Hello, On Wednesday, June 29, 2011 1:02 PM Hans Verkuil wrote: > On Wednesday, June 29, 2011 12:44:48 Laurent Pinchart wrote: > > Hi Marek, > > > > On Wednesday 29 June 2011 11:49:06 Marek Szyprowski wrote: > > > This patch introduces VB2_STREAMON_WITHOUT_BUFFERS io flag and changes > > > the order of operations during stream on operation. Now the buffer are > > > first queued to the driver and then the start_streaming method is > called. > > > This resolves the most common case when the driver needs to know buffer > > > addresses to enable dma engine and start streaming. For drivers that > can > > > handle start_streaming without queued buffers (mem2mem and 'one shot' > > > capture case) a new VB2_STREAMON_WITHOUT_BUFFERS io flag has been > > > introduced. Driver can set it to let videobuf2 know that it support > this > > > mode. > > > > Is starting/stopping DMA engines that expensive on most hardware ? > Several > > mails mentioned that drivers should keep one buffer around to avoid > stopping > > the DMA engine in case of buffer underrun. The OMAP3 ISP driver just > stops > the > > ISP when it runs out of buffers, and restart it when a new buffer is > queued. > > Yes, this can be expensive. For video capture (e.g. from HDMI) you never > want > to stop capturing when you run out of buffers. Starting it up again will > lead > to a 1 or 2 frame delay, which is unacceptable for e.g. video conferencing. > > And when I start the DMA engine I'd like to know whether only one buffer is > queued or if I have two or more. In the latter case I can setup both the > 'current' and 'next' pointers in the DMA engine which will make the first > frame available quicker (otherwise you will probably get an additional > frame > delay). > > > Switching the order of the start_streaming and __enqueue_in_driver calls > would > > make my life more difficult on the OMAP3 because I will have to check if > the > > queue is streaming in the qbuf callback. Your s5p-fimc driver has to > check > for > > that as well. I wonder if it really helps for other drivers. > > Why not add a 'is_streaming' boolean argument to enqueue_in_driver? Again, thanks for the idea! Best regards -- Marek Szyprowski Samsung Poland R&D Center -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
