[PATCH] Implement support for GPU timings of calls when proofiling with glretrace.

[Carl Worth]

Carl Worth <cworth at cworth.org> writes:
> It doesn't currently run with the Intel driver, but that looks like
> missing support for GL_TIMESTAMP in the driver, (which we should just
> add—it should be strictly easier than the existing support for
> GL_TIME_ELAPSED).

I stubbed things out so that I could actually run this branch with the
Intel driver. I found that things don't work for any trace that uses
glBegin/glEnd to draw things. (It ends up calling glBeginQuery between
glBegin and glEnd resulting in GL_INVALID_OPERATION errors).

My series avoids this problem by taking advantage of the new
CALL_FLAG_RENDER_BEGIN and CALL_FLAG_RENDER_END flags which I
added. What my code does is simply call glBeginQuery immediately before
glBegin and glEndQuery immediately after glEnd so that the whole block
is treated as a single drawing operation. (The GPU time is emitted in
the profile report with the call number of the glEnd call).

So that's one feature that would be nice to still have if we go with
James' code.

-Carl

-------------- 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/20120801/0acae06f/attachment.pgp>