From mboxrd@z Thu Jan  1 00:00:00 1970
X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me
X-Spam-Level: 
X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00 autolearn=ham
	autolearn_force=no version=3.4.4
X-Google-Thread: a07f3367d7,577ecbb6c5a53a1f
X-Google-Attributes: gida07f3367d7,public,usenet
X-Google-NewGroupId: yes
X-Google-Language: ENGLISH,ASCII-7-bit
Received: by 10.180.106.102 with SMTP id gt6mr2283863wib.0.1349942506747;
        Thu, 11 Oct 2012 01:01:46 -0700 (PDT)
Path: 
 q11ni134269320wiw.1!nntp.google.com!feeder2.cambriumusenet.nl!feed.tweaknews.nl!216.40.29.245.MISMATCH!novia!border4.nntp.dca.giganews.com!border2.nntp.dca.giganews.com!border3.nntp.dca.giganews.com!border1.nntp.dca.giganews.com!nntp.giganews.com!news.glorb.com!eternal-september.org!feeder.eternal-september.org!mx04.eternal-september.org!.POSTED!not-for-mail
From: Ludovic Brenta <ludovic@ludovic-brenta.org>
Newsgroups: comp.lang.ada
Subject: LLVM and GCC (was: Forking old GMGPL libs ?)
Date: Sat, 06 Oct 2012 13:27:48 +0200
Organization: A noiseless patient Spider
Message-ID: <87fw5rhmbv.fsf_-_@ludovic-brenta.org>
References: <5de9a47c-6594-4653-a083-38f90312e70e@googlegroups.com>
	<k4ouk2$731$1@dont-email.me>
Mime-Version: 1.0
Injection-Info: mx04.eternal-september.org;
 posting-host="bfc8171067f57009ca3c3b958f065a0f";
	logging-data="20374"; mail-complaints-to="abuse@eternal-september.org";
	posting-account="U2FsdGVkX1+XuoPoiawsqOl2WsLN+C2A"
User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/23.4 (gnu/linux)
Cancel-Lock: sha1:HsVdo6sprcn+AQ6mjcKalIGidNw=
	sha1:yV7HgFJfxq3CHiZMpn+IoCbDGxc=
Content-Type: text/plain; charset=us-ascii
Date: 2012-10-06T13:27:48+02:00
List-Id: <comp.lang.ada>

Brian Drummond writes on comp.lang.ada:
> I am currently trying to get GHDL to build with gcc 4.7; it last built
> successfully with gcc4.3.4 or so. Like Gnat, it is written mostly in
> Ada, with a thin layer of C between it and the GCC internals. However
> it is about an order of magnitude smaller than Gnat.

I'm glad you're doing this; the package ghdl in Debian unstable is stale
and is the last one that still requires gnat-4.4; all other Ada packages
have migrated to gnat-4.6 long ago.

> As to targetting Clang versus LLVM, I haven't looked at Clang. But
> LLVM looked to me like a virtual machine (asm-like) rather than a
> syntax tree (AST like Generic or Gimple in gcc terms). Which to me,
> would raise the question of whether the LLVM back end can do the
> optimisations you want.

LLVM is both the virtual machine and an intermediate representation[1].
Contrary to GCC, LLVM's IR is serializable: one can save it to disk and
re-load it into the virtual machine for ahead-of-time or just-in-time
compilation.  LLVM is also a good optimizer that works on that IR; just
not yet as good as GCC.  The IR is, I think, better designed than that
of GCC because it was designed from the onset as language-neutral and
very rich.  The IR of GCC was initially designed for C only and was
later extended in ad hoc ways for other languages.

[1] http://llvm.org/docs/LangRef.html

A hypothetical Ada front-end for LLVM needs only emit this
well-specified IR.  Duncan Sands once made such an Ada front-end based
on the Ada front-end in GCC 4.2.  I'm not sure how this worked at the
time; I suppose Duncan replaced the GCC interface you mentioned with an
LLVM interface.  This has now been superseded by DragonEgg, whereby the
GCC Ada front-end feeds its AST into LLVM, which then translates it to
its IR, runs its optimizers and emits the final assembly.

-- 
Ludovic Brenta.