Replace use of GrTHashTable in GrFontCache with SkTDynamicHash.

Replace use of GrTHashTable in GrFontCache with SkTDynamicHash.

Searching the font cache for existing text strikes was showing up
as a hotspot on Android. This change reduces that cost.

Committed: https://skia.googlesource.com/skia/+/dd6d22751a1ec858daaf78c3929c22737483e3db

[jvanverth1, 2014-07-17 14:57:12]

(Except it's measurably slower in Chrome, so clearly either SkTDynamicHash has a
problem, or I'm doing something dumb. Probably the latter.)

[jvanverth1, 2014-07-21 14:50:14]

Updated to include the hashtable of GrGlyphs in GrTextStrike, and use Murmur3 as
the hash function. This may be a shade faster, but on the whole it appears to be
about as fast (or slow) as the old code. PTAL.

[bsalomon, 2014-07-21 14:55:01]

lgtm. Too bad there isn't a speedup, but at least it is a step towards removing
GrTHashTable.

[jvanverth1, 2014-07-21 14:56:04]

On 2014/07/21 14:50:14, jvanverth1 wrote:
> Updated to include the hashtable of GrGlyphs in GrTextStrike, and use Murmur3
as
> the hash function. This may be a shade faster, but on the whole it appears to
be
> about as fast (or slow) as the old code. PTAL.

Oh, one other thing -- I used Murmur3 to hash the GrGlyph::PackedID because it
was better than using the raw ID. But it feels like overkill in this case, as on
a Western webpage most of the IDs will lie within the range of printable ASCII
characters. But I couldn't find a better yet simple hash.

[jvanverth1, 2014-07-22 13:14:04]

Adding Mike to make sure the addition of deleteAll() and freeAll() is good
(needed to avoid a memory leak). PTAL.

[mtklein, 2014-07-22 13:54:30]

https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash.h
File src/core/SkTDynamicHash.h (right):

https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash...
src/core/SkTDynamicHash.h:138: void deleteAll() {
I think you can do both of these just as easily from the outside with Iter.  I'd
be totally fine with these as helper standalone functions in this file that
operate on a SkTDynamicHash.  Just seems we don't need them as methods.

[jvanverth1, 2014-07-22 14:14:08]

https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash.h
File src/core/SkTDynamicHash.h (right):

https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash...
src/core/SkTDynamicHash.h:138: void deleteAll() {
Yup, just had that discussion here. Done.

[mtklein, 2014-07-22 14:16:35]

On 2014/07/22 14:14:08, jvanverth1 wrote:
>
https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash.h
> File src/core/SkTDynamicHash.h (right):
> 
>
https://codereview.chromium.org/390103002/diff/100001/src/core/SkTDynamicHash...
> src/core/SkTDynamicHash.h:138: void deleteAll() {
> Yup, just had that discussion here. Done.

LGTM.

Pull me over some time when I'm back to look at profiles?

[jvanverth1, 2014-07-22 20:07:16]

The CQ bit was checked by jvanverth@google.com

[commit-bot: I haz the power, 2014-07-22 20:08:24]

CQ is trying da patch. Follow status at
 https://skia-tree-status.appspot.com/cq/jvanverth@google.com/390103002/120001

[commit-bot: I haz the power, 2014-07-22 20:25:30]

Change committed as dd6d22751a1ec858daaf78c3929c22737483e3db