[sdcc-devel] [ sdcc-Feature Requests-966504 ] excessive push/pop between function calls

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

Feature Requests item #966504, was opened at 2004-06-04 15:17
Message generated for change (Comment added) made by maartenbrock
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=350599&aid=966504&group_id=599

Category: None
Group: None
Status: Open
Priority: 5
Submitted By: Nobody/Anonymous (nobody)
Assigned to: Nobody/Anonymous (nobody)
>Summary: excessive push/pop between function calls

Initial Comment:
I has found several mistakes in SDCC:

Source:

#include <8052.h>

unsigned char test;

void test1(void)
{
 //do something...
}

void test2(void)
{
 //do something
}

void test3(void)
{
 unsigned char tmp;
 for (tmp=0; tmp!=5; tmp++)
 {
  test1();
  test2();
  test1();
 }
}


void main(void)
{
test3();
}

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:
                           388 ;	---------------------------
--------------
                            389 ;	 function test3
                            390 ;	---------------------------
--------------
   0033                     391 _test3:
                            392 ;test.c:18: for (tmp=0; tmp!
=5; tmp++)
                            393 ;     genAssign
   0033 7A 00               394 	mov	r2,#0x00
   0035                     395 00101$:
                            396 ;     genCmpEq
   0035 BA 05 01            397 	cjne
	r2,#0x05,00109$
                            398 ;       Peephole 112.b  changed 
ljmp to sjmp
                            399 ;       Peephole 251.b  replaced 
sjmp to ret with ret
   0038 22                  400 	ret
   0039                     401 00109$:
                            402 ;test.c:20: test1();
                            403 ;     genCall
   0039 C0 02               404 	push	ar2
   003B 12s00r31            405 	lcall	_test1
   003E D0 02               406 	pop
	ar2                           <=====
                            407 ;test.c:21: test2
();                                |spare operations with stack
                            408 ;     
genCall                                       |
   0040 C0 02               409 	push
	ar2                           <=====
   0042 12s00r32            410 	lcall	_test2
   0045 D0 02               411 	pop
	ar2                           <=====
                            412 ;test.c:22: test1
();                                |spare operations with stack
                            413 ;     
genCall                                       |
   0047 C0 02               414 	push
	ar2                           <=====
   0049 12s00r31            415 	lcall	_test1
   004C D0 02               416 	pop	ar2
                            417 ;test.c:18: for (tmp=0; tmp!
=5; tmp++)
                            418 ;     genPlus
                            419 ;     genPlusIncr
   004E 0A                  420 	inc	r2
                            421 ;       Peephole 112.b  changed 
ljmp to sjmp
   004F 80 E4               422 	sjmp    00101$
   0051                     423 00105$:
   0051 22                  424 	ret


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

>Comment By: Maarten Brock (maartenbrock)
Date: 2004-08-11 21:35

Message:
Logged In: YES 
user_id=888171

Nice feature request, no bug

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

Comment By: Yee_Keat Phuah (kiwlm)
Date: 2004-07-12 07:47

Message:
Logged In: YES 
user_id=940150

tmp is assigned in r2 in test3, and all the functions called
by test3, might use r2 for other optimizations, hence test3
pushes r2 (or ar2) on the stack whenever it calls other
functions. So I think this is normal.

If you wishes to optimize it so that it only push/pop ar2
once in the whole loop, then its a different story, please
clarify.

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=350599&aid=966504&group_id=599