shebs@utah-cs.UUCP (Stanley Shebs) (03/14/85)
I have uncovered an article in the SIGSAM Bulletin of March 1973 in which Richard Fateman compares Maclisp with Fortran (the DEC-F40 compiler). He exhibits a program to compute (via iteration), one thousand cube roots, first written in Fortran, then in Maclisp, then in Macsyma. The Fortran version took 2.22 seconds, the Maclisp version 1.81. A recursive reformulation of the Maclisp code took 2.20 sec. The Macsyma version had exactly the same time as the Maclisp version (as might be expected). This is the entire article; obviously not a systematic study. Also, I found an abstract of a Guy Steele paper presented at the 1977 Macsyma Users Conference. It was entitled "Fast Arithmetic in Maclisp" - the abstract I saw was in the May 1977 SIGSAM, and made the speed claim for Maclisp. I don't have access to the paper; did Steele have any numbers? stan shebs