Hi Richard, Did you have a chance to look at my updated patch?
Any comments will be appreciated. Yuri. 2015-05-20 17:00 GMT+03:00 Yuri Rumyantsev <ysrum...@gmail.com>: > Hi All, > > Here is updated patch to optimize mask stores. The main goal of it is > to avoid execution of mask store if its mask is zero vector since > loads that follow it can be blocked. > The following changes were done: > 1. A test on sink legality was added - it simply prohibits to cross > statements with non-null vdef or vuse. > 2. New phi node is created in join block for moved MASK_STORE statements. > 3. Test was changed to check that 2 MASK_STORE statements are not > moved to the same block. > 4. New field was added to loop_vec_info structure to mark loops > having MASK_STORE's. > > Any comments will be appreciated. > Yuri. > > 2015-05-20 Yuri Rumyantsev <ysrum...@gmail.com> > > * config/i386/i386.c: Include files stringpool.h and tree-ssanames.h. > (ix86_vectorize_is_zero_vector): New function. > (TARGET_VECTORIZE_IS_ZERO_VECTOR): New target macro > * doc/tm.texi.in: Add @hook TARGET_VECTORIZE_IS_ZERO_VECTOR. > * doc/tm.texi: Updated. > * target.def (is_zero_vector): New DEFHOOK. > * tree-vect-stmts.c : Include tree-into-ssa.h. > (vectorizable_mask_load_store): Initialize has_mask_store field. > (is_valid_sink): New function. > (optimize_mask_stores): New function. > * tree-vectorizer.c (vectorize_loops): Invoke optimaze_mask_stores for > loops having masked stores. > * tree-vectorizer.h (loop_vec_info): Add new has_mask_store field and > correspondent macros. > (optimize_mask_stores): Update prototype. > > gcc/testsuite/ChangeLog: > * gcc.target/i386/avx2-vect-mask-store-move1.c: New test.
