Re: [RFC v2 2/3] drm/i915: duplicate i915_gem_ring_dispatch trace and add ctx parameter

["Ceraolo Spurio, Daniele" <daniele.ceraolospurio@intel.com>]

On 7/17/2014 5:25 PM, Chris Wilson wrote:
> On Wed, Jul 16, 2014 at 05:22:38PM +0100, daniele.ceraolospurio@intel.com wrote:
>> From: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
>>
>> The context used to execute a batchbuffer is becoming increasingly
>> important. Duplicating to avoid modifications to the original trace.
>
> I am sure we don't want both. The structure encoding is exposed to
> userspace so we are free to update the tracepoints within reason.

As you can see by the next patch in the series, I plan to add a callback 
inside the trace. My original patch modified the existing trace, but (if 
I've understood correctly) Daniel asked for a duplicated trace to avoid 
adding the callback into the existing one.

> I would also like a better ctx identifier than its pointer. Using the
> pointer for tracking objects makes it more difficult to read traces
> (although it is easy for scripts).

I use the VM pointer to track the ppgtt; that pointer is also printed by 
several other traces, including the ppgtt init/release ones that I've 
submitted for comments in this series. However, I don't mind changing 
the way we identify the ctx as long as I still have access to the VM 
pointer. I'll have a look at the possible ways of identifying the ctx 
and I'll try to find a better solution than the current one.

thanks,
Daniele