Description of problem: I have just noticed that 'cpuspeed' daemon ignores nice'd jobs. Unfortunately 'kernel-utils' package lacks and documentation on 'cpuspeed' so I had to search the net and found, to my great surprise, on Carl Thompson's (the author) WWW pages that this is a FEATURE. Well, I would argue that this is a BUG. Nice'd jobs are not meant to run *slow*, they are meant not to disturb other, short-lived jobs. Especially in local networks, where machines are used by several people at the same time, users are encouraged to nice their long-running jobs in order to make other people using editors, mailers etc, comfortable. Also, this behaviour is kind of self-contradiction to what 'cpuspeed' help says: "... it increases the CPU speed when the system is not idle...". Well, with a nice'd long-running job the system is surely not idle. I'd vote at least to add a configuration option making the 'cpuspeed' monitoring "idle" status and not "user" status of the CPU(s) (as 'top' reports). Version-Release number of selected component (if applicable): kernel-utils-2.4-9.1.131 How reproducible: always Steps to Reproduce: 1. make sure 'cpuspeed' is running and the CPU is supporting frequency changes 2. wait till CPU(s) goes into "low speed" status 3. run a nice'd job: "nice +19 command" 4. check CPU frequency - it will not increase Actual results: CPU freaquency stays "low" Expected results: It should go "high" Additional info:
In the meantime I have found this problem resolved in FC Development tree, as "cpuspeed-1.2.1-1.28" (still no manpage, also no docs on author's page so I could not find the new "-n" option before I reported in the original post). The source RPM builds w/o any problem on FC3, the only problem remaining that FC3 has 'cpuspeed' integrated into 'kernel-utils' so one needs "--force" option to install the new 'cpuspeed'. Would be profitable to get new 'cpuspeed' included into 'kernel-utils' for FC3 users. regards, Michal. *** This bug has been marked as a duplicate of 132383 ***