[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: OZ PCL performance lossage

    I actually did exchange mail with Kantrowicz...

OK, good.  At least he got an ACK.  It might be good if the rest of us saw
at least the first message in such exchanges, though you may want to take
it private if it gets intense.

    One factor here is the sparc/mach page translation lossage.

It would be interesting to run things on SunOS to see how much of a factor
this is.  I guess the current story on this paging problem is that Berman
is working on some changes that should help, right?

    Python is definitely slower than the Allegro compiler, and always will be.

No problem with compiler speed, it was the terrible runtime speed and size
that I was worried about.
    From MK's report, it seemed that running his application under CMU CL/PCL
    involved continual compilation.  That is more of a mystery.

Yeah, that's what we need to understand.  Maybe our PCL is broken in a way
that could be fixed fairly easily.  If he's not creating new code at
runtime, the compiler should be out of the loop.

-- Scott