Re: [Tack-devel] ACK compiles on NetBSD-macppc, sort of...

[Gregory T. (t. K. <gt...@di...>]

At 12:14 PM -0400 7/18/06, David Given wrote:
>Essential reading includes the em and ncg documents --- em describes the
>intermediate-code format, and ncg is the manual for the compiler
>generator. I strongly suspect it's worth getting fluent in em assembly
>first.

I've downloaded the PDFs and I will review them.  Thanks for the direction! =
 From a "stack" based virtual machine point of view, this doesn't sound all=
 that much different than Forth, which I find to have a lot of virtues. =20

>I'll admit that every time I look at a ncg input file I think, "Gosh,
>that looks complicated," and then find something else to do... ncg
>appears to be amazingly sophisticated and capable, and I'm sure it makes
>writing a decent code generator an absolute breeze, but it also
>possesses the ability to scare me deeply.

Has anyone done any code analysis on the output?  How efficient/fast/small=
 is the output? =20

How well does the debugger work?  Can it take EM and generate C or Pascal so=
urce?

>Er, that probably wasn't very reassuring!

I've come to understand that Andrew Tanenbaum knows what he's doing :-)  I=
 was pretty blown away to find not only had he written a microkernel OS, but=
 a toolchain to build it as well.  With Tanenbaum releasing everything under=
 a BSD license, it's pretty easy for me to choose to study his efforts when=
 trying to figure out where I need to go.

I think there's a lot of interest in BSD licensed open source compilers.  It=
 might just be a matter of getting the ACK tree close enough to get interest=
 by additional people.  I'm pretty convinced that if I knew what I was doing=
 the tree would compile fine on NetBSD.

>If you do tackle a code generator (remember you'll need to do an
>assembler too...) I'd probably suggest doing minimal work on integrating
>it into the build system for the time being --- not only is there a
>reasonable chance you'll stay sane that way, but it is, hopefully, being
>made obsolete.

I appreciate the advice.  My thinking was that if I managed to write a=
 back-end table for PowerPC I'd probably be able to roll with the changes to=
 the tree. =20

Perhaps my overview is too simplistic, but I was under the impression that=
 the programming languages are "compiled" into an intermediate step by the=
 "front-end," the intermediate step was the EM virtual machine, and the=
 "back-end" was what took the intermediate step and generated object code=
 for the particular architecture. I get this from reading http://www.cs.vu.n=
l/ack/=20

If this is true, then once a back-end table has been written all the=
 languages supported by the front-ends will also be supported.  Personally=
 I'm not nearly as concerned about compiling times as I am in the object=
 code generated, so this intermediate step (EM) isn't a drawback and I=
 actually think it is a great solution.  It also leaves open the possibility=
 that patches to source code could be distributed in EM, creating something=
 very close to binary patches.  Not everyone wants to have to build from=
 source, and there are probably more than a few sysadmins that really like=
 to make sure source doesn't reside on the host.

Now, if my understanding is flawed, I definately would appreciate some=
 pointers towards documentation that can clear up my confusion :-)

thanks,
tim


Gregory T. (tim) Kelly
Owner
Dialectronics.com

P.O. Box 606
Newberry, SC 29108

"Anything war can do, peace can do better."  -- Bishop Desmond Tutu