[Pixman] [PATCH] Add a lowlevel blitter benchmark

Siarhei Siamashka siarhei.siamashka at gmail.com
Wed Sep 15 06:09:24 PDT 2010

On Wednesday 15 September 2010 15:37:41 Dmitri Vorobiev wrote:
> From: Jonathan Morton <jonathan.morton at movial.com>
> This test is a modified version of Siarhei's compositor throughput
> benchmark.  It's expanded with explicit reporting of memory bandwidth
> consumption for the M-test, and with an additional 8x8-random test
> intended to determine peak ops/sec capability.  There are also quite a
> lot more operations tested for.
> We try to be portable as much as we can, and take care of systems
> that do not support gettimeofday() or memalign() calls. For this
> reason, we add checking for these calls, and update test utils to
> fall back to time() or malloc(), respectively.

I'm sorry for confusing between clock() and time() functions in my previous
reply. Surely it should have been clock() function, not time(). Because the
latter has only 1 second precision, which is rather bad.

Anyway, both clock() and time() are standard C functions, so there should be no
need to check for either of them in the configure script.

The header file 'malloc.h' is nonstandard and should not be included. Actually
the existing 'aligned_malloc' function already provides the needed
functionality and has been in use for a some time already (thus proven not to
cause problems). So just moving 'aligned_malloc' to 'utils.*' and using it from
all the test programs seems to be the right choice. And it can be done by a
separate patch before introducing the rest of changes.

Also this benchmark does not belong to TESTPROGRAMS. TESTPROGRAMS are run on
'make check' invocation and are testing correctness. Adding a benchmarking
program there is really not a good idea.

Best regards,
Siarhei Siamashka
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/pixman/attachments/20100915/c436e02a/attachment.pgp>

More information about the Pixman mailing list