[PATCH i-g-t v2 4/5] tests/xe_compute_preempt: adjust number of children according to ram size

[Dandamudi, Priyanka]

LGTM,
Reviewed-by: Priyanka Dandamudi <priyanka.dandamudi at intel.com>

> -----Original Message-----
> From: Dandamudi, Priyanka
> Sent: 07 April 2025 02:20 PM
> To: Kempczynski, Zbigniew <Zbigniew.Kempczynski at intel.com>; igt-
> dev at lists.freedesktop.org
> Cc: Dugast, Francois <Francois.Dugast at intel.com>
> Subject: RE: [PATCH i-g-t v2 4/5] tests/xe_compute_preempt: adjust number
> of children according to ram size
> 
> 
> 
> > -----Original Message-----
> > From: Kempczynski, Zbigniew <zbigniew.kempczynski at intel.com>
> > Sent: 04 April 2025 06:02 PM
> > To: igt-dev at lists.freedesktop.org
> > Cc: Kempczynski, Zbigniew <zbigniew.kempczynski at intel.com>; Dugast,
> > Francois <francois.dugast at intel.com>; Dandamudi, Priyanka
> > <priyanka.dandamudi at intel.com>
> > Subject: [PATCH i-g-t v2 4/5] tests/xe_compute_preempt: adjust number
> > of children according to ram size
> >
> > Setting arbitrary number or children is prone to oom scenario and
> > getting test to be killed.
> >
> > Single job with sip turned on takes ~100MB for Xe2 tasks so adjust
> > number of children to consume ~50% ram.
> >
> > Signed-off-by: Zbigniew Kempczyński <zbigniew.kempczynski at intel.com>
> > Cc: Francois Dugast <francois.dugast at intel.com>
> > Cc: Priyanka Dandamudi <priyanka.dandamudi at intel.com>
> > ---
> >  tests/intel/xe_compute_preempt.c | 20 ++++++++++++++------
> >  1 file changed, 14 insertions(+), 6 deletions(-)
> >
> > diff --git a/tests/intel/xe_compute_preempt.c
> > b/tests/intel/xe_compute_preempt.c
> > index 876e538237..f07e87d4c4 100644
> > --- a/tests/intel/xe_compute_preempt.c
> > +++ b/tests/intel/xe_compute_preempt.c
> > @@ -40,13 +40,18 @@ test_compute_preempt(int fd, struct
> > drm_xe_engine_class_instance *hwe, bool thre
> >  	igt_require_f(run_intel_compute_kernel_preempt(fd, hwe,
> > threadgroup_preemption), "GPU not supported\n");  }
> >
> > +#define CONTEXT_MB 100
> > +
> >  igt_main
> >  {
> >  	int xe;
> >  	struct drm_xe_engine_class_instance *hwe;
> > +	uint64_t ram_mb;
> >
> > -	igt_fixture
> > +	igt_fixture {
> >  		xe = drm_open_driver(DRIVER_XE);
> > +		ram_mb = igt_get_avail_ram_mb();
> > +	}
> >
> >  	igt_subtest_with_dynamic("compute-preempt") {
> >  		xe_for_each_engine(xe, hwe) {
> > @@ -64,13 +69,16 @@ igt_main
> >  				continue;
> >
> >  			igt_dynamic_f("engine-%s",
> > xe_engine_class_string(hwe->engine_class)) {
> > -				uint16_t dev_id = intel_get_drm_devid(xe);
> >  				int child_count;
> >
> > -				if (IS_PANTHERLAKE(dev_id))
> > -					child_count = 50;
> > -				else
> > -					child_count = 100;
> > +				/*
> > +				 * Get half of ram / 2, then divide by
> > +				 * CONTEXT_MB * 2 (long and short) job
> > +				 */
> > +				child_count = ram_mb / 2 / CONTEXT_MB / 2;
> Here it should be CONTEXT_MB* 2 or CONTEXT_MB/2 ??
> -- Priyanka
> > +
> > +				igt_debug("RAM: %zd, child count: %d\n",
> > +					  ram_mb, child_count);
> >
> >  				test_compute_preempt(xe, hwe, false);
> >  				igt_fork(child, child_count)
> > --
> > 2.34.1