Re: Patch: Simplification for i386 STACK_SIZE_MODE definition

[Richard Henderson <rth at redhat dot com>]

On Fri, Mar 30, 2007 at 01:55:45PM +0200, Kai Tietz wrote:
> ... because the word_mode is there still the SImode ...

This conclusion is incorrect ...

> reasoned by the fact, that a long is still 32-bit.

... because it is based on an unrelated predicate.

The size of any C type (including "long") is irrelevant to 
the size of word_mode.

>         * gcc/explow.ct: (emit_stack_probe): Replaced "word_mode" by 
> STACK_SIZE_MODE.

This change is unnecessary because this is the *probe*, i.e.
the load or store to the stack.  This could happen in *any*
mode -- it just needs to be any reasonably efficient type.

>         * gcc/config/i386/i386.h: Define STACK_SIZE_MODE as Pmode.

This is the same value appready produced by expr.h.  You really
need to understand how word_mode is defined.

Patch rejected.


r~