Fix for HostPrefixSum
In this commit we moved from _mm_extract_epi32
to _mm_storeu_si32
. This was done because SSE2 is available in all x86_64 processors, whereas SSE4 is not.
When compiling with gcc9 (Ubuntu OS) the _mm_storeu_si32
is not recognized, see here. HostPrefixSum.cpp
has been changed to use _mm_store_ss
instead.
FYI: @dcampora
Edited by Giovanni Bassi