From: Philip Reames <listmail@philipreames.com>
Date: Thu, 18 Apr 2019 17:01:19 +0000 (+0000)
Subject: [LoopPred] Fix a blatantly obvious bug in r358684
X-Git-Url: https://granicus.if.org/sourcecode?a=commitdiff_plain;h=382ffde1a121eb6828c24272c01d48e0830406c2;p=llvm

[LoopPred] Fix a blatantly obvious bug in r358684

The bug is that I didn't check whether the operand of the invariant_loads were themselves invariant.  I don't know how this got missed in the patch and review.  I even had an unreduced test case locally, and I remember handling this case, but I must have lost it in one of the rebases.  Oops.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@358688 91177308-0d34-0410-b5e6-96231b3b80d8
---

diff --git a/lib/Transforms/Scalar/LoopPredication.cpp b/lib/Transforms/Scalar/LoopPredication.cpp
index 0fc796eae18..15e5b6433af 100644
--- a/lib/Transforms/Scalar/LoopPredication.cpp
+++ b/lib/Transforms/Scalar/LoopPredication.cpp
@@ -510,7 +510,7 @@ bool LoopPredication::isLoopInvariantValue(const SCEV* S) {
   // TODO: This should be sunk inside SCEV.
   if (const SCEVUnknown *U = dyn_cast<SCEVUnknown>(S))
     if (const auto *LI = dyn_cast<LoadInst>(U->getValue()))
-      if (LI->isUnordered())
+      if (LI->isUnordered() && L->hasLoopInvariantOperands(LI))
         if (AA->pointsToConstantMemory(LI->getOperand(0)) ||
             LI->getMetadata(LLVMContext::MD_invariant_load) != nullptr)
           return true;
diff --git a/test/Transforms/LoopPredication/invariant_load.ll b/test/Transforms/LoopPredication/invariant_load.ll
index 543df0a23a2..ffdc38274ef 100644
--- a/test/Transforms/LoopPredication/invariant_load.ll
+++ b/test/Transforms/LoopPredication/invariant_load.ll
@@ -126,6 +126,70 @@ exit:
   ret i32 %result
 }
 
+; Case where we have an invariant load, but it's not loading from a loop
+; invariant location.
+define i32 @neg_varying_invariant_load_op(i32* %array, i32* %lengths, i32 %n) {
+; CHECK-LABEL: @neg_varying_invariant_load_op(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[N:%.*]], 0
+; CHECK-NEXT:    br i1 [[TMP5]], label [[EXIT:%.*]], label [[LOOP_PREHEADER:%.*]]
+; CHECK:       loop.preheader:
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[LOOP_ACC:%.*]] = phi i32 [ [[LOOP_ACC_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[I_NEXT:%.*]], [[LOOP]] ], [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[UNKNOWN:%.*]] = load volatile i1, i1* @UNKNOWN
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[UNKNOWN]]) [ "deopt"() ]
+; CHECK-NEXT:    [[LENGTH_ADDR:%.*]] = getelementptr i32, i32* [[LENGTHS:%.*]], i32 [[I]]
+; CHECK-NEXT:    [[LEN:%.*]] = load i32, i32* [[LENGTH_ADDR]], align 4, !invariant.load !0
+; CHECK-NEXT:    [[WITHIN_BOUNDS:%.*]] = icmp ult i32 [[I]], [[LEN]]
+; CHECK-NEXT:    call void (i1, ...) @llvm.experimental.guard(i1 [[WITHIN_BOUNDS]], i32 9) [ "deopt"() ]
+; CHECK-NEXT:    [[I_I64:%.*]] = zext i32 [[I]] to i64
+; CHECK-NEXT:    [[ARRAY_I_PTR:%.*]] = getelementptr inbounds i32, i32* [[ARRAY:%.*]], i64 [[I_I64]]
+; CHECK-NEXT:    [[ARRAY_I:%.*]] = load i32, i32* [[ARRAY_I_PTR]], align 4
+; CHECK-NEXT:    [[LOOP_ACC_NEXT]] = add i32 [[LOOP_ACC]], [[ARRAY_I]]
+; CHECK-NEXT:    [[I_NEXT]] = add nuw i32 [[I]], 1
+; CHECK-NEXT:    [[CONTINUE:%.*]] = icmp ult i32 [[I_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[CONTINUE]], label [[LOOP]], label [[EXIT_LOOPEXIT:%.*]]
+; CHECK:       exit.loopexit:
+; CHECK-NEXT:    [[LOOP_ACC_NEXT_LCSSA:%.*]] = phi i32 [ [[LOOP_ACC_NEXT]], [[LOOP]] ]
+; CHECK-NEXT:    br label [[EXIT]]
+; CHECK:       exit:
+; CHECK-NEXT:    [[RESULT:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[LOOP_ACC_NEXT_LCSSA]], [[EXIT_LOOPEXIT]] ]
+; CHECK-NEXT:    ret i32 [[RESULT]]
+;
+entry:
+  %tmp5 = icmp eq i32 %n, 0
+  br i1 %tmp5, label %exit, label %loop.preheader
+
+loop.preheader:
+  br label %loop
+
+loop:
+  %loop.acc = phi i32 [ %loop.acc.next, %loop ], [ 0, %loop.preheader ]
+  %i = phi i32 [ %i.next, %loop ], [ 0, %loop.preheader ]
+  %unknown = load volatile i1, i1* @UNKNOWN
+  call void (i1, ...) @llvm.experimental.guard(i1 %unknown) [ "deopt"() ]
+
+  %length.addr = getelementptr i32, i32* %lengths, i32 %i
+  %len = load i32, i32* %length.addr, align 4, !invariant.load !{}
+  %within.bounds = icmp ult i32 %i, %len
+  call void (i1, ...) @llvm.experimental.guard(i1 %within.bounds, i32 9) [ "deopt"() ]
+
+  %i.i64 = zext i32 %i to i64
+  %array.i.ptr = getelementptr inbounds i32, i32* %array, i64 %i.i64
+  %array.i = load i32, i32* %array.i.ptr, align 4
+  %loop.acc.next = add i32 %loop.acc, %array.i
+
+  %i.next = add nuw i32 %i, 1
+  %continue = icmp ult i32 %i.next, %n
+  br i1 %continue, label %loop, label %exit
+
+exit:
+  %result = phi i32 [ 0, %entry ], [ %loop.acc.next, %loop ]
+  ret i32 %result
+}
+
 ; This is a case where moving the load which provides the limit for the latch
 ; would be invalid, so we can't preform the tempting transform.  Loading the
 ; latch limit may fault since we could always fail the guard.