HW draw workarounds

[no sievers, 2013-05-16 00:09:55]

https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):

https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
gpu/command_buffer/service/gles2_cmd_decoder.cc:2845:
RestoreFramebufferBindings();
This should only be set on Vivante gpus. Why did you need this? I'm just asking
in case the workaround introduced a bug (and is not limited as it should be
somehow) :)

[joth, 2013-05-16 00:20:44]

https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):

https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
gpu/command_buffer/service/gles2_cmd_decoder.cc:2845:
RestoreFramebufferBindings();
On 2013/05/16 00:09:55, Daniel Sievers wrote:
> This should only be set on Vivante gpus. Why did you need this? I'm just
asking
> in case the workaround introduced a bug (and is not limited as it should be
> somehow) :)

Seems it's being set on Imagination gpus too?

gpu/command_buffer/service/feature_info.cc
    if (is_vivante || is_imagination) {
      workarounds_.unbind_fbo_on_context_switch = true;
    }

*IIRC* without this commented out, you'd get one frame but then would just
render blank (even with virtual context)

[no sievers, 2013-05-16 00:28:20]

On 2013/05/16 00:20:44, joth wrote:
>
https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
> File gpu/command_buffer/service/gles2_cmd_decoder.cc (right):
> 
>
https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/servic...
> gpu/command_buffer/service/gles2_cmd_decoder.cc:2845:
> RestoreFramebufferBindings();
> On 2013/05/16 00:09:55, Daniel Sievers wrote:
> > This should only be set on Vivante gpus. Why did you need this? I'm just
> asking
> > in case the workaround introduced a bug (and is not limited as it should be
> > somehow) :)
> 
> Seems it's being set on Imagination gpus too?
> 
> gpu/command_buffer/service/feature_info.cc
>     if (is_vivante || is_imagination) {
>       workarounds_.unbind_fbo_on_context_switch = true;
>     }
> 
> *IIRC* without this commented out, you'd get one frame but then would just
> render blank (even with virtual context)

Weird, I can't find that in the history of feature_info.cc but am probably doing
something wrong.

[boliu, 2013-05-16 00:32:55]

Eric added the workaround for imagination GPUs
On May 15, 2013 5:28 PM, <sievers@chromium.org> wrote:

> On 2013/05/16 00:20:44, joth wrote:
>
> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>
command_buffer/service/gles2_**cmd_decoder.cc<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc>
>
>> File gpu/command_buffer/service/**gles2_cmd_decoder.cc (right):
>>
>
>
> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>
command_buffer/service/gles2_**cmd_decoder.cc#newcode2845<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc#newcode2845>
>
>> gpu/command_buffer/service/**gles2_cmd_decoder.cc:2845:
>> RestoreFramebufferBindings();
>> On 2013/05/16 00:09:55, Daniel Sievers wrote:
>> > This should only be set on Vivante gpus. Why did you need this? I'm just
>> asking
>> > in case the workaround introduced a bug (and is not limited as it
>> should be
>> > somehow) :)
>>
>
>  Seems it's being set on Imagination gpus too?
>>
>
>  gpu/command_buffer/service/**feature_info.cc
>>      if (is_vivante || is_imagination) {
>>        workarounds_.unbind_fbo_on_**context_switch = true;
>>      }
>>
>
>  *IIRC* without this commented out, you'd get one frame but then would just
>> render blank (even with virtual context)
>>
>
> Weird, I can't find that in the history of feature_info.cc but am probably
> doing
> something wrong.
>
>
https://codereview.chromium.**org/14600002/<https://codereview.chromium.org/1...
>

[joth, 2013-05-16 00:33:00]

https://chromiumcodereview.appspot.com/14069008/diff/61001/gpu/command_buffer...
- landed Monday, so I don't think this was what Bo was seeing :-/



On 15 May 2013 17:28, <sievers@chromium.org> wrote:

> On 2013/05/16 00:20:44, joth wrote:
>
> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>
command_buffer/service/gles2_**cmd_decoder.cc<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc>
>
>> File gpu/command_buffer/service/**gles2_cmd_decoder.cc (right):
>>
>
>
> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>
command_buffer/service/gles2_**cmd_decoder.cc#newcode2845<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc#newcode2845>
>
>> gpu/command_buffer/service/**gles2_cmd_decoder.cc:2845:
>> RestoreFramebufferBindings();
>> On 2013/05/16 00:09:55, Daniel Sievers wrote:
>> > This should only be set on Vivante gpus. Why did you need this? I'm just
>> asking
>> > in case the workaround introduced a bug (and is not limited as it
>> should be
>> > somehow) :)
>>
>
>  Seems it's being set on Imagination gpus too?
>>
>
>  gpu/command_buffer/service/**feature_info.cc
>>      if (is_vivante || is_imagination) {
>>        workarounds_.unbind_fbo_on_**context_switch = true;
>>      }
>>
>
>  *IIRC* without this commented out, you'd get one frame but then would just
>> render blank (even with virtual context)
>>
>
> Weird, I can't find that in the history of feature_info.cc but am probably
> doing
> something wrong.
>
>
https://codereview.chromium.**org/14600002/<https://codereview.chromium.org/1...
>

[joth, 2013-05-16 00:37:47]

On 15 May 2013 17:32, Jonathan Dixon <joth@chromium.org> wrote:

>
>
https://chromiumcodereview.appspot.com/14069008/diff/61001/gpu/command_buffer...
> - landed Monday, so I don't think this was what Bo was seeing :-/
>
>
>
Opps correction - it landed over a week ago, Friday May 3rd. So yes sounds
like this was the reason :)



>
> On 15 May 2013 17:28, <sievers@chromium.org> wrote:
>
>> On 2013/05/16 00:20:44, joth wrote:
>>
>> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>>
command_buffer/service/gles2_**cmd_decoder.cc<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc>
>>
>>> File gpu/command_buffer/service/**gles2_cmd_decoder.cc (right):
>>>
>>
>>
>> https://codereview.chromium.**org/14600002/diff/13001/gpu/**
>>
command_buffer/service/gles2_**cmd_decoder.cc#newcode2845<https://codereview.chromium.org/14600002/diff/13001/gpu/command_buffer/service/gles2_cmd_decoder.cc#newcode2845>
>>
>>> gpu/command_buffer/service/**gles2_cmd_decoder.cc:2845:
>>> RestoreFramebufferBindings();
>>> On 2013/05/16 00:09:55, Daniel Sievers wrote:
>>> > This should only be set on Vivante gpus. Why did you need this? I'm
>>> just
>>> asking
>>> > in case the workaround introduced a bug (and is not limited as it
>>> should be
>>> > somehow) :)
>>>
>>
>>  Seems it's being set on Imagination gpus too?
>>>
>>
>>  gpu/command_buffer/service/**feature_info.cc
>>>      if (is_vivante || is_imagination) {
>>>        workarounds_.unbind_fbo_on_**context_switch = true;
>>>      }
>>>
>>
>>  *IIRC* without this commented out, you'd get one frame but then would
>>> just
>>> render blank (even with virtual context)
>>>
>>
>> Weird, I can't find that in the history of feature_info.cc but am
>> probably doing
>> something wrong.
>>
>>
https://codereview.chromium.**org/14600002/<https://codereview.chromium.org/1...
>>
>
>

[boliu, 2013-05-25 23:01:19]

Closing! Superceded by https://codereview.chromium.org/15860003/