John McCalpin wrote:
>
> In article <34582C66.347B@erols.com>, <kasmjs@erols.com> wrote:
> >What is the sustained floating point performance of the newer, or
> >newest (300 MHz?), Pentium II processors for problems like the following
> >performed using the 64-bit real variables a, b, c, and d:
> >
> > DO 10 i = 1,1000000
> > a(i) = b(i)*c(i)
> >10 CONTINUE
> >
> >and...
> >
> > DO 20 i = 1,1000000
> > a(i) = b(i)*c(i) + d(i)
> >20 CONTINUE
>
> These kernels are a lot like the STREAM benchmark.
>
> http://www.cs.virginia.edu/stream/
>
> I don't have any Pentium II numbers, but would welcome any contributions!
[ I pulled the numbers below from:
http://www.tdl.com/~netex/mb/440lxdata.html ]
gcc -O3 stream_d.c second_cpu.c -o stream_d -lm
Define N 3000000
ASUS P2L97 440LX w/128MB SDRAM 10ns:
[hws note: this is a 266 MHz system]
[...]
-------------------------------------------------------------
This system uses 8 bytes per DOUBLE PRECISION word.
--------------------------------------------------
Function Rate (MB/s) RMS time Min time Max time
Copy: 165.5172 0.2960 0.2900 0.3000
Scale: 165.5172 0.2960 0.2900 0.3000
Add: 189.4737 0.3860 0.3800 0.3900
Triad: 175.6098 0.4110 0.4100 0.4200
===========================
[hws note: it would also be interesting to test
this with egcs -O6 ... ]
This archive was generated by hypermail 2b29 : Tue Apr 18 2000 - 05:23:07 CDT