Merged ARM-Thumb branch...

Philip Blundell
Wed Feb 9 12:24:00 GMT 2000

>If you do submit it then they are almost certainly going to want to
>know why this patch is necessary...

Richard's analysis (see the archives) explains most of it from what I remember.

>Also - since most to the two functions are the same, why duplicate the
>code ?  Wouldn't it be easier to perform the addressof specific tests
>first and then just tailcall rtx_equal_p() if they all pass ?

Well, the `for_addressof' version actually performs a subset of the full 
checks.  I'd have to check the code again to remind myself exactly what goes 
on, but the basic idea was that two values were equal for addressof purposes 
even if they were in different modes.


More information about the Gcc mailing list