[sdcc-devel] [ sdcc-Bugs-2986039 ] PIC14: bit operations with wrong value in W register

[SourceForge.net <no...@so...>]

Bugs item #2986039, was opened at 2010-04-12 11:10
Message generated for change (Settings changed) made by tecodev
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=2986039&group_id=599

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: pic14 target
Group: fixed
Status: Closed
Resolution: Fixed
Priority: 5
Private: No
Submitted By: Jim Paris (jimatjtan)
Assigned to: Nobody/Anonymous (nobody)
Summary: PIC14: bit operations with wrong value in W register

Initial Comment:
Attached code miscompiles on pic14.
sdcc -v:
SDCC : mcs51/gbz80/z80/avr/ds390/pic16/pic14/TININative/xa51/ds400/hc08 2.9.0 #5416 (Jan 10 2010) (UNIX)

compile with:
sdcc --std-c89 -mpic14 -p16f628a --opt-code-speed -c -o bug.o bug.c

In the .lst file, you can see four occurances like this one:

               00192 ;       .line   11; "bug.c"     RB6 = (c & 8) ? 1 : 0;
001E 0000 0000 00193         BANKSEL r0x1001
0020 0C00      00194         RRF     r0x1001,W
0021 1803      00195         BTFSC   STATUS,0
0022 2800      00196         GOTO    _00001_DS_
0023 0000 0000 00197         BANKSEL _PORTB_bits
0025 1300      00198         BCF     _PORTB_bits,6
0026           00199 _00001_DS_
0026 1C03      00200         BTFSS   STATUS,0
0027 2800      00201         GOTO    _00002_DS_
0028 0000 0000 00202         BANKSEL _PORTB_bits
002A 1700      00203         BSF     _PORTB_bits,6
002B           00204 _00002_DS_

but the "W" register (used in "RRF r0x1001,W") is a stale value, in this case 0x7B.


----------------------------------------------------------------------

Comment By: Raphael Neider (tecodev)
Date: 2012-02-11 15:22

Message:
Fixed for PIC14 and PIC16 in r7302.

----------------------------------------------------------------------

Comment By: Jim Paris (jimatjtan)
Date: 2010-04-13 13:07

Message:
Sorry for the misdiagnosis, been dealing with too many assembly language
architectures lately :)
Your description sounds right and the workaround does the right thing. 
Thanks for looking into it.

----------------------------------------------------------------------

Comment By: Raphael Neider (tecodev)
Date: 2010-04-12 15:28

Message:
OK, I see the problem now:

RB6 = (c & 8) ? 1 : 0;

is handled incorrectly (also for c&4, c&2, and c&1). Instead of implicitly
doing

RB6 = (0 != (c & 8)) ? 1 : 0;

sdcc executes

struct {
   uint8_t b0:1;
} temp;
temp.b0 = (c & 8); // THIS ALWAYS CLEARS temp.b0
RB6 = ((uint8_t)temp.b0) ? 1: 0;

by mistaking the logic bit test on bit 3 for an assignment to a bitfield of
width 1, for which only bit 0 of the expression is relevant...

As a workaround, you can use

RB6 = (0 != (c & 8)) ? 1 : 0;

I'll try to fix this.

Best regards,
Raphael

----------------------------------------------------------------------

Comment By: Raphael Neider (tecodev)
Date: 2010-04-12 14:02

Message:
I do not see the problem: "RRF r0x1001, W" does not READ W at all -- it
overwrites W with "r0x1001 >> 1" and updates STATUS, extracting the least
signicant bit of r0x1001 (prior to shifting).
Could you give any more details? Observed vs. expected output?

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=2986039&group_id=599