[MBP] Add flags to disable the BadCFGConflict check in MachineBlockPlacement.

Some benchmarks have shown that this could lead to a potential
performance benefit, and so adding some flags to try to help measure the
difference.

A possible explanation. In diamond-shaped CFGs (A followed by either
B or C both followed by D), putting B and C both in between A and
D leads to the code being less dense than it could be. Always either
B or C have to be skipped increasing the chance of cache misses etc.
Moving either B or C to after D might be beneficial on average.

In the long run, but we should probably do a better job of analyzing the
basic block and branch probabilities to move the correct one of B or
C to after D. But even if we don't use this in the long run, it is
a good baseline for benchmarking.

Original patch authored by Daniel Jasper with test tweaks and a second
flag added by me.

Differential Revision: http://reviews.llvm.org/D6969

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@226034 91177308-0d34-0410-b5e6-96231b3b80d8

lib/CodeGen/MachineBlockPlacement.cpp

@@ -60,6 +60,17 @@ static cl::opt<unsigned> AlignAllBlock("align-all-blocks",
                                                 "blocks in the function."),
                                        cl::init(0), cl::Hidden);
 
+static cl::opt<bool> OnlyHotBadCFGConflictCheck(
+    "only-hot-bad-cfg-conflict-check",
+    cl::desc("Only check that a hot successor doesn't have a hot predecessor."),
+    cl::init(false), cl::Hidden);
+
+static cl::opt<bool> NoBadCFGConflictCheck(
+    "no-bad-cfg-conflict-check",
+    cl::desc("Don't check whether a hot successor has a more important "
+             "predecessor."),
+    cl::init(false), cl::Hidden);
+
 // FIXME: Find a good default for this flag and remove the flag.
 static cl::opt<unsigned>
 ExitBlockBias("block-placement-exit-block-bias",
@@ -374,28 +385,32 @@ MachineBasicBlock *MachineBlockPlacement::selectBestSuccessor(
         continue;
       }
 
-      // Make sure that a hot successor doesn't have a globally more important
+      if (!NoBadCFGConflictCheck) {
-      // predecessor.
+        // Make sure that a hot successor doesn't have a globally more
-      BlockFrequency CandidateEdgeFreq
+        // important predecessor.
-        = MBFI->getBlockFreq(BB) * SuccProb * HotProb.getCompl();
+        BlockFrequency CandidateEdgeFreq =
-      bool BadCFGConflict = false;
+            OnlyHotBadCFGConflictCheck
-      for (MachineBasicBlock::pred_iterator PI = (*SI)->pred_begin(),
+                ? MBFI->getBlockFreq(BB) * SuccProb
-                                            PE = (*SI)->pred_end();
+                : MBFI->getBlockFreq(BB) * SuccProb * HotProb.getCompl();
-           PI != PE; ++PI) {
+        bool BadCFGConflict = false;
-        if (*PI == *SI || (BlockFilter && !BlockFilter->count(*PI)) ||
+        for (MachineBasicBlock::pred_iterator PI = (*SI)->pred_begin(),
-            BlockToChain[*PI] == &Chain)
+                                              PE = (*SI)->pred_end();
-          continue;
+             PI != PE; ++PI) {
-        BlockFrequency PredEdgeFreq
+          if (*PI == *SI || (BlockFilter && !BlockFilter->count(*PI)) ||
-          = MBFI->getBlockFreq(*PI) * MBPI->getEdgeProbability(*PI, *SI);
+              BlockToChain[*PI] == &Chain)
-        if (PredEdgeFreq >= CandidateEdgeFreq) {
+            continue;
-          BadCFGConflict = true;
+          BlockFrequency PredEdgeFreq =
-          break;
+              MBFI->getBlockFreq(*PI) * MBPI->getEdgeProbability(*PI, *SI);
+          if (PredEdgeFreq >= CandidateEdgeFreq) {
+            BadCFGConflict = true;
+            break;
+          }
+        }
+        if (BadCFGConflict) {
+          DEBUG(dbgs() << "    " << getBlockName(*SI) << " -> " << SuccProb
+                       << " (prob) (non-cold CFG conflict)\n");
+          continue;
         }
-      }
-      if (BadCFGConflict) {
-        DEBUG(dbgs() << "    " << getBlockName(*SI) << " -> " << SuccProb
-                     << " (prob) (non-cold CFG conflict)\n");
-        continue;
       }
     }
 

test/CodeGen/X86/code_placement_bad_cfg_check.ll

@@ -0,0 +1,112 @@
+; RUN: llc -mcpu=corei7 -mtriple=x86_64-linux < %s | FileCheck %s -check-prefix=CHECK-BAD-CFG
+; RUN: llc -mcpu=corei7 -mtriple=x86_64-linux -no-bad-cfg-conflict-check < %s | FileCheck %s -check-prefix=CHECK-NO-BAD-CFG
+; RUN: llc -mcpu=corei7 -mtriple=x86_64-linux -only-hot-bad-cfg-conflict-check < %s | FileCheck %s -check-prefix=CHECK-HOT-BAD-CFG
+
+define void @foo(i32 %t) {
+; Test that we lift the call to 'c' up to immediately follow the call to 'b'
+; when we disable the cfg conflict check.
+;
+; CHECK-BAD-CFG-LABEL: foo:
+; CHECK-BAD-CFG: callq b
+; CHECK-BAD-CFG: callq a
+; CHECK-BAD-CFG: callq c
+;
+; CHECK-NO-BAD-CFG-LABEL: foo:
+; CHECK-NO-BAD-CFG: callq b
+; CHECK-NO-BAD-CFG: callq c
+; CHECK-NO-BAD-CFG: callq a
+;
+; CHECK-HOT-BAD-CFG-LABEL: foo:
+; CHECK-HOT-BAD-CFG: callq b
+; CHECK-HOT-BAD-CFG: callq c
+; CHECK-HOT-BAD-CFG: callq a
+
+entry:
+  %cmp = icmp eq i32 %t, 0
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  call void @a()
+  br label %if.end
+
+if.else:
+  call void @b()
+  br label %if.end
+
+if.end:
+  call void @c()
+  ret void
+}
+
+define void @bar(i32 %t1, i32 %t2, i32 %t3) {
+; Test that we lift the call to 'c' up to immediately follow the call to 'b'
+; when we disable the cfg conflict check.
+;
+; CHECK-BAD-CFG-LABEL: bar:
+; CHECK-BAD-CFG: callq a
+; CHECK-BAD-CFG: callq c
+; CHECK-BAD-CFG: callq d
+; CHECK-BAD-CFG: callq f
+; CHECK-BAD-CFG: callq b
+; CHECK-BAD-CFG: callq e
+; CHECK-BAD-CFG: callq g
+;
+; CHECK-NO-BAD-CFG-LABEL: bar:
+; CHECK-NO-BAD-CFG: callq a
+; CHECK-NO-BAD-CFG: callq c
+; CHECK-NO-BAD-CFG: callq g
+; CHECK-NO-BAD-CFG: callq d
+; CHECK-NO-BAD-CFG: callq f
+; CHECK-NO-BAD-CFG: callq b
+; CHECK-NO-BAD-CFG: callq e
+;
+; CHECK-HOT-BAD-CFG-LABEL: bar:
+; CHECK-HOT-BAD-CFG: callq a
+; CHECK-HOT-BAD-CFG: callq c
+; CHECK-HOT-BAD-CFG: callq d
+; CHECK-HOT-BAD-CFG: callq f
+; CHECK-HOT-BAD-CFG: callq g
+; CHECK-HOT-BAD-CFG: callq b
+; CHECK-HOT-BAD-CFG: callq e
+
+entry:
+  br i1 undef, label %if1.then, label %if1.else
+
+if1.then:
+  call void @a()
+  %cmp2 = icmp eq i32 %t2, 0
+  br i1 %cmp2, label %if2.then, label %if2.else
+
+if2.then:
+  call void @b()
+  br label %if.end
+
+if2.else:
+  call void @c()
+  br label %if.end
+
+if1.else:
+  call void @d()
+  %cmp3 = icmp eq i32 %t3, 0
+  br i1 %cmp3, label %if3.then, label %if3.else
+
+if3.then:
+  call void @e()
+  br label %if.end
+
+if3.else:
+  call void @f()
+  br label %if.end
+
+if.end:
+  call void @g()
+  ret void
+}
+
+declare void @a()
+declare void @b()
+declare void @c()
+declare void @d()
+declare void @e()
+declare void @f()
+declare void @g()