[sdcc-devel] [ sdcc-Feature Requests-1744193 ] built-in memcpy()

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

Feature Requests item #1744193, was opened at 2007-06-27 15:25
Message generated for change (Comment added) made by spth
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=350599&aid=1744193&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: None
Group: None
Status: Open
Priority: 5
Private: No
Submitted By: Frieder Ferlemann (frief)
Assigned to: Nobody/Anonymous (nobody)
Summary: built-in memcpy()

Initial Comment:
Setting up the parameters for the 8051 memcpy() uses about 24 byte (small(!) memory model):

                          160 ;       memcpy.c:23: memcpy(&x,&y,32);
   0070 75 09 00            161         mov     _memcpy_PARM_2,#_y
   0073 75 0A 00            162         mov     (_memcpy_PARM_2 + 1),#(_y >> 8)
   0076 75 0B 00            163         mov     (_memcpy_PARM_2 + 2),#0x0
   0079 75 0C 20            164         mov     _memcpy_PARM_3,#0x20
   007C E4                  165         clr     a
   007D F5 0D               166         mov     (_memcpy_PARM_3 + 1),a
   007F 90 00 08            167         mov     dptr,#_x
   0082 75 F0 40            168         mov     b,#0x40
   0085 12 00 8B            169         lcall   _memcpy


which is twice as long as the 12 byte f.e. for this
routine which directly copies from xdata to 
(i)data memory:

   0064 90 00 00            133           mov dptr,#_y
   0067 78 08               134           mov r0,#_x
   0069                     135         loop:
   0069 E0                  136           movx a,@dptr
   006A F6                  137           mov @r0,a
   006B A3                  138           inc dptr
   006C 08                  139           inc r0
   006D B8 28 F9            140           cjne r0,#_x+32,loop


Also it would be by far quicker and less registers are used.(At least if the pointer memory spaces and the number of bytes are known at compile time:)


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

Comment By: Philipp Krause (spth)
Date: 2007-06-27 16:52

Message:
Logged In: YES 
user_id=564030
Originator: NO

The situation is similar on the Z80.
A builtin memcpy could be used to implement structure assignment, too.
Of course we'd still need a normal one in case someone takes it's adress.

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=350599&aid=1744193&group_id=599