stream results, Compaq XP1000

From: Greg Lindahl (
Date: Tue Mar 30 1999 - 18:39:53 CST

This is the new single-cpu motherboard Compaq workstation:

Compiled with a modern Digital Fortran compiler (not sure what
version), f77 -O5 -tune ev6 -non_shared stream_d.f, then run under
Linux on a 500 mhz XP1000. Almost but not quite as good as the DS20.

gcc majorly sucks in this situation, even the latest egcs. I'm not
sure how it could fuck up something so simple. Sure, it can't issue
the 'write hint' instruction, but for the loops themselves, almost all
of your time is waiting for main memory, so you'd think even stupid
code would be sufficient. Ah well.

 Double precision appears to have 16 digits of accuracy
 Assuming 8 bytes per DOUBLE PRECISION word
 Array size = 2000000
 Offset = 0
 The total memory requirement is 45 MB
 You are running each test 10 times
 The *best* time for each test is used
 Your clock granularity/precision appears to be 1 microseconds
 The tests below will each take a time on the order
 of 24636 microseconds
    (= 24636 clock ticks)
 Increase the size of the arrays if this shows that
 you are not getting at least 20 clock ticks per test.
 WARNING -- The above is only a rough guideline.
 For best results, please be sure you know the
 precision of your system timer.
Function Rate (MB/s) RMS time Min time Max time
Copy: 896.7368 0.0358 0.0357 0.0364
Scale: 879.5568 0.0364 0.0364 0.0365
Add: 892.0275 0.0538 0.0538 0.0539
Triad: 888.3963 0.0541 0.0540 0.0541
 Sum of a is = 2.306601562591874E+018
 Sum of b is = 4.613203124856438E+017
 Sum of c is = 6.150937500141256E+017

This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:08 CDT