Note that FFTW has had a way of exposing gcc bugs, especially with SIMD (see http://www.fftw.org/faq/section2.html#compilerCrashes), and it sounds like you are seeing different results with different gcc versions...so that may be what is happening here. Ugh.
Michel Dänzer wrote:
Adding -mabi=altivec to build tests/bench doesn't seem to make a difference here (I do wonder whether all fftw3 dependants were or will be builtwith it though...).
Can you tell us whether -mabi=altivec is really required for programs calling FFTW? I never did figure out exactly what this flag does on Linux. (It is not needed for 16-byte stack alignment, apparently, since I think that is guaranteed by the SVR4 ABI.)
Steven