Bit-reverse every byte
b ~0,~1,~2,3,4,5,6,7,8
Turn vector of 8 u64 into 64 u8 (transpose 8x64)
b 6,7,8,0,1,2,3,4,5
Turn vector of 16 u32 into 32 u16 (transpose 16x32)
b 5,6,7,8,0,1,2,3,4
16x16 transpose in each 256-bit part
b 4,5,6,7,0,1,2,3,8
You can dump a full array of 512 indices into the query box (no prefix, comma-separated, spaces are allowed), or 9 indices-of-bits-of-indices (with the "b" prefix). If you use raw indices, the permutation must still correspond to a BPC permutation.
Put the least-significant index first, but the print-back shows the indices most-significant-first. Use source indices in raw mode, but destination indices in b mode. Yes, it's confusing. Sorry.
In b mode, the ~ operator is available to complement a bit. It does not complement the number after it, but the corresponding bit.
Consider this an alpha version for now. Expect bugs. You can open an issue.