Get rid of fastMallocGoodSize() and replace it with something generic.

Get rid of fastMallocGoodSize() and replace it with something generic.

- fastMallocGoodSize() was Mac-specific and it looks relatively slow and
non-inlineable. The replacement is based around a single lookup into a small
table.
- We now have allocation quantization on all platforms for Vector. I've
verified that many Vector instantiations are able to recover previously-wasted
memory as real element capacity. See bug for details.
- The quantization is also immediately suitable for use with the Partition
allocator, to avoid fragmentation due to excessive numbers of buckets. To be
followed up in a future patch series.


BUG=286477
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=157524

Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=157555

[abarth-chromium, 2013-09-07 05:08:23]

It would be good to explain where these rounding constants come from.  Is there
some ASSERT we can make to ensure they don't get out of date?

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h
File Source/wtf/Vector.h (right):

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h#newcod...
Source/wtf/Vector.h:265: RELEASE_ASSERT(newCapacity <=
WTF::QuantizedAllocation::kMaxUnquantizedAllocation / sizeof(T));
WTF::  <-- no need for WTF prefix inside WTF.

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h#newcod...
Source/wtf/Vector.h:266: size_t origSizeToAllocate = newCapacity * sizeof(T);
origSizeToAllocate ---> originalSizeToAllocate

Please use complete words in variable names.

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h
File Source/wtf/WTF.h (right):

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h#newcode56
Source/wtf/WTF.h:56: class QuantizedAllocation {
Why is this in WTF.h?  Please put it in its own header.

https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h#newcode62
Source/wtf/WTF.h:62: static const size_t kTableSize = kMaxAllocation /
kMinRoundingLimit;
Can you add some comments explaining where these numbers come from?

[Chris Evans, 2013-09-09 22:09:58]

PTAL!

On 2013/09/07 05:08:23, abarth wrote:
> It would be good to explain where these rounding constants come from.  Is
there
> some ASSERT we can make to ensure they don't get out of date?

I've added comments (to the new QuantizedAllocation.h) that explains the
constants and overall motivations a little. I can't think of a useful ASSERT.
It's more of a best-effort that just happens to work on top of any
sanely-implemented bucketing allocator if that makes sense? FWIW, I just
blind-validated the current strategy against the Windows LFH, and we're in good
shape.

> 
> https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h
> File Source/wtf/Vector.h (right):
> 
>
https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h#newcod...
> Source/wtf/Vector.h:265: RELEASE_ASSERT(newCapacity <=
> WTF::QuantizedAllocation::kMaxUnquantizedAllocation / sizeof(T));
> WTF::  <-- no need for WTF prefix inside WTF.

Done.

> 
>
https://codereview.chromium.org/23604048/diff/4001/Source/wtf/Vector.h#newcod...
> Source/wtf/Vector.h:266: size_t origSizeToAllocate = newCapacity * sizeof(T);
> origSizeToAllocate ---> originalSizeToAllocate
> 
> Please use complete words in variable names.

Done.

> 
> https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h
> File Source/wtf/WTF.h (right):
> 
> https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h#newcode56
> Source/wtf/WTF.h:56: class QuantizedAllocation {
> Why is this in WTF.h?  Please put it in its own header.

Yeah, fair call. Welcome to QuantizedAllocator.{h,cpp}

> 
> https://codereview.chromium.org/23604048/diff/4001/Source/wtf/WTF.h#newcode62
> Source/wtf/WTF.h:62: static const size_t kTableSize = kMaxAllocation /
> kMinRoundingLimit;
> Can you add some comments explaining where these numbers come from?

Done.

[abarth-chromium, 2013-09-10 16:11:33]

lgtm

[commit-bot: I haz the power, 2013-09-10 16:12:52]

Change committed as 157524

[Chris Evans, 2013-09-10 16:25:50]

On 2013/09/10 16:12:52, I haz the power (commit-bot) wrote:
> Change committed as 157524

CQ turbo'ed the change inappropriately. Sadness. Reverted at r157525.

[abarth-chromium, 2013-09-10 16:28:48]

boo

[commit-bot: I haz the power, 2013-09-10 21:30:43]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/cevans@chromium.org/23604048/24001

[commit-bot: I haz the power, 2013-09-10 22:51:13]

Change committed as 157555