Problems with FFT functions in IPP v5.0 for Xscale

[jjbunn]

I am trying to convert my FFT application which used IPP v3.0 to the new v5.0.

For some reason I am finding discrepancies between the documentation, and what is available in the include and library files.

I am wanting to use the not-in-place FFT on arrays of 32 bit floats for real and imaginary.

I declare:

#include "ipp.h"

Ipp32f rein[SAMPLESIZE], imin[SAMPLESIZE];
Ipp32f reout[SAMPLESIZE], imout[SAMPLESIZE];
Ipp8u *pBuffer, *pAllocBuffer;
IppsFFTSpec_R_32f FFTSpec;
IppHintAlgorithm hint = ippAlgHintAccurate;

Then, later,

ippsFFTFwd_CToC_32f(rein, imin, reout, imout, &FFTSpec, pBuffer);


And the VS 2005 compiler tells me:
1>.fft.cpp(168) : error C2079: 'FFTSpec' uses undefined struct 'FFTSpec_R_32f'
1>.fft.cpp(269) : error C3861: 'ippsFFTFwd_CToC_32f': identifier not found

Yet these are the structs and function names specified in the documentation.

I can see in the ipps.h include file that these identifiers are in an _OWN_BLDPCS clause.

Am I missing a #define that brings in what I need? Help, please!

Thanks!
Julian

[Ying_S_Intel]

Julian,

I am not clear which IPP packages in 3.0 and 5.0 you installed for the Xscale system you claimed in yoru subject line.

Please notice that the IPP for Intel PXA25x/26x/27x processors do not support float data type (32f) but fixed-point data type.

I would recommendyouprovide more clarifications such as which package you used in IPP3.0 and 5.0, what target system you are using and submit an issue with these deatils in Intel Premier Support where our support engineers will provide furtherassistances.

Thanks,
Ying S
Intel Corp.

[jjbunn]

Could you please tell me how to access Premier Support? I used to have an account, but cannot remember the password, and the password retrieval web page at the Premier Support web site seems to be broken.

I am targetting ARM processors e.g. on the Axim x51v. I am using IPP v5.0. The documentation I have with v5.0 does not say that the 32 bit floats are unsupported, so I am now unclear what I should be using.

In IPP v3.0 I was using the FFT for 16 bit shorts, but now need to upgrade to 32 bits of precision.

Julian