[PATCH] Middle-end dead code elimination

Mark Mitchell mark@codesourcery.com
Fri Apr 18 15:33:00 GMT 2003


O> code in the first place.  We have the following in
begin_destructor_body:
>   /* If it is not safe to avoid setting up the vtables, then
>      someone will change the condition to be boolean_true_node.
>      (Actually, for now, we do not have code to set the condition
>      appropriately, so we just assume that we always need to
>      initialize the vtables.)  */
>   finish_if_stmt_cond (boolean_true_node, if_stmt);
>   
> Presumably this sillyness is necessary (Jason???!?).  I'm nowhere near
> fluent enough to know what's really going on here.

There is a potential C++-specific optimization to be done here, but we
don't actually do it.

The right thing to do is to remove that code.  I'll try to get to that;
I'm trying to speed up something else that might win from this as well.

Thanks,

-- 
Mark Mitchell
CodeSourcery, LLC
mark@codesourcery.com



More information about the Gcc-patches mailing list