On Tue, 09 Mar 2004 14:29:56 -0600, arargh403NOSPAM@[EMAIL PROTECTED]
pushed the wrong button before I was finished. Plus, I wanted
to think a little more before finishing.
>On Tue, 09 Mar 2004 19:16:56 +1100, Johnathan
><zork_666@[EMAIL PROTECTED]
> wrote:
>
<snip>
>
>>I think I might even go back to my original idea of Modula-2/Pascal
>>subsets, especially now that you mention there's little scope for
>>optimisation of any kind (it's certainly the case my original will use
>>calls to the standard C library).
If it weren't for Pascal permitting nested SUBs/FUNCTIONs, I could
write a front end similar to BCET and use the rest of the system
pretty much as is. I don't think that Pascal would be all that
different (internally) from Basic. However, I don't have a good
grammar for Pascal, nor all that much interest.
>The BCET code generator emits pretty simple code, for:
> C = A + B
>you get:
> lod A
> lod B
> add
> sto C
>which BOPT changes to:
> lod A
> add B
> sto C
This should be closer to:
The BCET code generator emits pretty simple psedo code, for:
C = A + B
you get:
lod eax,A
lod ebx,B
add eax,ebx
sto eax,C
which BOPT changes to:
lod eax,A
add eax,B
sto eax,C
With different regs, probably.
There are several places where BCET could optimize more. Common
sub-expression, for one:
C = (A + B) * (A + B)
will calculate (A + B) twice. Hopefully, neither A or B are function
calls with side effects. :-)
Constant expression folding:
C = ASC("A")
actually generates the code to do it, but what should happen is:
mov C, 65
Turn SELECT CASE's into a jump table, in some cases.
But these things require a pass over the expression tree between BCET
& BGEN.
>>
>>> There isn't too much that can be done in terms of code optimaztion.
>>> Basic generally winds up doing so many library calls, that there isn't
>>> much left to optimize. Take a look at the intermediate files of a
>>> BCET compile.
Or a QuickBasic/PDS/VBDOS listing.
>>
>>You've worked on BCET since 1992? What was that paper you mentioned in
>>the FAQ which gave you a hint about code generation?
>I believe it was this one:
>http://www.cs.princeton.edu/software/iburg/iburg.pdf
>and the .PS files in this one:
>ftp://ftp.cs.princeton.edu/pub/packages/iburg.zip
>
>>
>>> Gee, that's where I was, 12 years ago.
>>
>>12 years... hmmm. Long road ahead. :)
>It was a part time project most of the time. If I had had any formal
>schooling on the subject, it would gone faster.
--
Arargh403 at [drop the 'http://www.'
from ->] http://www.arargh.com
BCET Basic Compiler Page: http://www.arargh.com/basic/index.html
To reply by email, remove the garbage from the reply address.
|
