Lines Matching defs:passes

47 through minimal IR layering, lean data structures and passes, and a careful
48 selection of fast optimization passes.  It has two optimization recipes: full
208 multiple translation threads).  The recipe for optimization passes depends on
388 - Do analysis passes and transformation passes on the high-level ICE, as
396 - Do more analysis and transformation passes on the low-level ICE, as desired.
438 optimization passes.  It also helps determine whether a code correctness problem
442 The implementation of target lowering also controls the recipe of passes used
460 analysis passes, mostly related to register allocation, stack slot assignment,
545 These passes are described in more detail below.
558 Optimization passes
691 allocation passes end up being extremely expensive.  This is because of the
693 small but ends up being quite large on subsequent passes.  Its performance can
909 On x86, with good register allocation, the Subzero passes described above
930 passes, it's not attractive to run it an extra time just for RMW analysis.
1034 There are some situations where passes need to use some `STL container class
1171   logical operations or passes of interest.  Basic timing information plus
1295 - 27% for liveness analysis (two liveness passes plus live range construction)
1376 objects are allocated from the CFG-local allocator, some passes use temporary