Results of PAPI_FP_OPS for BOTS benchmark

[anilkumarnandamuri]

Hi all,

I have been doing performance analysis of OpenMP tasks using BOTS(Barcelona OpenMP Tasking Suite) benchmark applications.I have used PAPI_FP_OPS to compute the floating point operations.I have got very less number of operations for some applications.The following are the results i have got:

Fibonacci-------size=10----------range of values(2 to 5)
Alignment------input file=prot.20.aa----------range of values(300 to 600)
Nqueens-------size=13----------range of values(2 to 5)
Sort--------------size=128000000----------range of values(2 to 5)
Strassen-------size=2048----------range of values(400000000 to 500000000)
Fft-----------------size=2048----------range of values(10000 to 30000)
Floorplan-------size=input.15----------range of values(2 to 5)
health------------size=small.input-------range of values(around 50000)

Are these results normal or am i doing something wrong?


Thanks,
Anil
University of Houston

[James Ralph]

Hi Anil,

Performance counters are not always precise,
so the preset event definitions have some slack associated with them.

What type of processor are you running on, also what OS and compiler version?
I am unfamiliar with the specifics of BOTS, what counts seemed low to you?

Fibonacci is probably an integer code so I wouldn't expect many floating point ops.
I also wouldn't expect flops in a sort code either.
For strasses, 2048^2.8 is O(2e9) operations so if its working on 32bit floats and using a SIMD instruction thats in the neighborhood ( 4 flops/instruction * 5e8)
FFT is O(n lg n) or ~20k

James

[anilkumarnandamuri]

Hi Ralph,

Thanks for the reply.I am concerned with the results of fibonacci,nqueens,sort and floorplan applications because i got single digits when i have used PAPI_FP_OPS event.
The following is the system information i am using:

processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 45
model name : Intel(R) Xeon(R) CPU E5-2640 0 @ 2.50GHz
stepping : 7
microcode : 0x710
cpu MHz : 2493.836
cache size : 15360 KB
physical id : 0
siblings : 12
core id : 0
cpu cores : 6
apicid : 0
initial apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes


I am using the linux kernel version 3.10.14-100.fc18.x86_64.I am using OpenUH compiler 3.0.36.


Thanks,
Anil
University of Houston