Re: PROBLEM: CD burning at 12x uses excessive CPU, although DMA is

[Markus Plail]

Hi Sergio!

* Sergio Bruder writes:
>On Sun, Apr 14, 2002 at 03:39:35PM +0300, Anssi Saari wrote:
>>(...)
>>
>>I have now tried the writer on an old Pentium motherboard with Intel
>>430HX chipset and PIIX3. The performance problems didn't happen there,
>>so I would guess this is more a problem with how Linux handles the VIA
>>686b southbridge.
>>
>>But what can I do to help fix this problem?

>We are now in 2.4.20-preBLAH time and appears that VIA 686b southbridge
>still has the same problem (and appears to be that dont surface with
>anothers controllers)

>Ive just burned a image (.cue/.bin) with cdrdao and my load went nuts,
>up to 5. Same motherboard chipset, same general conditions.  KT133a,
>VIA 686b southbridge, in that case with a LG combo drive (HL-DT-ST
>RW/DVD GCC-4120B), hard-disk as /dev/hda (with ReiserFS), combo drive
>as /dev/hdc.

>What is strange is that problem dont appears with ISO images
>(cdrecord-burned). There is any already-know solution for that problem
>with VIA686b IDE?

The problem isn't really the southbridge, the problem is, that the
kernel doesn't seem to use (I still didn't get a definitive answer)
DMA when doing things with bigger blocksizes (grabbing audio CDs,
writing DAO CDs). That's why you don't have a problem burning 'normal'
TAO CDs. It seems that older chipsets just deal way better with PIO
modes (the PIIX3 doesn't have DMA, does it?).
For me it helped at least a bit to do the following:
- Disable interrupt sharing for IDE devices
- hdparm -d1 -c3 -u0 -X34/66 /dev/burner

With this settings I have 30% processor usage when writing at speed 20,
as opposed to at least 60% I had before.

regards
Markus