[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: IRQ's (was Re: Linux Driver)

On Sat, Nov 04, 2000 at 08:12:00AM -0500, Rich Payne wrote:
> > This is a little off topic but I thought the comparison
> > was interesting. The lack of interrupts on the second cpu
> > of the UP2K I find a bit worrisome. Implies that the
> > interrupts aren't split so one processor is more loaded if you
> > do heavy I/O or network accesses. I'd prefer the overhead
> > be spread across both cpus. Thats because my codes assume 
> > both cpus are equal and by only putting overhead on one 
> > cpu, it slows BOTH cpus down to that slower level, essentially
> > wasting cycles on the second cpu because it can get it's
> > workload done more quickly than the other cpu servicing 
> > interrupts...
> This is fixed in 2.4. There may even be patches against 2.2 that
> distribute the interupts. I think it was Andrea Arcangeli who
> did the work.

Yes, the original patches were Andrea's, but they actually broadcast
the interrupt to all the CPUs. This is wasteful/inefficient, so
Richard Henderson completed the work by doling out interrupts to CPUs
on a sort-of static round-robin allocation at register_interrupt()
time, which however can be overridden via command-line args (and I
think the overriding code came from Andrea :-).


Jay A Estabrook                            Alpha Engineering - LINUX Project
Compaq Computer Corp. - MRO1-2/K20         (508) 467-2080
200 Forest Street, Marlboro MA 01752       Jay.Estabrook@compaq.com

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index] []