On Sun, Jul 14, 2013 at 01:28:11PM -0700, Ben Widawsky wrote:
> On Fri, Jul 12, 2013 at 06:08:27PM +0100, Chris Wilson wrote:
> > The TRACE_EVENT_CONDITION is supposed to generate more efficient code
> > than if (cond) trace(), which is what we are currently using inside the
> > register access functions.
> > 
> > v2: Rebase onto uncore
> > 
> > Signed-off-by: Chris Wilson <[email protected]>
> 
> One thing I've thought about this is a trace_on_once, or trace_on_poll
> would be cool for polled reads. Obviously we can post-process it, but we
> always risk losing events if we fill up too much of the ringbuffer with
> stuff we don't care about.

I see what you mean, having a singular entry would be better than none.
We could use a local variable and a new macro.

> In any case, I very much like the condition added to the trace itself,
> and reading probably the same docs as you agree it should be faster.
> 
> Looking over our code. It looks like I missed trace events when writing
> the PTEs in i915_gem_gtt.c. Can you get around to fixing those up?

Indeed, we need to do a review of what tracepoints we have and we need.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre
_______________________________________________
Intel-gfx mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/intel-gfx

Reply via email to