trgsmp.mac and translated files



>>>>> "Robert" == Robert Dodier  writes:

    Robert> the difficulty with the translation code is that it (at
    Robert> present) amounts to a reimplementation of the interpreter,
    Robert> which is now out of date.

Ok.  That seems to be a lot of work, just to get a translator that
isn't any faster or better than just running original mac code.

    Robert> for other functions, the translator emits something
    Robert> generic like MFUNCALL or MEVAL; in that case
    Robert> the effect is the same as the interpreter, but it is no faster.

I think that's fair.  If the translator doesn't know, you lose.

    Robert> a data point here -- a few months ago i spent a couple of
    Robert> hours trying to translate the test suite.
    Robert> i didn't get far (i couldn't get rtest1.mac translated).

It currently dies translating
t[n](x):=ratexpand(2*x*t[n-1](x)-t[n-2](x)).

    Robert> bug reports (there are about 10 of them) are all closed,
    Robert> interestingly enough.

Hmm.  Searching for a summary keyword of trans gives about 10 open
bugs.  I have a fix for a couple of them (because they're easy), but
perhaps it's not worth the effort, and any such effort is better spent
on a new translator or other bugs.

Ray