Split Convolver.SIMDVerification into two tests

Split Convolver.SIMDVerification into two tests

Split Convolver.SIMDVerification into two tests, one which tests all tricky edge cases and one
that tests that large scales don't end up distorting the results when using SIMD. Also enable
running these tests with DEBUG and TSAN.

On my machine, running with DEBUG and ASAN, this test went from 60 seconds to 10 seconds.
Hope that is fast enough, as the additional coverage would be nice to have.

BUG=134400
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=253409

[Stephen White, 2014-02-25 21:13:59]

LGTM.

(As a side note, we will be switching to the Skia implementation of high-quality
filtering in the not-too-distant future, so hopefully we can remove this code at
that point.)

https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest.cc
File skia/ext/convolver_unittest.cc (right):

https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest....
skia/ext/convolver_unittest.cc:322: int dest_sizes[][2] = { {1280, 1024}, {177,
123} };
I'm surprised these run in reasonable time in DEBUG. The big image combinations
were taking approximately 45 seconds before, and intermittently falling over the
45 second timeout, leading to flakiness. But perhaps that was only due to the
small (edge cases) being resized to the large destination sizes, or because
you've removed the 640x480 destination size. 

At any rate, please keep an eye on the runtimes on the bots when this lands,
since they may have weaker hardware than what you're running locally.

[hubbe, 2014-02-25 21:59:18]

https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest.cc
File skia/ext/convolver_unittest.cc (right):

https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest....
skia/ext/convolver_unittest.cc:322: int dest_sizes[][2] = { {1280, 1024}, {177,
123} };
On 2014/02/25 21:13:59, Stephen White wrote:
> I'm surprised these run in reasonable time in DEBUG. The big image
combinations
> were taking approximately 45 seconds before, and intermittently falling over
the
> 45 second timeout, leading to flakiness. But perhaps that was only due to the
> small (edge cases) being resized to the large destination sizes, or because
> you've removed the 640x480 destination size. 
> 
> At any rate, please keep an eye on the runtimes on the bots when this lands,
> since they may have weaker hardware than what you're running locally.

A good chunk of time was spent comparing the images I think.
With ASAN, on my machine, replacing the comparison with memcmp saved 50% of the
time...

[hubbe, 2014-02-25 21:59:27]

The CQ bit was checked by hubbe@chromium.org

[Stephen White, 2014-02-25 22:12:14]

On 2014/02/25 21:59:18, hubbe wrote:
>
https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest.cc
> File skia/ext/convolver_unittest.cc (right):
> 
>
https://codereview.chromium.org/178013006/diff/1/skia/ext/convolver_unittest....
> skia/ext/convolver_unittest.cc:322: int dest_sizes[][2] = { {1280, 1024},
{177,
> 123} };
> On 2014/02/25 21:13:59, Stephen White wrote:
> > I'm surprised these run in reasonable time in DEBUG. The big image
> combinations
> > were taking approximately 45 seconds before, and intermittently falling over
> the
> > 45 second timeout, leading to flakiness. But perhaps that was only due to
the
> > small (edge cases) being resized to the large destination sizes, or because
> > you've removed the 640x480 destination size. 
> > 
> > At any rate, please keep an eye on the runtimes on the bots when this lands,
> > since they may have weaker hardware than what you're running locally.
> 
> A good chunk of time was spent comparing the images I think.
> With ASAN, on my machine, replacing the comparison with memcmp saved 50% of
the
> time...

Ahh, I missed that in the diffs. Nice!

[commit-bot: I haz the power, 2014-02-25 22:33:03]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/hubbe@chromium.org/178013006/1

[commit-bot: I haz the power, 2014-02-25 23:43:02]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/hubbe@chromium.org/178013006/1

[Paweł Hajdan Jr., 2014-02-26 05:45:02]

The CQ bit was unchecked by phajdan.jr@chromium.org

[Paweł Hajdan Jr., 2014-02-26 06:01:04]

The CQ bit was checked by phajdan.jr@chromium.org

[commit-bot: I haz the power, 2014-02-26 06:14:29]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/hubbe@chromium.org/178013006/1

[commit-bot: I haz the power, 2014-02-26 11:43:40]

Change committed as 253409