From: Patrycjusz R. Łogiewa (silverdr_at_inet.com.pl)
Date: 2004-12-14 12:15:18
On 13 gru 2004, at 18:54, Greg King wrote: > From: Patrycjusz R. Łogiewa; on Date: December 13, 2004 05:57 AM -0500 >> >> AFAIR, all the others mentioned here weren't really compiling BASIC >> into >> machine code, but rather into so-called "speedcodes". That (AFAIR, >> again) >> was something like calling BASIC routines for the given keywords >> rather >> than having the interpreter to read and interpret the tokens. >> >> Corrections to the above, anyone? > > The speed-code output of the compilers is appended to a rather large > run-time module that interprets them, in the same way that the BASIC > ROM > interprets BASIC tokens. Those compiled programs run faster because > integer variables are combined directly by real integer operations, and > data is predigested: during the compilation, PETSCII numbers are > converted > into binary, the lengths of string-literals are measured, and the > variables > tables are pre-built -- varible-names are replaced by offsets into > those > tables. The compiled program doesn't need to spend time creating and > looking for its data. Yes. That's more or less what I thought. And the bottom line of this is that the speedcode compilers' outputs aren't good examples to study when searching the ML solution to a given BASIC algorithm, are they? > > By the way, those compilers are recursively-built. They are BASIC > programs > which compiled themselves. > Huh, never realised that! -- GNU - GNU's Not Unix - Richard Stallman? Message was sent through the cbm-hackers mailing list
Archive generated by hypermail pre-2.1.8.