feat: replace wcp with assembly implementation

[DavePearce]

This replaces the WCP module with an assembly implementation. The assembly implementation implements identical functionality, but in a rather more compact form.

NOTE: should not be merged until corresponding linea-tracer PR is ready, and updated release of go-corset is available.

[amkCha]

What happens if ARGUMENT_1 is not zero ?

[DavePearce]

Then we fall through to exit_0 and we have RESULT=0. An interesting question here is whether we should do anything with ARGUMENT_2. In fact, in the current wcp implementation the ISZERO instruction is handled simply as an equality ARGUMENT_1 == ARGUMENT_2 under the assumption that ARGUMENT_2==0 whenever wcp(ISZERO,...) is called.

[amkCha]

Assembly just knows how to split due to the types ? That's nice

[DavePearce]

Yes, that's the field agnosticity part. It uses the stuff from my report to make it work, adding carry lines and so forth.

Sorry, I misunderstood what you were saying. Yes, it can rely on the width of a register as the type of all registers is proven. Therefore, yeah, it can figure out how to split ARGUMENT_1 here based on the widths. In fact, if you provide an incorrect width for the left-hand side then it will complain. Also, it reads with the most significant bits on the left. So, in the above, arg1sign gets bit 31 of ARGUMENT_1 and arg1bits gets bits 30 .. 0.