Re: RFA: loader Elf64 loader & split insn/data accessors

["Frank Ch. Eigler" <fche at redhat dot com>]

Hi -

On Sat, Jan 13, 2001 at 11:00:32PM +1100, matthew green wrote:
: [...]
: this following patch splits the current elf `loader' module
: [...] this hack allows full 32 bit harvard arch machines to have
: a full normal address space, with insn being (at the object file format)
: from 0x.10000.0000 ... 0x1.ffff.ffff.  

Right.


: it should be removed when sid can deal with >32 bit address spaces.

How do you suggest that should be done?  We used to have a host_int_8
address option in the sid::bus class; maybe it's time to bring it back.
At another point in history, the host_int_4 address option didn't exist,
forcing all address arithmetic to be done on long-long types, which was
a performance loser.

(I hesitate to mention the other option of writing new sidutil classes
that grow accessors & buses on demand in order to represent a sparsely
populated 64-bit address space.)


: OK to commit?

Sure - it's a good first step.  

I would suggest though renaming the "insn"/"data" portions of the code and
accessor names names to something like "0" and "1", in an attempt to encode
the 33rd bit of the address in the name, rather than our specific
interpretation of that bit for a single target port.

Also, what about the "set-pc!" output signal?

Also, it would be nice to commit some elf32 and elf64 executables into
the test suite, and run the elf-loader on them.

- FChE

PGP signature