rmariani@watmum.UUCP (12/05/86)
This one was found easier because it produces an assembly time error
however, it is not so easy to determine the source...
main()
{
register int p;
int s[10];
s[p/4]/p;
}
Causes an assembly time error when compiled with the +l option but
works fine otherwise...
;main()
;{
public _main
_main:
link a5,#.2
movem.l .3,-(sp)
; register int p;
; int s[10];
;
; s[p/4]/p;
move.l d4,d0
move.l #4,d1 <---- you guys should generate a
jsr .divs# a shift please... don't call
asl.l #2,d0 the signed division thing...
lea -40(a5),a6
move.l d0,-(sp)
move.l (a6,.l),d0 <---- CHOKE HERE
move.l (sp)+,d3
move.l d4,d1
jsr .divs#
;}
.4
movem.l (sp)+,.3
unlk a5
rts
.2 equ -40
.3 reg d4
public .begin
dseg
cseg
end
Besides the fact that this code is not exactly ummmm optimal, there is
a slight (so slight) syntax error in the assembler, that line should read
move.l (a6,d0.l),d0
Sigh....
Can we try to do something about this one B4 Version 3.40 too?
By the way, I am a registered (and satisfied!) Commercial version owner
and I have not recieved anything from Aztec ever. I sent in my forms
last March.
Maybe they don't like Canadians or something? :-)
-Ricopage@ulowell.UUCP (Bob Page) (12/06/86)
rmariani@watmum.UUCP (Rico Mariani) wrote in article <717@watmum.UUCP>:
>This one was found easier because it produces an assembly time error
Fixed in V3.30e (beta version). It still generates the jsr .divs#
instruction, though.
Old code: New code: Bob's comments
-------------- -------------- --------------
move.l d4,d0 move.l #4,d1 reversed order
move.l #4,d1 move.l d4,d0
jsr .divs# jsr .divs# still no shift
asl.l #2,d0 asl.l #2,d0
lea -40(a5),a6 lea -40(a5),a0
move.l d0,-(sp) move.l d0,d2 a speedup here
move.l (a6,.l),d0 move.l (a0,d2.l),d0 fixed the bug
move.l (sp)+,d3 was excess code
move.l d4,d1 move.l d4,d1
jsr .divs# jsr .divs#
The rest of the code is about the same. It runs through the
assembler OK, too.
..Bob (not affiliated with Manx, by the way, except as beta tester)
--
Bob Page, U of Lowell CS Dept. ulowell!page, page@ulowell.CSNET