[PATCH] drm/i915/gvt: install gvt core threads with high priority

[Zhenyu Wang]

On 2018.02.01 16:16:12 +0800, Zhenyu Wang wrote:
> On 2018.02.01 14:11:05 +0800, Weinan Li wrote:
> > GVT-g core threads current use the default priority which is same as host
> > application, so if one application with heavy CPU workload running in host,
> > it will impact the latency of GVT-g core threads. This patch set the core
> > threads policy as FIFO and priority as high which is lower then i915 signal
> > thread for improving the scheduling performance.
> > 
> > Signed-off-by: Weinan Li <weinan.z.li at intel.com>
> > ---
> >  drivers/gpu/drm/i915/gvt/gvt.c       | 10 ++++++++++
> >  drivers/gpu/drm/i915/gvt/gvt.h       |  2 ++
> >  drivers/gpu/drm/i915/gvt/scheduler.c |  3 +++
> >  3 files changed, 15 insertions(+)
> > 
> > diff --git a/drivers/gpu/drm/i915/gvt/gvt.c b/drivers/gpu/drm/i915/gvt/gvt.c
> > index fac54f3..49006c5 100644
> > --- a/drivers/gpu/drm/i915/gvt/gvt.c
> > +++ b/drivers/gpu/drm/i915/gvt/gvt.c
> > @@ -33,6 +33,7 @@
> >  #include <linux/types.h>
> >  #include <xen/xen.h>
> >  #include <linux/kthread.h>
> > +#include <uapi/linux/sched/types.h>
> >  
> >  #include "i915_drv.h"
> >  #include "gvt.h"
> > @@ -253,6 +254,12 @@ static void init_device_info(struct intel_gvt *gvt)
> >  	info->msi_cap_offset = pdev->msi_cap;
> >  }
> >  
> > +void set_gvt_thread_priority(void)
> > +{
> > +	struct sched_param param = { .sched_priority = 10 };
> > +	sched_setscheduler_nocheck(current, SCHED_FIFO, &param);
> > +}
> 
> Let's use name as intel_gvt_set_thread_priority() and might inline it too.
> 
> > +
> >  static int gvt_service_thread(void *data)
> >  {
> >  	struct intel_gvt *gvt = (struct intel_gvt *)data;
> > @@ -260,6 +267,9 @@ static int gvt_service_thread(void *data)
> >  
> >  	gvt_dbg_core("service thread start\n");
> >  
> > +	/* Install ourselves with high priority to reduce scheduling latency */
> > +	set_gvt_thread_priority();
> > +
> >  	while (!kthread_should_stop()) {
> >  		ret = wait_event_interruptible(gvt->service_thread_wq,
> >  				kthread_should_stop() || gvt->service_request);
> > diff --git a/drivers/gpu/drm/i915/gvt/gvt.h b/drivers/gpu/drm/i915/gvt/gvt.h
> > index c88c489..262d281 100644
> > --- a/drivers/gpu/drm/i915/gvt/gvt.h
> > +++ b/drivers/gpu/drm/i915/gvt/gvt.h
> > @@ -337,6 +337,8 @@ static inline void intel_gvt_request_service(struct intel_gvt *gvt,
> >  	wake_up(&gvt->service_thread_wq);
> >  }
> >  
> > +void set_gvt_thread_priority(void);
> > +
> >  void intel_gvt_free_firmware(struct intel_gvt *gvt);
> >  int intel_gvt_load_firmware(struct intel_gvt *gvt);
> >  
> > diff --git a/drivers/gpu/drm/i915/gvt/scheduler.c b/drivers/gpu/drm/i915/gvt/scheduler.c
> > index b55b358..f71b917 100644
> > --- a/drivers/gpu/drm/i915/gvt/scheduler.c
> > +++ b/drivers/gpu/drm/i915/gvt/scheduler.c
> > @@ -847,6 +847,9 @@ static int workload_thread(void *priv)
> >  
> >  	gvt_dbg_core("workload thread for ring %d started\n", ring_id);
> >  
> > +	/* Install ourselves with high priority to reduce scheduling latency */
> > +	set_gvt_thread_priority();
> > +

Just to record my concern that setting high prio for workload thread which
is our in-kernel gfx client acting for vGPU, which might impact normal
host gfx process, do we have evaluated on that?

> >  	while (!kthread_should_stop()) {
> >  		add_wait_queue(&scheduler->waitq[ring_id], &wait);
> >  		do {
> > -- 
> > 1.9.1
> > 
> > _______________________________________________
> > intel-gvt-dev mailing list
> > intel-gvt-dev at lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev
> 
> -- 
> Open Source Technology Center, Intel ltd.
> 
> $gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827



> _______________________________________________
> intel-gvt-dev mailing list
> intel-gvt-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/intel-gvt-dev


-- 
Open Source Technology Center, Intel ltd.

$gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 195 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/intel-gvt-dev/attachments/20180201/4e3ab57e/attachment.sig>