[Patch, avr] Shrink interrupt vector table down to last used entry

[Senthil Kumar Selvaraj <senthil_kumar dot selvaraj at atmel dot com>]

Hi,

This patch tries to shrink the interrupt vector table by deleting unused 
entries at the end of the table as part of linker relaxation.

The motivation for this patch is that, currently, the full interrupt
vector table is being placed in the linked executable, regardless of the
number of interrupts actually used. This increases the size of the
executable for archs (XMEGAs, for example), that have lots of interrupts
(125 interrupts * 4 bytes for XMEGAs) if the user program only uses a
few of them.

At a high level, this patch figures out the size of the vector table and
the address of the last entry in the table that has a user defined
handler, and deletes everything in between.

It does it by hooking into linker relaxation for the .vectors section and
doing the following:

1. It goes through the global symbols in the bfd and looks for
those representing interrupt handlers.

2. For each such symbol found, it extracts the vector number from the
name. It also finds out whether there is a definition available for the
symbol from the symbol information available.

3. At the end of (2), symbol indices representing handlers for the last 
available interrupt for that arch, and the last used interrupt are
available - they are the symbols with the max vector number, and the max
vector number with a definition, respectively.

4. To find out the addresses of entries in the vector table for the last 
available interrupt and the last used interrupt, it walks the relocation 
entries for the section and looks for entries with symbol indices matching
what was found in (3). The corresponding relocation offsets represent
the addresses of the entries.

5. Bytes between the offsets obtained in (4) are deleted (excluding the
defined entry's size), as are relocation entries pointing into the deleted area.

Note that this ((1) and (2) relies on avr-libc's convention of naming interrupt
handlers __vector_<n>, with n representing the interrupt number. 

Also, the patch kicks in only if --shrink-ivt is passed along with --relax,
otherwise, the __vector_<n> symbols get defined to __bad_interrupt as
usual.

Thoughts?

Regards
Senthil

Attachment: avr_link_relax.patch
Description: Text document