Re: [Xpert]SOLVED- *Severe* time lossage with XF86 4.1.0 and S3VirgeMX LCD
On Sat, 5 Jan 2002, Vladimir Dergachev wrote: > Take a look in your linux kernel configuration - is "disable interrupts > during APM calls" option enabled ? No. But, it didn't make a difference, anyway. Evidently, calls into the APM BIOS are expensive, and for all I know, the BIOS is shutting IRQs off (I'm not that impressed with that laptop's BIOS overall- the config options are more limited than any machine I've seen in recent memory). -Kenny -- Kenneth R. Crudup Sr. SW Engineer, Scott County Consulting, Washington, D.C. Home1: PO Box 914 Silver Spring, MD 20910-0914 [EMAIL PROTECTED] Home2: 38010 Village Cmn. #217 Fremont, CA 94536-7525 (510) 745-0101 ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]SOLVED- *Severe* time lossage with XF86 4.1.0 and S3VirgeMX LCD
On Sat, 5 Jan 2002, Kenneth Crudup wrote: > > *grumble* > > (I *swear* I'd tested with only a clean "weave" only before, but ) > > So, I narrowed it down to the use of APM. If APM was enabled in the kernel, > *no matter* what options were set in XFree86, I'd get clock slowdowns. So, > I umounted /proc, and tried it again- no slowdown! I downloaded the code, > looked at the linux-specific APM stuff, and even binary edited XFree86 so > that it wouldn't try and open /proc/apm (and by extention, /dev/apm_bios) - > but that didn't work either. So, although I *swear* I started X in a known > minimal state when trying these tests before, I started looking at my > default X applications. > > Turns out that in my .xinitrc is this line: > > xapm -geometry 150x20+0-0 & > > and "man xapm" has this little gem: > > > -delay delay >Sets the number of seconds delay between each update. The default is 1. > > > Figuring that the BIOS call is expensive, I changed that default value to be > 120 seconds, and sure enough, that appears to have fixed things. > > (Doing a "fuser /proc/apm" makes it appear that the device is opened *every* > time it's read, which can't be good for thruput, as God only knows what the > apm proc driver has to do on that open()- so I'll be sending a note to > the "Xapm" maintainer about that and maybe changing the default polling > time, too if when I download the latest version things work the same.) > > My apologies to the Xfree project; your product in fact works very well. Take a look in your linux kernel configuration - is "disable interrupts during APM calls" option enabled ? Vladimir Dergachev > > -Kenny, glad this is over > > -- > Kenneth R. Crudup Sr. SW Engineer, Scott County Consulting, Washington, D.C. > Home1: PO Box 914 Silver Spring, MD 20910-0914 [EMAIL PROTECTED] > Home2: 38010 Village Cmn. #217 Fremont, CA 94536-7525 (510) 745-0101 > > ___ > Xpert mailing list > [EMAIL PROTECTED] > http://XFree86.Org/mailman/listinfo/xpert > ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert
Re: [Xpert]SOLVED- *Severe* time lossage with XF86 4.1.0 and S3VirgeMX LCD
*grumble* (I *swear* I'd tested with only a clean "weave" only before, but ) So, I narrowed it down to the use of APM. If APM was enabled in the kernel, *no matter* what options were set in XFree86, I'd get clock slowdowns. So, I umounted /proc, and tried it again- no slowdown! I downloaded the code, looked at the linux-specific APM stuff, and even binary edited XFree86 so that it wouldn't try and open /proc/apm (and by extention, /dev/apm_bios) - but that didn't work either. So, although I *swear* I started X in a known minimal state when trying these tests before, I started looking at my default X applications. Turns out that in my .xinitrc is this line: xapm -geometry 150x20+0-0 & and "man xapm" has this little gem: -delay delay Sets the number of seconds delay between each update. The default is 1. Figuring that the BIOS call is expensive, I changed that default value to be 120 seconds, and sure enough, that appears to have fixed things. (Doing a "fuser /proc/apm" makes it appear that the device is opened *every* time it's read, which can't be good for thruput, as God only knows what the apm proc driver has to do on that open()- so I'll be sending a note to the "Xapm" maintainer about that and maybe changing the default polling time, too if when I download the latest version things work the same.) My apologies to the Xfree project; your product in fact works very well. -Kenny, glad this is over -- Kenneth R. Crudup Sr. SW Engineer, Scott County Consulting, Washington, D.C. Home1: PO Box 914 Silver Spring, MD 20910-0914 [EMAIL PROTECTED] Home2: 38010 Village Cmn. #217 Fremont, CA 94536-7525 (510) 745-0101 ___ Xpert mailing list [EMAIL PROTECTED] http://XFree86.Org/mailman/listinfo/xpert