Re: [PATCH v2] drm/i915/trace: add hw_id to gem requests trace points

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Quoting Lionel Landwerlin (2017-12-15 18:06:03)
> When monitoring the GPU with i915 perf, reports are tagged with a hw
> id. When also tracking the requests using the kernel tracepoints, if
> we include the hw_id from i915_gem_context, this allows us to
> correlate a process with hw id fields in the OA reports.
> 
> v2: Place hw_id at the end of the tracepoint to not disrupt too much
>     existing tools (Chris)
> 
> Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@xxxxxxxxx>

This is a good excuse to reopen an old thread:
20170511130045.32271-1-tvrtko.ursulin@xxxxxxxxxxxxxxx
│     On 11/05/2017 14:16, Tvrtko Ursulin wrote:
│     >
│     > On 11/05/2017 14:07, Chris Wilson wrote:
│     >> On Thu, May 11, 2017 at 02:00:45PM +0100, Tvrtko Ursulin wrote:
│     >>> From: Tvrtko Ursulin <tvrtko.ursulin@xxxxxxxxx>
│     >>>
│     >>> For userspace receiving binary data it is easier if all related
│     >>> request tracepoints emit the binary data in the same order of
│     >>> dev, ring, ctx, seqno, ...
│     >>
│     >> We decided that dev, ctx, ring, seqno was the right heirachy last time.
│     >> After much debate :)
│     >> ctx is the logical view of the device for a user
│     >> ctx + ring = timeline
│     >
│     > I couldn't remember so I thought it must have been what is documented in
│     > TP_printk. Now I am confused. On one hand it's true, but on the other
│     > ctx/seqno is also a pair as opposed to engine being stuck in the middle.
│     > So ring-ctx-seqno is easier for humans I think. Even since per
│     > ctx/engine seqno space.
│     
│     [thread bump!]
│     
│     Can we agree what to do here? Printk's are all engine-ctx-seqno which
│     makes it sound like that was the order we agreed upon. But binary blobs
│     are inconsistent as you have noticed - do we care enough to go and
│     fiddle with that is the question? Is overlay the only userspace to your
│     knowledge that accesses the binary blobs?
│     
│     Regards,
│     
│     Tvrtko

Given that we are adding info here, we can also go through and make the
order consistent. dev, hw_id, engine, ctx, seqno?
-Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx




[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux