On Fri, 16 Sep 2011 11:07:29 +0200, Daniel Vetter <dan...@ffwll.ch> wrote: > On Fri, Sep 16, 2011 at 09:37:41AM +0100, Chris Wilson wrote: > > On Fri, 16 Sep 2011 09:58:53 +0200, Daniel Vetter <dan...@ffwll.ch> wrote: > > > We already have trace_i915_gem_request_add in i915_add_request which is > > > essentially giving out the same information (well, minus the hopefully > > > correct singal_mbox reg addresses). I think we can drop this one. > > > > I thought that as well, but then realised that a complete set of ring > > tracepoints would be useful independent of the request/object > > tracepoints. > > > > A couple of tracepoints on the gen6 paths is insufficients. ;-) > > Well, I've quickly reviewed them and I think we're mostly covered. Many of > the tracepoints are called from generic code i915_gem.c right before/after > calling into the chipset-specific ringbuffer functions.
It's just the difference of being able to watch only i915:intel_ring* to reduce the amount of noise if that is all you are interested in. And you don't have to remember the relationship between a particular i915_gem_request and its effect upon the ring. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
