On Tue, Apr 17, 2012 at 8:49 AM, Jan Hubicka <hubi...@ucw.cz> wrote: >> Hello, >> >> There is code in stmt.c since the initial checkin, that tries to >> balance a switch tree according to some ascii heuristics. I see a >> couple of problems with this code: >> >> 1. It doesn't seem to help much. With the attached patch to remove the >> code, I see no compile time changes to e.g. compile GCC itself. >> >> 2. It isn't clear what the heuristic is based on (no reference to any >> testing done, or a reference to a book or paper). >> >> 3. The heuristic is applied for case values in the range <-1,127> >> (inclusive) even if the type of the switch expression isn't char or >> int but e.g. an enum. This results in funny application of this >> heuristic in GCC itself to e.g. some cases of enum rtx_code and enum >> tree_code. > > Note that it would make a lot of sense to teach this heuristics predict.c > and properly identify chars.
Indeed this would be the proper place to implement this logic. > Also it is possble to get an historgrams from profile feedback into > switch expansion. I always wanted to do that once switch expansion code > is cleaned up and moved to gimple level... Indeed. At least the parts that expand switch stmts to (balanced) trees should be moved to the GIMPLE level, retaining only the table-jump-like expansions as switch stmts. >> >> >> The attached patch removes the heuristic. >> >> Bootstrapped and tested on powerpc-unknown-linux-gnu. OK for trunk? Ok. Thanks, Richard. >> Ciao! >> Steven > >
