Re: SNAPSHOT (stable) 980618
Derrick J Brashear wrote:
>On Thu, 18 Jun 1998, Derrick J Brashear wrote:
<snip>
>- backport of math-emu from sparc64 by Peter Maydell
<snip>
If people are going to roll this into kernel trees, could I suggest that
(a) they use version 4, not version 3
(b) they send me patches they make (I'm thinking of ashldi3.S in
particular here...)
(a) is partly my fault, as I don't think I ever properly announced
it. Anyway, you can get it from
http://www.chiark.greenend.org.uk/~pmaydell/sparc/ as usual.
It's completely rewritten to use inline functions rather than cpp
macros. This makes it much cleaner. The disadvantage is that the module
size has increased somewhat. It also has a number of bugfixes to various
of the emulation routines. Basically, I would say that you have very
little chance of getting accurate results from version 3, whereas v4
seems to be pretty much OK for all the single and double precision
operations. Quad precision is still not fully supported, though.
There are some other minor improvements which are mostly interesting
to people wanting to hack the code rather than use it :->
It also doesn't yet handle exceptions properly, which (as I have just
this minute discovered) causes the emacs20 compile process to SIGFPE :-<
<ASIDE>
OTOH, that's an IEEE_754_exception, which doesn't go near the emulation
code. It *ought* to SIGFPE then. Moreover, it's an 'inexact' exception.
Why on earth is emacs asking to be notified of those? They happen all
the time... Something has set every bit in the trap enable mask to 1.
Maybe it's emacs' fault after all :->
</ASIDE>
Peter Maydell
--
To UNSUBSCRIBE, email to debian-sparc-request@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listmaster@lists.debian.org
Reply to: