[PATCH] D57382: [LV] Move interleave count computation to LVP::plan().

Diego Caballero via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Feb 5 18:19:48 PST 2019


dcaballe added inline comments.


================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:6134
+      LLVM_DEBUG(dbgs() << "LV: Invalidate all interleaved groups due to "
+                           "fold-tail by masking "
+                           "which requires masked-interleaved support.\n");
----------------
Merge with next line?


================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:6154
+    if (MaxVF > 1) {
+      for (unsigned VF = 1; VF <= MaxVF; VF *= 2) {
+        // Collect Uniform and Scalar instructions after vectorization with VF.
----------------
Shouldn't `VF` now starts from 1 and 6160 be removed?


================
Comment at: llvm/lib/Transforms/Vectorize/LoopVectorize.cpp:6170
 
-  // Select the optimal vectorization factor.
-  return CM.selectVectorizationFactor(MaxVF);
+  // Build VPlans if we either have multiple possible vectorization factors or
+  // decided to interleave.
----------------
multiple? do you mean at least one?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57382/new/

https://reviews.llvm.org/D57382





More information about the llvm-commits mailing list