Here are a few bits of news about the C frontend:
1. I have ported it to compile and run under linux. On our research
machines, it is available here:
Note that this C compiler will still generate "64-bit" LLVM files, thus
the code will probably not work right if run through the C backend,
then compiled on x86. For middle-end work (optimizers) this shouldn't
matter too much. I did some VERY brief touch testing of LLI on Linux,
and it ran hello world correctly, so you may have good luck with that.
If you run into problems with LLI on Linux not running a program
correctly, please let me know.
2. I have hacked the Makefile.common and Makefile.$(UNAME) to
automatically pick the right gcc frontend based on the architecture
being compiled for, so on the research machines, the stock
Makefile.config will work regardless of whether you are building on
Linux or Sparc. You can still override the choice of GCC frontend in
the Makefile.config file if you feel the urge.
3. This rebuild of GCC is actually a correct, clean, cross compiler. I am
in the process of updating the Sparc version to be built with the same
process. As part of this change, however, IT IS VERY IMPORTANT that
you no longer refer to $(GCCINSTALL)/bin/gcc. Instead, use
$(GCCINSTALL)/bin/llvm-gcc. You will probably have to update your
local .cshrc to point to the correct location (the LLVM build tree has
been updated, so you only have to update aliases and stuff).
4. As far as the sparc version of the GCC frontend goes, I am going to
leave what is in /home/vadve/lattner/cvs/gcc_install alone for the time
being. It is not correctly built, has problems, and is (among other
things) huge (217M vs 21M for the x86 build). As soon as my LLVM tree
is done building on solaris, I will make a new sparc version of the GCC
frontend available in /home/vadve/lattner/cvs/gcc_install_sparc, with
all of the nice characteristics of the x86 build.
In a few weeks, or whenever I think about it, I will nuke the old
/home/vadve/lattner/cvs/gcc_install directory, so you all should
transition over sometime in the near future. As long as this tree is
alive, gcc_install/bin/gcc will be a valid executable.
5. I have a list of some simple instructions for how to configure and
build the C frontend. No hacks required. If anyone is interested, I
can make these available on the web site. This is only important if we
plan to distribute the GCC front-end source.
6. It's possible we could do a binary distro of the x86 version of the GCC
frontend, which would probably make several 426 people happy. Whether
or not we can do this, and how we go about it needs to be discussed
before we do so however.
Anyway, enjoy the new x86 version, this should make life simpler.