[sdcc-devel] indentation of assembly source

[Frieder F. <fri...@we...>]

Hi,

(sorry for the somewhat lengthy mail)
The subject of this mail could have been:

"improved readabily of generated assembly source"
"hierarchical indentation"
"white spaces in assembly generation"

The assembly source is somewhat hard to read, 
because peephole roules spring to the eye
whereas the higher levels (c-source, genXxx) 
seem somewhat hidden.

An example:

int tst_ib ( unsigned char t)
{
  if(t) return 0;
  else return 0x1234;
}

compiles to:

;------------------------------------------------------------
;       test_ret.c:20: int tst_ib ( unsigned char t)
;       -----------------------------------------
;        function tst_ib
;       -----------------------------------------
_tst_ib:
;       genReceive
        mov     r2,dpl
;       test_ret.c:22: if(t) return 0;
;       genIfx
        mov     a,r2
;       genIfxJump
; Peephole 110   removed ljmp by inverse jump logic
        jz  00102$
00107$:
;       genRet
; Peephole 181   used 16 bit load of dptr
        mov  dptr,#0x0000
; Peephole 132   changed ljmp to sjmp
;       test_ret.c:23: else return 0x1234;
;       genRet
; Peephole 182   used 16 bit load of dptr
; Peephole 237a  removed sjmp to ret
        ret
00102$:
        mov     dptr,#(((0x12)<<8) + 0x34)
00104$:
        ret



Which is hard to read if you read more lines, 
or stare at the same lines a little longer;)
(although I once was pleased that the 
peephole comments were so prominent)

Using another scheme for indentation (f.e. +2 columns for
each abstraction level), there is no "hierarchy inversion"
in white spacing.

So what about:



;------------------------------------------------------------
;       test_ret.c:20: int tst_ib ( unsigned char t)
;       -----------------------------------------
;        function tst_ib
;       -----------------------------------------
_tst_ib:
;     genReceive
        mov     r2,dpl
; test_ret.c:22: if(t) return 0;
;     genIfx
        mov     a,r2
;     genIfxJump
;       Peephole 110   removed ljmp by inverse jump logic
        jz  00102$
00107$:
;     genRet
;       Peephole 181   used 16 bit load of dptr
        mov  dptr,#0x0000
;       Peephole 132   changed ljmp to sjmp
; test_ret.c:23: else return 0x1234;
;     genRet
;       Peephole 182   used 16 bit load of dptr
;       Peephole 237a  removed sjmp to ret
        ret
00102$:
        mov     dptr,#(((0x12)<<8) + 0x34)
00104$:
        ret


The PIC port already uses a similar indentation scheme.
I'd submit a patch (mcs51) if this scheme makes 
sense and no tools are broken by it?

Regards,

Frieder