> > mt = per*(bsize/secsize)/100 * 1000 / secsps/s; > >to > > mt = (per-50)*(bsize/secsize)/100 * 1000 / secsps/s; > > Did it help?
Yes, it seems so... BTW I've added call to cdr_buffer_cap right after usleep and here is what I get (I print the return value from cdr_buffer_cap). with "mt = per*..." and -dao [results in underrun!!!]: Track 01: 1 of 32 MB written (fifo 100%) [buf 97%] |300 826ms 826ms| after sleep 100% 0.4x with "mt = per*..." without -dao [also results in underrun!!!]: Track 01: 1 of 32 MB written (fifo 96%) [buf 72%] |140 57ms 613ms| after sleep 72% 32.5x Track 01: 2 of 32 MB written (fifo 100%) [buf 98%] |307 825ms 836ms| after sleep 2% 5.0x with "mt = (per-50)*..." and -dao [works!!!]: Track 01: 1 of 32 MB written (fifo 100%) [buf 97%] |300 400ms 400ms| after sleep 49% 0.4x Track 01: 2 of 32 MB written (fifo 96%) [buf 89%] |249 316ms 332ms| after sleep 51% 10.5x Track 01: 3 of 32 MB written (fifo 100%) [buf 88%] |243 267ms 324ms| after sleep 56% 12.1x Track 01: 4 of 32 MB written (fifo 100%) [buf 96%] |294 284ms 392ms| after sleep 61% 13.8x with "mt = (per-50)*..." without -dao [also works!!!]: Track 01: 1 of 32 MB written (fifo 100%) [buf 72%] |140 54ms 186ms| after sleep 72% 34.2x Track 01: 2 of 32 MB written (fifo 100%) [buf 97%] |300 400ms 400ms| after sleep 50% 5.0x Track 01: 3 of 32 MB written (fifo 98%) [buf 87%] |236 294ms 314ms| after sleep 52% 10.7x Track 01: 4 of 32 MB written (fifo 100%) [buf 91%] |262 264ms 349ms| after sleep 59% 13.2x The drive is Ricoh MP2125 based. It reports 1470 bytes for buffer space in reply to READ BUFFER CAPACITY. Sleeping for 800ms results in 1200 bytes at 10x which is pretty much the whole buffer capacity [at least with one audio block accuracy]. The fact that you see "after sleep 100%" in first line is not a "lie," but indication that unit has already suffered from underrun at exit from usleep. So that it's hardly drive or usleep's fault. But another question [related to -v bug/feature!] is following. Is it intentional that it considers to fall asleep after every 1MB written? I mean 1MB is 1/2 of buffer and it probably should consider usleep more often so that unit buffer wouldn't be drained that much? A. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
