-08-17 7:08 ` [igt-dev] ✓ Fi.CI.BAT: success " Patchwork
2023-08-17 18:07 ` [igt-dev] ✓ Fi.CI.IGT: " Patchwork
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=596f4301-2088-6976-db5d-91dbe43fd86e@intel.com \
--to=sujaritha.sundaresan@intel.com \
--cc=badal.nilawar@intel.com \
--cc=igt-dev@lists.freedesktop.org \
--cc=riana.tauro@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