[Intel-gfx] [PATCH] Ring buffer dumping (Working)

[Martin Olsson]

Ben Gamari wrote:
> After reimplementing the batch buffer dumping semantics (thanks to Chris Wilson
> for his suggestion), it seems that batch buffer dumping is finally pretty
> stable. Despite what are probably pretty bad inadequacies in my parsing script,
> the output looks pretty reasonable. Moreover, we now are dumping as many batch
> buffers as are still in memory, so we should be guaranteed to dump any
> batchbuffers which cause a gpu crash. Anyways, I'll be putting up a git
> repository soon with all of this. Finally, I've included the latest revision of
> my parsing script below, so dumping functionality should even be usable. Let me
> know what you think. Thanks,

Nice work!

This might be overkill but if you anticipate more updates to this script, then
maybe you should add an automatic update on startup sort of like how
the alsa-info.sh script updates itself from hg if you run an old version.

For details, check out the update() function at the top of:
http://hg.alsa-project.org/alsa/file/86421c9031db/alsa-info.sh

This may or may not be releated to _your_ script but in general I think it would be
nice with some kind of script that script that dumped _all_ relevant info
into a single file that could be attached to bug reports. For example, dmesg,
contents of "i915_gem_interrupt" sampled twice with 1 sec in between to see
if interrupts are still firing and even xorg_log, xorg_conf and maybe versions
of packages etc.



		Martin