[vorbis] cdparanoia mailing list ?

Ed Sweetman ed.sweetman at wmich.edu
Fri Mar 15 15:00:29 PST 2002



On Fri, 2002-03-15 at 17:24, Segher Boessenkool wrote:
> > But you cant beat the ide cdrom driver for ide
> > drives, especially with the DMA transfer patch.  Ripping with dma means
> > you use about 2% to negligable cpu to read audio cds.
> 
> IDE DMA read is actually _slower_ than IDE PIO mode read, in most cases
> (the notable exception being really high system loads).  PIO mode uses
> more system time, DMA uses more user time (because of cache loads), and
> DMA mode uses more wall clock time in most cases.  PIO mode reads hide
> the cache latency in i/o latency.

data
 with dma
hdparm -t /dev/cdrom
/dev/cdrom:
 Timing buffered disk reads:  64 MB in 10.04 seconds =  6.37 MB/sec

without dma (pio mode 4)
hdparm -t /dev/cdrom
/dev/cdrom:
 Timing buffered disk reads:  64 MB in 32.17 seconds =  1.99 MB/sec
  (and that's with 100% cpu usage)

audio 
   without dma (pio mode 4)
        cdparanoia -Z 1 /dev/null

        real    0m23.549s
        user    0m0.130s
        sys     0m0.310s

   with dma 
        cdparanoia -Z 1 /dev/null

        real    0m21.518s
        user    0m0.130s
        sys     0m0.320s

<p>Without dma caused the system to crawl and cpu to hit 100% (that's
increasing heat and slows everything else down)  With dma didn't make
any noticable change, it was like i wasn't even ripping.  This
discussion wasn't about whether dma was good to use or not, of course
it's preferred, it was just on who's implimentation was better, scsi or
the ide patch i posted.  Pio mode whores the computer's cpu and is
wholefully disgusting to use when you have something like dma mode at
your disposal. 
 

You can stop spreading fud about dma now. 

> 
> Segher
> 
> 

<p><p>--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.



More information about the Vorbis mailing list