From mboxrd@z Thu Jan  1 00:00:00 1970
From: Bart Veer <bartv@cygnus.co.uk>
To: lunn@ma.tech.ascom.ch
Cc: ecos-discuss@sourceware.cygnus.com
Subject: [ECOS] Re: stacks as globel char[]
Date: Fri, 07 May 1999 04:11:00 -0000
Message-id: <199905071110.MAA10183@sheesh.cygnus.co.uk>
References: <199905070702.JAA09921@biferten.ma.tech.ascom.ch> <199905070702.JAA09921@biferten.ma.tech.ascom.ch>
X-SW-Source: 1999-05/msg00027.html

>>>>> "Andrew" == Andrew Lunn <lunn@ma.tech.ascom.ch> writes:

    Andrew> This question comes out of ignorance as to how the
    Andrew> compiler/loader works. In the example code you have the
    Andrew> stacks as a global char stack[x][y]. Is this a good idea
    Andrew> for real applications? The same applies to memory for
    Andrew> memory pools etc.

    Andrew> Im thinking about code that is for a ROM, or downloading
    Andrew> over a serial line. If i looked at this code would i find
    Andrew> lots of zeros which are the stack and memory for the pool,
    Andrew> or are these variables created during run time startup?

Uninitialized data like

  char stacks[x][y];

ends up in the bss segment rather than the data segment. During
startup the system will automatically zero the entire bss segment.
There is no need to transfer lots of zeroes down the serial line or to
store them anywhere in the executable image.

If the code were changed to something like:

  char stacks[x][y] = { '1' };

then the entire array would end up in the data segment and would
appear in the executable. In case of doubt it is fairly easy to be
sure: simply compile a code snippet to a .o file and either just look
at the size of the output file or use powerpc-eabi-objdump -d for a
disassembly.

Bart Veer // eCos net maintainer