[patch] [PR25413] fix natural-alignment assumption in the vectorizer

[Dorit Nuzman]

David Edelsohn <dje@watson.ibm.com> wrote on 10/07/2007 18:21:03:

>    I do not understand the intention of this patch for rs6000.  Is
> the new hook only implemented for supported vector types?  ALIGN_POWER is
> not natural alignment for 64-bit double, so I do not understand why the
> hook returns true for MASK_ALIGN_POWER under TARGET_MACHO and does not
> include a test for AIX.
>

this is exactly what I meant by "I'm not 100% sure about the rs6000
implementation"... :-) ("but even if it's not accurate enough, it is still
an improvement over the current situation."). Is the following ok, or do we
need additional specific tests for AIX?

+ static bool
+ rs6000_vector_alignment_reachable (tree type, bool is_packed)
+ {
+   if (is_packed)
+     return false;
+
+   if (TARGET_32BIT)
+     {
+       if (rs6000_alignment_flags == MASK_ALIGN_NATURAL)
+         return true;
+
+       if (rs6000_alignment_flags ==  MASK_ALIGN_POWER)
+         return true;
+     }
+   else
+     {
+       if (TARGET_MACHO)
+         return false;
+
+       /* Assuming that all other types are naturally aligned. CHECKME!
*/
+       return true;
+     }
+
+ }

thanks,
dorit

> Thanks, David
>