hi: 2011/8/2 Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>: > On Mon, 1 Aug 2011, loody wrote: > >> hi: >> >> 2011/8/1 Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>: >> > On Mon, 1 Aug 2011, loody wrote: >> > >> >> sorry for picking this old thread, since the subject is the same. >> >> my question is: >> >> In high speed device and an urb->inerval is less than 8, take 2 or 4 >> >> for example, the number of iso transfer descriptors will be >> >> 1+(span+7)/8. >> >> >> >> But the formula above seems to be used for urb-> interval =1, right? >> >> Why the # of iso transfer descriptors not to be span/interval but 1+(span+7)/8? >> > >> > Because each iTD provides information for an entire frame, i.e., 8 >> > microframes. >> The each micro-frame of iso transfer descriptors will not be filled >> except urb->interval is 1, right? > > True. > >> if the urb->interval = 2, that mean the total number of iso transfer >> descriptors we need are (1+(span+7)/8)*2 not 1+(span+7)/8, since only >> half of iso transfer descriptors are filled with packets in iso sched. > > I don't follow your reasoning, but regardless, it is wrong. Let's > consider a simple example: urb->interval = 2 and > urb->number_of_packets = 12. Then span = 12 * 2 = 24 uframes. > > Suppose the URB's first packet is scheduled for uframe 5 of frame 61. > Then the twelve packets will be scheduled for: > > uframe 5 of frame 61 > uframe 7 of frame 61 > uframe 1 of frame 62 > ... > uframe 1 of frame 64 > uframe 3 of frame 64 > > There are 4 frames involved here: 61 through 64. Thus 4 iTDs are > required, since each iTD handles a single frame. This agrees with the > formula in the driver: 1+(24+7)/8 = 4. But your formula would give > (1+(24+7)/8)*2 = 8, which is incorrect. > > Alan Stern I found what my problem is. I confuse the meaning of parameters. -- Appreciate your clear explanation, -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
