[Top][All Lists]

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

Re: [Patch] PVM low-level cleanup and optimization, good version [Was: R

From: Jose E. Marchesi
Subject: Re: [Patch] PVM low-level cleanup and optimization, good version [Was: Re: [Patch, first version] PVM low-level cleanup and optimization]
Date: Sun, 01 Dec 2019 13:40:17 +0100
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/26.1 (gnu/linux)

    I agree that the general versions of slide and bulge are not very
    Forthish, and the parameters are distracting.
    What about only adopting the one common version of each?  "slide" could
    mean "slide 1,2" and "bulge" could mean "bulge 1".  Those are very
    intuitive to me.
    slide ( a b c -- b c )
    bulge ( a b -- a a b )
    In fact I think I only used these versions in my change set.

That would be better, but still I won't be adopting these instructions
yet.  It's good that jitter provides them though.
    About GCD I really like my version better, but okay.
    JITTER_ARGUn is like JITTER_ARGNn, but the type is unsigned.
    In my changeset the purpose of JITTER_ARGUn is catching buts more
    easily.  With an unsigned argument if you pass a negative by mistake the
    operation will fail catastrophically instead of executing a change which
    leads to a subtly incorrect stack.  No performance difference.

Thats much better yes.
    I gave specialized arguments to revn, for performance since in practice
    you always reverse to a shallow depth, and in those cases the code is
    very fast.  revn 3 is not any slower than swap.


    I am waiting for your approval before porting and preparing a final


reply via email to

[Prev in Thread] Current Thread [Next in Thread]