I am using the Intel XEON processor with Quadcore (8 Thread) VPX3-1220 SBC E3 1505L CPU. We are using intel IPP libraries for signal processing. Please confirm how many parallel FFTs and max. FFT length, I can run using parallel threads. Is there any timing analysis??
The max of IPP FFT length depends on the precisions and real or complex data types, namely:
for double precision complex DFT (64fc) the length upper bound is 67108863 (2^26 - 1).
for single precision complex DFT (32fc) the length upper bound is 134217727 (2^27 - 1).
for double precision complex FFT (64fc) the length upper bound is 2^27.
for double precision complex FFT (32fc) the length upper bound is 2^28.
There are no published perf results, as you may take the IPP Perf system, chose the list of functions you like to test ( e.x. ippsFFTFwd_CToC_32f), and run the perf system on your specific machine with your environment. The output performance results in clock per element would answer your question.
Thank you sir!!
My soubt is as i am using quad core Processor(8 threads) and i am running FFT 64k parallely four separate application threads. Catch is each FFT consumes two cores of hardware?? In that case i can not run more than two FFTs parallely!! . Hence, requesting you to confirm!!
However as you suggested, i will check with iperf system and comeback to you sir.
“running FFT 64k parallely four separate application threads.”
In that case, you may see the thread oversubscription which might lead to performance degradation.The general recommendation to set 1 OpenMP thread call ( ippSetNumThreads(1))
Or explicitly link your application against sequential versions of IPP and didn’t modify your code.
welcome:). The issue is closing and we will no longer respond to this thread. If you require additional assistance from Intel, please start a new thread. Any further interaction in this thread will be considered community only.