[PATCH] D50822: [BFI] Use rounding while computing profile counts.

[Easwaran Raman via Phabricator via llvm-commits]

eraman created this revision.
eraman added a reviewer: davidxl.

Profile count of a block is computed by multiplying its block frequency
by entry count and dividing the result by entry block frequency. Do
rounded division in the last step and update test cases appropriately.


Repository:
  rL LLVM

https://reviews.llvm.org/D50822

Files:
  lib/Analysis/BlockFrequencyInfoImpl.cpp
  test/Analysis/BlockFrequencyInfo/irreducible_pgo.ll
  test/Instrumentation/cgprofile.ll
  unittests/Analysis/BlockFrequencyInfoTest.cpp


Index: unittests/Analysis/BlockFrequencyInfoTest.cpp
===================================================================
--- unittests/Analysis/BlockFrequencyInfoTest.cpp
+++ unittests/Analysis/BlockFrequencyInfoTest.cpp
@@ -77,8 +77,10 @@
 
   EXPECT_EQ(BFI.getBlockProfileCount(&BB0).getValue(), UINT64_C(100));
   EXPECT_EQ(BFI.getBlockProfileCount(BB3).getValue(), UINT64_C(100));
-  EXPECT_EQ(BFI.getBlockProfileCount(BB1).getValue(), 100 * BB1Freq / BB0Freq);
-  EXPECT_EQ(BFI.getBlockProfileCount(BB2).getValue(), 100 * BB2Freq / BB0Freq);
+  EXPECT_EQ(BFI.getBlockProfileCount(BB1).getValue(),
+            (100 * BB1Freq + BB0Freq / 2) / BB0Freq);
+  EXPECT_EQ(BFI.getBlockProfileCount(BB2).getValue(),
+            (100 * BB2Freq + BB0Freq / 2) / BB0Freq);
 
   // Scale the frequencies of BB0, BB1 and BB2 by a factor of two.
   SmallPtrSet<BasicBlock *, 4> BlocksToScale({BB1, BB2});
Index: test/Instrumentation/cgprofile.ll
===================================================================
--- test/Instrumentation/cgprofile.ll
+++ test/Instrumentation/cgprofile.ll
@@ -38,4 +38,4 @@
 ; CHECK: ![[e3]] = !{void (i1)* @freq, i32 ()* @func3, i64 150}
 ; CHECK: ![[e4]] = !{void (i1)* @freq, i32 ()* @func1, i64 10}
 ; CHECK: ![[e5]] = !{void (i1)* @freq, void ()* @a, i64 11}
-; CHECK: ![[e6]] = !{void (i1)* @freq, void ()* @b, i64 20}
+; CHECK: ![[e6]] = !{void (i1)* @freq, void ()* @b, i64 21}
Index: test/Analysis/BlockFrequencyInfo/irreducible_pgo.ll
===================================================================
--- test/Analysis/BlockFrequencyInfo/irreducible_pgo.ll
+++ test/Analysis/BlockFrequencyInfo/irreducible_pgo.ll
@@ -146,19 +146,19 @@
 ; CHECK-NEXT: block-frequency-info: _Z11irreduciblePh
 ; CHECK-NEXT: - entry: {{.*}} count = 1
 ; CHECK-NEXT: - for.cond1: {{.*}} count = 1
-; CHECK-NEXT: - dispatch_op: {{.*}} count = 201
-; CHECK-NEXT: - dispatch_op.sw.bb6_crit_edge: {{.*}} count = 200
-; CHECK-NEXT: - sw.bb: {{.*}} count = 0
-; CHECK-NEXT: - TARGET_1: {{.*}} count = 299
-; CHECK-NEXT: - sw.bb6: {{.*}} count = 500, irr_loop_header_weight = 501
-; CHECK-NEXT: - if.then: {{.*}} count = 299
-; CHECK-NEXT: - TARGET_2: {{.*}} count = 98
-; CHECK-NEXT: - sw.bb15: {{.*}} count = 99, irr_loop_header_weight = 100
-; CHECK-NEXT: - if.then18: {{.*}} count = 99
+; CHECK-NEXT: - dispatch_op: {{.*}} count = 202
+; CHECK-NEXT: - dispatch_op.sw.bb6_crit_edge: {{.*}} count = 201
+; CHECK-NEXT: - sw.bb: {{.*}} count = 1
+; CHECK-NEXT: - TARGET_1: {{.*}} count = 300
+; CHECK-NEXT: - sw.bb6: {{.*}} count = 501, irr_loop_header_weight = 501
+; CHECK-NEXT: - if.then: {{.*}} count = 300
+; CHECK-NEXT: - TARGET_2: {{.*}} count = 99
+; CHECK-NEXT: - sw.bb15: {{.*}} count = 100, irr_loop_header_weight = 100
+; CHECK-NEXT: - if.then18: {{.*}} count = 100
 ; CHECK-NEXT: - unknown_op: {{.*}} count = 0
 ; CHECK-NEXT: - sw.default: {{.*}} count = 0
 ; CHECK-NEXT: - exit: {{.*}} count = 1
-; CHECK-NEXT: - indirectgoto: {{.*}} count = 399, irr_loop_header_weight = 400
+; CHECK-NEXT: - indirectgoto: {{.*}} count = 400, irr_loop_header_weight = 400
 
 ; Missing some irr loop annotations.
 ; Function Attrs: noinline norecurse nounwind uwtable
Index: lib/Analysis/BlockFrequencyInfoImpl.cpp
===================================================================
--- lib/Analysis/BlockFrequencyInfoImpl.cpp
+++ lib/Analysis/BlockFrequencyInfoImpl.cpp
@@ -573,7 +573,9 @@
   APInt BlockFreq(128, Freq);
   APInt EntryFreq(128, getEntryFreq());
   BlockCount *= BlockFreq;
-  BlockCount = BlockCount.udiv(EntryFreq);
+  // Rounded division of BlockCount by EntryFreq. Since EntryFreq is unsigned
+  // lshr by 1 gives EntryFreq/2.
+  BlockCount = (BlockCount + EntryFreq.lshr(1)).udiv(EntryFreq);
   return BlockCount.getLimitedValue();
 }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D50822.160948.patch
Type: text/x-patch
Size: 3766 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180816/3ba2b657/attachment.bin>