| Re: Q: IDL benchmarks [message #5802] |
Sat, 24 February 1996 00:00  |
Ken Knighton
Messages: 44
Registered: May 1995
|
Member |
|
|
rivers@cars3.uchicago.edu (Mark Rivers) wrote:
> In article <4gkdde$9nm@reznor.larc.nasa.gov>, zawodny@arbd0.larc.nasa.gov (Joseph M Zawodny) writes:
>> In article <sterner.825034453@warble.jhuapl.edu> sterner@warble.jhuapl.edu (Ray Sterner) writes:
>>> Bringfried Stecklum <stecklum@gwaihir.astro.uni-jena.de> writes:
>>>
>>>> Is there a collection of benchmark test results which shows how IDL behaves
>>>> on different platforms? There is no reference to this issue in the FAQ.
> Folks,
>
> In doing this compilation, please use TIME_TEST2, not TIME_TEST. TIME_TEST has
> a number of bugs, particularly for disk I/O on VMS. TIME_TEST2 also runs tests
> for longer so that the times are more meaningful.
Something else to keep in mind: Is the disk local or remote (nfs)? We use an NFS
file server and when I changed to a local area, I reduced the disk i/o time by a
factor of 100 (as one would expect).
HP-800 T500 w/ 3 CPUs, 512M memory (64M process size), HP-UX 9.04:
IDL> time_test2
1 1.55186 Empty For loop, 2000000 times
2 1.03341 Call empty procedure (1 param) 100,000 times
3 0.662755 Add 100,000 integer scalars and store
4 0.611896 25,000 scalar loops each of 5 ops, 2 =, 1 if)
5 0.615466 Mult 512 by 512 byte by constant and store, 10 times
6 0.486863 Shift 512 by 512 byte and store, 100 times
7 1.07607 Add constant to 512 x 512 byte array and store, 50 times
8 0.738948 Add two 512 by 512 byte images and store, 30 times
9 1.91268 Mult 512 by 512 floating by constant and store, 30 times
10 1.76016 Add constant to 512 x 512 floating and store, 30 times
11 2.78627 Add two 512 by 512 floating images and store, 30 times
12 0.531054 Generate 225000 random numbers
13 0.509610 Invert a 150 by 150 random matrix
14 0.496082 LU Decomposition of a 150 by 150 random matrix
15 0.740078 Transpose 256 x 256 byte, FOR loops
16 0.0969341 Transpose 256 x 256 byte, row and column ops
17 0.00765991 Transpose 256 x 256 byte, transpose function
18 1.57419 Log of 100,000 numbers, FOR loop
19 0.166916 Log of 100,000 numbers, vector ops
20 0.687387 131072 point forward plus inverse FFT
21 2.96173 Smooth 512 by 512 byte array, 5x5 boxcar, 10 times
22 0.252672 Smooth 512 by 512 floating array, 5x5 boxcar, 2 times
23 0.648114 Write and read 512 by 512 byte array x 20
21.9088=Total Time, 0.60454475=Geometric mean, 23 tests.
Using NFS disk, test 23 completed in 65 seconds. This skewed my geometric mean
up to 0.72.
Ken Knighton knighton@gav.gat.com knighton@cts.com
General Atomics
San Diego, CA
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
