HFischer%USC-ECLB@sri-unix.UUCP (03/23/84)
From: Herm Fischer <HFischer@USC-ECLB> [Forwarded from the Prolog Digest by Laws@SRI-AI.] [...] IBM was kind enough to let us have PC/IX for today, and we brought up UNSW Prolog. With a minor exception the code and makefiles were compatible with PC/IX. (They have frustrated me for a whole year, being incompatible every PCDOS "C" compiler from Lattice onward.) PC/IX and Prolog are neatly integrated; all Unix features, and even shell calls, can be made within the Prolog environment. Even help files are included. It is kind of nice to be tracing away and browse and modify your prolog code within the interpretive environment, using the INed (nee rand) editor and all the other Unix stuff. The 64 K limitation of PC/IX bothers me, more emotionally than factually, because only one of my programs couldn't be run today. I'm sure I will get really upset unless I find some hack around this limitation. A benchmark really surprises me. The Zebra problem (using Pereira's solution) provides the following statistics: DEC-2040 6 seconds (if compiled) (Timed on TOPS-20) 42 seconds (if interpreted) ( " " " ) VAX-11/780 204 secs (interpreted) (UNSW) (Timed on Unix Sys III) IBM PC/XT 544 secs (interpreted) ( " ) (Timed on " " " ) The latter 2 times are wall-clock with no other jobs or users running, and these two Prologs were compiled from the same source code and make file! The PC/IX was CPU-bound, and its disk never blinked during the execution of the test. -- Herm Fischer