[PATCH i-g-t] tests/xe_exec_threads: Skip Compute engine for legacy mode
Upadhyay, Tejas
tejas.upadhyay at intel.com
Mon Apr 1 17:18:21 UTC 2024
> -----Original Message-----
> From: De Marchi, Lucas <lucas.demarchi at intel.com>
> Sent: Monday, April 1, 2024 9:31 PM
> To: Upadhyay, Tejas <tejas.upadhyay at intel.com>
> Cc: igt-dev at lists.freedesktop.org; intel-xe at lists.freedesktop.org; Brost,
> Matthew <matthew.brost at intel.com>
> Subject: Re: [PATCH i-g-t] tests/xe_exec_threads: Skip Compute engine for
> legacy mode
>
> On Mon, Apr 01, 2024 at 08:18:34PM +0530, Tejas Upadhyay wrote:
> >RCS/CCS are dependent engines as they are sharing reset domain.
> >Whenever there is reset from CCS, all the exec queues running on RCS
> >are victimised.
> >
> >Lets skip parallel execution on CCS with RCS.
> >
> >Fixes: VLK-55941, VLK-52341
>
> these shouldn't be here.
>
> Also in the title of this commit, what do you mean by "legacy mode"? It
> seems you are just blocking CCS regardless of any flags.
Legacy meaning non-compute mode, skipping for compute indirectly meaning going in legacy mode and not compute mode.
Thanks,
Tejas
>
> Lucas De Marchi
>
> >Signed-off-by: Tejas Upadhyay <tejas.upadhyay at intel.com>
> >---
> > tests/intel/xe_exec_threads.c | 8 ++++++++
> > 1 file changed, 8 insertions(+)
> >
> >diff --git a/tests/intel/xe_exec_threads.c
> >b/tests/intel/xe_exec_threads.c index e9588f9b9..b1c65e1c5 100644
> >--- a/tests/intel/xe_exec_threads.c
> >+++ b/tests/intel/xe_exec_threads.c
> >@@ -996,6 +996,14 @@ static void threads(int fd, int flags)
> > }
> >
> > xe_for_each_engine(fd, hwe) {
> >+ /* RCS/CCS sharing reset domain hence dependent engines.
> >+ * When CCS is doing reset, all the contexts of RCS are
> >+ * victimised, so skip the compute engine avoiding
> >+ * parallel execution with RCS
> >+ */
> >+ if (hwe->engine_class ==
> DRM_XE_ENGINE_CLASS_COMPUTE)
> >+ continue;
> >+
> > threads_data[i].mutex = &mutex;
> > threads_data[i].cond = &cond;
> > #define ADDRESS_SHIFT 39
> >--
> >2.25.1
> >
More information about the igt-dev
mailing list