While you're waiting for the 2008-10 snapshot, here's something new you can play with: SSE2 support for CMUCL. You can grab the (large) binary for Linux from
<http://common-lisp.net/project/cmucl/downloads/experimental/cmucl-2008-10-06-sse2-x86-linux.tar.bz2> <http://common-lisp.net/project/cmucl/downloads/experimental/cmucl-2008-10-06-sse2-x86-linux.extra.tar.bz2> This includes support for both x87 and sse2. (This idea was suggested by Carl.) The C runtime will try to determine if you machine supports sse2 and loads the sse2 core if it does. Otherwise, it loads the x87 core. There's a new command-line switch to allow you to select between the x87 or sse2 core if you want to force a certain core. "-fpu x87" selects the x87 core, and "-fpu sse2" selects the sse2 core. I've run some numerical tests and the sse2 core works just fine. I do not know if the sse2 core is faster or slower than the x87 core, but it does fix some issues with double-float support for x87. (See trac ticket #15, which now returns the correct value.) If you have a machine that does NOT support sse2, I would appreciate it if you could try running this to see if the detection works. There are some rough edges, but if all goes well, SSE2 support should be available for the next snapshot. A related question: Is there anyone who doesn't have sse2? If everyone has sse2, we may eventually drop support for x87. (Except, perhaps for long-float support, if that ever happens.) Ray P.S. Darwin binaries will be uploaded in a bit.
