Re: [igt-dev] [PATCH i-g-t] tests/i915_pm_rc6_residency: Add mutli-gt functionality to rc6_idle and rc6_fence test

["Sundaresan, Sujaritha" <sujaritha.sundaresan@intel.com>]

On 8/22/2023 11:06 AM, Riana Tauro wrote:
> Hi Suja
>
> typo in the patch header
Will change
>
> On 8/21/2023 6:27 PM, Sujaritha Sundaresan wrote:
>> Adding multi-gt capabilty to rc6_idle and rc6_fence tests
>>
>> Signed-off-by: Sujaritha Sundaresan <sujaritha.sundaresan@intel.com>
>> ---
>>   tests/i915/i915_pm_rc6_residency.c | 28 +++++++++++++++++++---------
>>   1 file changed, 19 insertions(+), 9 deletions(-)
>>
>> diff --git a/tests/i915/i915_pm_rc6_residency.c 
>> b/tests/i915/i915_pm_rc6_residency.c
>> index b266680ac..d727dab12 100644
>> --- a/tests/i915/i915_pm_rc6_residency.c
>> +++ b/tests/i915/i915_pm_rc6_residency.c
>> @@ -376,7 +376,7 @@ static void kill_children(int sig)
>>       signal(sig, old);
>>   }
>>   -static void rc6_idle(int i915, uint32_t ctx_id, uint64_t flags)
>> +static void rc6_idle(int i915, uint32_t ctx_id, uint64_t flags, 
>> unsigned int gt)
>>   {
>>       const int64_t duration_ns = SLEEP_DURATION * 
>> (int64_t)NSEC_PER_SEC;
>>       const int tolerance = 20; /* Some RC6 is better than none! */
>> @@ -397,7 +397,7 @@ static void rc6_idle(int i915, uint32_t ctx_id, 
>> uint64_t flags)
>>       struct igt_power gpu;
>>       int fd;
>>   -    fd = open_pmu(i915, I915_PMU_RC6_RESIDENCY);
>> +    fd = open_pmu(i915, __I915_PMU_RC6_RESIDENCY(gt));
>>       igt_drop_caches_set(i915, DROP_IDLE);
>>       igt_require(__pmu_wait_for_rc6(fd));
>>       igt_power_open(i915, &gpu, "gpu");
>> @@ -471,12 +471,13 @@ static void rc6_idle(int i915, uint32_t ctx_id, 
>> uint64_t flags)
>>       }
>>   }
>>   -static void rc6_fence(int i915, const intel_ctx_t *ctx)
>> +static void rc6_fence(int i915, unsigned int gt)
>>   {
>>       const int64_t duration_ns = SLEEP_DURATION * 
>> (int64_t)NSEC_PER_SEC;
>>       const int tolerance = 20; /* Some RC6 is better than none! */
>>       const unsigned int gen = intel_gen(intel_get_drm_devid(i915));
>>       const struct intel_execution_engine2 *e;
>> +    const intel_ctx_t *ctx;
>>       struct power_sample sample[2];
>>       unsigned long slept;
>>       uint64_t rc6, ts[2], ahnd;
>> @@ -485,7 +486,7 @@ static void rc6_fence(int i915, const intel_ctx_t 
>> *ctx)
>>         igt_require_sw_sync();
>>   -    fd = open_pmu(i915, I915_PMU_RC6_RESIDENCY);
>> +    fd = open_pmu(i915, __I915_PMU_RC6_RESIDENCY(gt));
>>       igt_drop_caches_set(i915, DROP_IDLE);
>>       igt_require(__pmu_wait_for_rc6(fd));
>>       igt_power_open(i915, &gpu, "gpu");
>> @@ -509,6 +510,7 @@ static void rc6_fence(int i915, const intel_ctx_t 
>> *ctx)
>>       assert_within_epsilon(rc6, ts[1] - ts[0], 5);
>>         /* Submit but delay execution, we should be idle and 
>> conserving power */
>> +    ctx = intel_ctx_create_for_gt(i915, gt);
>>       ahnd = get_reloc_ahnd(i915, ctx->id);
>>       for_each_ctx_engine(i915, ctx, e) {
>>           igt_spin_t *spin;
>> @@ -550,6 +552,7 @@ static void rc6_fence(int i915, const intel_ctx_t 
>> *ctx)
>>           gem_quiescent_gpu(i915);
>>       }
>>       put_ahnd(ahnd);
>> +    intel_ctx_destroy(i915, ctx);
>>         igt_power_close(&gpu);
>>       close(fd);
>> @@ -558,6 +561,7 @@ static void rc6_fence(int i915, const intel_ctx_t 
>> *ctx)
>>   igt_main
>>   {
>>       int i915 = -1;
>> +    int dir, gt;
>>       const intel_ctx_t *ctx;
>>
> Remove intel_ctx_create_all_physical from igt_fixture since we are not
> using it anywhere
Sure will remove the redundant code
>>       /* Use drm_open_driver to verify device existence */
>> @@ -572,10 +576,14 @@ igt_main
>>           igt_require_gem(i915);
>>           gem_quiescent_gpu(i915);
>>   -        for_each_ctx_engine(i915, ctx, e) {
>> -            if (e->instance == 0) {
>> -                igt_dynamic_f("%s", e->name)
>> -                    rc6_idle(i915, ctx->id, e->flags);
>> +        i915_for_each_gt(i915, dir, gt) {
>> +            ctx = intel_ctx_create_for_gt(i915, gt);
>> +            for_each_ctx_engine(i915, ctx, e) {
>> +                if (e->instance == 0) {
>> +                    igt_dynamic_f("%s", gt, e->name)
> format specifier missing for gt.
Will fix
>> +                        rc6_idle(i915, ctx->id, e->flags, gt);
>> +                }
>> +            intel_ctx_destroy(i915, ctx);
>>               }
>>           }
>>       }
>> @@ -584,7 +592,9 @@ igt_main
>>           igt_require_gem(i915);
>>           gem_quiescent_gpu(i915);
> This should be changed from igt_subtest to igt_subtest_with_dynamic.
>
> Thanks
> Riana

I have a feeling I sent an old patch instead of the latest version. 
Because the test would not pass with this patch.

Will double check and resend

Thanks for the review,

Suja

>>   -        rc6_fence(i915, ctx);
>> +        i915_for_each_gt(i915, dir, gt)
>> +            igt_dynamic_f("gt%u", gt)
>> +                rc6_fence(i915, gt);
>>       }
>>         igt_subtest_group {