Hi all.
I've a question (maybe a stupid question!) about I/O access on ISA card...
I'm writing a real-time module/driver for an ISA card that simply put some
data to a 16 bit DAC... what I can't explain
myself is why the number of cpu cycle for outb and/or writeb is not
constant... if linux kernel is not preempive
why this variability ? maybe interrupt ? I think no cause things doesn't
change even if I add a cli() before the write.
Is there a way to reduce this variability ?
Thanks in advance and excuse my english!
Paolo
----------------------------------------------------------------------------
System : Rtlinux 3.2 pre 1 on Kernel 2.4.19, RH 7.2
CPU model name : AMD-K6(tm) 3D processor
cpu MHz : 451.029
cache size : 64 KB
bogomips : 897.84
----------------------------------------------------------------------------
TEST PROGRAM:
void * thread_code(int fd)
{
pthread_make_periodic_np(pthread_self(), gethrtime(), 1000000000); //
schedulo ogni 1s
printk(KERN_INFO "MIO: realtime thread is started...\n");
while (1)
{
unsigned int counta1,counta2,fixa;
pthread_wait_np();
rdtscl(counta1);
rdtscl(counta2);
fixa = counta2-counta1;
printk("rdtsc took %d cycles\n",fixa);
rdtscl(counta1);
outb(0, 0x330);
rdtscl(counta2);
printk("outb() took %d cycles\n",counta2-counta1-fixa);
rdtscl(counta1);
writeb(0, mio_base+OFF_DAC16_2_EN); // DAC 2 a 16 bit enable
mb();
rdtscl(counta2);
printk("writeb() took %d cycles\n",counta2-counta1-fixa);
return 0;
}
}
----------------------------------------------------------------------------
RESULTS in normal conditions:
Jan 12 14:49:29 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:29 ds4 kernel: outb() took 622 cycles
Jan 12 14:49:29 ds4 kernel: writeb() took 393 cycles
Jan 12 14:49:30 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:30 ds4 kernel: outb() took 632 cycles
Jan 12 14:49:30 ds4 kernel: writeb() took 601 cycles
Jan 12 14:49:31 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:31 ds4 kernel: outb() took 623 cycles
Jan 12 14:49:31 ds4 kernel: writeb() took 444 cycles
Jan 12 14:49:32 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:32 ds4 kernel: outb() took 622 cycles
Jan 12 14:49:32 ds4 kernel: writeb() took 440 cycles
Jan 12 14:49:33 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:33 ds4 kernel: outb() took 627 cycles
Jan 12 14:49:33 ds4 kernel: writeb() took 454 cycles
Jan 12 14:49:34 ds4 kernel: rdtsc took 9 cycles
Jan 12 14:49:34 ds4 kernel: outb() took 796 cycles
Jan 12 14:49:34 ds4 kernel: writeb() took 456 cycles
Jan 12 14:49:35 ds4 kernel: rdtsc took 9 cycles
----------------------------------------------------------------------------
RESULTS with kernel compiling and ping flood active:
Jan 12 15:16:47 ds4 kernel: writeb() took 613 cycles
Jan 12 15:16:48 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:48 ds4 kernel: outb() took 676 cycles
Jan 12 15:16:48 ds4 kernel: writeb() took 615 cycles
Jan 12 15:16:49 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:49 ds4 kernel: outb() took 756 cycles
Jan 12 15:16:49 ds4 kernel: writeb() took 773 cycles
Jan 12 15:16:50 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:50 ds4 kernel: outb() took 666 cycles
Jan 12 15:16:50 ds4 kernel: writeb() took 767 cycles
Jan 12 15:16:51 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:51 ds4 kernel: outb() took 667 cycles
Jan 12 15:16:51 ds4 kernel: writeb() took 544 cycles
Jan 12 15:16:52 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:52 ds4 kernel: outb() took 663 cycles
Jan 12 15:16:52 ds4 kernel: writeb() took 660 cycles
Jan 12 15:16:54 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:54 ds4 kernel: outb() took 665 cycles
Jan 12 15:16:54 ds4 kernel: writeb() took 655 cycles
Jan 12 15:16:54 ds4 kernel: rdtsc took 9 cycles
Jan 12 15:16:54 ds4 kernel: outb() took 663 cycles
Jan 12 15:16:54 ds4 kernel: writeb() took 566 cycles
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Thu Jan 23 2003 - 22:00:13 EST