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: PR 32919 - value numbering versus computed goto


On Wed, Aug 01, 2007 at 12:33:35PM -0400, Daniel Berlin wrote:
> This is fine.
> 
> It may actually be easier to simply initialize them all to their own
> values in init_sccvn (where we init them to VN_TOP), and mark them
> visited (IE VN_INFO (name)->visited = true).
> 
> This would prevent them from ever being visited in the first place,
> and will also stop it from needing to iterate for abnormal phis (IE
> because the value will have been vn_top, and is now not, it will
> consider it changed)
> 
> However, from a correctness perspective, what you have done below will
> work fine as well, so if you don't want to do the above, just throw a
> todo comment and i'll get to it in a week or so.

Thanks!  I added this:

  /* TODO: We could check for this in init_sccvn, and replace this
     with a gcc_assert.  */

-- 
Daniel Jacobowitz
CodeSourcery


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