SoftLight with SkPMFloat

SoftLight with SkPMFloat

SSE speeds up about 4.5x over existing integer SSE,
NEON speeds up about 3x over serial integer code.

We expect 1-2 bit component diffs in the usual GMs.

Still guarded by SK_SUPPORT_LEGACY_XFERMODES,
which I'll now try to lift in Chrome.


BUG=skia:

Committed: https://skia.googlesource.com/skia/+/3e47d49b46b3ab62071218ef3dd44642c9713e04

CQ_EXTRA_TRYBOTS=client.skia:Test-ChromeOS-GCC-Daisy-CPU-NEON-Arm7-Debug-Trybot

Committed: https://skia.googlesource.com/skia/+/823b2a76e24bcac23e984dff025a7d53fc4aa584

[mtklein, 2015-06-27 17:08:51]

The CQ bit was checked by mtklein@google.com to run a CQ dry run

[commit-bot: I haz the power, 2015-06-27 17:09:01]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1221493002/80001

[mtklein, 2015-06-27 17:12:29]

The CQ bit was checked by mtklein@google.com to run a CQ dry run

[commit-bot: I haz the power, 2015-06-27 17:12:40]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1221493002/100001

[mtklein_C, 2015-06-27 17:14:49]

mtklein@chromium.org changed reviewers:
+ reed@google.com
- mtklein@google.com

[commit-bot: I haz the power, 2015-06-27 17:18:21]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-06-27 17:18:22]

Dry run: This issue passed the CQ dry run.

[reed1, 2015-06-29 15:32:09]

lgtm

[mtklein, 2015-06-29 15:33:11]

The CQ bit was checked by mtklein@google.com

[commit-bot: I haz the power, 2015-06-29 15:33:18]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1221493002/100001

[commit-bot: I haz the power, 2015-06-29 15:40:33]

Committed patchset #6 (id:100001) as
https://skia.googlesource.com/skia/+/3e47d49b46b3ab62071218ef3dd44642c9713e04

[mtklein, 2015-06-29 16:37:20]

A revert of this CL (patchset #6 id:100001) has been created in
https://codereview.chromium.org/1221683002/ by mtklein@google.com.

The reason for reverting is: xfermodes and xfermodes2 show major diffs on Nexus
5 and Daisy (both ARMv7 w/NEON).  Nexus 9 and SSE all look fine....

[mtklein, 2015-06-29 16:43:25]

On 2015/06/29 16:37:20, mtklein wrote:
> A revert of this CL (patchset #6 id:100001) has been created in
> https://codereview.chromium.org/1221683002/ by mailto:mtklein@google.com.
> 
> The reason for reverting is: xfermodes and xfermodes2 show major diffs on
Nexus
> 5 and Daisy (both ARMv7 w/NEON).  Nexus 9 and SSE all look fine....

This points to sqrt(), where SSE and ARMv8 have a native sqrt(), and on ARMv7
it's emulated via approximate inverse sqrt().

[mtklein, 2015-06-29 16:46:28]

On 2015/06/29 16:43:25, mtklein wrote:
> On 2015/06/29 16:37:20, mtklein wrote:
> > A revert of this CL (patchset #6 id:100001) has been created in
> > https://codereview.chromium.org/1221683002/ by mailto:mtklein@google.com.
> > 
> > The reason for reverting is: xfermodes and xfermodes2 show major diffs on
> Nexus
> > 5 and Daisy (both ARMv7 w/NEON).  Nexus 9 and SSE all look fine....
> 
> This points to sqrt(), where SSE and ARMv8 have a native sqrt(), and on ARMv7
> it's emulated via approximate inverse sqrt().

SkPMColorAssert() is also failing on those same bots, e.g.

https://uberchromegw.corp.google.com/i/client.skia.android/builders/Test-Andr...

[mtklein, 2015-06-29 18:09:40]

On 2015/06/29 16:46:28, mtklein wrote:
> On 2015/06/29 16:43:25, mtklein wrote:
> > On 2015/06/29 16:37:20, mtklein wrote:
> > > A revert of this CL (patchset #6 id:100001) has been created in
> > > https://codereview.chromium.org/1221683002/ by mailto:mtklein@google.com.
> > > 
> > > The reason for reverting is: xfermodes and xfermodes2 show major diffs on
> > Nexus
> > > 5 and Daisy (both ARMv7 w/NEON).  Nexus 9 and SSE all look fine....
> > 
> > This points to sqrt(), where SSE and ARMv8 have a native sqrt(), and on
ARMv7
> > it's emulated via approximate inverse sqrt().
> 
> SkPMColorAssert() is also failing on those same bots, e.g.
> 
>
https://uberchromegw.corp.google.com/i/client.skia.android/builders/Test-Andr...

Looks like it was different treatment of NaNs.  Solved, I think, by paying
better attention to < vs. <=.

[mtklein, 2015-06-29 18:10:04]

The CQ bit was checked by mtklein@google.com

[mtklein, 2015-06-29 18:10:05]

The patchset sent to the CQ was uploaded after l-g-t-m from reed@google.com
Link to the patchset: https://codereview.chromium.org/1221493002/#ps120001
(title: "Edges matter.")

[commit-bot: I haz the power, 2015-06-29 18:10:14]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1221493002/120001

[commit-bot: I haz the power, 2015-06-29 19:16:29]

Committed patchset #7 (id:120001) as
https://skia.googlesource.com/skia/+/823b2a76e24bcac23e984dff025a7d53fc4aa584