Talk:CMU Common Lisp

Latest comment: 19 years ago by BlakeStone in topic "there is no interpreter"?

"there is no interpreter"?

edit

I think whoever wrote this was thinking of SBCL. CMUCL compiles to interpreted bytecode as well as native code; SBCL doesn't have bytecode. I don't think the CMUCL bytecode is used for much(I don't have the manual handy, and the site is down.), but it does exist.

Aha, here's the part I was looking for:Byte Coded Compilation--BlakeStone 01:43, 12 August 2005 (UTC)Reply

Numerical performance

edit
And it's not just non-assignment based functional languages where you can see supposedly less-efficient high level languages crushing the performance of C/C++. CMU CommonLisp can beat C/C++ on numeric code. There was a paper a few years back documenting it: using a Sun SPARC workstation, if you use the optional type declarations, and write scientific/numeric code in Lisp, using vectors (Lisp arrays) and assignments to implement exactly the same algorithm as C, the CMU CommonLisp code will perform better than C code generated by either the Solaris C compiler or GCC with maximum optimization.[1]

Can't seem to find the paper he references; all the CMUCL benchmarks I've dug up were inter-Lisp implementation benchmarks.