Hello everyone,

I disassembled Roberts examples and they are available at

<ftp://ftp.futurebasic.org/FutureBasic/dropbox/OVERHEAD_Disassembled.sea>


Now before it begins, I can tell you that while I understand how to
disassemble,
my knowledge of PPC assembly language is still very shallow.  I *may* be
able to answer
some topical questions but any question that require any real depth of
knowledge would 
be deferred to someone with a true understanding.

Just in case you are wondering how I can disassemble but not be fluent in
assembly, its
pretty much like learning a spoken language.  First you learn the rules of
the language and
then you practice reading, writing and conversing in it.  At this stage,
I've assembled
(no pun intended) the rules of the language and have written a program that
will interpret
it.  You know that you really have a grasp on a language when you can write
and speak it.
I'm just not at that level at this time.

Robert,I havent downloaded the file yet but I know it will enrich us.


Thanks,


W.

Assembly as a second language.



-----Original Message-----
From: Robert Purves [mailto:robert.purves@...]
Sent: Thursday, May 03, 2001 9:35 PM
To: futurebasic@...
Subject: Re: [FB] Re: Sylvain suggests fast subroutines



> > I feel like I'm asking a silly question but if the epilog is
> > not performed and the return is being performed early, wouldn't
> > references to variables on the stack be incorrect?  It would
> > seems to me that either a crash would occur because incorrect
> > variable offset references or stack buildup.
>
>The "l" in the PPC ops "bl" an "blr" refers to something called a
>link register which is not the stack pointer I fear.
>
>But yours is by no means a silly question.  It's surely related to
>the tragic fragility of Robert's method.
>
>What would be most helpful would be to have a clear PPC decompilation of
>FN <name> USING <address> in the parameterless case. Could you
>provide that?  And the decompilation of END FN too.



I have placed in the FB.org dropbox a small file that explores various
issues relating to FN overhead and the "tragic fragility" of the strange
fast method. This may help answer some of the questions that have been
posted on the topic.

<ftp://ftp.futurebasic.org/FutureBasic/dropbox/FN_Overhead_studies.sit>


Robert P.



--
To unsubscribe, send ANY message to <futurebasic-unsubscribe@...>