LHC@home

A while back I became curious about the difference between the Windows 32 bit SSE3 and PNI programs for sixtrack so I checksummed them and found out that they have the same hash:

sixtrack_win32_4517_sse3.exe

md5-----2465e81d861214f8cd6e3f317dc39ff6
SHA1----a8b4533ab58791f3d64b70239e8ac70d0ee31da3
SHA256-45a24f0555a65576d3134def8adc952502af072676a51f4add93a683a20ac8fb
CRC32---ed8516f6

sixtrack_win32_4517_pni.exe

md5-----2465e81d861214f8cd6e3f317dc39ff6
SHA1----a8b4533ab58791f3d64b70239e8ac70d0ee31da3
SHA256-45a24f0555a65576d3134def8adc952502af072676a51f4add93a683a20ac8fb
CRC32---ed8516f6

I never knew Windows had a 32 and 64 bit version of the program on this project so out of curiosity I checksummed the 32 bit and 64 bit programs and discovered that they too have exactly the same checksum.

I know that just like different SIMD instructions (like SSE2 vs. AVX) don't always equal faster or more efficient problem solving, bitness doesn't either. So is there a difference between the functionality of the 32 bit and 64 bit Windows programs?

'PNI' stands for Prescott New Instructions, and was Intel's trade name for what became SSE3 (http://en.wikipedia.org/wiki/SSE3). So the instruction sets are the same, and the same application has indeed been deployed twice so that BOINC's feature detection works correctly on both Intel and AMD CPUs.

I don't know the thinking on why a separate 64-bit deployment was necessary too.

As stated before the executables are identical.
It is thought we had to have a 64-bit named executable for
64-bit system. AVS coming soon.

---