Re: GCC Projects?

Re: GCC Projects?

[Jeffrey A Law]

  > As for bytecode generation, there are serious maintenance problems with
  > what has been done so far.
I don't think there's any question the bytecode stuff is horribly
broken and a maintenance nightmare.

I wouldn't lose sleep if it just disappeared one day.

jeff

Re: GCC Projects?

[Jim Wilson]

	I see that there are two main unfinished projects in the egcs
	source tree: pre-compiled include files and bytecode generation.
	Is anyone working on one of these?

Per Bothner already commented on pre-compiled header files, and I agree
with his comments that we should get cpplib working first, and that the
current implementation of pre-compiled header files does not seem to be
the right approach.

As for bytecode generation, there are serious maintenance problems with
what has been done so far.  The current support was added in between and
along side the tree->RTL conversions.  The result is that there are almost
a hundred different places that have
	if (output_bytecode) { emit bytecodes } else { emit RTL }
The bytecode support was working at one time, but it now badly broken because
this is too difficult to maintain.  I think the implementation needs to be
completely rethought.  In the long term, the best solution might be to
clean up the front-end/middle-end interface.  If the front-end ends convert
an entire function to trees before calling the middle-end, then we only need
one place in the middle-end that checks for whether we are generating bytecodes
or RTL.  This will allow us to separate the bytecode middle-end from the RTL
middle-end, which should make it easier to maintain both of them.

Jim

Re: GCC Projects?

[Jim Wilson]

	I'd be happy to provide a patch to make it disappear. Do you want one?

I already happen to have a bytecode-removal patch.

Jim

Re: GCC Projects?

[Jeffrey A Law]

  In message <Pine.SOL.3.90.970818102506.302I-100000@hutch.informatik.rwth-aach
en.de>you write:
  > > 
  > > I wouldn't lose sleep if it just disappeared one day.
  > 
  > I'd be happy to provide a patch to make it disappear. Do you want one?
Not quite yet.

Ripping out the bytecode will make gcc2 -> egcs merging more difficult in
the short term, so unless there's a noteworthy benefit I'd like to avoid
gratutious changes right now.

Looking a little into the future, I see it being ripped out unless someone
objects.

jeff

Re: GCC Projects?

[Bernd Schmidt]

> 
> I wouldn't lose sleep if it just disappeared one day.

I'd be happy to provide a patch to make it disappear. Do you want one?

Bernd

Re: GCC Projects?

[Joel Sherrill]

On Sun, 17 Aug 1997, Jeffrey A Law wrote:

>   > I see that there are two main unfinished projects in the egcs
>   > source tree: pre-compiled include files and bytecode generation.
>   > Is anyone working on one of these?
> Nobody is working on these as far as I know.  Lots of other stuff
> is probably more important -- cpplib, lots of optimization work,
> fold in ada & pascal front ends, packaging & release issues, docs,
> etc etc.

I would like to put a plug in for architectural simulators here.  There
are a number of targets --especially embedded ones -- which are difficult
and/or time-consuming to test.  Of the RTEMS configurations (about 8
different ones), I can only do automated tests on 2 of them (powerpc and
sparc).  The others I can only verify that they can successfully build a
full toolchain and support libraries.  Periodically, I run code on actual
target boards.  

Long term, if the testing of egcs (and other support libraries) is to be
full and complete, we really need to find/develop good simulators for a
handful of cpus. 

--joel
Joel Sherrill                    Sr. Computer Scientist
joel@OARcorp.com                 On-Line Applications Research
Ask me about RTEMS: a free RTOS  Huntsville AL 35805
   Support Available             (205) 722-9985

Re: GCC Projects?

[Jeffrey A Law]

  In message <k2hgcovl4a.fsf@zero.aec.at>you write:
  > Hi,
  > 
  > I see that there are two main unfinished projects in the egcs
  > source tree: pre-compiled include files and bytecode generation.
  > Is anyone working on one of these?
Nobody is working on these as far as I know.  Lots of other stuff
is probably more important -- cpplib, lots of optimization work,
fold in ada & pascal front ends, packaging & release issues, docs,
etc etc.

jeff

Re: GCC Projects?

[Per Bothner]

> I see that there are two main unfinished projects in the egcs
> source tree: pre-compiled include files and bytecode generation.
> Is anyone working on one of these?

As the author of cpplib, I would rather we first finish cpplib
as the default cpp (see my previous message), before we even start
on pre-compiled header files.  RMS has a model for pre-compiled header
files which would not mesh as well with cpplib, and it is unclear
how much of a win it would be anyway.

Bytecode generation is another of RMS's pet projects that no-one
has shown much interst in in years, and it is arguable how useful
it is.  (For one: it does not support C++.)

	--Per Bothner
Cygnus Solutions     bothner@cygnus.com     http://www.cygnus.com/~bothner

GCC Projects?

[Andi Kleen]

Hi,

I see that there are two main unfinished projects in the egcs
source tree: pre-compiled include files and bytecode generation.
Is anyone working on one of these?

-Andi