[Piglit] A couple of python bugs found in logic.

[Dylan Baker]

On Thursday 19 September 2013 20:12:34 Ken Phillis Jr wrote:
> I probably should be a little bit more price about issue #3... I do expect
> the times to be different, but there is a problem of seeing less than one
> tenth of the total time when I only have four processors.
> 
> 
> Good example:
> ~/Documents/piglit $ grep \"time\" mesa-9.2.0-devel-quick/main | awk '{SUM
> += $2} END {print SUM}' | tail
> 1499.51
> ~/Documents/piglit $ grep \"time_elapsed\" mesa-9.2.0-devel-quick/main |awk
> '{print $2}'
> 302.1552939414978
> 
> Two bad examples:
> $ grep \"time_elapsed\" crash-example-1/main |awk '{print $2}'
> 632.9559738636017
> $ grep \"time\" crash-example-1/main | awk '{SUM += $2} END {print SUM}' |
> tail
> 7719.78
> $ grep \"time_elapsed\" crash-example-2/main |awk '{print $2}'
> 693.4222819805145
> $ grep \"time\" crash-example-2/main | awk '{SUM += $2} END {print SUM}'
> 9741.47
> 
> NOTE: The the good example had threading enabled, and gave an somewhat
> expected result since there is four threads to allow for concurrent tasks.
> However when you look at non-threaded results in contrast the result
> becomes extra exaggerated. So it may be a good idea to add better handling
> of these time stamps.
> 
> On Thu, Sep 19, 2013 at 5:57 PM, Ken Phillis Jr <kphillisjr at gmail.com>wrote:
> > I noticed that there is three bugs found within piglit framework that are
> > not what they should be.
> > 
> > 1) resuming tests is a lot more complex than it should be... These two
> > commands are examples of what should be used since the test profile is
> > always written to the test file. ( "profile" )
> > 
> > piglit-run.py -r My-Results-Directory
> > 
> > piglit-run.py -r My-Results-Directory2/main
> > 2) Resuming the piglit run with "-x some-test-to-skip" does not work. It
> > is simply ignored and the test that should be skipped is ran anyways.
> > 
> > 3) The time time_elapsed value may not reflect the actual elapsed time...
> > Especially if you need to resume tests for some reason.
> > Actual result:
> > $ grep \"time_elapsed\" example-piglit-run/main
> > 
> >     "time_elapsed": 693.4222819805145
> > 
> > Approximate result ( using awk )
> > $ grep \"time\" example-piglit-run/main | awk '{SUM += $2} END {print
> > SUM}' | tail
> > 9741.47

Taking this on list.

1) Resume support in piglit has always been more of a hack than anything else. 
it is on a rather long list of things that I would like to rewrite, but I have 
other things on my plate right now. Patches are welcome

2) this is by design. A resume is meant to pick up exactly where the previous 
run left off, since a -x argument is compiled to regex it would make 
reproducing a test impossible of *some* tests that matched that regex where 
filtered, but others were not.

3) I have no idea.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20130920/ded53095/attachment.pgp>