On Thu, Nov 13, 2014 at 05:49:47PM -0700, Jeff Law wrote: > On 11/12/14 22:55, tsaund...@mozilla.com wrote: > >From: Trevor Saunders <tsaund...@mozilla.com> > > > >Hi, > > > >The only user of this is splay_tree. We only have a couple splay trees in > >ggc > >memory, and it wasn't clear to me any of them were tree based instead of hash > >based for performance reasons, so I chose to just convert them to hash_map > >rather than writing a templated splay tree. > > > >bootstrapped + regtested x86_64-unknown-linux-gnu, ok? > > > >trev > > > >gcc/ > > > > * hash-map.h (hash_map::iterator): New class. > > (hash_map::begin): New method. > > (hash_map::end): Likewise. > > * alias.c, config/alpha/alpha.c, dwarf2asm.c, omp-low.c, tree.h: > > replace splay_tree with hash_map. > Splay trees in the alias code are very very old. They came in with the > initial "restrict" support from Mark Mitchell many many years ago. I don't > recall the reasoning behind using splay trees instead of hash tables or some > other data structure.
fair enough, I could believe it was a case of what was handy, splay_tree does make a nicer map than htab. > In cases where you're removing the last remnants of a splay tree in a file, > please make sure to remove the splay-tree include. sure > OK with the splay-tree include removal, but if we start to get reports that > these data structures have become a bottleneck, we'll be looking to you to > address that problem ;-) of course, a C++ splay tree shouldn't be too hard, but the only use for it I can see is ggc stuff where std::map won't really work. thanks! Trev > > jeff >
