[Libva] memcpy VA image buffer, very poor performance!

[hank peng]

Do you think "dri_bo_get_subdata" which call DRM_IOCTL_I915_GEM_PREAD
ioctl can give me fast copy speed?

2011/7/9 Bian, Jonathan <jonathan.bian at intel.com>:
> Did you get the VAImage from vaDeriveImage()? Typically this points to uncached memory and copying from it would be slow.
>
> Regards,
> Jonathan
>
> -----Original Message-----
> From: libva-bounces+jonathan.bian=intel.com at lists.freedesktop.org [mailto:libva-bounces+jonathan.bian=intel.com at lists.freedesktop.org] On Behalf Of hank peng
> Sent: Wednesday, July 06, 2011 12:05 AM
> To: libva at lists.freedesktop.org
> Subject: [Libva] memcpy VA image buffer, very poor performance!
>
> Hi:
> I use vaMapBuffer function to map VA image buffer, then copy data from it to my allocated memory using malloc function. I found it's performace is very poor, then, I do a test program on same platform between two allocated memory whose size is same as VA image buffer, the performance is much better, Is there any difference between the VA image mapped buffer and normal buffer allocated by malloc function?
>
> --
> The simplest is not all best but the best is surely the simplest!
> _______________________________________________
> Libva mailing list
> Libva at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/libva
>



-- 
The simplest is not all best but the best is surely the simplest!