topic Bo, in Intel® oneAPI Math Kernel Library

output length of cosine transform

Bo_Q_ — Mon, 21 Jul 2014 16:03:50 GMT

I have a seemingly naive question: why the length of the output of cosine transform might be longer than n? For example, for staggered cosine transform, the output length is 3n/2. But just looking at its equation, the output should have the same length as the input, which is n. Thanks!

Bo,

Chao_Y_Intel — Tue, 22 Jul 2014 02:11:55 GMT

Bo,

Are you talking to some specific MKL questions, or this just a general question regarding the DCT transform?

Thanks,
Chao

Thanks for your reply. I am

Bo_Q_ — Tue, 22 Jul 2014 02:37:30 GMT

Thanks for your reply. I am only interested in using DCT but the question is applicable to sine transforms as well. Here are the defination of the transforms:

https://software.intel.com/sites/products/documentation/hpc/mkl/mklman/GUID-48C88175-7E59-46C7-99E1-A78BCEE0E981.htm

For instance, for the staggered cosine transform, its output is F(k) with k = 0...n-1, that is n elements in total. But if you look at the documentation here:

https://software.intel.com/sites/products/documentation/hpc/mkl/mklman/GUID-3DA88A75-AC3D-45F1-B647-7D020451D010.htm

You will find the output has 3n/2 elements. Why?

Bo,

Chao_Y_Intel — Thu, 24 Jul 2014 03:11:06 GMT

Bo,
I think you are talking the dpar/spar parameter for these functions. Actually, they are not used for the computed result. They are just used for store the intermediate sin/cos data that are used in the TT computation.

For the computation function:
d_forward_trig_transform(double f[], DFTI_DESCRIPTOR_HANDLE *handle, MKL_INT ipar[], double dpar[], MKL_INT *stat);

Both the input and output is F[], which is the same size.

Thanks
Chao