home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Workbench Add-On
/
Workbench Add-On - Volume 1.iso
/
BBS-Archive
/
Dev
/
Obrn-A_1.6_lib.lha
/
oberon-a
/
source3.lha
/
source
/
Kernel
/
LMath.asm
< prev
next >
Wrap
Assembly Source File
|
1995-06-29
|
7KB
|
267 lines
**********************************************************************
*
* $RCSfile: LMath.asm $
* Description: Runtime support for the Oberon-A compiler
*
* Created by: fjc (Frank Copeland)
* $Revision: 1.4 $
* $Author: fjc $
* $Date: 1995/06/29 19:03:32 $
*
* Copyright © 1994, Frank Copeland.
* This file is part of the Oberon-A Library.
* See Oberon-A.doc for conditions of use and distribution.
*
* Log entries are at the end of the file.
*
**********************************************************************
*
* Acknowledgements
* ----------------
*
* The 32-bit multiply and divide procedures are from the runtime
* library of Patrick Quaid's PCQ freeware Pascal compiler, which in
* turn came from the runtime library of Sozobon C.
*
**********************************************************************
**********
* lmath.s
**********
* Copyright (c) 1988 by Sozobon, Limited. Author: Johann Ruegg
*
* Permission is granted to anyone to use this software for any purpose
* on any computer system, and to redistribute it freely, with the
* following restrictions:
* 1) No charge may be made other than reasonable charges for
* reproduction.
* 2) Modified versions must be clearly marked as such.
* 3) The authors are not responsible for any harmful consequences
* of using this software, even if they result from defects in it.
*
*
* For PCQ Pascal:
* These are the 32-bit math functions from Sozobon-C,
* as noted above. I changed the names of the routines to
* be more similar to the rest of my library, and handle the
* divide by zero condition differently. Other than that I
* haven't changed the code a bit.
*
* For Oberon-A:
* I have changed the names (again) and modified the
* routines to accept parameters passed in registers instead of
* on the stack, in keeping with the conventions I use in the
* rest of the compiler.
*
**********************************************************************
;---------------------------------------------------------------------
; Program unit hunk name
TTL Kernel
;---------------------------------------------------------------------
;----------------------------------------------------------------
; PROCEDURE Kernel_Mul32 (
; l1 {D0} : LONGINT;
; l2 {D1} : LONGINT)
; : LONGINT;
;
; Calculates l1 * l2, returning the result in D0.
;----------------------------------------------------------------
SECTION Kernel,CODE
XDEF Kernel_Mul32
XREF Kernel_Halt
Kernel_Mul32:
movem.l d2-d4,-(a7)
tst.l d0
smi d4
bpl lm1
neg.l d0
lm1:
tst.l d1
bpl lm2
not.b d4
neg.l d1
lm2:
move.w d1,d2
mulu d0,d2 /* d2 = Al * Bl */
move.l d1,d3
swap d3
mulu d0,d3 /* d3 = Al * Bh */
swap d0
mulu d1,d0 /* d0 = Ah * Bl */
add.l d3,d0 /* d0 = (Ah*Bl + Al*Bh) */
swap d0
clr.w d0 /* d0 = (Ah*Bl + Al*Bh) << 16 */
add.l d2,d0 /* d0 = A*B */
tst.b d4
beq lm3
neg.l d0
lm3:
movem.l (a7)+,d2-d4
rts
;---------------------------------------------------------------------
;----------------------------------------------------------------
; PROCEDURE Kernel_Div32
; l1 {D0} : LONGINT;
; l2 {D1} : LONGINT)
; : LONGINT;
;
; Calculates l1 DIV l2, returning the result in D0 (quotient) and
; D1 (remainder).
;----------------------------------------------------------------
SECTION Kernel,CODE
XDEF Kernel_Div32
; XREF Kernel.Halt
Kernel_Div32:
movem.l d2-d5,-(a7)
tst.l d0
smi d4
bpl ld1
neg.l d0
ld1:
tst.l d1
smi d5
bpl ld2
neg.l d1
ld2:
tst.l d1
bne.s nz1
* divide by zero
move.l #105,d0
lea module,a0
move.l (146*$10000)+19,d1
jsr Kernel_Halt
nz1:
cmp.l d1,d0
bhi norm
beq is1
* A<B, so ret 0, rem A
move.l d0,d1
clr.l d0
bra.s ld5
* A==B, so ret 1, rem 0
is1:
moveq.l #1,d0
clr.l d1
bra.s ld5
* A>B and B is not 0
norm:
cmp.l #1,d1
bne.s not1
* B==1, so ret A, rem 0
clr.l d1
bra.s ld5
* check for A short (implies B short also)
not1:
cmp.l #$ffff,d0
bhi slow
* A short and B short -- use 'divu'
divu d1,d0 /* d0 = REM:ANS */
swap d0 /* d0 = ANS:REM */
clr.l d1
move.w d0,d1 /* d1 = REM */
clr.w d0
swap d0
bra.s ld5
* check for B short
slow:
cmp.l #$ffff,d1
bhi slower
* A long and B short -- use special stuff from gnu
move.l d0,d2
clr.w d2
swap d2
divu d1,d2 /* d2 = REM:ANS of Ahi/B */
clr.l d3
move.w d2,d3 /* d3 = Ahi/B */
swap d3
move.w d0,d2 /* d2 = REM << 16 + Alo */
divu d1,d2 /* d2 = REM:ANS of stuff/B */
move.l d2,d1
clr.w d1
swap d1 /* d1 = REM */
clr.l d0
move.w d2,d0
add.l d3,d0 /* d0 = ANS */
bra.s ld5
* A>B, B > 1
slower:
move.l #1,d2
clr.l d3
moreadj:
cmp.l d0,d1
bhi.s adj
add.l d2,d2
add.l d1,d1
bpl moreadj
* we shifted B until its >A or sign bit set
* we shifted #1 (d2) along with it
adj:
cmp.l d0,d1
bhi.s ltuns
or.l d2,d3
sub.l d1,d0
ltuns:
lsr.l #1,d1
lsr.l #1,d2
bne adj
* d3=answer, d0=rem
move.l d0,d1
move.l d3,d0
ld5:
cmp.b d4,d5
beq ld3
neg.l d0
ld3:
tst.b d4
beq ld4
neg.l d1
ld4:
movem.l (a7)+,d2-d5
rts
module:
DC.B "Kernel_Div32",0
;---------------------------------------------------------------------
END ; Kernel
**********************************************************************
*
* $Log: LMath.asm $
;; Revision 1.4 1995/06/29 19:03:32 fjc
;; - Release 1.6
;;
;; Revision 1.3 1995/01/26 00:37:31 fjc
;; - Release 1.5
;;
;; Revision 1.3 1995/01/26 00:37:31 fjc
;; - Release 1.5
;;
**********************************************************************