Re: NEC-LIST: NEC Execution Times Under Windows

Jim Moser (jmoser_at_kintronic.com) wrote:
> ..., we are seeing much less of an improvement than we expected.
> The execution times are only a factor of ~3 faster on a 200 MHz
> Pentium MMX With 128 MB of SDRAM (10 ns) in DOS box on Win 95 OSR2
> than on a 90 MHz Pentium with 16 MB of RAM (70 ns) in DOS box on Win
> 3.1. We expected an improvement closer to 10 times faster.

I don't know why you expected that, but let's solve this
scientifically.

Please state exactly what the execution time was, and also the number
of segments in the job.

The execution time should be (at least for large jobs:)

  8/3*N^3 / F

with N the number of segments and F the Flops (floating point
operations) per second your processor is doing.

Approximate practical values: a Pentium 200 should do about 30M
flops. A Pentium II about 40 M, a Pentium Pro about 50M (all at
200MHz.) The Pentium MMX you mention will behave like a normal
pentium, MMX has nothing to do with the numerical calculations NEC is
doing.

If the F in the above equation is much less than the value we should
expect, then either:

- The algorithm in NEC 4 is not optimal (it also wasn't in NEC 2).
- Memory access on your system is too slow.
- Internal cache of processor is too small (>16*N would be nice.)
- The compiler you used did a bad job (compare with other compilers)

Please note that the equation given applies only to large N (>~1000),
where the time is dominated by the solver of the system of
equations. For a clean test, the job also should not use symmetry
tricks or elaborate field patterns.

Greetings,
Jos

----
    Dr. Jozef R. Bergervoet Electromagnetism and EMC
    Philips Research Laboratories, Eindhoven, The Netherlands
    E-mail: bergervo_at_natlab.research.philips.com Phone: +31-40-742403
Received on Fri Oct 10 1997 - 08:59:37 EDT