so my software started crashing, eventually pinpointed ippsThreshold_GT_64f - unlike normal comparison or ippsThreshold_GT_32f, it doesn't work with NANs. It just keeps them NAN.
it's customer's responsibility to provide correct FP input - result of compare operation with NAN according to IEEE is undefined.
Aaaah, ok then, thanks for the answer. It was just quite surprising since in 32-bit mode it works as expected.
For more complete information about compiler optimizations, see our Optimization Notice.