a few more comments ...

[Mohammed Abdul Azeem]

Hi Michael,

On Tue, Jun 6, 2017 at 12:10 AM, Michael Meeks <michael.meeks at collabora.com>
wrote:

> Hi Mohammed,
>
>         Now I read it again, I'm rather skeptical that:
>
> +        if( !maUsedBuffers.empty() )
> +        {
> +            pProducedBuffer = maUsedBuffers.front();
> +            maUsedBuffers.pop();
> +        }
>
>         is helpful. I'm not sure that we can re-use these buffers
> efficiently
> here - perhaps its better just to allocate new ones in the stream read
> method; can you profile with and without that (would love to get rid of
> the extra complexity if possible).
>
> Here you go.

with UsedBuffers:
https://demo.collaboracloudsuite.com/tdf/index.php/s/Hbsp5rI8FE7CHmF
without UsedBuffers:
https://demo.collaboracloudsuite.com/tdf/index.php/s/PIcWD261zVcwpWj

Regards,
Azeem
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice/attachments/20170606/58618ce1/attachment.html>