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]
Other format: [Raw text]

Re: [tree-ssa] More aliasing fixes


On Thu, 2003-12-18 at 10:05, law@redhat.com wrote:
> In message <1071759097.4996.17.camel@frodo.toronto.redhat.com>, Diego Novillo w
> rites:
>  >On Thu, 2003-12-18 at 09:31, law@redhat.com wrote:
>  >
>  >> He's referring to the SGI paper which talks about how to handle aliasing
>  >> and memory operands in SSA form.
>  >> 
>  >Dunno about a paper, but I do have a tutorial the ORC team presented in
>  >PACT'02
>  >
>  >http://people.redhat.com/dnovillo/pub/ORC-PACT02-tutorial.pdf
> I'm referring to:
> 
> F. C. Chow, S. Chan, S.-M. Liu, R. Lo, and M. Streich, ``Effective 
> Representation of Aliases and Indirect Memory Operations in SSA Form
> 
Ah, OK.  Yes, it's essentially the same info that they presented in
PACT.  The key differences with our approach are:

- We do not allow mixing real and virtual operands.  So, if we wanted to
do zero-versioning we would not need to jump through hoops like they do.

- We do not have multi-level pointers, so we just need to have a single
virtual variable (memory tag) per pointer.

- We don't build the hashed SSA form.  I haven't really looked into it
in detail, though nothing in particular stood out when I read what they
do.


Diego.


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