Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
Jan Waclawek
05/27/09 16:18
Read: 239 times
Bratislava
Slovakia


 
#165636 - findings
Responding to: Matthias Arndt's previous message
Find complete results of compilation with v2.6.0 here - I will remove it after a week.

Looking at 8052.mem reveals, that "data" area (DSEG) is filled up to 0x77, leaving 8 bytes in this segment.

Unfortunately, the .map nor .sym files are not very informative as long as automatic variables are concerned - i.e. space the compiler allocates to local variables when it cannot locate them to registers, and "spills", i.e. space for intermediate calculation results (again if they won't fit to registers). You have to go to .lst files, to the DSEG description, to have detailed information on these.

Now take the 2.9.x, cut down from the buff[] array (I simply decreased BUFFSIZE to 16+1, but I don't guarantee it will work), which is the single biggest variable in "data", and comparing the DSEG in .lst you can see, where the newer version "looses space". It's slightly more than the 8 bytes "spare".

JW

List of 24 messages in thread
TopicAuthorDate
SDCC Linker problem with CMON51      Matthias Arndt      05/27/09 11:51      
   a wild, wild guess      Erik Malund      05/27/09 12:04      
      I already tried a typecast      Matthias Arndt      05/27/09 12:07      
         Why not take this to the SDCC forum?      Richard Erlacher      05/27/09 13:01      
            Already considered this...      Matthias Arndt      05/27/09 14:06      
      another possibility which is easy to check if ...      Erik Malund      05/27/09 15:21      
         I assume so...      Matthias Arndt      05/27/09 15:40      
            findings      Jan Waclawek      05/27/09 16:18      
               findings2      Jan Waclawek      05/27/09 16:43      
                  Temporary fix      Matthias Arndt      05/28/09 11:08      
                     back to the drawing board...      Matthias Arndt      05/28/09 11:43      
                        it's much less for DATA      Jan Waclawek      05/28/09 16:29      
                  fix for --funsigned-char      Jan Waclawek      06/01/09 03:34      
                     I'd like to but...      Matthias Arndt      06/01/09 12:24      
                        how to patch      Jan Waclawek      06/01/09 15:35      
                           Patch worked but SDCC doesn't compile anymore...      Matthias Arndt      06/01/09 15:58      
                              HOW doesn't compile?      Jan Waclawek      06/01/09 16:07      
               Buffer cut      Matthias Arndt      05/27/09 16:45      
   not enough direct-addressable memory      Jan Waclawek      05/27/09 13:07      
      Nice idea but....      Matthias Arndt      05/27/09 14:04      
         I can try with the versions I have here...      Jan Waclawek      05/27/09 14:26      
            Sure, why not      Matthias Arndt      05/27/09 14:59      
   Solution found      Matthias Arndt      05/29/09 11:51      
   SDCC Linker problem with CMON51      Yathish Kumar.N.S      06/19/09 08:29      

Back to Subject List