* [PATCH i-g-t] tests/xe_gt_freq: Use sync reset
@ 2024-09-25 22:09 Vinay Belgaumkar
2024-09-26 14:36 ` Rodrigo Vivi
2024-10-03 21:27 ` Dixit, Ashutosh
0 siblings, 2 replies; 3+ messages in thread
From: Vinay Belgaumkar @ 2024-09-25 22:09 UTC (permalink / raw)
To: intel-gfx, igt-dev; +Cc: Vinay Belgaumkar, Badal Nilawar
Some recent WAs reduce the GT freq during driver load/reset. Use sync
reset so that we give enough time for GT frequency to be restored after
reset has completed.
Also, stash/restore frequencies per GT as they can be different.
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2676
Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2711
Cc: Badal Nilawar <badal.nilawar@intel.com>
Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
---
tests/intel/xe_gt_freq.c | 24 +++++++++++++++++-------
1 file changed, 17 insertions(+), 7 deletions(-)
diff --git a/tests/intel/xe_gt_freq.c b/tests/intel/xe_gt_freq.c
index 427e12c55..de4d111ea 100644
--- a/tests/intel/xe_gt_freq.c
+++ b/tests/intel/xe_gt_freq.c
@@ -341,7 +341,8 @@ static void test_reset(int fd, int gt_id, int cycles)
igt_assert_f(get_freq(fd, gt_id, "cur") == rpn,
"Failed after %d good cycles\n", i);
- xe_force_gt_reset_async(fd, gt_id);
+ xe_force_gt_reset_sync(fd, gt_id);
+
usleep(SLPC_FREQ_LATENCY_US);
igt_assert_f(get_freq(fd, gt_id, "min") == rpn,
@@ -433,16 +434,23 @@ igt_main
int fd;
int gt;
struct drm_xe_engine_class_instance *hwe;
- uint32_t stash_min;
- uint32_t stash_max;
+ uint32_t *stash_min, *stash_max;
+ int num_gts;
igt_fixture {
fd = drm_open_driver(DRIVER_XE);
igt_require(xe_sysfs_gt_has_node(fd, 0, "freq0"));
+ num_gts = xe_number_gt(fd);
+
/* The defaults are the same. Stashing the gt0 is enough */
- stash_min = get_freq(fd, 0, "min");
- stash_max = get_freq(fd, 0, "max");
+ stash_min = (uint32_t *) malloc(sizeof(uint32_t) * num_gts);
+ stash_max = (uint32_t *) malloc(sizeof(uint32_t) * num_gts);
+
+ xe_for_each_gt(fd, gt) {
+ stash_min[gt] = get_freq(fd, gt, "min");
+ stash_max[gt] = get_freq(fd, gt, "max");
+ }
}
igt_subtest("throttle_basic_api") {
@@ -517,9 +525,11 @@ igt_main
igt_fixture {
xe_for_each_gt(fd, gt) {
- set_freq(fd, gt, "min", stash_min);
- set_freq(fd, gt, "max", stash_max);
+ set_freq(fd, gt, "max", stash_max[gt]);
+ set_freq(fd, gt, "min", stash_min[gt]);
}
+ free(stash_min);
+ free(stash_max);
drm_close_driver(fd);
}
}
--
2.38.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [PATCH i-g-t] tests/xe_gt_freq: Use sync reset
2024-09-25 22:09 [PATCH i-g-t] tests/xe_gt_freq: Use sync reset Vinay Belgaumkar
@ 2024-09-26 14:36 ` Rodrigo Vivi
2024-10-03 21:27 ` Dixit, Ashutosh
1 sibling, 0 replies; 3+ messages in thread
From: Rodrigo Vivi @ 2024-09-26 14:36 UTC (permalink / raw)
To: Vinay Belgaumkar; +Cc: intel-gfx, igt-dev, Badal Nilawar
On Wed, Sep 25, 2024 at 03:09:23PM -0700, Vinay Belgaumkar wrote:
> Some recent WAs reduce the GT freq during driver load/reset. Use sync
> reset so that we give enough time for GT frequency to be restored after
> reset has completed.
>
> Also, stash/restore frequencies per GT as they can be different.
>
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2676
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2711
>
> Cc: Badal Nilawar <badal.nilawar@intel.com>
> Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> ---
> tests/intel/xe_gt_freq.c | 24 +++++++++++++++++-------
> 1 file changed, 17 insertions(+), 7 deletions(-)
>
> diff --git a/tests/intel/xe_gt_freq.c b/tests/intel/xe_gt_freq.c
> index 427e12c55..de4d111ea 100644
> --- a/tests/intel/xe_gt_freq.c
> +++ b/tests/intel/xe_gt_freq.c
> @@ -341,7 +341,8 @@ static void test_reset(int fd, int gt_id, int cycles)
> igt_assert_f(get_freq(fd, gt_id, "cur") == rpn,
> "Failed after %d good cycles\n", i);
>
> - xe_force_gt_reset_async(fd, gt_id);
> + xe_force_gt_reset_sync(fd, gt_id);
> +
> usleep(SLPC_FREQ_LATENCY_US);
>
> igt_assert_f(get_freq(fd, gt_id, "min") == rpn,
> @@ -433,16 +434,23 @@ igt_main
> int fd;
> int gt;
> struct drm_xe_engine_class_instance *hwe;
> - uint32_t stash_min;
> - uint32_t stash_max;
> + uint32_t *stash_min, *stash_max;
> + int num_gts;
>
> igt_fixture {
> fd = drm_open_driver(DRIVER_XE);
>
> igt_require(xe_sysfs_gt_has_node(fd, 0, "freq0"));
> + num_gts = xe_number_gt(fd);
> +
> /* The defaults are the same. Stashing the gt0 is enough */
> - stash_min = get_freq(fd, 0, "min");
> - stash_max = get_freq(fd, 0, "max");
> + stash_min = (uint32_t *) malloc(sizeof(uint32_t) * num_gts);
> + stash_max = (uint32_t *) malloc(sizeof(uint32_t) * num_gts);
> +
> + xe_for_each_gt(fd, gt) {
> + stash_min[gt] = get_freq(fd, gt, "min");
> + stash_max[gt] = get_freq(fd, gt, "max");
> + }
> }
>
> igt_subtest("throttle_basic_api") {
> @@ -517,9 +525,11 @@ igt_main
>
> igt_fixture {
> xe_for_each_gt(fd, gt) {
> - set_freq(fd, gt, "min", stash_min);
> - set_freq(fd, gt, "max", stash_max);
> + set_freq(fd, gt, "max", stash_max[gt]);
> + set_freq(fd, gt, "min", stash_min[gt]);
> }
> + free(stash_min);
> + free(stash_max);
> drm_close_driver(fd);
> }
> }
> --
> 2.38.1
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH i-g-t] tests/xe_gt_freq: Use sync reset
2024-09-25 22:09 [PATCH i-g-t] tests/xe_gt_freq: Use sync reset Vinay Belgaumkar
2024-09-26 14:36 ` Rodrigo Vivi
@ 2024-10-03 21:27 ` Dixit, Ashutosh
1 sibling, 0 replies; 3+ messages in thread
From: Dixit, Ashutosh @ 2024-10-03 21:27 UTC (permalink / raw)
To: Vinay Belgaumkar; +Cc: intel-gfx, igt-dev, Badal Nilawar
On Wed, 25 Sep 2024 15:09:23 -0700, Vinay Belgaumkar wrote:
>
> Some recent WAs reduce the GT freq during driver load/reset. Use sync
> reset so that we give enough time for GT frequency to be restored after
> reset has completed.
>
> Also, stash/restore frequencies per GT as they can be different.
>
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2676
> Closes: https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/2711
>
No empty line here.
> Cc: Badal Nilawar <badal.nilawar@intel.com>
> Signed-off-by: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Please use 'intel' in the commit title, intel is not the only vendor
contributing to igt.
I have fixed these up this time and merged this.
Thanks.
--
Ashutosh
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2024-10-03 21:27 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2024-09-25 22:09 [PATCH i-g-t] tests/xe_gt_freq: Use sync reset Vinay Belgaumkar
2024-09-26 14:36 ` Rodrigo Vivi
2024-10-03 21:27 ` Dixit, Ashutosh
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox