My current TODO list for apitrace
Carl Worth
cworth at cworth.org
Mon Aug 27 12:37:18 PDT 2012
Hi folks,
I've got a bunch of things that I want to with apitrace, so I thought I
would share these ideas with everyone on the list for better
coordination.
These are sorted according to my current priority. I plan to tackle
these roughly in the order presented:
1. Fix a few "apitrace trim" bugs
People have given me traces where trim fails. I'll feed these to
the stress tester and fix the bugs.
2. Add a "GPU" row next to the "CPU" row in the profile timeline view.
As discussed in my previous mail. If James (or anyone else)
wants to be me to this, that would be great!
3. Faster trimming
I've already got some changes implemented that dramatically
speed up "apitrace trim", (5x faster in one case I
measured). These will be easier to submit after the trim code is
merged upstream.
4. On-demand tracing
Trimming is sufficiently expensive that it would be nice to
avoid the need to do it by simply capturing the desired portion
of an application in the first place. So I'd like to have the
ability to interactively start tracing when desired.
My current thought is to take the state extraction of "glretrace
-D" and write code that reads that JSON and sets that state into
the OpenGL state machine.
5. Compile traces to C
I'm envisioning a modified version of "apitrace dump" that
provides code that can be compiled to get a program that
recreates the behavior of a trace.
6. Move trace decompression to a separate thread
To make it easier to get good numbers out of profiling trace
replay.
7. Even faster trimming
Step 3 above has some first, obvious steps to make my current
trim code faster, (replacing std::set<unsigned>), that I've
already implemented. There are still some obvious slow spots
that could be further improved.
-Carl
--
carl.d.worth at intel.com
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: not available
URL: <http://lists.freedesktop.org/archives/apitrace/attachments/20120827/2ce5c349/attachment.pgp>
More information about the apitrace
mailing list