[PATCH] D43813: [Machine Combiner] Valid use of OptSize

Andrew V. Tischenko via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Feb 27 02:33:51 PST 2018


avt77 created this revision.
avt77 added reviewers: Gerolf, RKSimon, spatel, dtemirbulatov.

Currently we're able to select longer alternative code sequence if it has better sched numbers but we should always select the shortest code sequence if OptSize == true. This patch fixes the given issue.


https://reviews.llvm.org/D43813

Files:
  lib/CodeGen/MachineCombiner.cpp


Index: lib/CodeGen/MachineCombiner.cpp
===================================================================
--- lib/CodeGen/MachineCombiner.cpp
+++ lib/CodeGen/MachineCombiner.cpp
@@ -407,8 +407,11 @@
 /// \returns true when new instruction sequence should be generated
 /// independent if it lengthens critical path or not
 bool MachineCombiner::doSubstitute(unsigned NewSize, unsigned OldSize) {
-  if (OptSize && (NewSize < OldSize))
-    return true;
+  if (OptSize) { // First of all check OptSize option
+    if (NewSize < OldSize)
+      return true; // Do substitute in any case if new size < old size
+    return false;  // We should not do substitute if new size >= old size
+  }
   if (!TSchedModel.hasInstrSchedModelOrItineraries())
     return true;
   return false;
@@ -588,7 +591,7 @@
         // Eagerly stop after the first pattern fires.
         Changed = true;
         break;
-      } else {
+      } else if (!(OptSize && (NewInstCount > OldInstCount))) {
         // For big basic blocks, we only compute the full trace the first time
         // we hit this. We do not invalidate the trace, but instead update the
         // instruction depths incrementally.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D43813.136054.patch
Type: text/x-patch
Size: 1183 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180227/6f0dad08/attachment.bin>


More information about the llvm-commits mailing list