Path: utzoo!utgpu!news-server.csri.toronto.edu!mailrus!uunet!willett!ForthNet
From: ForthNet@willett.UUCP (ForthNet articles from GEnie)
Newsgroups: comp.lang.forth
Subject: Optimization
Message-ID: <1226.UUL1.3#5129@willett.UUCP>
Date: 27 Jun 90 02:30:07 GMT
Organization: Latest link in the ForthNet chain.  (Pgh, PA)
Lines: 38


 Date: 06-25-90 (06:32)              Number: 1805 (Echo)
   To: MARK SMILEY                   Refer#: NONE
 From: PETE KOZIAR                     Read: NO
 Subj: MORE EFFICIENT 0<             Status: PUBLIC MESSAGE

 Another little optimization for F83 and/or F-PC.  Using the magic CWD 
 instruction, we can gain quite a few cycles, making the definition for 
 0< being: 

 CODE 0<  POP AX  CWD  PUSH DX  NEXT  END-CODE 

 We can extend this to other words in a similar fashion: 

 CODE 0>  POP AX  NEG AX  CWD  PUSH DX  NEXT  END-CODE 
 CODE 0<> POP AX  CWD   MOV BX, DX 
          NEG AX  CWD   AND BX, DX  PUSH BX  NEXT  END-CODE 

 This saves the Jxx instruction, as well as the immediate move 
 instructions.  I think it's used often enough to justify "kernel 
 twiddling." 
 ---
  * Via Qwikmail 2.01  The Baltimore Sun 

 Date: 06-25-90 (06:54)              Number: 1806 (Echo)
   To: MARK SMILEY                   Refer#: NONE
 From: PETE KOZIAR                     Read: NO
 Subj: MISTAKE IN LAST NOTE          Status: PUBLIC MESSAGE

 Oops.  That's what I get for entering a message the first thing in the 
 morning!  The AND BX, DX in the definition for 0<> always results in 
 zero, since the number cannot be both positive and negative at the same 
 time.  The correct instruction would be OR BX, DX which returns true if 
 the number was either > 0 or < 0, and false if it was the only number 
 that is neither positive nor negative, i.e., 0.
-----
This message came from GEnie via willett through a semi-automated process.
Report problems to: uunet!willett!dwp or willett!dwp@hobbes.cert.sei.cmu.edu