On Fri, 4 Jun 2010, Richard Guenther wrote: > The mark_addressable_vector change was at least semi-ugly. I think > it is reasonable to require a constant index for register qualified vectors.
What does the Cell document describing the extension being implemented by the original patch say here? Is this implementing the extension more accurately, or deliberately deviating from it? Either may be OK - if properly justified. > Artem is one of our Goole Summer of Code students btw., and I > requested these changes. Details of the request? Again, I'd like (once the copyright assignment is complete) to understand what changed from the previous patch and why. This patch submission does not make things easy to review; exact URLs for the previous rounds of discussion are helpful when dealing with a patch that has undergone multiple past reviews long ago. (The opposite error is only giving URLs without a self-contained explanation of what the resubmission is about; this submission avoids that particular problem, although the Cell document would still seem to be relevant in understanding the rationale for the patch.) -- Joseph S. Myers jos...@codesourcery.com