Agner`s CPU blog

May I suggest a couple of instructions:
-Bit reverse: Reverse all the bits of a register.

LUT base logical ops:

These instructions implement arbitrary functions for boolean vectors. However, the function is applied only to the least significant bit of each vector element. There are two reasons for this limitation. (1) This is the way boolean vectors are represented in ForwardCom. The boolean value is stored in the least significant bit. The remaining bits are option bits, which can be propagated to the result and used for specifying options to e.g. a floating point instruction using the LUT result as a mask. (2) The 3-input LUT can be implemented in hardware by using an 8-bit barrel shifter where the 8 bits of the LUT are shifted right by the count defined by the three input bits. The hardware allready has an 8-bit barrel shifter for each 8-bit vector element so that the implementation is cheap. A LUT function that applies to all bits of a vector will require one 8-bit barrel shifter for each vector bit, which would be more expensive in terms of silicon space.

I am not sure I understand what you want to do with the msb, lsb inputs?

MultiCarryOps:
Rd ops Rs under Rc (Carry mask): Ops can be +, - >>, <<, rot The carry mask work as follow
Each bit set in the Rc indicate the msb of a sub bit string. i.e. the carry does not propagate to the left bit of a set bit in the carry mask.

BTW, I have made the move_bits instruction for efficient insertion and extraction of bit fields.