Begin moving locking out of GrResourceCache

This CL starts the process of moving locking out of the Resource Cache. It isn't complete but this CL alone should be plenty to consider. It also removes GrResourceCache::attach (which was no longer used).

It originally also removed GrResourceCache::LockType (which was only used for stencil buffer and isn't compatible with where we are going) but naively removing it caused huge differences in the cache usage.

[bsalomon, 2012-08-30 19:09:57]

LGTM

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp
File src/gpu/GrResourceCache.cpp (right):

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:207: this->attachToHead(entry, false);
Just curious, was this reorder necessary or just for readability? (No objection,
just wondering if there is some fragility that isn't obvious.)

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:444: for ( ; NULL != entry; entry = iter.next()) {
I didn't think of this in when the list was first introduced but it is kinda
nice how the other skia iterator classes require an initial next call so you can
write

while (entry = iter->next()) {
 ...
}

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:450: EntryList::Iter::kHead_IterStart);
doesn't fit on one line?

[robertphillips, 2012-08-30 19:22:53]

committed as r5354

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp
File src/gpu/GrResourceCache.cpp (right):

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:207: this->attachToHead(entry, false);
findAndLock is going to go away - replaced with a call to find followed
(temporarily) by a call to lock in GrContext::findAndLockTexture. This
re-ordering was to move the lock to a better place from which to move it out.

Note that this CL also makes it so the only place GrResourceEntry::lock is
called is from within GrResourceCache::lock.

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:444: for ( ; NULL != entry; entry = iter.next()) {
That's a bit problematic in this case since the DLinkedList allows both forward
& reverse iteration. For 'next' we wouldn't know if we had hit the end of the
list or if it was the first call & we should snap to the start.

http://codereview.appspot.com/6498064/diff/4001/src/gpu/GrResourceCache.cpp#n...
src/gpu/GrResourceCache.cpp:450: EntryList::Iter::kHead_IterStart);
On 2012/08/30 19:09:57, bsalomon wrote:
> doesn't fit on one line?

Done.