[patch] gcc/reload1.c

[Kazu Hirata]

Richard and Jeff,

> A more correct solution follows.  I've checked this against the
> test case, but not with a full simulator bootstrap, but I do not
> expect problems.

Just a little thing. Richard's code outputs things like "sub #-1, r0", which
is illegal and confuses gas. I am wondering if there is a quick and easy way
of flipping the sign of an integer operand in a machine description file. If
not, do we need something like the attached patch? I myself don't really
like it, but I hope my intention is clear... The patch assumes that
Richard's patch has been applied.

You can get a test case for the problem from L_divdi3 in libgcc2.c.

Thanks,

Kazu Hirata

-------------- next part --------------
2000-05-18  Kazu Hirata  <kazu@hxi.com>

	* h8300.c (output_adds_subs): New function.
	* h8300-proto.h (output_adds_subs): The prototype for the above.
	* h8300.md: Change adds and subs to use the C style instead of the
	templates.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: h8adds.patch
Type: text/x-diff
Size: 4095 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20000517/7b592edc/attachment.bin>