[RFC] [PATCH 1/2] Add a basic support of logging cpu and gpu time

[Zack Rusin]

On Thursday, August 25, 2011 04:43:13 AM Yuanhan Liu wrote:
> On Thu, 2011-08-25 at 06:06 +0800, Jose Fonseca wrote:
> > Yes. Although there are use cases for measuring the real application, I
> > do feel that measuring times in retrace time is more useful, as
> > everything becomes reproducible & easily automated.
> 
> Yeah, right, Jose talked to me about this. While, one issue:
> 
> There are too overhead while tracing or retracing the app. Say, here I
> have an app:
> 
> fps     test-plane
> 182     normally run
> 70      by glxtrace
> 100	by glretrace
> 
> 175     by glxtrace, a separate thread to compress and write data to
> disk.

What branch are those timings based on? I don't think I've seen that code. I 
remember an email with gzwrite commented out and that was only 174fps back 
then, so it's surprising that not compressing/not writing to disk is slower 
than doing it in a separate thread.

> In the glxtrace side, we can delay the real write and even open another
> separate thread to write the trace data. So that the overhead while
> tracing can be reduced lots.
> 
> While, we can not delay the read while retracing, right? So that it's
> hard to reduce the big overhead.

Yea, we can. In fact it'd be probably a lot easier, since we control the 
entire app. We create a ringbuffer of calls and decompress in a separate 
thread. The rendering thread just pops the calls from the buffer and executes 
them. We were just never optimizing the retracer because it wasn't a priority. 
It's really fast as it is so there has never been a need for it. Plus with 
snappy it's even faster.

z