Optimize flash memory handling a little bit by ignoring 0xff data that

resides above the last non-0xff data value in the address space.  Only
do this for flash memory since writing a 0xff to flash is a no-op.
This has the affect of creating smaller output files when dumping
memory contents from flash if the program in flash does not consume
the whole memory space.  It also results in shorter programming times
when avrdude is asked to load a file into flash that has lots of 0xff
filled data past the last non-0xff data value.

I think this is basically where Alexey was going with his s-record
routine, but this should have a similar affect for all the I/O
routines.  The main difference is that Alexey's also optimized 0xff
from the beginning of the address space and was not limited to flash.

I think that these optimizations should be limited to the flash since
it is currently the only memory that treats 0xff as special.


git-svn-id: svn://svn.savannah.nongnu.org/avrdude/trunk/avrdude@331 81a1dc3b-b13d-400b-aceb-764788c761c2

fileio.c

@@ -933,9 +933,7 @@ int fileio(int op, char * filename, FILEFMT format,
 
   if (fio.op == FIO_READ) {
     /* 0xff fill unspecified memory */
-    for (i=0; i<size; i++) {
-      buf[i] = 0xff;
-    }
+    memset(buf, 0xff, size);
   }
 
   using_stdio = 0;
@@ -1011,6 +1009,16 @@ int fileio(int op, char * filename, FILEFMT format,
       return -1;
   }
 
+  if (rc > 0) {
+    if ((op == FIO_READ) && (strcasecmp(mem->desc, "flash") == 0)) {
+      /*
+       * if we are reading flash, just mark the size as being the
+       * highest non-0xff byte
+       */
+      rc = avr_mem_hiaddr(mem);
+    }
+  }
+
   return rc;
 }