vaapidecode GPU to CPU download performance

[Arun Raghavan]

On 27 November 2014 at 09:50, Arun Raghavan <arun at accosted.net> wrote:
[...]
>> Are you:
>> 1. Using vaGetImage() + map the resulting pixels + direct read ; or
>> 2. Using vaDeriveImage() + map buffer + use Uncacheable Speculative
>> Write Combining (USWC) memory copy?
>
> I just filed a bug about what looks like the same issue:
> https://bugzilla.gnome.org/show_bug.cgi?id=740774
>
> The summary is filesrc ! demux ! parse ! vaapidecode ! xvimagesink is
> incredibly slow -- is that expected? I'm on a Ivybridge-based desktop.

Brainfart, it's Haswell (i5 4570, HD 4600).

-- Arun