Mike Hamburg

Thanks for the report. I will leave it open for now because I don't have a powerpc machine to try to diagnose this on.

Thanks for the patch! Merged.

Fix posix_memalign invocations

This makes sense though I will change the 06f628 to have a different solution since w=0 anyway, it's not about shifting it in the other direction. But for 7e6a38 do zero-length scalars cause failure? They ideally ought to work.

Thanks, sorry for the slow response.

Build with gcc

Fix windows build

Merged, sorry to be slow. I also changed the cast in bit_to_mask to be explicit, since otherwise you have to know the C promotion rules to verify correctness ... hope that doesn't break anything in Windows.