From: Matthew Simpson Date: Fri, 30 Sep 2016 15:13:52 +0000 (+0000) Subject: [LV] Build all scalar steps for non-uniform induction variables X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=7808833e28be784dfad2f66ed0bc3cdc05e30db2;p=platform%2Fupstream%2Fllvm.git [LV] Build all scalar steps for non-uniform induction variables When building the steps for scalar induction variables, we previously attempted to determine if all the scalar users of the induction variable were uniform. If they were, we would only emit the step corresponding to vector lane zero. This optimization was too aggressive. We generally don't know the entire set of induction variable users that will be scalar. We have isScalarAfterVectorization, but this is only a conservative estimate of the instructions that will be scalarized. Thus, an induction variable may have scalar users that aren't already known to be scalar. To avoid emitting unused steps, we can only check that the induction variable is uniform. This should fix PR30542. Reference: https://llvm.org/bugs/show_bug.cgi?id=30542 llvm-svn: 282863 --- diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp index 897d14a..c228270 100644 --- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp +++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp @@ -2326,22 +2326,11 @@ void InnerLoopVectorizer::buildScalarSteps(Value *ScalarIV, Value *Step, assert(ScalarIVTy->isIntegerTy() && ScalarIVTy == Step->getType() && "Val and Step should have the same integer type"); - auto scalarUserIsUniform = [&](User *U) -> bool { - auto *I = cast(U); - return !OrigLoop->contains(I) || !Legal->isScalarAfterVectorization(I) || - Legal->isUniformAfterVectorization(I); - }; - // Determine the number of scalars we need to generate for each unroll - // iteration. If EntryVal is uniform or all it's scalar users are uniform, we - // only need to generate the first lane. Otherwise, we generate all VF - // values. We are essentially determining if the induction variable has no - // "multi-scalar" (non-uniform scalar) users. + // iteration. If EntryVal is uniform, we only need to generate the first + // lane. Otherwise, we generate all VF values. unsigned Lanes = - Legal->isUniformAfterVectorization(cast(EntryVal)) || - all_of(EntryVal->users(), scalarUserIsUniform) - ? 1 - : VF; + Legal->isUniformAfterVectorization(cast(EntryVal)) ? 1 : VF; // Compute the scalar steps and save the results in VectorLoopValueMap. ScalarParts Entry(UF); diff --git a/llvm/test/Transforms/LoopVectorize/induction.ll b/llvm/test/Transforms/LoopVectorize/induction.ll index 1866263..4ae4a34 100644 --- a/llvm/test/Transforms/LoopVectorize/induction.ll +++ b/llvm/test/Transforms/LoopVectorize/induction.ll @@ -287,6 +287,103 @@ for.end: ret void } +; PR30542. Ensure we generate all the scalar steps for the induction variable. +; The scalar induction variable is used by a getelementptr instruction +; (uniform), and a udiv (non-uniform). +; +; int sum = 0; +; for (int i = 0; i < n; ++i) { +; int x = a[i]; +; if (c) +; x /= i; +; sum += x; +; } +; +; CHECK-LABEL: @scalarize_induction_variable_05( +; CHECK: vector.body: +; CHECK: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue2 ] +; CHECK: %[[I0:.+]] = add i32 %index, 0 +; CHECK: %[[I1:.+]] = add i32 %index, 1 +; CHECK: getelementptr inbounds i32, i32* %a, i32 %[[I0]] +; CHECK: pred.udiv.if: +; CHECK: udiv i32 {{.*}}, %[[I0]] +; CHECK: pred.udiv.if1: +; CHECK: udiv i32 {{.*}}, %[[I1]] +; +; UNROLL-NO_IC-LABEL: @scalarize_induction_variable_05( +; UNROLL-NO-IC: vector.body: +; UNROLL-NO-IC: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue11 ] +; UNROLL-NO-IC: %[[I0:.+]] = add i32 %index, 0 +; UNROLL-NO-IC: %[[I1:.+]] = add i32 %index, 1 +; UNROLL-NO-IC: %[[I2:.+]] = add i32 %index, 2 +; UNROLL-NO-IC: %[[I3:.+]] = add i32 %index, 3 +; UNROLL-NO-IC: getelementptr inbounds i32, i32* %a, i32 %[[I0]] +; UNROLL-NO-IC: getelementptr inbounds i32, i32* %a, i32 %[[I2]] +; UNROLL-NO-IC: pred.udiv.if: +; UNROLL-NO-IC: udiv i32 {{.*}}, %[[I0]] +; UNROLL-NO-IC: pred.udiv.if6: +; UNROLL-NO-IC: udiv i32 {{.*}}, %[[I1]] +; UNROLL-NO-IC: pred.udiv.if8: +; UNROLL-NO-IC: udiv i32 {{.*}}, %[[I2]] +; UNROLL-NO-IC: pred.udiv.if10: +; UNROLL-NO-IC: udiv i32 {{.*}}, %[[I3]] +; +; IND-LABEL: @scalarize_induction_variable_05( +; IND: vector.body: +; IND: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue2 ] +; IND: %[[I1:.+]] = or i32 %index, 1 +; IND: %[[E0:.+]] = sext i32 %index to i64 +; IND: getelementptr inbounds i32, i32* %a, i64 %[[E0]] +; IND: pred.udiv.if: +; IND: udiv i32 {{.*}}, %index +; IND: pred.udiv.if1: +; IND: udiv i32 {{.*}}, %[[I1]] +; +; UNROLL-LABEL: @scalarize_induction_variable_05( +; UNROLL: vector.body: +; UNROLL: %index = phi i32 [ 0, %vector.ph ], [ %index.next, %pred.udiv.continue11 ] +; UNROLL: %[[I1:.+]] = or i32 %index, 1 +; UNROLL: %[[I2:.+]] = or i32 %index, 2 +; UNROLL: %[[I3:.+]] = or i32 %index, 3 +; UNROLL: %[[E0:.+]] = sext i32 %index to i64 +; UNROLL: %[[G0:.+]] = getelementptr inbounds i32, i32* %a, i64 %[[E0]] +; UNROLL: getelementptr i32, i32* %[[G0]], i64 2 +; UNROLL: pred.udiv.if: +; UNROLL: udiv i32 {{.*}}, %index +; UNROLL: pred.udiv.if6: +; UNROLL: udiv i32 {{.*}}, %[[I1]] +; UNROLL: pred.udiv.if8: +; UNROLL: udiv i32 {{.*}}, %[[I2]] +; UNROLL: pred.udiv.if10: +; UNROLL: udiv i32 {{.*}}, %[[I3]] + +define i32 @scalarize_induction_variable_05(i32* %a, i32 %x, i1 %c, i32 %n) { +entry: + br label %for.body + +for.body: + %i = phi i32 [ 0, %entry ], [ %i.next, %if.end ] + %sum = phi i32 [ 0, %entry ], [ %tmp4, %if.end ] + %tmp0 = getelementptr inbounds i32, i32* %a, i32 %i + %tmp1 = load i32, i32* %tmp0, align 4 + br i1 %c, label %if.then, label %if.end + +if.then: + %tmp2 = udiv i32 %tmp1, %i + br label %if.end + +if.end: + %tmp3 = phi i32 [ %tmp2, %if.then ], [ %tmp1, %for.body ] + %tmp4 = add i32 %tmp3, %sum + %i.next = add nuw nsw i32 %i, 1 + %cond = icmp slt i32 %i.next, %n + br i1 %cond, label %for.body, label %for.end + +for.end: + %tmp5 = phi i32 [ %tmp4, %if.end ] + ret i32 %tmp5 +} + ; Ensure we generate both a vector and a scalar induction variable. In this ; test, the induction variable is used by an instruction that will be ; vectorized (trunc) as well as an instruction that will remain in scalar form