This is the mail archive of the
mailing list for the GCC project.
Re: PATCH to rs6000.md, comments?
- To: law at cygnus dot com
- Subject: Re: PATCH to rs6000.md, comments?
- From: David Edelsohn <dje at watson dot ibm dot com>
- Date: Tue, 21 Jul 1998 11:11:37 -0400
- Cc: Franz Sirl <Franz dot Sirl-kernel at lauterbach dot com>, Richard Henderson <rth at cygnus dot com>, egcs-patches at cygnus dot com
I don't want to remove save_stack_function and
restore_stack_function from rs6000.md. This is a prototyping problem, not
a problem with the machine description. The patterns in rs6000.md are
useful because they document some magic going on behind the scenes.
As far as the CONSTANT_P_RTX changes, Richard already commented on
the unnecessary "return 0;" cases. I personally do not like the style of
most of the changes. The CONSTANT_P_RTX test should be or'ed into the
other tests for consistency, not a separate "if ... return 1".
Richard: my comment about movdi as well was referring to 64-bit
mode in which case DImode is an integer. movdi *does* need to be changes
as well symmetrically with movsi.
Jeff: Do you want me to redo all of this? Integrate it into the
egcs-1.1 release branch as well?