Commit Graph

7 Commits

Author SHA1 Message Date
fmalita
5776508126 Add a SkRWBuffer reserve mechanism
Currently, Chromium stores segmented data in a SharedBuffer and appends
to SkRWBuffer one segment at a time:

        const char* segment = 0;
        for (size_t length = data->getSomeData(segment, m_rwBuffer->size());
            length; length = data->getSomeData(segment, m_rwBuffer->size())) {
            m_rwBuffer->append(segment, length, remaining);
        }

This can yield a bunch of just-above-4k allocations => wasted RAM due to
internal fragmentation.

Ideally, we'd want a SkRWBuffer::reserve(size_t bytes) API, but the
current internals don't support that trivially.

Alternatively, the caller can pass a reserve hint at append() time.

BUG=chromium:651698
R=scroggo@google.com,reed@google.com
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2385803002

Review-Url: https://codereview.chromium.org/2385803002
2016-09-30 13:34:19 -07:00
scroggo
6351640285 Fixes for SkRWBuffer
Do not call SkBufferHead::validate in SkROBuffer's destructor, which
may be called in a separate thread from SkRWBuffer::append. validate()
reads SkBufferBlock::fUsed, and append() writes to it, resulting in
a data race.

Update some comments to be more clear about how it is safe to use
these classes across threads.

Test the readers in separate threads.

In addition, make sure it is safe to create a reader even when no
data has been appended. Add tests for this case.

Mark a parameter to SkBufferHead::validate() as const, reflecting
its use.

BUG=chromium:601578
BUG=chromium:605479

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1871953002

Review URL: https://codereview.chromium.org/1871953002
2016-04-22 06:59:01 -07:00
bungeman
a3760992c9 Revert of Fixes for SkRWBuffer (patchset #5 id:80001 of https://codereview.chromium.org/1871953002/ )
Reason for revert:
Making MSAN and TSAN rather unhappy.

https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Debug-MSAN/builds/1586

https://build.chromium.org/p/client.skia/builders/Test-Ubuntu-GCC-GCE-CPU-AVX2-x86_64-Release-TSAN/builds/5922

Original issue's description:
> Fixes for SkRWBuffer
>
> Do not call SkBufferHead::validate in SkROBuffer's destructor, which
> may be called in a separate thread from SkRWBuffer::append. validate()
> reads SkBufferBlock::fUsed, and append() writes to it, resulting in
> a data race.
>
> Update some comments to be more clear about how it is safe to use
> these classes across threads.
>
> Test the readers in separate threads.
>
> In addition, make sure it is safe to create a reader even when no
> data has been appended. Add tests for this case.
>
> Mark a parameter to SkBufferHead::validate() as const, reflecting
> its use.
>
> BUG=chromium:601578
> GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1871953002
>
> Committed: https://skia.googlesource.com/skia/+/d06920a29fe11c68bde2b93948ec99f277bb8459

TBR=mtklein@google.com,reed@google.com,scroggo@google.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:601578

Review URL: https://codereview.chromium.org/1882853004
2016-04-14 14:57:01 -07:00
scroggo
d06920a29f Fixes for SkRWBuffer
Do not call SkBufferHead::validate in SkROBuffer's destructor, which
may be called in a separate thread from SkRWBuffer::append. validate()
reads SkBufferBlock::fUsed, and append() writes to it, resulting in
a data race.

Update some comments to be more clear about how it is safe to use
these classes across threads.

Test the readers in separate threads.

In addition, make sure it is safe to create a reader even when no
data has been appended. Add tests for this case.

Mark a parameter to SkBufferHead::validate() as const, reflecting
its use.

BUG=chromium:601578
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1871953002

Review URL: https://codereview.chromium.org/1871953002
2016-04-14 11:40:48 -07:00
reed
377add7426 Fix race condition in SkROBuffer.
SkBufferBlock::fUsed may be updated by the writer while a reader is
attempting to read it.

BUG=chromium:601578
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1872853002

Review URL: https://codereview.chromium.org/1872853002
2016-04-08 12:47:14 -07:00
scroggo
af0ba653bf Add SK_API to SkROBuffer::Iter
Necessary for crrev.com/1812273003 in Chromium.

Also remove trailing whitespace.

GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1856033002

Review URL: https://codereview.chromium.org/1856033002
2016-04-04 10:05:23 -07:00
scroggo
77899d1220 Move SkRWBuffer.h to include/ and add SK_API
Needed for use in Blink in crrev.com/1812273003
GOLD_TRYBOT_URL= https://gold.skia.org/search2?unt=true&query=source_type%3Dgm&master=false&issue=1838463003

Review URL: https://codereview.chromium.org/1838463003
2016-03-25 12:00:15 -07:00