Re: i860 code generation

jason@convex.com (Jason L. Eckhardt) writes:
>[re generating l floating point code on the i860]
>I wish now to have the cg generate pipelined versions of the instructions.
>Should I do this in the code selection phase or is it better to select
>scalar instruction as I've been doing and transform these into pipelined
>versions in a later phase?


I can't really say which approach would be best, but I strongly suspect
that only the latter approach will be within reason as far as implementation
complexity goes.


(And by the way, don't forget about the interesting way the piplined FP
instructions work... you have to have another "dummy" FP instruction two
cycles down from the instruction where you start the operation, and that
"dummy" instruction is the one which "catches" the results of the earlier
operation as those results fall out of the end of the FP pipe.)
--
-- Ronald F. Guilmette ------------------------------------------------------
------ domain address: rfg@netcom.com ---------------------------------------
------ uucp address: ...!uunet!netcom.com!rfg -------------------------------
--

Return to the comp.compilers page.
Search the comp.compilers archives again.