Date: Sat, 12 Nov 2005 16:18:15 +0900 From: Pyun YongHyeon <pyunyh@gmail.com> To: freebsd-hackers@freebsd.org Subject: poor fdc(4) performance Message-ID: <20051112071815.GD18405@rndsoft.co.kr>
next in thread | raw e-mail | index | archive | help
Hi, I encountered a P3 SMP system that shows poor fdc(4) performance. But when I use larger block size, say 36b, it seems it works as expected. #dd if=boot.flp of=/dev/fd0 2880+0 records in 2880+0 records out 1474560 bytes transferred in 607.483808 secs (2427 bytes/sec) #dd if=boot.flp of=/dev/fd0 bs=36b 80+0 records in 80+0 records out 1474560 bytes transferred in 48.482513 secs (30414 bytes/sec) On other P4 machine I got: #dd if=boot.flp of=/dev/fd0 2880+0 records in 2880+0 records out 1474560 bytes transferred in 51.007538 secs (28909 bytes/sec) #dd if=boot.flp of=/dev/fd0 bs=36b 80+0 records in 80+0 records out 1474560 bytes transferred in 47.972662 secs (30738 bytes/sec) This type of poor dd performance is the same symptom I'd like to solve on sparc64. On ebus based Ultrasparcs it shows almost the same phenomenon. Since I can't see any DMA or retry errors on fdc I have no idea what caused this. In addition, it works well on sbus based ultrasparcs which don't support DMA. Any ideas? -- Regards, Pyun YongHyeon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20051112071815.GD18405>