diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp --- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp +++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp @@ -5137,12 +5137,11 @@ if (Worklist.count(I)) return; - // If all users of the pointer will be memory accesses and scalar, place the - // pointer in ScalarPtrs. Otherwise, place the pointer in - // PossibleNonScalarPtrs. - if (llvm::all_of(I->users(), [&](User *U) { - return (isa(U) || isa(U)) && - isScalarUse(cast(U), Ptr); + // If the use of the pointer will be a scalar use, and all users of the + // pointer are memory accesses, place the pointer in ScalarPtrs. Otherwise, + // place the pointer in PossibleNonScalarPtrs. + if (isScalarUse(MemAccess, Ptr) && llvm::all_of(I->users(), [&](User *U) { + return isa(U) || isa(U); })) ScalarPtrs.insert(I); else diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll --- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll @@ -1,68 +1,113 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; REQUIRES: asserts ; RUN: opt -loop-vectorize -scalable-vectorization=on -S -mtriple=aarch64 -mattr=+sve -debug-only=loop-vectorize < %s 2>&1 | FileCheck %s target triple = "aarch64-unknown-linux-gnu" -; In the test below the PHI instruction: -; %0 = phi i8* [ %incdec.ptr190, %loop.body ], [ %src, %entry ] -; has multiple uses, i.e. -; 1. As a uniform address for the load, and -; 2. Non-uniform use by the getelementptr + store, which leads to replication. - -; CHECK-LABEL: LV: Checking a loop in "phi_multiple_use" -; CHECK-NOT: LV: Found new scalar instruction: %incdec.ptr190 = getelementptr inbounds i8, i8* %0, i64 1 +; CHECK-LABEL: LV: Checking a loop in "pointer_induction_used_as_vector" +; CHECK-NOT: LV: Found {{.*}} scalar instruction: %ptr.iv.2.next = getelementptr inbounds i8, i8* %ptr.iv.2, i64 1 ; ; CHECK: VPlan 'Initial VPlan for VF={vscale x 2},UF>=1' { ; CHECK-NEXT: loop.body: -; CHECK-NEXT: WIDEN-INDUCTION %index = phi 0, %index.next -; CHECK-NEXT: WIDEN-PHI %curchar = phi %curchar.next, %curptr -; CHECK-NEXT: WIDEN-PHI %0 = phi %incdec.ptr190, %src -; CHECK-NEXT: WIDEN-GEP Var[Inv] ir<%incdec.ptr190> = getelementptr ir<%0>, ir<1> -; CHECK-NEXT: WIDEN store ir<%curchar>, ir<%incdec.ptr190> -; CHECK-NEXT: WIDEN ir<%1> = load ir<%0> -; CHECK-NEXT: WIDEN ir<%2> = add ir<%1>, ir<1> -; CHECK-NEXT: WIDEN store ir<%0>, ir<%2> +; CHECK-NEXT: WIDEN-INDUCTION %iv = phi 0, %iv.next +; CHECK-NEXT: WIDEN-PHI %ptr.iv.1 = phi %start.1, %ptr.iv.1.next +; CHECK-NEXT: WIDEN-PHI %ptr.iv.2 = phi %start.2, %ptr.iv.2.next +; CHECK-NEXT: WIDEN-GEP Var[Inv] ir<%ptr.iv.2.next> = getelementptr ir<%ptr.iv.2>, ir<1> +; CHECK-NEXT: WIDEN store ir<%ptr.iv.1>, ir<%ptr.iv.2.next> +; CHECK-NEXT: WIDEN ir<%lv> = load ir<%ptr.iv.2> +; CHECK-NEXT: WIDEN ir<%add> = add ir<%lv>, ir<1> +; CHECK-NEXT: WIDEN store ir<%ptr.iv.2>, ir<%add> ; CHECK-NEXT: No successors ; CHECK-NEXT: } -define void @phi_multiple_use(i8** noalias %curptr, i8* noalias %src, i64 %N) #0 { -; CHECK-LABEL: @phi_multiple_use( +; In the test below the pointer phi %ptr.iv.2 is used as +; 1. As a uniform address for the load, and +; 2. Non-uniform use by the getelementptr which is stored. This requires the +; vector value. +define void @pointer_induction_used_as_vector(i8** noalias %start.1, i8* noalias %start.2, i64 %N) { +; CHECK-LABEL: @pointer_induction_used_as_vector( +; CHECK-NEXT: entry: +; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2 +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]] +; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] +; CHECK: vector.ph: +; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 2 +; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]] +; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] +; CHECK-NEXT: [[IND_END:%.*]] = getelementptr i8*, i8** [[START_1:%.*]], i64 [[N_VEC]] +; CHECK-NEXT: [[IND_END3:%.*]] = getelementptr i8, i8* [[START_2:%.*]], i64 [[N_VEC]] +; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] ; CHECK: vector.body: -; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ 0, %vector.ph ], [ {{.*}}, %vector.body ] -; CHECK-NEXT: {{.*}} = add i64 [[INDEX1]], 0 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX1]], 0 -; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8*, i8** %curptr, i64 [[TMP1]] -; CHECK-NEXT: [[TMP2:%.*]] = call @llvm.experimental.stepvector.nxv2i64() -; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement poison, i64 [[INDEX1]], i32 0 +; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0 +; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], 0 +; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8*, i8** [[START_1]], i64 [[TMP5]] +; CHECK-NEXT: [[TMP6:%.*]] = call @llvm.experimental.stepvector.nxv2i64() +; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement poison, i64 [[INDEX]], i32 0 ; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector [[DOTSPLATINSERT]], poison, zeroinitializer -; CHECK-NEXT: [[TMP3:%.*]] = add shufflevector ( insertelement ( poison, i64 0, i32 0), poison, zeroinitializer), [[TMP2]] -; CHECK-NEXT: [[TMP4:%.*]] = add [[DOTSPLAT]], [[TMP3]] -; CHECK-NEXT: [[NEXT_GEP6:%.*]] = getelementptr i8, i8* %src, [[TMP4]] -; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, [[NEXT_GEP6]], i64 1 -; CHECK: store [[TMP5]], * -; CHECK-NEXT: [[TMP6:%.*]] = extractelement [[NEXT_GEP6]], i32 0 -; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, i8* [[TMP6]], i32 0 -; CHECK-NEXT: [[TMP8:%.*]] = bitcast i8* [[TMP7]] to * -; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load , * [[TMP8]] -; CHECK-NEXT: [[TMP9:%.*]] = add [[WIDE_LOAD]], -; CHECK: store [[TMP9]], * +; CHECK-NEXT: [[TMP7:%.*]] = add shufflevector ( insertelement ( poison, i64 0, i32 0), poison, zeroinitializer), [[TMP6]] +; CHECK-NEXT: [[TMP8:%.*]] = add [[DOTSPLAT]], [[TMP7]] +; CHECK-NEXT: [[NEXT_GEP4:%.*]] = getelementptr i8, i8* [[START_2]], [[TMP8]] +; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, [[NEXT_GEP4]], i64 1 +; CHECK-NEXT: [[TMP10:%.*]] = getelementptr i8*, i8** [[NEXT_GEP]], i32 0 +; CHECK-NEXT: [[TMP11:%.*]] = bitcast i8** [[TMP10]] to * +; CHECK-NEXT: store [[TMP9]], * [[TMP11]], align 8 +; CHECK-NEXT: [[TMP12:%.*]] = extractelement [[NEXT_GEP4]], i32 0 +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, i8* [[TMP12]], i32 0 +; CHECK-NEXT: [[TMP14:%.*]] = bitcast i8* [[TMP13]] to * +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load , * [[TMP14]], align 1 +; CHECK-NEXT: [[TMP15:%.*]] = add [[WIDE_LOAD]], shufflevector ( insertelement ( poison, i8 1, i32 0), poison, zeroinitializer) +; CHECK-NEXT: [[TMP16:%.*]] = bitcast i8* [[TMP13]] to * +; CHECK-NEXT: store [[TMP15]], * [[TMP16]], align 1 +; CHECK-NEXT: [[TMP17:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP18:%.*]] = mul i64 [[TMP17]], 2 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP18]] +; CHECK-NEXT: [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK: middle.block: +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] +; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] +; CHECK: scalar.ph: +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] +; CHECK-NEXT: [[BC_RESUME_VAL1:%.*]] = phi i8** [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[START_1]], [[ENTRY]] ] +; CHECK-NEXT: [[BC_RESUME_VAL2:%.*]] = phi i8* [ [[IND_END3]], [[MIDDLE_BLOCK]] ], [ [[START_2]], [[ENTRY]] ] +; CHECK-NEXT: br label [[LOOP_BODY:%.*]] +; CHECK: loop.body: +; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_BODY]] ] +; CHECK-NEXT: [[PTR_IV_1:%.*]] = phi i8** [ [[BC_RESUME_VAL1]], [[SCALAR_PH]] ], [ [[PTR_IV_1_NEXT:%.*]], [[LOOP_BODY]] ] +; CHECK-NEXT: [[PTR_IV_2:%.*]] = phi i8* [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[PTR_IV_2_NEXT:%.*]], [[LOOP_BODY]] ] +; CHECK-NEXT: [[PTR_IV_1_NEXT]] = getelementptr inbounds i8*, i8** [[PTR_IV_1]], i64 1 +; CHECK-NEXT: [[PTR_IV_2_NEXT]] = getelementptr inbounds i8, i8* [[PTR_IV_2]], i64 1 +; CHECK-NEXT: store i8* [[PTR_IV_2_NEXT]], i8** [[PTR_IV_1]], align 8 +; CHECK-NEXT: [[LV:%.*]] = load i8, i8* [[PTR_IV_2]], align 1 +; CHECK-NEXT: [[ADD:%.*]] = add i8 [[LV]], 1 +; CHECK-NEXT: store i8 [[ADD]], i8* [[PTR_IV_2]], align 1 +; CHECK-NEXT: [[IV_NEXT]] = add nuw i64 [[IV]], 1 +; CHECK-NEXT: [[C:%.*]] = icmp ne i64 [[IV_NEXT]], [[N]] +; CHECK-NEXT: br i1 [[C]], label [[LOOP_BODY]], label [[EXIT]], !llvm.loop [[LOOP2:![0-9]+]] +; CHECK: exit: +; CHECK-NEXT: ret void +; + entry: br label %loop.body loop.body: ; preds = %loop.body, %entry - %index = phi i64 [ 0, %entry ], [ %index.next, %loop.body ] - %curchar = phi i8** [ %curchar.next, %loop.body ], [ %curptr, %entry ] - %0 = phi i8* [ %incdec.ptr190, %loop.body ], [ %src, %entry ] - %incdec.ptr190 = getelementptr inbounds i8, i8* %0, i64 1 - %curchar.next = getelementptr inbounds i8*, i8** %curchar, i64 1 - store i8* %incdec.ptr190, i8** %curchar, align 8 - %1 = load i8, i8* %0, align 1 - %2 = add i8 %1, 1 - store i8 %2, i8* %0, align 1 - %index.next = add nuw i64 %index, 1 - %3 = icmp ne i64 %index.next, %N - br i1 %3, label %loop.body, label %exit, !llvm.loop !0 + %iv = phi i64 [ 0, %entry ], [ %iv.next, %loop.body ] + %ptr.iv.1 = phi i8** [ %start.1, %entry ], [ %ptr.iv.1.next, %loop.body ] + %ptr.iv.2 = phi i8* [ %start.2, %entry ], [ %ptr.iv.2.next, %loop.body ] + %ptr.iv.1.next = getelementptr inbounds i8*, i8** %ptr.iv.1, i64 1 + %ptr.iv.2.next = getelementptr inbounds i8, i8* %ptr.iv.2, i64 1 + store i8* %ptr.iv.2.next, i8** %ptr.iv.1, align 8 + %lv = load i8, i8* %ptr.iv.2, align 1 + %add = add i8 %lv, 1 + store i8 %add, i8* %ptr.iv.2, align 1 + %iv.next = add nuw i64 %iv, 1 + %c = icmp ne i64 %iv.next, %N + br i1 %c, label %loop.body, label %exit, !llvm.loop !0 exit: ; preds = %loop.body ret void