|SSA Versus DU-chains: Is the worst case size really different? email@example.com (Bageshri Sathe) (2005-07-26)|
|Re: SSA Versus DU-chains: Is the worst case size really different? firstname.lastname@example.org (Mayan Moudgill) (2005-07-28)|
|Re: SSA Versus DU-chains: Is the worst case size really different? email@example.com (Jeremy Singer) (2005-08-05)|
|From:||Mayan Moudgill <firstname.lastname@example.org>|
|Date:||28 Jul 2005 02:30:47 -0400|
|Organization:||Posted via Supernews, http://www.supernews.com|
|Posted-Date:||28 Jul 2005 02:30:43 EDT|
Bageshri Sathe wrote:
> Hello everyone,
> All articles which I referred to (Including Wegman and Zadeck's conditional
> constant propagation using SSA and many other technical papers) state that
> SSA is better than DU-chains since its worst case as well as practical
> size is less than number of DU-chains.
> But I am unable understand why - especially the worst case.
I'd suggest looking at Richar Johnson's thesis
His thesis compares sparse representations (and the CFG), on both
worst-case and actual programs. Net result - (my interpretation) SSAs
don't give very much benefit.
Return to the
Search the comp.compilers archives again.