I have a question regarding the "expected" time to complete a call to ibwrite.
Here is a code snippet... This code works well however the question I have is in regard to the time required to complete the call...
The timing data below was collected using an HP XW8600 with dual NI488.2 installed running RedHat (RHEL) 5.1. with the NI-488.2 2.5 - Linux - Beta 32 Bit driver with gcc 4.3.3.
I would be very interested in anyone elses performance on Win32/Linux platforms while using NI and/or 3rd party drivers.
Code snippet which produced this timing...
Note variable mIEEE488SpirentGPS.dataToWriteOut is a string which is approx 200 chars in length.
case IEEE_488_IBWRT:
// start write timer
StartTimer( 2, writeCaseTimerName ); // Starting a timer (0.6us to start/stop)
// write data to GPIB
if( ( status = ibwrt( mIEEE488SpirentGPS.unitDescriptor,
(char *)mIEEE488SpirentGPS.dataToWriteOut,
mIEEE488SpirentGPS.writeCount ) ) & ERR )
{
// place error in member error message parameter & report error to user
HandleIbwrtError( SPIRENT_GPS_BOARD_IDENTIFIER, status,
(char *)mIEEE488SpirentGPS.dataToWriteOut,
fileAndFunctionName,
(char*)mIEEE488SpirentGPS.errorMessage );
break;
}
// stop write timer
StopTimer( 2 ); // Stop timer (0.6us to start/stop)
Here is the timer data dump after the run...
Note an avg, min, max time of 38.9ms, 1.4ms, and a 67.6ms (respectively) to complete a call to ibwrite.
See below for a histogram of the call times...
#### avg/min/max timer data for timer(2) ####
Name of timer: IEEE_488_B_Local::RealTimeSpirentGPSTask - Write case
Average looping time (seconds): 0.038902582920
Minimum looping time (seconds): 0.001380000000
Maximum looping time (seconds): 0.067578000000
Average rate for Power Supply real time loop: 25.705234 Hz
Number of loops: 1007
Error happened in StartTimer: false
Error happened in StartTimer: false
The histogram data indicates that 999 calls (out of 1007) took between 10-50ms and
8 calls took between 50-100ms with the worst case being 67.6ms.
All of this is consistent however this is way too slow for the planned update rate to this device.
Are these results consistent with other peoples experience/timing on other platforms ?
I need to have a worst case duration of 3ms (with 1ms of margin) in order to achieve a 250hz message rate. All comments or suggestions are welcome...
Do I need to switch to scramnet ? That is my only other option for this device.
#### histogram data for timer(2) ####
sim_uut_timer.report histo t00,c00 : 0.000000000 : 0
sim_uut_timer.report histo t01,c01 : 0.000000100 : 0
sim_uut_timer.report histo t02,c02 : 0.000000500 : 0
sim_uut_timer.report histo t03,c03 : 0.000001000 : 0
sim_uut_timer.report histo t04,c04 : 0.000005000 : 0
sim_uut_timer.report histo t05,c05 : 0.000010000 : 0
sim_uut_timer.report histo t06,c06 : 0.000050000 : 0
sim_uut_timer.report histo t07,c07 : 0.000100000 : 0
sim_uut_timer.report histo t08,c08 : 0.000500000 : 0
sim_uut_timer.report histo t09,c09 : 0.001000000 : 0
sim_uut_timer.report histo t10,c10 : 0.005000000 : 0
sim_uut_timer.report histo t11,c11 : 0.010000000 : 999
sim_uut_timer.report histo t12,c12 : 0.050000000 : 8
sim_uut_timer.report histo t13,c13 : 0.100000000 : 0
sim_uut_timer.report histo t14,c14 : 1.000000000 : 0
