Fwd: [Beowulf] Re: Why is communication so expensive for very small messages
Many of your questions may have already been answered in earlier discussions or in the FAQ. The search results page will indicate current discussions as well as past list serves, articles, and papers.
Mark Hahn hahn at mcmaster.caThu Apr 26 09:41:37 PDT 2007
- Previous message: Fwd: [Beowulf] Re: Why is communication so expensive for very small messages
- Next message: Fwd: Re: [Beowulf] Why is communication so expensive for very small messages?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
> zero, but MPI_Wtime() only has a resolution of 4ms.
why use MPI_Wtime? I use:
#include <sys/time.h>
double gtod() {
struct timeval tv;
gettimeofday(&tv,0);
return tv->tv_sec + 1e-6 * tv->tv_usec;
}
portable, high-resolution. you might be thinking "oh, but MPI_Wtime() is
based on clock(3) or times(2), which have to be better because they are
mandated by more standards and provide the tempting distinctions between
wall, user and system times. alas, the resolution or accuracy of such
accounting is never of much importance to the kernel, so has to be ignored.
in short, use wallclock time from gettimeofday...
if you want, I can send my rdtsc-based timing routines, but they need
to be calibrated and you have to consider cpu binding on anything other
than single-core systems...
regards, mark hahn.
- Previous message: Fwd: [Beowulf] Re: Why is communication so expensive for very small messages
- Next message: Fwd: Re: [Beowulf] Why is communication so expensive for very small messages?
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Beowulf mailing list
