|
Centre county Pennsylvania, USA | Folks that use ffts need to download and install Python's Pylab fft package. Execution times for pylab's fft running on this ancient 32 bit win XP OS for a test I just ran are copy/pasted below. Notice that pylab fft's run as fast for non power of 2 fft sizes as they do for power of 2 fft sizes. Matlab's fft execution times are same as pylab fft execution times on this ancient win OS. Only difference is python and pylab are free and matlab is $10K+. Execution times are about 10x faster that those shown when I run same fft execution time test on a modern 64 bit Linux machine with 16GB ram running python/pylab . The magic of modern programming language makes very large software FFTs the preferred way to do many signal processing applications, even on Win machines !!
Python's Pylab fft execution time test: 2009-04-16
Platform for this test: win32
fft size execution time, sec
8 0.000100292076228
16 8.82793762894e-005
32 9.91746157682e-005
64 0.000104761918065
128 0.000143873034143
256 0.000183822245565
512 0.00025282542893
1024 0.000438882595414
2048 0.000862679474626
4096 0.0017337399027
8192 0.00392368303793
16384 0.00930621070555
32768 0.0237884982589
65536 0.0510252001302
131072 0.111424776054
262144 0.234120537666
524288 0.478524225844
1048576 1.0406020877
2097152 2.94105898934
>>> ================================ RESTART ================================
>>>
Python's Pylab fft execution time test, Date: 2009-04-16
Platform for this test: win32
fft size execution time, sec
5 0.000107276204099
10 8.82793762894e-005
20 9.58222343901e-005
40 0.000162311131722
80 0.000114819062199
160 0.000163428592181
320 0.000217066694231
640 0.000344736551713
1280 0.000606222299203
2560 0.00121579697978
5120 0.00212596852393
10240 0.00475814663596
20480 0.0125242174634
40960 0.0270928288372
81920 0.0600584711185
163840 0.131555826229
327680 0.277765470192
655360 0.560451956883
1310720 1.17311473944
2621440 3. | |
|
|
Computers, past and present
