Move npot resizing out of GrContext and simplify GrContext texture functions.

Move npot resizing out of GrContext and simplify GrContext texture functions.

Committed: https://skia.googlesource.com/skia/+/8a8100349105c8c6de39fcb34e47679da7a67f54

Committed: https://skia.googlesource.com/skia/+/6c96672491b04cb782bce8fee778124df66524a0

Committed: https://skia.googlesource.com/skia/+/37f9a2694c15f08e361ebda74fe9f0fffbf452aa

[bsalomon, 2015-01-28 22:23:16]

bsalomon@google.com changed reviewers:
+ robertphillips@google.com

[bsalomon, 2015-01-28 22:23:16]

This also removes the cpu upscaling code path which we used to use when the
config wasn't renderable. Now I upgrade the config to a renderable config (e.g.
a8->rgba8 when a8 isn't a renderable config). This lets us rescale compressed
textures as well. We also can now rescale images that don't have a content key
(e.g. volatile bitmaps).

We no longer needed tex-params or keys as paramters to GrContext texture
creation/cache functions so I removed them

There is still plenty of cleanup to be done but the CL was getting big so I
stopped.

[bsalomon, 2015-01-28 22:24:42]

On 2015/01/28 22:23:16, bsalomon wrote:
> This also removes the cpu upscaling code path which we used to use when the
> config wasn't renderable. Now I upgrade the config to a renderable config
(e.g.
> a8->rgba8 when a8 isn't a renderable config). This lets us rescale compressed
> textures as well. We also can now rescale images that don't have a content key
> (e.g. volatile bitmaps).
> 

The CPU scaling code only did nearest sampling. So now when the params indicate
bilerp and the original config isn't renderable the rescaled texture looks much
better. However, it may use more texture memory.

[robertphillips, 2015-01-30 14:47:21]

first pass - I would like to go through it again though.

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h
File include/gpu/GrContext.h (right):

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h...
include/gpu/GrContext.h:216: /**
Remove "specified key and" ?

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h...
include/gpu/GrContext.h:276: /**
Remove "the specified" ?

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
File src/effects/SkBlurMaskFilter.cpp (right):

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
src/effects/SkBlurMaskFilter.cpp:929: unsigned int texSide = smallRectSide +
2*blurRadius;
I guess we could move the creation of 'texDesc' down closer to its use in
createTexture.

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
src/effects/SkBlurMaskFilter.cpp:956: 
2 overlength lines

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp
File src/gpu/SkGr.cpp (right):

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:114: if (origKey.isValid() && kNo_Stretch != stretch) {
Why not use the domain from the original key?

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:180: 
or two -> of two ?

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:182: // width or height. If optionalKey is valid it will be set
on the new texture. stretch
neartest -> nearest ?

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:190: 
it's

[bsalomon, 2015-01-30 15:37:40]

new patch uploaded

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h
File include/gpu/GrContext.h (right):

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h...
include/gpu/GrContext.h:216: /**
On 2015/01/30 14:47:21, robertphillips wrote:
> Remove "specified key and" ?

Done.

https://codereview.chromium.org/882223003/diff/120001/include/gpu/GrContext.h...
include/gpu/GrContext.h:276: /**
On 2015/01/30 14:47:21, robertphillips wrote:
> Remove "the specified" ?

Done.

I'd like to remove this function and just have a generic "is this pixel config
supported" query, but held back on doing that in this cl.

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
File src/effects/SkBlurMaskFilter.cpp (right):

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
src/effects/SkBlurMaskFilter.cpp:929: unsigned int texSide = smallRectSide +
2*blurRadius;
On 2015/01/30 14:47:21, robertphillips wrote:
> I guess we could move the creation of 'texDesc' down closer to its use in
> createTexture.

Done.

https://codereview.chromium.org/882223003/diff/120001/src/effects/SkBlurMaskF...
src/effects/SkBlurMaskFilter.cpp:956: 
On 2015/01/30 14:47:21, robertphillips wrote:
> 2 overlength lines

Done.

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp
File src/gpu/SkGr.cpp (right):

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:114: if (origKey.isValid() && kNo_Stretch != stretch) {
On 2015/01/30 14:47:21, robertphillips wrote:
> Why not use the domain from the original key?

That could cause a key collision.

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:180: 
On 2015/01/30 14:47:21, robertphillips wrote:
> or two -> of two ?

Done.

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:182: // width or height. If optionalKey is valid it will be set
on the new texture. stretch
On 2015/01/30 14:47:21, robertphillips wrote:
> neartest -> nearest ?

Done.

https://codereview.chromium.org/882223003/diff/120001/src/gpu/SkGr.cpp#newcod...
src/gpu/SkGr.cpp:190: 
On 2015/01/30 14:47:21, robertphillips wrote:
> it's

Done.

[robertphillips, 2015-01-30 20:16:19]

lgtm

[bsalomon, 2015-01-30 20:36:40]

The CQ bit was checked by bsalomon@google.com

[commit-bot: I haz the power, 2015-01-30 20:37:36]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/882223003/160001

[commit-bot: I haz the power, 2015-01-30 20:43:48]

Committed patchset #9 (id:160001) as
https://skia.googlesource.com/skia/+/8a8100349105c8c6de39fcb34e47679da7a67f54

[fmalita_google_do_not_use, 2015-01-31 14:59:12]

On 2015/01/30 20:43:48, I haz the power (commit-bot) wrote:
> Committed patchset #9 (id:160001) as
> https://skia.googlesource.com/skia/+/8a8100349105c8c6de39fcb34e47679da7a67f54

This is causing webgl_conformance failures, blocking the roll:

https://codereview.chromium.org/892773003/
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...

failures:
WebglConformance.conformance_textures_tex_image_and_sub_image_2d_with_video
WebglConformance.conformance_textures_texture_npot_video

[f(malita), 2015-01-31 15:01:37]

A revert of this CL (patchset #9 id:160001) has been created in
https://codereview.chromium.org/868233005/ by fmalita@chromium.org.

The reason for reverting is: webGL conformance failures:

WebglConformance.conformance_textures_tex_image_and_sub_image_2d_with_video
WebglConformance.conformance_textures_texture_npot_video

https://codereview.chromium.org/892773003/
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...
.

[mtklein, 2015-01-31 16:21:09]

On 2015/01/31 15:01:37, f(malita) wrote:
> A revert of this CL (patchset #9 id:160001) has been created in
> https://codereview.chromium.org/868233005/ by mailto:fmalita@chromium.org.
> 
> The reason for reverting is: webGL conformance failures:
> 
> WebglConformance.conformance_textures_tex_image_and_sub_image_2d_with_video
> WebglConformance.conformance_textures_texture_npot_video
> 
> https://codereview.chromium.org/892773003/
>
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...
> .

This was also a mega perf regression: https://perf.skia.org/#2031

[mtklein, 2015-01-31 16:22:01]

On 2015/01/31 16:21:09, mtklein wrote:
> On 2015/01/31 15:01:37, f(malita) wrote:
> > A revert of this CL (patchset #9 id:160001) has been created in
> > https://codereview.chromium.org/868233005/ by mailto:fmalita@chromium.org.
> > 
> > The reason for reverting is: webGL conformance failures:
> > 
> > WebglConformance.conformance_textures_tex_image_and_sub_image_2d_with_video
> > WebglConformance.conformance_textures_texture_npot_video
> > 
> > https://codereview.chromium.org/892773003/
> >
>
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...
> > .
> 
> This was also a mega perf regression: https://perf.skia.org/#2031

Even crazier: https://perf.skia.org/#2033

[bsalomon, 2015-02-01 03:20:53]

Patchset #10 (id:180001) has been deleted

[bsalomon, 2015-02-01 03:21:23]

On 2015/01/31 16:22:01, mtklein wrote:
> On 2015/01/31 16:21:09, mtklein wrote:
> > On 2015/01/31 15:01:37, f(malita) wrote:
> > > A revert of this CL (patchset #9 id:160001) has been created in
> > > https://codereview.chromium.org/868233005/ by mailto:fmalita@chromium.org.
> > > 
> > > The reason for reverting is: webGL conformance failures:
> > > 
> > >
WebglConformance.conformance_textures_tex_image_and_sub_image_2d_with_video
> > > WebglConformance.conformance_textures_texture_npot_video
> > > 
> > > https://codereview.chromium.org/892773003/
> > >
> >
>
http://build.chromium.org/p/tryserver.chromium.mac/builders/mac_chromium_rel_...
> > > .
> > 
> > This was also a mega perf regression: https://perf.skia.org/#2031
> 
> Even crazier: https://perf.skia.org/#2033

I accidentally put the lines that inserted bitmap textures into the cache inside
an SkASSERT() instead of SkAssertResult()! Argh!

I'm going to reland, hoping the cr. test failures don't occur with the fix.

[bsalomon, 2015-02-01 03:21:28]

The CQ bit was checked by bsalomon@google.com

[commit-bot: I haz the power, 2015-02-01 03:21:39]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/882223003/200001

[commit-bot: I haz the power, 2015-02-01 03:27:59]

Committed patchset #10 (id:200001) as
https://skia.googlesource.com/skia/+/6c96672491b04cb782bce8fee778124df66524a0

[bsalomon, 2015-02-01 04:09:53]

A revert of this CL (patchset #10 id:200001) has been created in
https://codereview.chromium.org/887303002/ by bsalomon@google.com.

The reason for reverting is: perf fix didn't fix the cr webgl conformance tests.

[bsalomon, 2015-02-02 20:48:38]

The CQ bit was checked by bsalomon@google.com

[commit-bot: I haz the power, 2015-02-02 20:48:52]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/882223003/260001

[commit-bot: I haz the power, 2015-02-02 21:00:14]

Committed patchset #13 (id:260001) as
https://skia.googlesource.com/skia/+/37f9a2694c15f08e361ebda74fe9f0fffbf452aa