This llvm should have newgvn which allows better auto vectorization. I'm curious to see if panic=abort allows some of the nbody implementations at benchmark gam

This llvm should have newgvn which allows better auto vectorization. I'm curious to see if panic=abort allows some of the nbody implementations at benchmark game to autovec. Might not have to wait on SIMD to beat c++. The fastest nobody entry is FORTRAN and it has no intrinsics, just autovec.