[llvm-commits] CVS: llvm/docs/ReleaseNotes.html

[Chris Lattner]

On Thu, 12 Feb 2004, Alkis Evlogimenos wrote:
> > Index: llvm/docs/ReleaseNotes.html
> > diff -u llvm/docs/ReleaseNotes.html:1.119 llvm/docs/ReleaseNotes.html:1.120
> > --- llvm/docs/ReleaseNotes.html:1.119	Mon Feb  9 23:22:23 2004
> > +++ llvm/docs/ReleaseNotes.html	Thu Feb 12 15:21:17 2004
> > @@ -145,6 +145,7 @@
> >  <ol>
> >  <li><a href="http://llvm.cs.uiuc.edu/PR187">[loopsimplify] Many pointless phi nodes are created</a></li>
> >  <li>The X86 backend didn't generate <tt>fchs</tt> to negate floating point numbers</li>
> > +<li>The X86 backend didn't expand memcpy() into the <tt>rep movs</tt> instruction</li>
> >  </ol>
>
> Does gcc do this? Is it faster than the normal code due to
> implementation or because the code is much more compact?

Yes, GCC does it.  Both.  Also it's typical for the hardware to implement
prefetching for large copies...

-Chris

-- 
http://llvm.cs.uiuc.edu/
http://www.nondot.org/~sabre/Projects/