Hello, As Victor and Cort asked in their emails RE my previous test results on scheduling precision (2.4.18 kernel, rtl 3.1, with and without xfs patches), I now measured rtlinux sceduling on 2.4.4 kernel rtlinux v3.1. The architecture and hardware is almost identical; the only difference being that the 2.4.4 kernel was on a different harddrive, but the kernel config was the same, processor is P4, everything unchanged.
The bottom line is that scheduling is VERY PRECISE with the 2.4.4 kernel, with NO values at all above 15000 (output of monitor program, I guess it means 15us), WHILE with the 2.4.18 kernel (with NO xfs, just the pure rtl-3.1 patch) the scheduling has values up to 1e8 (that is 1e5 us = 0.1sec?!). 2.4.18-rtl-3.1: Mean: 5438.097 Median: 1568.000 SD: 548203.807 Minimum:1056.000 Maximum:114589920.000 2.4.4-rtl-3.1: Mean: 1674.800 Median: 1664.000 SD: 400.981 Minimum: 1056.000 Maximum: 11584.000 So: what is happening? Am I doing sth wrong with the 2.4.18 kernel? Or it is not really realtime with the patch? (ftp://ftp.fsmlabs.at/pub/rtlinux/contrib/gearheart/) Gaspar p.s.: Updated histograms at the same place: http://www-cfa.harvard.edu/~gbakos/rtlinux-stat/ see rtl*.7.* + ------------------------------------------------------------------------ + Gaspar A. Bakos Predoctoral Fellow, Solar and Stellar Physics Division Harvard-Smithsonian Center for Astrophysics 60 Garden Street, Cambridge, MA 02138 (USA) Phone(office): 617-495-7410 Fax: 617-495-7049 email: [EMAIL PROTECTED] or [EMAIL PROTECTED] homepage: http://cfa-www.harvard.edu/~gbakos + ------------------------------------------------------------------------ + _______________________________________________ Rtl mailing list [EMAIL PROTECTED] http://www2.fsmlabs.com/mailman/listinfo.cgi/rtl
