Igor <rych...@gmail.com> writes: > Hi Andreas, > I know it's subject to rounding etc. but ReductionKernel often finds a > max value at a different position than both numpy.argmax and > thrust::max_element, the latter two agreeing always. > http://dev.math.canterbury.ac.nz/home/pub/26/
Thrust takes care to always return the first occurrence of the max--that might be the reason: https://code.google.com/p/thrust/source/browse/thrust/system/detail/generic/extrema.inl Andreas
pgp6BR99NbeqA.pgp
Description: PGP signature
_______________________________________________ PyCUDA mailing list PyCUDA@tiker.net http://lists.tiker.net/listinfo/pycuda