| bug in systime(1) on IDL 5.2 Linux/Intel? [message #14404] |
Tue, 23 February 1999 00:00  |
woodford
Messages: 10
Registered: June 1996
|
Junior Member |
|
|
systime(1) on IDL 5.2 Linux/Intel seems to always yield whole numbers.
For instance,
print, systime(1) - long(systime(1))
always returns 0.0000... In contrast, this statement returns various
fractions in IDL 5.2 WinNT/Intel and IDL 5.0 Mac. Has anyone else
observed this behavior? It wreaks havoc with trying to use time_test3 and
graphics_times3 for benchmarking.
--
Paul Woodford
woodford@essexcorp.com
|
|
|
|
| Re: bug in systime(1) on IDL 5.2 Linux/Intel? [message #14470 is a reply to message #14404] |
Tue, 02 March 1999 00:00  |
Kevin P. Reardon
Messages: 2
Registered: January 1999
|
Junior Member |
|
|
"Woodford, Paul" wrote:
>
> systime(1) on IDL 5.2 Linux/Intel seems to always yield whole
> numbers.
I just ran into this problem while installing a IDL 5.2 on a new Linux
machine. Desperate to benchmark the beast, I coded up a quick workaround
using the information contained in /proc/uptime (which contains the
uptime in hundredths of a second). A simple program to read and format
this information in the same way as systime is given below. Then in
$IDL_DIR/lib/time_test.pro (or rather, a copy thereof), I substituted
all occurences of "systime" with "uptime". Millisecond accuracy it
ain't, but for running time_test[123], it seems sufficient (the fastest
of the individual tests on my machine takes 0.07 seconds). Of course, a
real fix for systime would be better.
Of course, this only works if you have the /proc filesystem enabled and
you machine doesn't reboot during the time_test.
I hope it helps,
kevin reardon
Osservatorio Astronomico di Capodimonte
; a program to read the /proc/uptime "file" in a systime(1)-like
; format
function uptime,return_type
; select the first of the two fields in the file
spawn, 'cat /proc/uptime | cut -f 1 -d " "' , uptime
uptime=double(uptime[0])
return,uptime
end
|
|
|
|
| Re: bug in systime(1) on IDL 5.2 Linux/Intel? [message #14488 is a reply to message #14404] |
Mon, 01 March 1999 00:00 |
J.D. Smith
Messages: 214
Registered: August 1996
|
Senior Member |
|
|
Ivan Zimine wrote:
>
> I have the same behavior...
>
> IDL> print, !version
> { x86 linux unix 5.2 Oct 30 1998}
> IDL> t=systime(1) & f=dialog_pickfile() & print, systime(1)-t
> 3.0000000
> IDL> t=systime(1) & f=dialog_pickfile() & print, systime(1)-t
> 1.0000000
> etc 10 times
>
> Couldn't get any fractional part. I doubt that my internal timing is so
> perfect.
>
> "Woodford, Paul" wrote:
>>
>> systime(1) on IDL 5.2 Linux/Intel seems to always yield whole numbers.
>> For instance,
>>
>> print, systime(1) - long(systime(1))
>>
>> always returns 0.0000... In contrast, this statement returns various
>> fractions in IDL 5.2 WinNT/Intel and IDL 5.0 Mac. Has anyone else
>> observed this behavior? It wreaks havoc with trying to use time_test3 and
>> graphics_times3 for benchmarking.
>>
>> --
>> Paul Woodford
>> woodford@essexcorp.com
>
> --
> Ivan Zimine
> Dpt. of Radiology (MRI), Geneva University Hospitals
> email: ivan.zimine@physics.unige.ch
> tel. : (+41 22) 372 70 70
This is a known bug, and has been reported. A consequence of this I
must delay a 2nd generation idl speed survey. Without accurate timing
information, this is useless on affected architechtures. I guess we'll
just have to wait for 5.2.1.... (and even then I'll have to code around
this brokenness :{ ).
JD
--
J.D. Smith |*| WORK: (607) 255-5842
Cornell University Dept. of Astronomy |*| (607) 255-6263
304 Space Sciences Bldg. |*| FAX: (607) 255-5875
Ithaca, NY 14853 |*|
|
|
|
|
| Re: bug in systime(1) on IDL 5.2 Linux/Intel? [message #14500 is a reply to message #14404] |
Mon, 01 March 1999 00:00 |
Ivan Zimine
Messages: 40
Registered: February 1999
|
Member |
|
|
I have the same behavior...
IDL> print, !version
{ x86 linux unix 5.2 Oct 30 1998}
IDL> t=systime(1) & f=dialog_pickfile() & print, systime(1)-t
3.0000000
IDL> t=systime(1) & f=dialog_pickfile() & print, systime(1)-t
1.0000000
etc 10 times
Couldn't get any fractional part. I doubt that my internal timing is so
perfect.
"Woodford, Paul" wrote:
>
> systime(1) on IDL 5.2 Linux/Intel seems to always yield whole numbers.
> For instance,
>
> print, systime(1) - long(systime(1))
>
> always returns 0.0000... In contrast, this statement returns various
> fractions in IDL 5.2 WinNT/Intel and IDL 5.0 Mac. Has anyone else
> observed this behavior? It wreaks havoc with trying to use time_test3 and
> graphics_times3 for benchmarking.
>
> --
> Paul Woodford
> woodford@essexcorp.com
--
Ivan Zimine
Dpt. of Radiology (MRI), Geneva University Hospitals
email: ivan.zimine@physics.unige.ch
tel. : (+41 22) 372 70 70
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
