On 8/18/05, John Que <qwejohn@xxxxxxxxx> wrote: > Hello, > Thanks. > I did tried nanosleep and it gave really more than 1 micorsecond. > > also: > from "man nanosleep": > ..... > ..... > BUGS > The current implementation of nanosleep is based on the normal kernel > timer mechanism, which has a resolution of 1/HZ s (i.e, 10 ms on > Linux/i386 and 1 ms on Linux/Alpha). > ....... > ....... > > I assume that from the same reasons, also setitimer() will not do the job. > > As I understand from you I am not permitted to use "outb" from > userspace, and this trick with 0x80 is not good for user space. > (In my case it does not stall the machine, but not much runs on it). > > So is there a way to achieve 1 microsec delay from user space ? I hope I am not missing anything, but If you can afford to make the process run with real time priority, then the same man page has the answer - "If the process is scheduled under a real-time policy like SCHED_FIFO or SCHED_RR, then pauses of up to 2 ms will be performed as busy waits with *microsecond* precision." Also one more thing is that if all you want is some code to execute after 1 microsecond, then you might be able to use the BogoMIPS value (not directly though) and loop appropriately. > Regads, > John Bhanu > On 8/18/05, Erik Mouw <J.A.K.Mouw@xxxxxxxxxxxxxx> wrote: > > On Thu, Aug 18, 2005 at 09:22:55AM -0400, John Que wrote: > > > I want to use delays of 1 microsend on linux 2.4 or linux 2.6 kernel > > > on x386 platform. > > > > > > As we know, usleep depends on the HZ of the system. > > > > It is? > > > > > So both in 2.4 and in 2.6 we cannot achieve 1 microsec delay calling > > > udelay (the man page of usleep says itslef that > > > "Probably not accurate on many machines down to the microsecond."). > > > > You certainly can. RTFM, usleep(2) says in the NOTES section: > > > > This function is obsolete. Use nanosleep(2) or setitimer(2) instead. > > > > > According to "Linux I/O port programming mini-HOWTO" > > > (http://www.tldp.org/HOWTO/IO-P ort-Programming-2.html) > > > you can use outb on 0x80 to create a dealy. > > > > The outb() trick can't be used in userland, you're stalling the > > complete machine by doing so! > > > > > So I added this snippest of code > > > and trird to use outb('a',0x80) for creating a sleep time in > > > resolution of microseconds thus: > > > ............ > > > > > > ioperm(0x80, 1, 1); > > > > > > for (i=0;i<10000;i++) > > > outb('a',0x80); > > > ............ > > > > > > On 3 machines (each of them had a different strength, one of the with > > > 2.6 > > > kernel and the other two with 2.6 kernek) it gave a result of > > > about 1.4 microsec for every such call, while on a differnet macine it > > > gave a much > > > higher average of 2.16 microsec. > > > > That doesn't surprise me, the kernel is allowed to reschedule your > > process during the for() loop, so depending on the system load, the > > actual delay varies a lot. > > > > > > Erik > > > > -- > > Erik Mouw > > J.A.K.Mouw@xxxxxxxxxxxxxx mouw@xxxxxxxxxxxx > > > > > > -----BEGIN PGP SIGNATURE----- > > Version: GnuPG v1.4.1 (GNU/Linux) > > > > iD8DBQFDBJh6/PlVHJtIto0RAiYfAJ9NuG1M8qwu4U34I2Iv12eS7eur5gCggojz > > /R9JT5rqug1QhV1ZFBI0Yds= > > =y6Xk > > -----END PGP SIGNATURE----- > > > > > > > > -- > Kernelnewbies: Help each other learn about the Linux kernel. > Archive: http://mail.nl.linux.org/kernelnewbies/ > FAQ: http://kernelnewbies.org/faq/ > > -- The difference between Theory and Practice is more so in Practice than in Theory. -- Kernelnewbies: Help each other learn about the Linux kernel. Archive: http://mail.nl.linux.org/kernelnewbies/ FAQ: http://kernelnewbies.org/faq/
