------- Comment #37 from jakub at gcc dot gnu dot org 2009-07-27 11:10 ------- Oh, and on 64-bit arches and float or 32-bit arches and double there is another complication - the comparison has different mode size from the cond_expr for pos. For 32-bit pos and 64-bit double it could perhaps just do the computation in 64-bit integers (vector of 2 (resp. 4 for avx)), for the other case it would need to shuffle the max and compute the pos in 2 vectors, or e.g. the Fortran FE could hand in the common case, emit a likely look for the case where array size is smaller than 4GB, using 32-bit pos zero extended to 64-bits at the end.
-- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31067
