[PATCH i-g-t v2 4/4] tests/intel/xe_pmu: Add multi client engine activity test
Riana Tauro
riana.tauro at intel.com
Tue Jun 3 06:37:30 UTC 2025
On 5/29/2025 11:18 PM, Umesh Nerlige Ramappa wrote:
> On Thu, May 29, 2025 at 12:23:59PM +0530, Riana Tauro wrote:
>> Add a test to validate engine activity by opening multiple
>> PMU clients and check that they do not interfere with
>> each other
>>
>> Signed-off-by: Riana Tauro <riana.tauro at intel.com>
>> ---
>> tests/intel/xe_pmu.c | 60 ++++++++++++++++++++++++++++++++++++++++++++
>> 1 file changed, 60 insertions(+)
>>
>> diff --git a/tests/intel/xe_pmu.c b/tests/intel/xe_pmu.c
>> index 21b294167..225b3b42e 100644
>> --- a/tests/intel/xe_pmu.c
>> +++ b/tests/intel/xe_pmu.c
>> @@ -46,6 +46,10 @@
>> * SUBTEST: engine-activity-suspend
>> * Description: Test to validate engine activity on all engines before
>> and after s2idle
>> *
>> + * SUBTEST: engine-activity-multi-client
>> + * Description: Test to validate engine activity with multiple PMU
>> clients and check that
>> + * they do not interfere with each other
>> + *
>> * SUBTEST: engine-activity-after-load-start
>> * Description: Validates engine activity when PMU is opened after
>> load started
>> *
>> @@ -612,6 +616,58 @@ static void engine_activity_load_start(int fd,
>> struct drm_xe_engine_class_instan
>> assert_within_epsilon(engine_active_ticks, engine_total_ticks,
>> tolerance);
>> }
>>
>> +static void engine_activity_multi_client(int fd, struct
>> drm_xe_engine_class_instance *eci,
>> + int num_clients)
>> +{
>> + uint64_t before[num_clients][2], after[num_clients][2],
>> pmu_fd[num_clients][2];
>
> This can be
>
> struct client {
> uint64_t before;
> uint64_t after;
> int pmu_fd;
> }[num_clients];
>
> Also, if num_clients is only ever going to be 2, you could drop the
> parameter to this function and just use 2 internally (maybe as a macro).
Sure will do that. Thank you for the review
Thanks
Riana
>
> Otherwise, the test/logic LGTM, with or without the above change, this is
>
> Reviewed-by: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
>
> Thanks,
> Umesh
>
>
>> + uint64_t ahnd, config, engine_active_ticks, engine_total_ticks;
>> + struct xe_cork *cork = NULL;
>> + uint32_t vm;
>> + int i = 0;
>> +
>> + vm = xe_vm_create(fd, 0, 0);
>> + ahnd = intel_allocator_open(fd, 0, INTEL_ALLOCATOR_RELOC);
>> +
>> + for (i = 0; i < num_clients; i++) {
>> + config = get_event_config(eci->gt_id, eci, "engine-active-
>> ticks");
>> + pmu_fd[i][0] = open_group(fd, config, -1);
>> + config = get_event_config(eci->gt_id, eci, "engine-total-
>> ticks");
>> + pmu_fd[i][1] = open_group(fd, config, pmu_fd[i][0]);
>> + }
>> +
>> + cork = xe_cork_create_opts(fd, eci, vm, 1, 1, .ahnd = ahnd);
>> + xe_cork_sync_start(fd, cork);
>> +
>> + for (i = 0; i < num_clients; i++)
>> + pmu_read_multi(pmu_fd[i][0], 2, before[i]);
>> +
>> + usleep(SLEEP_DURATION * USEC_PER_SEC);
>> +
>> + for (i = 0; i < num_clients; i++)
>> + pmu_read_multi(pmu_fd[i][0], 2, after[i]);
>> +
>> + end_cork(fd, cork);
>> + xe_cork_destroy(fd, cork);
>> + xe_vm_destroy(fd, vm);
>> + put_ahnd(ahnd);
>> +
>> + for (i = 0; i < num_clients; i++) {
>> + engine_active_ticks = after[i][0] - before[i][0];
>> + engine_total_ticks = after[i][1] - before[i][1];
>> +
>> + igt_debug("Client %d: Engine active ticks: after %ld, before
>> %ld delta %ld\n",
>> + i + 1, after[i][0], before[i][0], engine_active_ticks);
>> +
>> + igt_debug("Client %d Engine total ticks: after %ld, before
>> %ld delta %ld\n",
>> + i + 1, after[i][1], before[i][1], engine_total_ticks);
>> +
>> + close(pmu_fd[i][0]);
>> + close(pmu_fd[i][1]);
>> +
>> + assert_within_epsilon(engine_active_ticks,
>> engine_active_ticks, tolerance);
>> + }
>> +}
>> +
>> static void test_gt_c6_idle(int xe, unsigned int gt)
>> {
>> int pmu_fd;
>> @@ -889,6 +945,10 @@ igt_main
>> test_each_engine("engine-activity-after-load-start", fd, eci)
>> engine_activity_load_start(fd, eci);
>>
>> + igt_describe("Validate multiple PMU clients do not interfere with
>> each other");
>> + test_each_engine("engine-activity-multi-client", fd, eci)
>> + engine_activity_multi_client(fd, eci, 2);
>> +
>> igt_subtest_group {
>> int render_fd;
>>
>> --
>> 2.47.1
>>
More information about the igt-dev
mailing list