[PATCH] D137220: [LV][IVDescriptors] Fix recurrence identity element for FMin and FMax reductions.

[Karthik Senthil via Phabricator via llvm-commits]

karthiksenthil created this revision.
karthiksenthil added a reviewer: spatel.
Herald added a subscriber: hiraditya.
Herald added a project: All.
karthiksenthil requested review of this revision.
Herald added subscribers: llvm-commits, pcwang-thead.
Herald added a project: LLVM.

For a min and max reduction idioms, the identity (i.e. neutral) element
should be datatype's highest and lowest possible values respectively.
Current implementation in IVDescriptors incorrectly returns -Inf for FMin
reduction and +Inf for FMax reduction. This patch fixes this bug which
was causing incorrect reduction computation results in loops vectorized
by LV.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D137220

Files:
  llvm/lib/Analysis/IVDescriptors.cpp
  llvm/test/Transforms/LoopVectorize/reduction-inloop-cond.ll


Index: llvm/test/Transforms/LoopVectorize/reduction-inloop-cond.ll
===================================================================
--- llvm/test/Transforms/LoopVectorize/reduction-inloop-cond.ll
+++ llvm/test/Transforms/LoopVectorize/reduction-inloop-cond.ll
@@ -168,7 +168,7 @@
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE6]]
 ; CHECK:       pred.load.continue6:
 ; CHECK-NEXT:    [[TMP25:%.*]] = phi <4 x float> [ [[TMP19]], [[PRED_LOAD_CONTINUE4]] ], [ [[TMP24]], [[PRED_LOAD_IF5]] ]
-; CHECK-NEXT:    [[TMP26:%.*]] = select fast <4 x i1> [[TMP2]], <4 x float> [[TMP25]], <4 x float> <float 0xFFF0000000000000, float 0xFFF0000000000000, float 0xFFF0000000000000, float 0xFFF0000000000000>
+; CHECK-NEXT:    [[TMP26:%.*]] = select fast <4 x i1> [[TMP2]], <4 x float> [[TMP25]], <4 x float> <float 0x7FF0000000000000, float 0x7FF0000000000000, float 0x7FF0000000000000, float 0x7FF0000000000000>
 ; CHECK-NEXT:    [[TMP27:%.*]] = call fast float @llvm.vector.reduce.fmin.v4f32(<4 x float> [[TMP26]])
 ; CHECK-NEXT:    [[TMP28]] = call fast float @llvm.minnum.f32(float [[TMP27]], float [[VEC_PHI]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
Index: llvm/lib/Analysis/IVDescriptors.cpp
===================================================================
--- llvm/lib/Analysis/IVDescriptors.cpp
+++ llvm/lib/Analysis/IVDescriptors.cpp
@@ -1111,9 +1111,9 @@
     return ConstantInt::get(Tp,
                             APInt::getSignedMinValue(Tp->getIntegerBitWidth()));
   case RecurKind::FMin:
-    return ConstantFP::getInfinity(Tp, true);
+    return ConstantFP::getInfinity(Tp, false /*Negative*/);
   case RecurKind::FMax:
-    return ConstantFP::getInfinity(Tp, false);
+    return ConstantFP::getInfinity(Tp, true /*Negative*/);
   case RecurKind::SelectICmp:
   case RecurKind::SelectFCmp:
     return getRecurrenceStartValue();


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D137220.472470.patch
Type: text/x-patch
Size: 1862 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221102/f0cacec3/attachment.bin>