https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116615
--- Comment #8 from Palmer Dabbelt <palmer at gcc dot gnu.org> --- (In reply to Andrew Pinski from comment #7) > History on LOGICAL_OP_NON_SHORT_CIRCUIT being able to defined differently > from BRANCH_COST. It was originally added for powerpc (2002/2003ish) which > had expensive cset at the time; and there was no possibility to use the > crand/cror (etc.) instructions. Though now with ccmp is this is no longer > true but powerpc has not implemented that yet. > > Anyways back to RISCV since defining it to 0 might seem wrong since cset (at > least for integers) is cheap. OK, sounds like this is just the wrong way to go on RISC-V then? I just sent a patch, maybe someone will remember why things are this way. Thanks for the help here, I didn't know any of this ;)
