[sdcc-devel] [ sdcc-Bugs-966505 ] SDCC MCS51 code generation bug 2

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

Bugs item #966505, was opened at 2004-06-04 13:18
Message generated for change (Comment added) made by frief
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=966505&group_id=599

Category: None
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
Summary: SDCC MCS51 code generation bug 2

Initial Comment:
I has found several mistakes in SDCC:

Source:
#include <8052.h>

#define HL0 P0_0
#define HL1 P0_1

unsigned char test;

void test1(unsigned char arg)
{
 switch (arg)
 {
 case 0:
 {
  HL0=1;
  HL1=1;
  break;
 }
 case 1:
 {
  HL0=0;
  HL1=0;
  break;
 }
 case 2:
 {
  HL0=1;
  HL1=0;
  break;
 }
 }
}

void main(void)
{
test1(0);
}

SDCC run command:

sdcc --verbose --model-small --peep-asm -mmcs51 --
iram-size 256 
--xram-size 0 --code-size 12288 --nojtbound test.c

SDCC version:

SDCC : 
mcs51/gbz80/z80/avr/ds390/pic16/pic14/TININative/xa51
/ds400/hc08 2.4.1 (Jun  3 2004) (MINGW32)


Asm list:
                            353 ;test.c:8: void test1(unsigned 
char arg)
                            354 ;	---------------------------
--------------
                            355 ;	 function test1
                            356 ;	---------------------------
--------------
   0031                     357 _test1:
                    0002    358 	ar2 = 0x02
                    0003    359 	ar3 = 0x03
                    0004    360 	ar4 = 0x04
                    0005    361 	ar5 = 0x05
                    0006    362 	ar6 = 0x06
                    0007    363 	ar7 = 0x07
                    0000    364 	ar0 = 0x00
                    0001    365 	ar1 = 0x01
                            366 ;     genReceive
                            367 ;test.c:10: switch (arg)
                            368 ;     genJumpTab
                            369 ;       Peephole 244.a  moving 
first to a instead of r2
   0031 E5 82               370 	mov	a,dpl
   0033 FA                  371 	mov	r2,a
   0034 25 E0               372 	add	a,acc
   0036 2A                  373 	add	a,r2
   0037 90s00r3B            374 	mov
	dptr,#00108$
   003A 73                  375 	jmp	@a+dptr
   003B                     376 00108$:
   003B 02s00r47            377 	ljmp	00101$
   003E 02s00r4D            378 	ljmp	00102$
   0041 02s00r53            379 	ljmp	00103$
   0044 02s00r58            380 	ljmp	00104$
                            381 ;test.c:12: case 0:
   0047                     382 00101$:
                            383 ;test.c:14: HL0=1;
                            384 ;     genAssign
   0047 D2 80               385 	setb	_P0_0
                            386 ;test.c:15: HL1=1;
                            387 ;     genAssign
   0049 D2 81               388 	setb	_P0_1
                            389 ;test.c:16: break;
                            390 ;test.c:18: case 1:
                            391 ;       Peephole 112.b  changed 
ljmp to sjmp
   004B 80 0F               392 	sjmp    00106
$                             <=========== peephole 251.b 
is absent!
   004D                     393 00102$:
                            394 ;test.c:20: HL0=0;
                            395 ;     genAssign
   004D C2 80               396 	clr	_P0_0
                            397 ;test.c:21: HL1=0;
                            398 ;     genAssign
   004F C2 81               399 	clr	_P0_1
                            400 ;test.c:22: break;
                            401 ;test.c:24: case 2:
                            402 ;       Peephole 112.b  changed 
ljmp to sjmp
   0051 80 09               403 	sjmp    00106
$                             <=========== peephole 251.b 
is absent!
   0053                     404 00103$:
                            405 ;test.c:26: HL0=1;
                            406 ;     genAssign
   0053 D2 80               407 	setb	_P0_0
                            408 ;test.c:27: HL1=0;
                            409 ;     genAssign
   0055 C2 81               410 	clr	_P0_1
                            411 ;test.c:28: break;
                            412 ;test.c:30: case 3:
                            413 ;       Peephole 112.b  changed 
ljmp to sjmp
                            414 ;       Peephole 251.b  replaced 
sjmp to ret with ret <=== OK!
   0057 22                  415 	ret
   0058                     416 00104$:
                            417 ;test.c:32: HL0=0;
                            418 ;     genAssign
   0058 C2 80               419 	clr	_P0_0
                            420 ;test.c:33: HL1=1;
                            421 ;     genAssign
   005A D2 81               422 	setb	_P0_1
                            423 ;test.c:36: }
   005C                     424 00106$:
   005C 22                  425 	ret


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

>Comment By: Frieder Ferlemann (frief)
Date: 2004-06-04 17:44

Message:
Logged In: YES 
user_id=589052

Sorry, I should have looked closer, I at first thought it
Peephole 251.b had the conditional "labelRefCount". Instead
it has "labelIsReturnOnly" so it really should have been
applied in case 0 and case 1.
If you add more cases to the switch statement Peephole 251.b
is only applied to the last but one switch case.


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

Comment By: Frieder Ferlemann (frief)
Date: 2004-06-04 17:10

Message:
Logged In: YES 
user_id=589052

Peephole 251.b does not apply in this case because label
00106$ has a labelRefCount of 1 instead of 0 (although it is
not referenced).
If this would be fixed then Peephole 253.b would be applied
more often too.
(about 200 times more in the regression test)

Btw I'd consider this missed optimization rather a Feature
Request than a Bug:)


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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=966505&group_id=599