From 491ca2425d4a58bf1d732c5a051e5714a23ca198 Mon Sep 17 00:00:00 2001 From: Dorit Nuzman Date: Wed, 14 Aug 2019 05:22:20 +0000 Subject: [LV] Fold-tail flag This is the compiler-flag equivalent of the Predicate pragma (https://reviews.llvm.org/D65197), to direct the vectorizer to fold the remainder-loop into the main-loop using predication. Differential Revision: https://reviews.llvm.org/D66108 Reviewers: Ayal, hsaito, fhahn, SjoerdMeije llvm-svn: 368801 --- llvm/lib/Transforms/Vectorize/LoopVectorize.cpp | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) (limited to 'llvm/lib/Transforms') diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp index 1f8cbf7f340..20305416079 100644 --- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp +++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp @@ -177,6 +177,14 @@ static cl::opt TinyTripCountVectorThreshold( "value are vectorized only if no scalar iteration overheads " "are incurred.")); +// Indicates that an epilogue is undesired, predication is preferred. +// This means that the vectorizer will try to fold the loop-tail (epilogue) +// into the loop and predicate the loop body accordingly. +static cl::opt PreferPredicateOverEpilog( + "prefer-predicate-over-epilog", cl::init(false), cl::Hidden, + cl::desc("Indicate that an epilogue is undesired, predication should be " + "used instead.")); + static cl::opt MaximizeBandwidth( "vectorizer-maximize-bandwidth", cl::init(false), cl::Hidden, cl::desc("Maximize bandwidth when selecting vectorization factor which " @@ -906,7 +914,7 @@ enum ScalarEpilogueLowering { CM_ScalarEpilogueNotAllowedLowTripLoop, // Loop hint predicate indicating an epilogue is undesired. - CM_ScalarEpilogueNotNeededPredicatePragma + CM_ScalarEpilogueNotNeededUsePredicate }; /// LoopVectorizationCostModel - estimates the expected speedups due to @@ -4804,9 +4812,9 @@ Optional LoopVectorizationCostModel::computeMaxVF() { switch (ScalarEpilogueStatus) { case CM_ScalarEpilogueAllowed: return computeFeasibleMaxVF(TC); - case CM_ScalarEpilogueNotNeededPredicatePragma: + case CM_ScalarEpilogueNotNeededUsePredicate: LLVM_DEBUG( - dbgs() << "LV: vector predicate hint found.\n" + dbgs() << "LV: vector predicate hint/switch found.\n" << "LV: Not allowing scalar epilogue, creating predicated " << "vector loop.\n"); break; @@ -7298,8 +7306,8 @@ getScalarEpilogueLowering(Function *F, Loop *L, LoopVectorizeHints &Hints, (F->hasOptSize() || llvm::shouldOptimizeForSize(L->getHeader(), PSI, BFI))) SEL = CM_ScalarEpilogueNotAllowedOptSize; - else if (Hints.getPredicate()) - SEL = CM_ScalarEpilogueNotNeededPredicatePragma; + else if (PreferPredicateOverEpilog || Hints.getPredicate()) + SEL = CM_ScalarEpilogueNotNeededUsePredicate; return SEL; } -- cgit v1.2.3