Remove SSE2 ColorRect32 code/files

Remove SSE2 ColorRect32 code/files

Removes the disabled SSE2 optimization of ColorRect32 and deletes
the two files containing the code.
Measured on both Core Haswell and Atom Silvermont, and only got
some miniscule improvement compared to the default implementation.

Also tried to write a new, ultimate, version of this optimization,
but only got ~5% improvement on ColorRect32-heavy tests.

Signed-off-by: Henrik Smiding <henrik.smiding@intel.com>;

Committed: https://skia.googlesource.com/skia/+/3704df347a585fdde78117fcedbee3c289b63e33

[henrik.smiding, 2015-02-24 16:24:49]

henrik.smiding@intel.com changed reviewers:
+ joakim.landberg@intel.com, mtklein@google.com

[henrik.smiding, 2015-02-24 16:26:15]

Please tell me if I need to add something to the chrome_changes file as well,
since I remove a cpp file from the build.

[mtklein, 2015-02-24 16:34:23]

Please just put back the PlatformColorRectProcFactory declaration before
submitting.  Otherwise, this LGTM.  Chrome should pick this up automatically now
that we use opts.gypi.

https://codereview.chromium.org/957433002/diff/1/src/opts/opts_check_x86.cpp
File src/opts/opts_check_x86.cpp (left):

https://codereview.chromium.org/957433002/diff/1/src/opts/opts_check_x86.cpp#...
src/opts/opts_check_x86.cpp:265: SkBlitRow::ColorRectProc
PlatformColorRectProcFactory(); // suppress warning
I think you need to keep this.  Otherwise PlatformColorRectProcFactory isn't
declared here, and some of our compilers are helpful / anal about that.

[mtklein, 2015-02-24 16:36:29]

(Another option is to completely remove all the references to
PlatformColorRectProcFactory, given that it always returns NULL.  If you don't
do this, I'll follow up.)

[henrik.smiding, 2015-02-25 09:41:46]

On 2015/02/24 16:36:29, mtklein wrote:
> (Another option is to completely remove all the references to
> PlatformColorRectProcFactory, given that it always returns NULL.  If you don't
> do this, I'll follow up.)

You mean remove all of them, in all files? Including the factory in
SkBlitRow_D32.cpp?

[henrik.smiding, 2015-02-25 09:41:56]

https://codereview.chromium.org/957433002/diff/1/src/opts/opts_check_x86.cpp
File src/opts/opts_check_x86.cpp (left):

https://codereview.chromium.org/957433002/diff/1/src/opts/opts_check_x86.cpp#...
src/opts/opts_check_x86.cpp:265: SkBlitRow::ColorRectProc
PlatformColorRectProcFactory(); // suppress warning
On 2015/02/24 16:34:23, mtklein wrote:
> I think you need to keep this.  Otherwise PlatformColorRectProcFactory isn't
> declared here, and some of our compilers are helpful / anal about that.

Ok, I was hoping it was an old leftover, since the corresponding arm-file didn't
have it.

[mtklein, 2015-02-25 13:23:31]

> Ok, I was hoping it was an old leftover, since the corresponding arm-file
didn't
> have it.

Sometimes we can get away with things in ARM builds because we know MSVC won't
ever be compiling it.  This might be one of those things.

Let's leave the SkBlitRow_*.cpp cleanup for later.  Looking through there,
there's lots of weird plumbing and specialization that I want to take a separate
look at.

[henrik.smiding, 2015-02-25 15:18:49]

On 2015/02/25 13:23:31, mtklein wrote:
> > Ok, I was hoping it was an old leftover, since the corresponding arm-file
> didn't
> > have it.
> 
> Sometimes we can get away with things in ARM builds because we know MSVC won't
> ever be compiling it.  This might be one of those things.
> 
> Let's leave the SkBlitRow_*.cpp cleanup for later.  Looking through there,
> there's lots of weird plumbing and specialization that I want to take a
separate
> look at.

Ok, unremoved the declaration.

[mtklein, 2015-02-25 15:30:20]

The CQ bit was checked by mtklein@google.com

[mtklein, 2015-02-25 15:30:21]

lgtm

[commit-bot: I haz the power, 2015-02-25 15:30:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/957433002/20001

[commit-bot: I haz the power, 2015-02-25 15:37:19]

Committed patchset #2 (id:20001) as
https://skia.googlesource.com/skia/+/3704df347a585fdde78117fcedbee3c289b63e33