|
|
| Re: AMD CPU - pentium way faster? (should be IDL on win2000 faster than IDL on linux) [message #29093 is a reply to message #29085] |
Thu, 31 January 2002 12:35   |
Robert Stockwell
Messages: 74
Registered: October 2001
|
Member |
|
|
Rick Towler wrote:
> How much of this is from differences in processor architecture, platform
> configuration, and OS? You are comparing apples and oranges. Install Win2k
> on your AMD system (or linux on your laptop) and try again.
>
> -Rick
>
Of course you are right, so I went to a dual boot machine.
Both running idl 5.5 (in demo mode, we don't have a license on it)
I ran the time tests about 15 time each, and took an average
(after it read from file and compiled the code)
the computer: 1.4 Ghz AMD athlon, 512 Ram.
OS 1) Win2000 SP 2
Average time for time_test.pro = 0.596 seconds
OS 2) Slackware 8 2.4.17 kernel with X3864.1 running fvw2
Average time for time_test.pro = 0.725 seconds
0s 3) Slackware 8 2.4.17 kernel with no xwindows (i.e. command line only)
Average time for time_test.pro = 0.721 seconds
So, that is .725/.596 = 23.5% longer on linux.
(and the command line version is slightly quicker, probably in
printing out the 10 or twnety lines of text)
The difference that I attributed to the pentium over the athlon,
is the same as the difference from win2000 to linux, so it seems
that is is merely the OS, and not the processor that causes the
difference.
I do satellite data crunching on this machine, so the code runs
for about 120 days of CPU time to do the entire set (spread
over many machines of course). The difference is 120 days on linux,
or 100 days on win2000. I would prefer the 100 days.
I'd guess that there are compiler optimizations or different math
libraries available on win2000 that are not used in linux under
my current version and kernel options. I'll take this up with RSI.
Cheers,
bob
PS anyone else see this difference in speed between operating systems?
|
|
|
|
| Re: AMD CPU - pentium way faster? (should be IDL on win2000 faster than IDL on linux) [message #29121 is a reply to message #29085] |
Tue, 05 February 2002 15:14 |
John-David T. Smith
Messages: 384
Registered: January 2000
|
Senior Member |
|
|
Mark Hadfield wrote:
>
> "Robert Stockwell" <rgs1967@hotmail.com> wrote in message
> news:3C59AAA6.4040802@hotmail.com...
>> Average time for time_test.pro = ...
>
> Isn't time_test deprecated these days in favour of time_test2? (The reason
> being, I believe, that the tests in the former are too small to be
> meaningful.) Time_test2 is the only one mentioned in the help file. There's
> also a time_test3 that seems to be the same as time_test2 except that some
> tests have been increased in size. JD's IDLSPEC2 benchmark uses time_test3
> so perhaps we should standardise on this.
The problem with time_test3 is that it's already antiquated. I've
several times sounded the battle cry for developing an updated IDL
benchmark which taxes machine subsystems adequately and independently
(not that it shouldn't use time_test* as a basis). For instance, the
sole I/O test in time_test* amounts to 256KB array read/writes -- well
within the cache of all modern drives and OS's. An entire dimension on
the IDLSPEC2 graph is thus rendered almost useless. I have collected a
list of volunteers for constructing this community benchmark, and
eagerly solicit more. Depending on interest, an expanded 3D Graphics
benchmark might also be developed in parallel.
IDLSPEC3 has been dragging along primarily due to this stumbling block.
If people are interested in contributing their thoughts and ideas
concerning a more robust, community-developed test suite, then by all
means, contact me.
Having just completed a 3000 mile move, it may be some time yet before
you see the light in IDLSPEC3's eyes, but if we pull together, it will
be worth it. And for those whose arms I twisted as volunteers, don't
think I've lost your names ;)
JD
|
|
|
|
| Re: AMD CPU - pentium way faster? (should be IDL on win2000 faster than IDL on linux) [message #29170 is a reply to message #29085] |
Fri, 01 February 2002 06:23 |
Robert Stockwell
Messages: 74
Registered: October 2001
|
Member |
|
|
Mark Hadfield wrote:
> "Robert Stockwell" <rgs1967@hotmail.com> wrote in message
> news:3C59AAA6.4040802@hotmail.com...
>
>> Average time for time_test.pro = ...
>>
>
> Isn't time_test deprecated these days in favour of time_test2? (The reason
> being, I believe, that the tests in the former are too small to be
> meaningful.) Time_test2 is the only one mentioned in the help file.
I did both sets of tests. Same results. I only mentioned
the time_test ones because I am lazy.
> There's
> also a time_test3 that seems to be the same as time_test2 except that some
> tests have been increased in size. JD's IDLSPEC2 benchmark uses time_test3
> so perhaps we should standardise on this.
Sure, I'll find that and get the results.
Cheers,
bob
> ---
> Mark Hadfield
> m.hadfield@niwa.co.nz http://katipo.niwa.co.nz/~hadfield
> National Institute for Water and Atmospheric Research
>
>
>
>
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
