Add matrix configs to DM

Add matrix configs to DM

BUG=skia:832
Committed: https://skia.googlesource.com/skia/+/e3aeaec7ddda15faa923d34b67b4cdba71f5d119

[borenet, 2015-05-05 15:39:43]

borenet@google.com changed reviewers:
+ mtklein@google.com

[borenet, 2015-05-05 15:39:44]

Am I anywhere close on this?  The Android bot is failing, presumably because an
assert thinks it's living in 2D.

[mtklein, 2015-05-05 15:51:43]

mtklein@google.com changed reviewers:
+ reed@google.com

[mtklein, 2015-05-05 15:51:44]

I actually think we may want to use upright-8888/upright-gpu on the bots. 
That's designed to manually undo the effects of certain matrices, allowing us to
actually compare the results of our matrix transformation with the vanilla draw.

Adding Mike... ViaUpright's his baby.

https://codereview.chromium.org/1125793002/diff/1/dm/DM.cpp
File dm/DM.cpp (right):

https://codereview.chromium.org/1125793002/diff/1/dm/DM.cpp#newcode398
dm/DM.cpp:398: VIA("upright", ViaUpright, m, wrapped);
I suspect ViaUpright won't really be a useful thing to have here with a general
matrix.  It's really geared to work with only scale/skew matrices.

https://codereview.chromium.org/1125793002/diff/1/tools/dm_flags.py
File tools/dm_flags.py (right):

https://codereview.chromium.org/1125793002/diff/1/tools/dm_flags.py#newcode34
tools/dm_flags.py:34: if '-CPU-' in bot:
You ought to be able to just put 'matrix-8888' and 'matrix-gpu' in configs
unconditionally.  The --nogpu / --nocpu we pass to the bots will filter the
irrelevant parts out, just like they filter out either 565 and 8888 or gpu.

[borenet, 2015-05-05 17:47:39]

Okay, I backed off a bit and switched to upright-* and a 2x2 matrix.

[mtklein, 2015-05-05 17:59:17]

On 2015/05/05 17:47:39, borenet wrote:
> Okay, I backed off a bit and switched to upright-* and a 2x2 matrix.

So, paging this back in, I think we've got to do it something like this:

  --matrix 0 1 1 0 --config upright-matrix-8888

matrix does the forward transformation, then upright undoes it.  upright only
works with a particular subset of matrices (must be rect-stays-rect, 90,180,270
degree rotations okay), and further we probably want to have no extra scale or
we won't be able to compare against the vanilla 8888.

[borenet, 2015-05-05 19:04:25]

On 2015/05/05 17:59:17, mtklein wrote:
> On 2015/05/05 17:47:39, borenet wrote:
> > Okay, I backed off a bit and switched to upright-* and a 2x2 matrix.
> 
> So, paging this back in, I think we've got to do it something like this:
> 
>   --matrix 0 1 1 0 --config upright-matrix-8888
> 
> matrix does the forward transformation, then upright undoes it.  upright only
> works with a particular subset of matrices (must be rect-stays-rect,
90,180,270
> degree rotations okay), and further we probably want to have no extra scale or
> we won't be able to compare against the vanilla 8888.

Done. I assume we'll want to test perspective at some point?  We'll need a new
3x3 matrix config for that.

[mtklein, 2015-05-05 19:58:16]

lgtm

Eh, maybe... not many (any?) folks use perspective today, and it's much harder
for us to triage perspective images without the assist of something like upright
to undo the transformation.  Once it gets beyond 90 degree rotations, and maybe
scales, even something like upright becomes nearly useless.

[mtklein, 2015-05-05 19:58:24]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2015-05-05 19:58:38]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1125793002/40001

[commit-bot: I haz the power, 2015-05-05 20:11:34]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/e3aeaec7ddda15faa923d34b67b4cdba71f5d119