DAE performance
This is more to satisfy my curiosity than anything else, but maybe I
can help some folks out and learn something in the process. I've
recently started using my 450MHz G4 Cube (running woody with a
2.4.20-xfs-ben5 kernel) as a CD-ripping toaster. I'm using cdparanoia
III with ide-scsi to extract the audio, with jack as a front end. My
question is this: is it normal to see performance at around ~1.2 -
2.0x when ripping on a Cube? cdparanoia is using less than 4% of CPU,
so it doesn't seem to be swamped by I/O processing. I've done a bunch
of tweaking (upped the default DMA buffer, turned on Direct I/O), and
it doesn't seem to make much difference.
The combination of cdparanoia and LAME (--r3mix --nohist) results in
ripping and encoding taking ~110% the time that it would to play the
CD. I know that cdparanoia is, well, paranoid, but it seems a little
odd that I can get MP3s that sound just about the same out of iTunes
in about 10% of the time on my 500MHz original TiBook. Anyone have any
explanations or persuasive lies to account for the difference?
And if I play music through the Cube's USB speakers on top of ripping
and encoding, sound gets kind of rough (and the amount of CPU time
spent in kernel-land gets much higher). Is there any set of patches
that would improve this situation without harming ripping and encoding
performance overly?
(Details for the curious:
1. Kernel is a stock 2.4.20-ben5 with only the XFS patches
applied.
2. lsmod prints (most stuff is compiled into the kernel itself,
followed xiph.org recommendations and didn't compile ide-cd
support):
Module Size Used by Not tainted
ide-scsi 9824 15
sg 30788 1 (autoclean)
af_packet 11400 2 (autoclean)
sd_mod 12732 0 (autoclean) (unused)
hid 15024 0 (unused)
sr_mod 17588 28 (autoclean)
cdrom 29728 0 (autoclean) [sr_mod]
scsi_mod 64128 4 (autoclean) [ide-scsi sg sd_mod sr_mod]
3. cdparanoia -vsQ prints:
cdparanoia III release 9.8 (March 23, 2001)
(C) 2001 Monty <monty@xiph.org> and Xiphophorus
Report bugs to paranoia@xiph.org
http://www.xiph.org/paranoia/
Checking /dev/cdrom for cdrom...
Testing /dev/cdrom for cooked ioctl() interface
/dev/scsi/host0/bus0/target0/lun0/cd is not a cooked ioctl CDROM.
Testing /dev/cdrom for SCSI interface
generic device: /dev/scsi/host0/bus0/target0/lun0/generic
ioctl device: /dev/scsi/host0/bus0/target0/lun0/cd
Found an accessible SCSI CDROM drive.
Looking at revision of the SG interface in use...
SG interface version 3.1.24; OK.
CDROM model sensed sensed: MATSHITA DVD-ROM SR-8186 F113
Checking for SCSI emulation...
Drive is ATAPI (using SCSI host adaptor emulation)
Checking for MMC style command set...
Drive is MMC style
DMA scatter/gather table entries: 256
table entry size: 65536 bytes
maximum theoretical transfer: 7133 sectors
Setting default read size to 13 sectors (30576 bytes).
Verifying CDDA command set...
Expected command set reads OK.)
yrz,
Forrest
--
. . . the self-reflecting image of a narcotized mind . . .
ozymandias G desiderata ogd@aoaioxxysz.net desperate, deathless
(415)823-6356 http://www.pushby.com/forrest/ ::AOAIOXXYSZ::
Reply to: