Yurij Sysoev wrote:
Hi,
Iñaki wrote:
I'm a student of the Informatic University of Madrid. I'm
delivering a processing motion pictures software in Linux. I need
low-latency field-rate capture and i've read in the video4linux list
all messages about this.
I only want to ask if finally it's possible to capture fields
instead of frames (obtaining a half time period to process) in
video4linux. I've studied video4linux files (videodev), system files
(ioctl.h, raw.c) and drivers files (bttv driver). The most similar
thing i've seen in v4l is the use of VBI device because v4l video
seems work only with frames.
For me your question isn't clear - why do you think, that capturing of
frames will be more CPU-costly unlike capturing of fields?
No no. I haven't explain right. I know what's DMA. The average time
between frames is 40 us and i need reduce it to 20 us because i need
precision. I capture images of things which are in motion. Using DMA, is
the device which interrupts CPU when has finished his work (and CPU only
have to spend time to take the image of the buffer passed as interchange
place against programer transmision where is CPU which has to transfer
data between device and memory).
I need start to process images (despite images are fields) each 20
us. I want interruptions are each 20 us.
Actually, the capturing process is done by DMA-engine, without notable
CPU's eating (i hope, your code is enough optimised by compiler and
fitted to the CPU cache :-). The interval between fields also the same
as between frames - and you will have enough time for processing just
half lines from captured frames.
Regards,
Yurij
PS: the English isn't my native language, sorry :-)
PS: isn't it mine ;-)
--
video4linux-list mailing list
Unsubscribe
mailto:video4linux-list-request@xxxxxxxxxx?subject=unsubscribe
https://listman.redhat.com/mailman/listinfo/video4linux-list