Use STDArray to for tracked resources in vulkan command buffer

Use STDArray to for tracked resources in vulkan command buffer

This is a minor speed improvement to help with the number of times we
are freeing and re-allocating the tracked resource arrays

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2338963004

Committed: https://skia.googlesource.com/skia/+/594739c7924f46e1632d2f97a9b0822082b1169d

[egdaniel, 2016-09-14 17:24:55]

Description was changed from

==========
Use STDArray to for tracked resources in vulkan command buffer

This is a minor speed improvement to help with the number of times we
are freeing and re-allocating the tracked resource arrays

BUG=skia:
==========

to

==========
Use STDArray to for tracked resources in vulkan command buffer

This is a minor speed improvement to help with the number of times we
are freeing and re-allocating the tracked resource arrays

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2338963004
==========

[egdaniel, 2016-09-14 17:59:29]

egdaniel@google.com changed reviewers:
+ brianosman@google.com, jvanverth@google.com

[egdaniel, 2016-09-14 18:02:55]

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
File src/gpu/vk/GrVkCommandBuffer.h (right):

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
src/gpu/vk/GrVkCommandBuffer.h:175: static const int kMaxNumberofRewindResets =
8;
also if anyone has any better guesses of what this should be I'm all ears.

[egdaniel, 2016-09-19 17:20:57]

On 2016/09/14 18:02:55, egdaniel wrote:
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> File src/gpu/vk/GrVkCommandBuffer.h (right):
> 
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> src/gpu/vk/GrVkCommandBuffer.h:175: static const int kMaxNumberofRewindResets
=
> 8;
> also if anyone has any better guesses of what this should be I'm all ears.

ping

[Brian Osman, 2016-09-19 18:51:04]

Hmm. If this is/was a performance problem, doesn't this
strategy just make it less frequent? I'd be worried that
now it's going to be harder to identify, and introduce
more noise into profiling.

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
File src/gpu/vk/GrVkCommandBuffer.h (right):

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
src/gpu/vk/GrVkCommandBuffer.h:175: static const int kMaxNumberofRewindResets =
8;
On 2016/09/14 18:02:55, egdaniel wrote:
> also if anyone has any better guesses of what this should be I'm all ears.

kNumRewindResetsBeforeFullReset ?

[jvanverth1, 2016-09-19 18:51:04]

This is outside of this change, but one thing I've been wondering is how many
duplicates we have in the resource lists. I.e., how often do we call addResource
for the same resource in a given command buffer execution. If it happens a lot,
then we are doing multiple unrefs for a single resource -- in which case we
might want to experiment with a hashtable.

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
File src/gpu/vk/GrVkCommandBuffer.cpp (right):

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
src/gpu/vk/GrVkCommandBuffer.cpp:77:
fTrackedRecycledResources.setReserve(kInitialTrackedResourcesCount);
fNumResets = 0; ?

[egdaniel, 2016-09-19 19:52:37]

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
File src/gpu/vk/GrVkCommandBuffer.cpp (right):

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
src/gpu/vk/GrVkCommandBuffer.cpp:77:
fTrackedRecycledResources.setReserve(kInitialTrackedResourcesCount);
On 2016/09/19 18:51:04, jvanverth1 wrote:
> fNumResets = 0; ?

Done.

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
File src/gpu/vk/GrVkCommandBuffer.h (right):

https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
src/gpu/vk/GrVkCommandBuffer.h:175: static const int kMaxNumberofRewindResets =
8;
On 2016/09/19 18:51:04, Brian Osman wrote:
> On 2016/09/14 18:02:55, egdaniel wrote:
> > also if anyone has any better guesses of what this should be I'm all ears.
> 
> kNumRewindResetsBeforeFullReset ?

Done.

[egdaniel, 2016-09-20 14:08:26]

On 2016/09/19 19:52:37, egdaniel wrote:
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> File src/gpu/vk/GrVkCommandBuffer.cpp (right):
> 
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> src/gpu/vk/GrVkCommandBuffer.cpp:77:
> fTrackedRecycledResources.setReserve(kInitialTrackedResourcesCount);
> On 2016/09/19 18:51:04, jvanverth1 wrote:
> > fNumResets = 0; ?
> 
> Done.
> 
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> File src/gpu/vk/GrVkCommandBuffer.h (right):
> 
>
https://codereview.chromium.org/2338963004/diff/20001/src/gpu/vk/GrVkCommandB...
> src/gpu/vk/GrVkCommandBuffer.h:175: static const int kMaxNumberofRewindResets
=
> 8;
> On 2016/09/19 18:51:04, Brian Osman wrote:
> > On 2016/09/14 18:02:55, egdaniel wrote:
> > > also if anyone has any better guesses of what this should be I'm all ears.
> > 
> > kNumRewindResetsBeforeFullReset ?
> 
> Done.

thoughts?

[jvanverth1, 2016-09-20 15:42:35]

lgtm

[egdaniel, 2016-09-20 17:05:13]

The CQ bit was checked by egdaniel@google.com

[commit-bot: I haz the power, 2016-09-20 17:05:21]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2016-09-20 19:39:27]

Description was changed from

==========
Use STDArray to for tracked resources in vulkan command buffer

This is a minor speed improvement to help with the number of times we
are freeing and re-allocating the tracked resource arrays

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2338963004
==========

to

==========
Use STDArray to for tracked resources in vulkan command buffer

This is a minor speed improvement to help with the number of times we
are freeing and re-allocating the tracked resource arrays

BUG=skia:
GOLD_TRYBOT_URL= https://gold.skia.org/search?issue=2338963004

Committed:
https://skia.googlesource.com/skia/+/594739c7924f46e1632d2f97a9b0822082b1169d
==========

[commit-bot: I haz the power, 2016-09-20 19:39:27]

Committed patchset #3 (id:40001) as
https://skia.googlesource.com/skia/+/594739c7924f46e1632d2f97a9b0822082b1169d