This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: PATCH: reload register collision


On Wed, 3 Oct 2001, Dale Johannesen wrote:

> OK, the following fixes the problem by making the register
> interference checking understand that multiple output reloads are
> emitted in reverse order.  Bootstrapped on Darwin.
> 
> Wed Oct  3 12:12:06 PDT 2001  Dale Johannesen  <dalej@apple.com>
> 
>          * reload1.c (reload_reg_free_p):  teach register interference
>              checking that multiple output reloads are emitted in
>              reverse order.
>            reload1.c (reload_reg_reaches_end_p):  Ditto.
>            reload1.c (reloads_conflict):  Ditto.

This one does look correct, thanks.  Since you apparently aren't in the
MAINTAINERs file yet, I assume I'll have to check it in for you.  Two
things need to happen first; gcc.gnu.org has to accept my connections again,
and I'll need to know what the copyright assignment situation is with Apple.

A few comments for future patches (coding style violations only):

> @@ -4415,11 +4415,13 @@

Please use the "-p" option with diff, this will make these lines contain the
name of the function that is being patched which is quite useful when reading
the patch.

> +        later operand.  Note that multiple output operands are
> +         emitted in reverse order, so the conflicting ones are
> +        those with lower indices. */

Two spaces after a ".".  Also, the indentation seems to be a bit off in
one line.

> -      for (i = opnum; i < reload_n_operands; i++)
> +      for (i=0; i<=opnum; i++)

Spaces around operators.


Bernd


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]