Doug Sondak wrote:
>A few years ago we performed timing tests on f90 intrinsics. We wrote
>f77-style do loops to perform the same functionality as the intrinsics
>and compared timings. These results were not obtained using the
>current compiler (we used SGI 7.2), but they address some of the
>questions in the original posting in this thread.
>
>Most timings were about the same for the intrinsics and the do loops
>with a few significant exceptions. Here some selected results that
>showed large differences
[results snipped]
>Some of these f90 intrinsics could certainly contribute to "f90 being
>slower than f77."
>
>A full list, including the source code used to produce the timings, is
>available at
>http://scv.bu.edu/SCV/Origin2000/intrinsics/F90_serial_times_7.2.html.
How do these results compare with the latest SGI compiler? Compiler techniques
improve over the years. And in particular, I would expect all compiler vendors
would improve their vectoring code for F90/F95.
I use VMS and their compiler is probably similar for TRU64 and CVF in respect
to the optimisations that they have included. My experience is that the
default optimisation is similar enough to (what I said in a previous post)
waste my own code for less maintenance. OK, code written and debugged
extensively 10/20 years ago will need no further maintenance -- but who knows?
I have down-loaded all your source. In my spare time (ha-ha), I shall try to
look at some. I am particularly interested in the MATMUL for which I have
written a myriad of "F77 optimised" routines, using as many different
"optimising" techniques as I could think of.
An enjoyable thread -- thanks Walt for starting it. And I hope that you get
enough for your seminar/discussion paper. Will you be able to make your final
paper available via your web-site? Will you let us know?
Regards, Paddy
Paddy O'Brien,
Transmission Development,
TransGrid,
PO Box A1000, Sydney South,
NSW 2000, Australia
(Street address, 201 Elizabeth Street)
Tel: +61 2 9284-3063
Fax: +61 2 9284-3050
Email: [log in to unmask]
Either "\'" or "\s" (to escape the apostrophe) seems to work for most people,
but that little whizz-bang apostrophe gives me little spam.
|