Re: Contributing to autoconf, automake, etc...

[Ben Bergen]

Here is a link to the development tools for the cell:
http://www.bsc.es/projects/deepcomputing/linuxoncell/cbexdev.html?S_TACT=105AGX16&S_CMP=DWPA

Overview of the processor:
http://www-128.ibm.com/developerworks/power/library/pa-cellperf

General link to simulator and other tools:
http://domino.research.ibm.com/comm/research_projects.nsf/pages/cellcompiler.resources.html

The development kit provides a cross-compiler with C and C++ compilers
for the PPE and SPEs, e.g., ppu-gcc ( C compiler for the PPE ),
spu-g++ ( C++ compiler for the SPEs).  These are really just ports of
the GNU C/C++ compilers to the PPE and SPEs.  However, they need to be
defined and tested separately from each other and preferably from the
host C/C++ compiler.  A cell simulator that runs on Fedora Core 4 is
also available.  I have had a look at erlang.m4 and c.m4 in the "lib"
directory of the autoconf cvs source.  It looks like it would be
fairly easy to just introduce new language defines like PPUC and SPUC,
etc...  Does this seem reasonable given that two different C/C++
compilers need to be configured at the same time?  If this is a good
approach, would I want to define separate pre-processing flags as
well, e.g., PPUCPPFLAGS, or should all C/C++ compilers share the same
CPPFLAGS?

Ben

--
  Benjamin Karl Bergen
  CCS-2 Continuum Dynamics
  Los Alamos National Laboratory