[xine-user] Re: [directfb-users] PERFORMANCE: DirectFB vs XvMC for VIA C3 Unichrome

[Lucian M. <lu...@us...>]

Denis Oliver Kropp wrote:
> Quoting JP Dong:
> 
>>   Dear all,
>>
>>   I evaluated the performance for DirectFB and XvMC driver for VIA C3
>>   Unichrome display. The performance from the former was not as good as the
>>   later. On a 1GHz machine, the CPU usage from the former was almost about
>>   70-80% for a clip like mission impossible II; however, it was about 30-40%
>>   for the same clip if the later driver was used.
> 
> 
> XvMC uses hardware motion compensation and therefor has less CPU usage.
> 
> 
>>   I am not sure whether this was roughly right, since in theory (??) frame
>>   buffer has less overhead from the X services. Does this mean that the
>>   implementation of the frame buffer driver is not as good? Are there any
>>   other comparison data indicating otherwise? If so, what can I do to
>>   improve the performance of the DirectFB?
> 
> 
> I have the integration of VIA's HW MPEG support for DirectFB in my pipeline.

Coooool! Does this mean there is hope for HW MPEG decoding support in 
DirectFB *without* X?? Do you also have a rough estimation, when?

That would finally make some VIA motherboard worth to be considered for 
a HTPC, such a pity they only do mini-ITX. But if I understood well, 
some other VIA chipsest for P4 or AMD64 also support this HW decoding? 
I'm asking because I'm not satisfied with a mini-ITX which has at best 
only 2 PCI slots, an AMD64 can also be made quiet enough for a HTPC 
these days, but still able to receive more than 1 or 2 DVB-cards for 
example...

Lucian