SYNOPSIS

bw_file_rd [ -P <parallelism> ] [ -W <warmups> ] [ -N <repetitions> ] size file

DESCRIPTION

bw_file_rd times the read of the specified file in 64KB blocks. Each block is summed up as a seried of 4 byte integers in an unrolled loop. Results are reported in megabytes read per second.

The data is not accessed in the user program; the benchmark relies on the operating systems read interface to have actually moved the data. Systems that implement page flipping may fool this benchmark.

The benchmark is intended to be used on a file that is in memory, i.e., the benchmark is a reread benchmark. Other file benchmarking can be done with lmdd(8).

The size specification may end with ``k'' or ``m'' to mean kilobytes (* 1024) or megabytes (* 1024 * 1024).

OUTPUT

Output format is "%0.2f %.2f\n", megabytes, megabytes_per_second, i.e.,

8.00 25.33

MEMORY UTILIZATION

This benchmark can move up to three times the requested memory. Most Unix systems implement the read system call as a bcopy from kernel space to user space. Bcopy will use 2-3 times as much memory bandwidth: there is one read from the source and a write to the destionation. The write usually results in a cache line read and then a write back of the cache line at some later point. Memory utilization might be reduced by 1/3 if the processor architecture implemented ``load cache line'' and ``store cache line'' instructions (as well as ``getcachelinesize'').

ACKNOWLEDGEMENT

Funding for the development of this tool was provided by Sun Microsystems Computer Corporation.

AUTHOR

Carl Staelin and Larry McVoy

Comments, suggestions, and bug reports are always welcome.