wip-rtl disassembler

wip-rtl disassembler

[Andy Wingo]

A brief note to say that (system vm disassembler) in wip-rtl now has a
functioning "disassemble-program" that finds function bounds using the
ELF symbol table.  It's not great yet; would like to work on it a bit
more this evening, but it's better than what we had ;)

Andy
-- 
http://wingolog.org/

Re: wip-rtl disassembler

[Noah Lavine]

[-- Attachment #1: Type: text/plain, Size: 449 bytes --]

That's great! Thanks a lot for doing so much RTL stuff.

Noah


On Thu, May 2, 2013 at 4:11 AM, Andy Wingo <wingo@pobox.com> wrote:

> A brief note to say that (system vm disassembler) in wip-rtl now has a
> functioning "disassemble-program" that finds function bounds using the
> ELF symbol table.  It's not great yet; would like to work on it a bit
> more this evening, but it's better than what we had ;)
>
> Andy
> --
> http://wingolog.org/
>
>

[-- Attachment #2: Type: text/html, Size: 913 bytes --]

Re: wip-rtl disassembler

[Andy Wingo]

On Thu 02 May 2013 10:11, Andy Wingo <wingo@pobox.com> writes:

> A brief note to say that (system vm disassembler) in wip-rtl now has a
> functioning "disassemble-program" that finds function bounds using the
> ELF symbol table.  It's not great yet; would like to work on it a bit
> more this evening, but it's better than what we had ;)

At this point the disassembler is working much better, and works at the
REPL now via ,x.  Also, procedure-name now works.

I'm now moving to implement source line information and other metadata
via .debug_info / .debug_line / etc sections.

I think debugging things will be ready in a few weeks.  Noah, how's the
compiler coming?  I think we could reasonably be looking at compiling
everything with RTL by July or August if all goes well.

Andy
-- 
http://wingolog.org/

Re: wip-rtl disassembler

[Noah Lavine]

[-- Attachment #1: Type: text/plain, Size: 1310 bytes --]

On Sun, May 5, 2013 at 4:17 PM, Andy Wingo <wingo@pobox.com> wrote:

> At this point the disassembler is working much better, and works at the
> REPL now via ,x.  Also, procedure-name now works.
>
> I'm now moving to implement source line information and other metadata
> via .debug_info / .debug_line / etc sections.
>

Wow, that's great!


> I think debugging things will be ready in a few weeks.  Noah, how's the
> compiler coming?  I think we could reasonably be looking at compiling
> everything with RTL by July or August if all goes well.
>

Not as quickly as I'd like, but that's always going to be true. :-)

The status is still what it was in my last email - almost all language
features are done, but I would like to check edge cases much more
thoroughly than I have. What I have been working on most recently is a
patch to the test suite to add a command-line option to choose a different
compiler. I realized that my strategy of writing a separate test suite for
CPS will never get close to our 35,000+ Guile tests, and also can't be
reused the next time we want to test a new compiler. So my goal is to let
the test suite use the CPS compiler, and then just systematically fix any
bugs I find until things pass.

If anyone else wants to work on this, they are more than welcome to jump in.

Noah

[-- Attachment #2: Type: text/html, Size: 1952 bytes --]

Re: wip-rtl disassembler

[Ludovic Courtès]

Noah Lavine <noah.b.lavine@gmail.com> skribis:

> I realized that my strategy of writing a separate test suite for
> CPS will never get close to our 35,000+ Guile tests, and also can't be
> reused the next time we want to test a new compiler.

Ouch, indeed.

> So my goal is to let the test suite use the CPS compiler, and then
> just systematically fix any bugs I find until things pass.

Note that test cases are actually eval’d, except those that use the
‘c&e’ variants.

Thanks for working on this!

Ludo’.