AnsweredAssumed Answered

clAmdFft in-place vs. out-of-place performance

Question asked by cipoint on Mar 9, 2013
Latest reply on Mar 11, 2013 by bragadeesh



I need the input data to be conserved. There are two solutions for this: Either make a copy of the input data in the device memory and use the copy as input for a in-place fft or directly use the out-of-place fft.


All transformations are large complex to complex.


Which option will be faster?


Another question: Can I change the placeness with clAmdFftSetResultLocation after baking the plan?