Re: Info on the mvs source volumes

"kerravon86" <kerravon86-/[email protected]> Sun, 24 Oct 2010 14:56:56 -0000
Newsgroups gmane.comp.emulators.turnkey-mvs
Message-ID <[email protected]>
--- In [email protected], Mike Schwab <Mike.A.Schwab@...> wrote:
>
> Well, on hercules-os380 there is a guy trying to 
> port his PDOS to the mainframe.
> http://tech.groups.yahoo.com/group/hercules-os380/

That's more-or-less a complete rewrite, in fact.

> He has assmassed (or recreated) all the source for 
> TK3 and brought up to date with a disassembler.

No, it's far from being up-to-date. However, it is
"nicely baselined", as in a single (latest) version
of each source file, and under version control,
and indeed, some modules have been brought up-to-date
via disassembly.

Currently on hold waiting for DISASM bug fixes and
enhancements required to complete the IFOX matching.

Also there are associated tools.

>> I was considering fairly radical action to improve 
>> performance.  If the entire source were available 
>> I was thinking of cross compiling to gcc and 

Presumably after first writing such a thing?

Pre-compilation of the MVS instructions may be
doable as a performance boost.

> After the info I have gotten on the state of 
> the source I am now leaning towards a maybe 
> wackier idea and modifing Hercules to be able 
> to switch between emulated 370 instructions 
> and segments of native linux gcc.

Certainly worth thinking about.

However! Almost all the computation time on an
MVS system is running applications, not OS
overhead.  ie you don't need the source code
for MVS. You need it for the applications.

What CPU-intensive applications are you running?
What language are they written in?

If the answer is "xyz high level language",
then perhaps you don't want to run the app
under MVS in the first place, and may be more
interested in Jason Winter's native file access.

Or perhaps interested in getting an EBCDIC
version of NetBSD/Linux operational.

> If I could get it to work I could recode some 
> of the source that is available and leave the 
> 370 stuff without source alone, piece by piece.

As Mika alluded to, I do have visions of taking
the C code from PDOS/370 and putting it into
MVS 3.8j. Or vice-versa. I haven't really spent
much time considering that, as I have been
focussed on getting PDOS to work in its own
right (which it is).

BFN.  Paul.