On Wednesday, March 10, 2021 at 9:51:28 AM UTC-8, Robert Boyer wrote:
> I had something far simpler than a C compiler to do and just for fun
> looked into LLVM when it first became "a thing" (IE when Apple went that
> way), I ended up just doing what I "usually do" = LISP or Haskell type
> thing/semi DSL for the translation since the input syntax was under my
> own control but ever since then I have been curious as LLVM looks like
> it's a very powerful tool.

There's been an effort to port GCC to TOPS-20 in the past. Got somewhat far along too, but it's not being worked on now [not crediting the author in case he wants to remain anonymous ;), but it looked like a lot of great thought was put into the effort]. I recently thought about looking at LLVM as a potential target for porting, but not sure how memory intensive it (vs GCC) is. I suppose it could start out life as a sort of cross compiler, running external to TOPS-20, and then cross-compile itself. Lots of issues would have to be solved, like figuring out how to provide access (if needed) to the include files that may reside on the 20, and how to write the .rel files back to the 20. To make things simpler to start, one might require that the compiler user have the includes present in the local file system, and that the outputs (.rel and maybe other map or listing type files) go back to the local file system. And then build some kind of automated file transfer/virtual file system (I know there is some NFS support for Twenex, but it's been a little wonky when I half-heartedly tried it).

For extra fun (easier setup and use), perhaps the compiler and necessary related tools could be packaged in a Docker image, so it's trivial and error free to get something up and running. [also toying with the idea of packaging the panda distribution in a docker image, including a way to externally to TOPS-20 define the networking info, and have that information loaded into the appropriate SYSTEM: and other files at system startup, and customizing the klt20.ini to automagically include the proper network device definition.]

Steve

sbyrne007 wrote:
> There's been an effort to port GCC to TOPS-20 in the past. Got
> somewhat far along too, but it's not being worked on now

I was contracted by a Redmond-based company to do the PDP-10 backend.
It seemed to work pretty well when I handed it over. It was not running
on TOPS-20 at that point, but I don't know what happened after that.

> Lots of issues would have to be solved, like figuring out how to
> provide access (if needed) to the include files that may reside on the
> 20, and how to write the .rel files back to the 20.

I used a set of scripts which emulated a cross assembler, linker, etc by
transparently FTPing files to TOPS-20 (running locally on KLH10),
processing them there, and FTPing the result back to my Linux host.

https://github.com/larsbrinkhoff/pdp10-small-libc/tree/master/bin

On Saturday, September 4, 2021 at 11:28:47 AM UTC-7, lars...@nocrew.org wrote:
> sbyrne007 wrote:
> > There's been an effort to port GCC to TOPS-20 in the past. Got
> > somewhat far along too, but it's not being worked on now
> I was contracted by a Redmond-based company to do the PDP-10 backend.
> It seemed to work pretty well when I handed it over. It was not running
> on TOPS-20 at that point, but I don't know what happened after that.
> > Lots of issues would have to be solved, like figuring out how to
> > provide access (if needed) to the include files that may reside on the
> > 20, and how to write the .rel files back to the 20.
> I used a set of scripts which emulated a cross assembler, linker, etc by
> transparently FTPing files to TOPS-20 (running locally on KLH10),
> processing them there, and FTPing the result back to my Linux host.
>
> https://github.com/larsbrinkhoff/pdp10-small-libc/tree/master/bin

Wow -- I had no idea these were available (ftp-do, telnet-do)! Thanks for the tip Lars!

Steve

Lars, 2-3 years ago I continued your work on that ported gcc, adding several back-end optimizations, and getting a good bit of the gcc test suite working. The scripts are still used.

To elaborate on Lars' explanation, the compiler, running on Linux, generates assembly language. The assembly files are copied to TOPS-20 and assembled and linked there.

-+- Sid

On Sunday, September 5, 2021 at 9:40:05 AM UTC-7, srmax...@gmail.com wrote:
> Lars, 2-3 years ago I continued your work on that ported gcc, adding several back-end optimizations, and getting a good bit of the gcc test suite working. The scripts are still used.
>
> To elaborate on Lars' explanation, the compiler, running on Linux, generates assembly language. The assembly files are copied to TOPS-20 and assembled and linked there.
>
> -+- Sid

Hey Sid -- this is great news! Do you have a GH repo where your work is kept that you can share?

Thanks,
Steve

> Hey Sid -- this is great news! Do you have a GH repo where your work is kept that you can share?

I'm sorry, no. The work was done under contract, and the company holds the rights.

-+- Sid

On Sunday, September 5, 2021 at 3:02:00 PM UTC-7, srmax...@gmail.com wrote:
> > Hey Sid -- this is great news! Do you have a GH repo where your work is kept that you can share?
> I'm sorry, no. The work was done under contract, and the company holds the rights.
>
> -+- Sid

Ok -- thanks for the quick reply! Good to know it is doable (existence proof).

Steve

Sid Maxwell <srmaxwell3@gmail.com> writes:
> I'm sorry, no. The work was done under contract, and the company
> holds the rights.

Um ok, but it's GCC which is GPL, so I hope the company releases
whatever there is of the port.

Paul Rubin <no.email@nospam.invalid> writes:
>Sid Maxwell <srmaxwell3@gmail.com> writes:
>> I'm sorry, no. The work was done under contract, and the company
>> holds the rights.
>
>Um ok, but it's GCC which is GPL, so I hope the company releases
>whatever there is of the port.

If they're using it internally, they have no obligation to release
the port to anyone.

If they provide the gcc binary to anyone, then the GPL kicks in.