diff --git a/clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp b/clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp --- a/clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp +++ b/clang/test/CodeGenCXX/microsoft-abi-dynamic-cast.cpp @@ -31,8 +31,9 @@ // CHECK-NEXT: [[VBOFFP:%.*]] = getelementptr inbounds i32, ptr [[VBTBL]], i32 1 // CHECK-NEXT: [[VBOFFS:%.*]] = load i32, ptr [[VBOFFP]], align 4 // CHECK-NEXT: [[DELTA:%.*]] = add nsw i32 [[VBOFFS]], 4 -// CHECK-NEXT: [[ADJ:%.*]] = getelementptr inbounds i8, ptr %x, i32 [[DELTA]] -// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTDynamicCast(ptr [[ADJ]], i32 [[DELTA]], ptr nonnull @"??_R0?AUB@@@8", ptr nonnull @"??_R0?AUT@@@8", i32 1) +// CHECK-NEXT: [[ADJ1:%.*]] = getelementptr i8, ptr %x, i32 [[VBOFFS]] +// CHECK-NEXT: [[ADJ2:%.*]] = getelementptr inbounds i8, ptr [[ADJ1]], i32 4 +// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTDynamicCast(ptr nonnull [[ADJ2]], i32 [[DELTA]], ptr nonnull @"??_R0?AUB@@@8", ptr nonnull @"??_R0?AUT@@@8", i32 1) // CHECK-NEXT: ret ptr [[CALL]] T* test4(V* x) { return dynamic_cast(x); } @@ -62,8 +63,9 @@ // CHECK-NEXT: [[VBOFFP:%.*]] = getelementptr inbounds i32, ptr [[VBTBL]], i32 1 // CHECK-NEXT: [[VBOFFS:%.*]] = load i32, ptr [[VBOFFP]], align 4 // CHECK-NEXT: [[DELTA:%.*]] = add nsw i32 [[VBOFFS]], 4 -// CHECK-NEXT: [[ADJ:%.*]] = getelementptr inbounds i8, ptr %x, i32 [[DELTA]] -// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTDynamicCast(ptr nonnull [[ADJ]], i32 [[DELTA]], ptr {{.*}}@"??_R0?AUB@@@8", ptr {{.*}}@"??_R0?AUT@@@8", i32 0) +// CHECK-NEXT: [[ADJ1:%.*]] = getelementptr i8, ptr %x, i32 [[VBOFFS]] +// CHECK-NEXT: [[ADJ2:%.*]] = getelementptr inbounds i8, ptr [[ADJ1]], i32 4 +// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTDynamicCast(ptr nonnull [[ADJ2]], i32 [[DELTA]], ptr {{.*}}@"??_R0?AUB@@@8", ptr {{.*}}@"??_R0?AUT@@@8", i32 0) // CHECK-NEXT: br label // CHECK: [[RET:%.*]] = phi ptr // CHECK-NEXT: ret ptr [[RET]] @@ -94,9 +96,9 @@ // CHECK-NEXT: [[VBTBL:%.*]] = load ptr, ptr [[VBPTR]], align 4 // CHECK-NEXT: [[VBOFFP:%.*]] = getelementptr inbounds i32, ptr [[VBTBL]], i32 1 // CHECK-NEXT: [[VBOFFS:%.*]] = load i32, ptr [[VBOFFP]], align 4 -// CHECK-NEXT: [[DELTA:%.*]] = add nsw i32 [[VBOFFS]], 4 -// CHECK-NEXT: [[ADJ:%.*]] = getelementptr inbounds i8, ptr %x, i32 [[DELTA]] -// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTCastToVoid(ptr nonnull [[ADJ]]) +// CHECK-NEXT: [[ADJ1:%.*]] = getelementptr i8, ptr %x, i32 [[VBOFFS]] +// CHECK-NEXT: [[ADJ2:%.*]] = getelementptr inbounds i8, ptr [[ADJ1]], i32 4 +// CHECK-NEXT: [[CALL:%.*]] = tail call ptr @__RTCastToVoid(ptr nonnull [[ADJ2]]) // CHECK-NEXT: br label // CHECK: [[RET:%.*]] = phi ptr // CHECK-NEXT: ret ptr [[RET]] diff --git a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp --- a/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp +++ b/llvm/lib/Transforms/InstCombine/InstructionCombining.cpp @@ -2329,7 +2329,29 @@ return CastInst::CreatePointerBitCastOrAddrSpaceCast(Y, GEPType); } } - + if (LI && (GEP.getNumIndices() == 1) && !GEP.getType()->isVectorTy()) { + auto *Idx = dyn_cast(GEP.getOperand(1)); + + // Try to reassociate loop invariant index calculations to enable LICM. + if (Idx && (Idx->getOpcode() == Instruction::Add)) { + Value *Ptr = GEP.getOperand(0); + Value *InvIdx = Idx->getOperand(0); + Value *NonInvIdx = Idx->getOperand(1); + + // rewrite: + // %idx = add i64 %invariant, %indvars.iv + // %gep = getelementptr i32, i32* %ptr, i64 %idx + // as: + // %newptr = getelementptr i32, i32* %ptr, i64 %invariant + // %newgep = getelementptr i32, i32* %newptr, i64 %indvars.iv + auto *NewPtr = GetElementPtrInst::Create(GEP.getResultElementType(), + Ptr, InvIdx, "", &GEP); + auto *NewGEP = GetElementPtrInst::Create(GEP.getResultElementType(), + NewPtr, NonInvIdx); + NewGEP->setIsInBounds(GEP.isInBounds()); + return NewGEP; + } + } // We do not handle pointer-vector geps here. if (GEPType->isVectorTy()) return nullptr; diff --git a/llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll b/llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll --- a/llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll +++ b/llvm/test/Transforms/InstCombine/gep-combine-loop-invariant.ll @@ -182,6 +182,90 @@ br label %loop } +; Test that ADD->GEP chains get reassociated to separate invariants and +; thus provide more LICM opportunities. +define void @test1(float* %a, float* %b, i64 %disp) { +; CHECK-LABEL: @test1( +entry: + br label %for.body + +for.body: +; CHECK: for.body: + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] + ; CHECK: %[[INV:.*]] = getelementptr float, ptr %a, i64 %indvars.iv +; CHECK: %arrayidx = getelementptr inbounds float, ptr %[[INV]], i64 %disp + %idx = add i64 %indvars.iv, %disp + %arrayidx = getelementptr inbounds float, float* %a, i64 %idx + %0 = load float, float* %arrayidx + %div = fdiv float 1.500000e+00, %0 +; CHECK: %[[INV1:.*]] = getelementptr float, ptr %b, i64 %indvars.iv +; CHECK: %arrayidx1 = getelementptr inbounds float, ptr %[[INV1]], i64 %disp + %arrayidx1 = getelementptr inbounds float, float* %b, i64 %idx + store float %div, float* %arrayidx1 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp ne i64 %indvars.iv.next, 1024 + br i1 %exitcond, label %for.body, label %for.cond.cleanup + +for.cond.cleanup: + ret void +} + +; As test1 but esnure no transformation when we cannot prove the original ADD +; will become redundant, because it's used by a non-GEP. +define i64 @test2(float* %a, float* %b, i64 %disp) { +; CHECK-LABEL: @test2( +entry: + br label %for.body + +for.body: +; CHECK: for.body: + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] +; CHECK: %[[VAR1:.*]] = getelementptr float, ptr %a, i64 %indvars.iv +; CHECK: %arrayidx = getelementptr inbounds float, ptr %[[VAR1]], i64 %disp + %idx = add i64 %indvars.iv, %disp + %arrayidx = getelementptr inbounds float, float* %a, i64 %idx + %0 = load float, float* %arrayidx + %div = fdiv float 1.500000e+00, %0 +; CHECK: %[[VAR2:.*]] = getelementptr float, ptr %b, i64 %indvars.iv +; CHECK: %arrayidx1 = getelementptr inbounds float, ptr %[[VAR2]], i64 %disp + %arrayidx1 = getelementptr inbounds float, float* %b, i64 %idx + store float %div, float* %arrayidx1 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp ne i64 %indvars.iv.next, 1024 + br i1 %exitcond, label %for.body, label %for.cond.cleanup + +for.cond.cleanup: + ret i64 %idx +} + +; As test1 but esnure no transformation when we cannot prove the original ADD +; will become redundant, because the second GEP's ptr is not invariant. +define void @test3(float* %a, float* %b, i64 %disp) { +; CHECK-LABEL: @test3( +entry: + br label %for.body + +for.body: +; CHECK: for.body: + %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] +; CHECK: %[[VAR1:.*]] = getelementptr float, ptr %a, i64 %indvars.iv +; CHECK: %arrayidx = getelementptr inbounds float, ptr %[[VAR1]], i64 %disp + %idx = add i64 %indvars.iv, %disp + %arrayidx = getelementptr inbounds float, float* %a, i64 %idx + %0 = load float, float* %arrayidx + %div = fdiv float 1.500000e+00, %0 +; CHECK: %[[VAR2:.*]] = getelementptr float, ptr %arrayidx, i64 %indvars.iv +; CHECK: %arrayidx1 = getelementptr inbounds float, ptr %[[VAR2]], i64 %disp + %arrayidx1 = getelementptr inbounds float, float* %arrayidx, i64 %idx + store float %div, float* %arrayidx1 + %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 + %exitcond = icmp ne i64 %indvars.iv.next, 1024 + br i1 %exitcond, label %for.body, label %for.cond.cleanup + +for.cond.cleanup: + ret void +} + ; This would crash because we did not expect to be able to constant fold a GEP. define void @PR51485(<2 x i64> %v) { diff --git a/llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll b/llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll --- a/llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll +++ b/llvm/test/Transforms/LoopUnroll/runtime-unroll-remainder.ll @@ -35,22 +35,24 @@ ; CHECK-NEXT: [[EPIL_ITER_CMP_NOT:%.*]] = icmp eq i64 [[XTRAITER]], 1 ; CHECK-NEXT: br i1 [[EPIL_ITER_CMP_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT_EPILOG_LCSSA:%.*]], label [[FOR_BODY_EPIL_1:%.*]] ; CHECK: for.body.epil.1: -; CHECK-NEXT: [[INDVARS_IV_NEXT_EPIL:%.*]] = add nuw nsw i64 [[INDVARS_IV_UNR]], 1 -; CHECK-NEXT: [[ARRAYIDX_EPIL_1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV_NEXT_EPIL]] -; CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[ARRAYIDX_EPIL_1]], align 4 -; CHECK-NEXT: [[ARRAYIDX2_EPIL_1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT_EPIL]] -; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[ARRAYIDX2_EPIL_1]], align 4 -; CHECK-NEXT: [[MUL_EPIL_1:%.*]] = mul nsw i32 [[TMP4]], [[TMP3]] +; CHECK-NEXT: [[TMP3:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDVARS_IV_UNR]] +; CHECK-NEXT: [[ARRAYIDX_EPIL_1:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i64 1 +; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[ARRAYIDX_EPIL_1]], align 4 +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDVARS_IV_UNR]] +; CHECK-NEXT: [[ARRAYIDX2_EPIL_1:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i64 1 +; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[ARRAYIDX2_EPIL_1]], align 4 +; CHECK-NEXT: [[MUL_EPIL_1:%.*]] = mul nsw i32 [[TMP6]], [[TMP4]] ; CHECK-NEXT: [[ADD_EPIL_1:%.*]] = add nsw i32 [[MUL_EPIL_1]], [[ADD_EPIL]] ; CHECK-NEXT: [[EPIL_ITER_CMP_1_NOT:%.*]] = icmp eq i64 [[XTRAITER]], 2 ; CHECK-NEXT: br i1 [[EPIL_ITER_CMP_1_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT_EPILOG_LCSSA]], label [[FOR_BODY_EPIL_2:%.*]] ; CHECK: for.body.epil.2: -; CHECK-NEXT: [[INDVARS_IV_NEXT_EPIL_1:%.*]] = add nuw nsw i64 [[INDVARS_IV_UNR]], 2 -; CHECK-NEXT: [[ARRAYIDX_EPIL_2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV_NEXT_EPIL_1]] -; CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX_EPIL_2]], align 4 -; CHECK-NEXT: [[ARRAYIDX2_EPIL_2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT_EPIL_1]] -; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[ARRAYIDX2_EPIL_2]], align 4 -; CHECK-NEXT: [[MUL_EPIL_2:%.*]] = mul nsw i32 [[TMP6]], [[TMP5]] +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDVARS_IV_UNR]] +; CHECK-NEXT: [[ARRAYIDX_EPIL_2:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i64 2 +; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX_EPIL_2]], align 4 +; CHECK-NEXT: [[TMP9:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDVARS_IV_UNR]] +; CHECK-NEXT: [[ARRAYIDX2_EPIL_2:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 2 +; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX2_EPIL_2]], align 4 +; CHECK-NEXT: [[MUL_EPIL_2:%.*]] = mul nsw i32 [[TMP10]], [[TMP8]] ; CHECK-NEXT: [[ADD_EPIL_2:%.*]] = add nsw i32 [[MUL_EPIL_2]], [[ADD_EPIL_1]] ; CHECK-NEXT: br label [[FOR_COND_CLEANUP_LOOPEXIT_EPILOG_LCSSA]] ; CHECK: for.cond.cleanup.loopexit.epilog-lcssa: @@ -67,31 +69,31 @@ ; CHECK-NEXT: [[C_010:%.*]] = phi i32 [ 0, [[FOR_BODY_LR_PH_NEW]] ], [ [[ADD_3]], [[FOR_BODY]] ] ; CHECK-NEXT: [[NITER:%.*]] = phi i64 [ 0, [[FOR_BODY_LR_PH_NEW]] ], [ [[NITER_NEXT_3:%.*]], [[FOR_BODY]] ] ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]] -; CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 +; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]] -; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4 -; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP8]], [[TMP7]] +; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4 +; CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP12]], [[TMP11]] ; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[MUL]], [[C_010]] ; CHECK-NEXT: [[INDVARS_IV_NEXT:%.*]] = or i64 [[INDVARS_IV]], 1 ; CHECK-NEXT: [[ARRAYIDX_1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV_NEXT]] -; CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[ARRAYIDX_1]], align 4 +; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX_1]], align 4 ; CHECK-NEXT: [[ARRAYIDX2_1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT]] -; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX2_1]], align 4 -; CHECK-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[TMP10]], [[TMP9]] +; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX2_1]], align 4 +; CHECK-NEXT: [[MUL_1:%.*]] = mul nsw i32 [[TMP14]], [[TMP13]] ; CHECK-NEXT: [[ADD_1:%.*]] = add nsw i32 [[MUL_1]], [[ADD]] ; CHECK-NEXT: [[INDVARS_IV_NEXT_1:%.*]] = or i64 [[INDVARS_IV]], 2 ; CHECK-NEXT: [[ARRAYIDX_2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV_NEXT_1]] -; CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX_2]], align 4 +; CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX_2]], align 4 ; CHECK-NEXT: [[ARRAYIDX2_2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT_1]] -; CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX2_2]], align 4 -; CHECK-NEXT: [[MUL_2:%.*]] = mul nsw i32 [[TMP12]], [[TMP11]] +; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX2_2]], align 4 +; CHECK-NEXT: [[MUL_2:%.*]] = mul nsw i32 [[TMP16]], [[TMP15]] ; CHECK-NEXT: [[ADD_2:%.*]] = add nsw i32 [[MUL_2]], [[ADD_1]] ; CHECK-NEXT: [[INDVARS_IV_NEXT_2:%.*]] = or i64 [[INDVARS_IV]], 3 ; CHECK-NEXT: [[ARRAYIDX_3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV_NEXT_2]] -; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX_3]], align 4 +; CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX_3]], align 4 ; CHECK-NEXT: [[ARRAYIDX2_3:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV_NEXT_2]] -; CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX2_3]], align 4 -; CHECK-NEXT: [[MUL_3:%.*]] = mul nsw i32 [[TMP14]], [[TMP13]] +; CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX2_3]], align 4 +; CHECK-NEXT: [[MUL_3:%.*]] = mul nsw i32 [[TMP18]], [[TMP17]] ; CHECK-NEXT: [[ADD_3]] = add nsw i32 [[MUL_3]], [[ADD_2]] ; CHECK-NEXT: [[INDVARS_IV_NEXT_3]] = add nuw nsw i64 [[INDVARS_IV]], 4 ; CHECK-NEXT: [[NITER_NEXT_3]] = add i64 [[NITER]], 4 diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll --- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-interleaved-accesses.ll @@ -1402,11 +1402,11 @@ ; CHECK-NEXT: br label [[FOR_BODY:%.*]] ; CHECK: for.body: ; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] -; CHECK-NEXT: [[I_MINUS_1:%.*]] = add i64 [[I]], -1 -; CHECK-NEXT: [[I_MINUS_3:%.*]] = add i64 [[I]], -3 ; CHECK-NEXT: [[A_I:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]] -; CHECK-NEXT: [[A_I_MINUS_1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_MINUS_1]] -; CHECK-NEXT: [[A_I_MINUS_3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_MINUS_3]] +; CHECK-NEXT: [[TMP19:%.*]] = getelementptr i32, ptr [[A]], i64 [[I]] +; CHECK-NEXT: [[A_I_MINUS_1:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 -1 +; CHECK-NEXT: [[TMP20:%.*]] = getelementptr i32, ptr [[A]], i64 [[I]] +; CHECK-NEXT: [[A_I_MINUS_3:%.*]] = getelementptr inbounds i32, ptr [[TMP20]], i64 -3 ; CHECK-NEXT: store i32 [[X]], ptr [[A_I_MINUS_1]], align 4 ; CHECK-NEXT: store i32 [[Y]], ptr [[A_I_MINUS_3]], align 4 ; CHECK-NEXT: store i32 [[Z]], ptr [[A_I]], align 4 @@ -1459,11 +1459,11 @@ ; CHECK: vector.memcheck: ; CHECK-NEXT: [[TMP4:%.*]] = shl i64 [[N]], 1 ; CHECK-NEXT: [[TMP5:%.*]] = and i64 [[TMP4]], -4 -; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP5]], 4 -; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP6]] +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP5]] +; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP6]], i64 4 ; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 2 -; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[TMP5]], 6 -; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP7]] +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]] +; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP7]], i64 6 ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ugt ptr [[SCEVGEP2]], [[B]] ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP1]], [[SCEVGEP]] ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -1531,7 +1531,8 @@ ; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[IV1]] ; CHECK-NEXT: [[LOAD1:%.*]] = load i16, ptr [[GEP1]], align 4 ; CHECK-NEXT: [[CONV1:%.*]] = sext i16 [[LOAD1]] to i32 -; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[IV2]] +; CHECK-NEXT: [[TMP33:%.*]] = getelementptr i16, ptr [[A]], i64 [[IV]] +; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i16, ptr [[TMP33]], i64 2 ; CHECK-NEXT: [[LOAD2]] = load i16, ptr [[GEP2]], align 4 ; CHECK-NEXT: [[CONV2:%.*]] = sext i16 [[LOAD2]] to i32 ; CHECK-NEXT: [[MUL01:%.*]] = mul nsw i32 [[CONV]], [[CONV1]] diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll --- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-vector-reverse.ll @@ -27,8 +27,8 @@ ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] ; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[INDEX]], -1 -; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[TMP4]], [[N]] -; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds double, ptr [[B:%.*]], i64 [[TMP5]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i64 [[N]] ; CHECK-NEXT: [[TMP7:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP8:%.*]] = shl i64 [[TMP7]], 3 ; CHECK-NEXT: [[TMP9:%.*]] = sub i64 1, [[TMP8]] @@ -43,24 +43,25 @@ ; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load , ptr [[TMP16]], align 8 ; CHECK-NEXT: [[TMP17:%.*]] = fadd [[WIDE_LOAD]], shufflevector ( insertelement ( poison, double 1.000000e+00, i64 0), poison, zeroinitializer) ; CHECK-NEXT: [[TMP18:%.*]] = fadd [[WIDE_LOAD1]], shufflevector ( insertelement ( poison, double 1.000000e+00, i64 0), poison, zeroinitializer) -; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds double, ptr [[A:%.*]], i64 [[TMP5]] -; CHECK-NEXT: [[TMP20:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP21:%.*]] = shl i64 [[TMP20]], 3 -; CHECK-NEXT: [[TMP22:%.*]] = sub i64 1, [[TMP21]] -; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds double, ptr [[TMP19]], i64 [[TMP22]] -; CHECK-NEXT: store [[TMP17]], ptr [[TMP23]], align 8 -; CHECK-NEXT: [[TMP24:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP25:%.*]] = shl i64 [[TMP24]], 3 -; CHECK-NEXT: [[TMP26:%.*]] = sub i64 0, [[TMP25]] -; CHECK-NEXT: [[TMP27:%.*]] = sub i64 1, [[TMP25]] -; CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds double, ptr [[TMP19]], i64 [[TMP26]] -; CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds double, ptr [[TMP28]], i64 [[TMP27]] -; CHECK-NEXT: store [[TMP18]], ptr [[TMP29]], align 8 -; CHECK-NEXT: [[TMP30:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP31:%.*]] = shl i64 [[TMP30]], 4 -; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP31]] -; CHECK-NEXT: [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK-NEXT: [[TMP19:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds double, ptr [[TMP19]], i64 [[N]] +; CHECK-NEXT: [[TMP21:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP22:%.*]] = shl i64 [[TMP21]], 3 +; CHECK-NEXT: [[TMP23:%.*]] = sub i64 1, [[TMP22]] +; CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds double, ptr [[TMP20]], i64 [[TMP23]] +; CHECK-NEXT: store [[TMP17]], ptr [[TMP24]], align 8 +; CHECK-NEXT: [[TMP25:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP26:%.*]] = shl i64 [[TMP25]], 3 +; CHECK-NEXT: [[TMP27:%.*]] = sub i64 0, [[TMP26]] +; CHECK-NEXT: [[TMP28:%.*]] = sub i64 1, [[TMP26]] +; CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds double, ptr [[TMP20]], i64 [[TMP27]] +; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds double, ptr [[TMP29]], i64 [[TMP28]] +; CHECK-NEXT: store [[TMP18]], ptr [[TMP30]], align 8 +; CHECK-NEXT: [[TMP31:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP32:%.*]] = shl i64 [[TMP31]], 4 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP32]] +; CHECK-NEXT: [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_MOD_VF]], 0 ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]] @@ -74,10 +75,12 @@ ; CHECK: for.body: ; CHECK-NEXT: [[I_08_IN:%.*]] = phi i64 [ [[I_08:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] ; CHECK-NEXT: [[I_08]] = add nsw i64 [[I_08_IN]], -1 -; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[I_08]] -; CHECK-NEXT: [[TMP33:%.*]] = load double, ptr [[ARRAYIDX]], align 8 -; CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP33]], 1.000000e+00 -; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[I_08]] +; CHECK-NEXT: [[TMP34:%.*]] = getelementptr double, ptr [[B]], i64 [[I_08_IN]] +; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP34]], i64 -1 +; CHECK-NEXT: [[TMP35:%.*]] = load double, ptr [[ARRAYIDX]], align 8 +; CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP35]], 1.000000e+00 +; CHECK-NEXT: [[TMP36:%.*]] = getelementptr double, ptr [[A]], i64 [[I_08_IN]] +; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds double, ptr [[TMP36]], i64 -1 ; CHECK-NEXT: store double [[ADD]], ptr [[ARRAYIDX1]], align 8 ; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i64 [[I_08_IN]], 1 ; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] @@ -132,8 +135,8 @@ ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] ; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[INDEX]], -1 -; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP10]], [[N]] -; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP11]] +; CHECK-NEXT: [[TMP11:%.*]] = getelementptr i64, ptr [[B]], i64 [[TMP10]] +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP11]], i64 [[N]] ; CHECK-NEXT: [[TMP13:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP14:%.*]] = shl i64 [[TMP13]], 3 ; CHECK-NEXT: [[TMP15:%.*]] = sub i64 1, [[TMP14]] @@ -148,24 +151,25 @@ ; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load , ptr [[TMP22]], align 8 ; CHECK-NEXT: [[TMP23:%.*]] = add [[WIDE_LOAD]], shufflevector ( insertelement ( poison, i64 1, i64 0), poison, zeroinitializer) ; CHECK-NEXT: [[TMP24:%.*]] = add [[WIDE_LOAD3]], shufflevector ( insertelement ( poison, i64 1, i64 0), poison, zeroinitializer) -; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP11]] -; CHECK-NEXT: [[TMP26:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP27:%.*]] = shl i64 [[TMP26]], 3 -; CHECK-NEXT: [[TMP28:%.*]] = sub i64 1, [[TMP27]] -; CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds i64, ptr [[TMP25]], i64 [[TMP28]] -; CHECK-NEXT: store [[TMP23]], ptr [[TMP29]], align 8 -; CHECK-NEXT: [[TMP30:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP31:%.*]] = shl i64 [[TMP30]], 3 -; CHECK-NEXT: [[TMP32:%.*]] = sub i64 0, [[TMP31]] -; CHECK-NEXT: [[TMP33:%.*]] = sub i64 1, [[TMP31]] -; CHECK-NEXT: [[TMP34:%.*]] = getelementptr inbounds i64, ptr [[TMP25]], i64 [[TMP32]] -; CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds i64, ptr [[TMP34]], i64 [[TMP33]] -; CHECK-NEXT: store [[TMP24]], ptr [[TMP35]], align 8 -; CHECK-NEXT: [[TMP36:%.*]] = call i64 @llvm.vscale.i64() -; CHECK-NEXT: [[TMP37:%.*]] = shl i64 [[TMP36]], 4 -; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP37]] -; CHECK-NEXT: [[TMP38:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP38]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK-NEXT: [[TMP25:%.*]] = getelementptr i64, ptr [[A]], i64 [[TMP10]] +; CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds i64, ptr [[TMP25]], i64 [[N]] +; CHECK-NEXT: [[TMP27:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP28:%.*]] = shl i64 [[TMP27]], 3 +; CHECK-NEXT: [[TMP29:%.*]] = sub i64 1, [[TMP28]] +; CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds i64, ptr [[TMP26]], i64 [[TMP29]] +; CHECK-NEXT: store [[TMP23]], ptr [[TMP30]], align 8 +; CHECK-NEXT: [[TMP31:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP32:%.*]] = shl i64 [[TMP31]], 3 +; CHECK-NEXT: [[TMP33:%.*]] = sub i64 0, [[TMP32]] +; CHECK-NEXT: [[TMP34:%.*]] = sub i64 1, [[TMP32]] +; CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds i64, ptr [[TMP26]], i64 [[TMP33]] +; CHECK-NEXT: [[TMP36:%.*]] = getelementptr inbounds i64, ptr [[TMP35]], i64 [[TMP34]] +; CHECK-NEXT: store [[TMP24]], ptr [[TMP36]], align 8 +; CHECK-NEXT: [[TMP37:%.*]] = call i64 @llvm.vscale.i64() +; CHECK-NEXT: [[TMP38:%.*]] = shl i64 [[TMP37]], 4 +; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP38]] +; CHECK-NEXT: [[TMP39:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP39]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_MOD_VF]], 0 ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]] @@ -179,10 +183,12 @@ ; CHECK: for.body: ; CHECK-NEXT: [[I_09_IN:%.*]] = phi i64 [ [[I_09:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] ; CHECK-NEXT: [[I_09]] = add nsw i64 [[I_09_IN]], -1 -; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[I_09]] -; CHECK-NEXT: [[TMP39:%.*]] = load i64, ptr [[ARRAYIDX]], align 8 -; CHECK-NEXT: [[ADD:%.*]] = add i64 [[TMP39]], 1 -; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[I_09]] +; CHECK-NEXT: [[TMP40:%.*]] = getelementptr i64, ptr [[B]], i64 [[I_09_IN]] +; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[TMP40]], i64 -1 +; CHECK-NEXT: [[TMP41:%.*]] = load i64, ptr [[ARRAYIDX]], align 8 +; CHECK-NEXT: [[ADD:%.*]] = add i64 [[TMP41]], 1 +; CHECK-NEXT: [[TMP42:%.*]] = getelementptr i64, ptr [[A]], i64 [[I_09_IN]] +; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i64, ptr [[TMP42]], i64 -1 ; CHECK-NEXT: store i64 [[ADD]], ptr [[ARRAYIDX2]], align 8 ; CHECK-NEXT: [[CMP:%.*]] = icmp sgt i64 [[I_09_IN]], 1 ; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP_LOOPEXIT]], !llvm.loop [[LOOP6:![0-9]+]] diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll --- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-phi.ll @@ -35,8 +35,8 @@ ; CHECK-NEXT: [[TMP5:%.*]] = call i64 @llvm.vscale.i64() ; CHECK-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP5]], 5 ; CHECK-NEXT: [[TMP7:%.*]] = shl i64 [[INDEX]], 3 -; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[TMP6]], [[TMP7]] -; CHECK-NEXT: [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[C]], i64 [[TMP8]] +; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i8, ptr [[C]], i64 [[TMP6]] +; CHECK-NEXT: [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[TMP8]], i64 [[TMP7]] ; CHECK-NEXT: [[WIDE_VEC:%.*]] = load , ptr [[NEXT_GEP]], align 4 ; CHECK-NEXT: [[WIDE_VEC3:%.*]] = load , ptr [[NEXT_GEP2]], align 4 ; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { , } @llvm.experimental.vector.deinterleave2.nxv8i32( [[WIDE_VEC]]) diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll b/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll --- a/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll +++ b/llvm/test/Transforms/LoopVectorize/AArch64/vector-reverse-mask4.ll @@ -32,8 +32,8 @@ ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] ; CHECK-NEXT: [[TMP0:%.*]] = xor i64 [[INDEX]], -1 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[N]] -; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds double, ptr [[COND:%.*]], i64 [[TMP1]] +; CHECK-NEXT: [[TMP1:%.*]] = getelementptr double, ptr [[COND:%.*]], i64 [[TMP0]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i64 [[N]] ; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds double, ptr [[TMP2]], i64 -3 ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP3]], align 8 ; CHECK-NEXT: [[REVERSE:%.*]] = shufflevector <4 x double> [[WIDE_LOAD]], <4 x double> poison, <4 x i32> @@ -42,20 +42,21 @@ ; CHECK-NEXT: [[REVERSE2:%.*]] = shufflevector <4 x double> [[WIDE_LOAD1]], <4 x double> poison, <4 x i32> ; CHECK-NEXT: [[TMP5:%.*]] = fcmp une <4 x double> [[REVERSE]], zeroinitializer ; CHECK-NEXT: [[TMP6:%.*]] = fcmp une <4 x double> [[REVERSE2]], zeroinitializer -; CHECK-NEXT: [[TMP7:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] -; CHECK-NEXT: [[TMP8:%.*]] = getelementptr double, ptr [[TMP7]], i64 -3 +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT: [[TMP8:%.*]] = getelementptr double, ptr [[TMP7]], i64 [[N]] +; CHECK-NEXT: [[TMP9:%.*]] = getelementptr double, ptr [[TMP8]], i64 -3 ; CHECK-NEXT: [[REVERSE3:%.*]] = shufflevector <4 x i1> [[TMP5]], <4 x i1> poison, <4 x i32> -; CHECK-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP8]], i32 8, <4 x i1> [[REVERSE3]], <4 x double> poison) -; CHECK-NEXT: [[TMP9:%.*]] = getelementptr double, ptr [[TMP7]], i64 -7 +; CHECK-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP9]], i32 8, <4 x i1> [[REVERSE3]], <4 x double> poison) +; CHECK-NEXT: [[TMP10:%.*]] = getelementptr double, ptr [[TMP8]], i64 -7 ; CHECK-NEXT: [[REVERSE5:%.*]] = shufflevector <4 x i1> [[TMP6]], <4 x i1> poison, <4 x i32> -; CHECK-NEXT: [[WIDE_MASKED_LOAD6:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP9]], i32 8, <4 x i1> [[REVERSE5]], <4 x double> poison) -; CHECK-NEXT: [[TMP10:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD]], -; CHECK-NEXT: [[TMP11:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD6]], -; CHECK-NEXT: call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP10]], ptr [[TMP8]], i32 8, <4 x i1> [[REVERSE3]]) -; CHECK-NEXT: call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP11]], ptr [[TMP9]], i32 8, <4 x i1> [[REVERSE5]]) +; CHECK-NEXT: [[WIDE_MASKED_LOAD6:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP10]], i32 8, <4 x i1> [[REVERSE5]], <4 x double> poison) +; CHECK-NEXT: [[TMP11:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD]], +; CHECK-NEXT: [[TMP12:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD6]], +; CHECK-NEXT: call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP11]], ptr [[TMP9]], i32 8, <4 x i1> [[REVERSE3]]) +; CHECK-NEXT: call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP12]], ptr [[TMP10]], i32 8, <4 x i1> [[REVERSE5]]) ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 -; CHECK-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] +; CHECK-NEXT: [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[N]] ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]] @@ -69,14 +70,16 @@ ; CHECK: for.body: ; CHECK-NEXT: [[I_08_IN:%.*]] = phi i64 [ [[I_08:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] ; CHECK-NEXT: [[I_08]] = add nsw i64 [[I_08_IN]], -1 -; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[COND]], i64 [[I_08]] -; CHECK-NEXT: [[TMP13:%.*]] = load double, ptr [[ARRAYIDX]], align 8 -; CHECK-NEXT: [[TOBOOL:%.*]] = fcmp une double [[TMP13]], 0.000000e+00 +; CHECK-NEXT: [[TMP14:%.*]] = getelementptr double, ptr [[COND]], i64 [[I_08_IN]] +; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP14]], i64 -1 +; CHECK-NEXT: [[TMP15:%.*]] = load double, ptr [[ARRAYIDX]], align 8 +; CHECK-NEXT: [[TOBOOL:%.*]] = fcmp une double [[TMP15]], 0.000000e+00 ; CHECK-NEXT: br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_INC]] ; CHECK: if.then: -; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[I_08]] -; CHECK-NEXT: [[TMP14:%.*]] = load double, ptr [[ARRAYIDX1]], align 8 -; CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP14]], 1.000000e+00 +; CHECK-NEXT: [[TMP16:%.*]] = getelementptr double, ptr [[A]], i64 [[I_08_IN]] +; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds double, ptr [[TMP16]], i64 -1 +; CHECK-NEXT: [[TMP17:%.*]] = load double, ptr [[ARRAYIDX1]], align 8 +; CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP17]], 1.000000e+00 ; CHECK-NEXT: store double [[ADD]], ptr [[ARRAYIDX1]], align 8 ; CHECK-NEXT: br label [[FOR_INC]] ; CHECK: for.inc: diff --git a/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll b/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll --- a/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll +++ b/llvm/test/Transforms/LoopVectorize/if-conversion-nest.ll @@ -16,11 +16,12 @@ ; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[N]], -1 ; CHECK-NEXT: [[TMP2:%.*]] = zext i32 [[TMP1]] to i64 ; CHECK-NEXT: [[TMP3:%.*]] = shl nuw nsw i64 [[TMP2]], 2 -; CHECK-NEXT: [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 4 -; CHECK-NEXT: [[UGLYGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP4]] -; CHECK-NEXT: [[UGLYGEP1:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP4]] -; CHECK-NEXT: [[BOUND0:%.*]] = icmp ugt ptr [[UGLYGEP1]], [[A]] -; CHECK-NEXT: [[BOUND1:%.*]] = icmp ugt ptr [[UGLYGEP]], [[B]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP4]], i64 4 +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP5]], i64 4 +; CHECK-NEXT: [[BOUND0:%.*]] = icmp ugt ptr [[SCEVGEP1]], [[A]] +; CHECK-NEXT: [[BOUND1:%.*]] = icmp ugt ptr [[SCEVGEP]], [[B]] ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] ; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] ; CHECK: vector.ph: @@ -28,23 +29,23 @@ ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] -; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope !0, !noalias !3 -; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] -; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope !3 -; CHECK-NEXT: [[TMP7:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]] -; CHECK-NEXT: [[TMP8:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], -; CHECK-NEXT: [[TMP9:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD2]], -; CHECK-NEXT: [[TMP10:%.*]] = select <4 x i1> [[TMP9]], <4 x i32> , <4 x i32> -; CHECK-NEXT: [[TMP11:%.*]] = and <4 x i1> [[TMP7]], [[TMP8]] -; CHECK-NEXT: [[TMP12:%.*]] = xor <4 x i1> [[TMP8]], -; CHECK-NEXT: [[TMP13:%.*]] = and <4 x i1> [[TMP7]], [[TMP12]] -; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP11]], <4 x i32> , <4 x i32> -; CHECK-NEXT: [[PREDPHI3:%.*]] = select <4 x i1> [[TMP13]], <4 x i32> [[TMP10]], <4 x i32> [[PREDPHI]] -; CHECK-NEXT: store <4 x i32> [[PREDPHI3]], ptr [[TMP5]], align 4, !alias.scope !0, !noalias !3 +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]] +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope !0, !noalias !3 +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]] +; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP7]], align 4, !alias.scope !3 +; CHECK-NEXT: [[TMP8:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]] +; CHECK-NEXT: [[TMP9:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], +; CHECK-NEXT: [[TMP10:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD2]], +; CHECK-NEXT: [[TMP11:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> , <4 x i32> +; CHECK-NEXT: [[TMP12:%.*]] = and <4 x i1> [[TMP8]], [[TMP9]] +; CHECK-NEXT: [[TMP13:%.*]] = xor <4 x i1> [[TMP9]], +; CHECK-NEXT: [[TMP14:%.*]] = and <4 x i1> [[TMP8]], [[TMP13]] +; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP12]], <4 x i32> , <4 x i32> +; CHECK-NEXT: [[PREDPHI3:%.*]] = select <4 x i1> [[TMP14]], <4 x i32> [[TMP11]], <4 x i32> [[PREDPHI]] +; CHECK-NEXT: store <4 x i32> [[PREDPHI3]], ptr [[TMP6]], align 4, !alias.scope !0, !noalias !3 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 -; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] +; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] ; CHECK: middle.block: ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N_VEC]], [[TMP0]] ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]] @@ -54,16 +55,16 @@ ; CHECK: for.body: ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[IF_END14:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]] -; CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 +; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDVARS_IV]] -; CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4 -; CHECK-NEXT: [[CMP3:%.*]] = icmp sgt i32 [[TMP15]], [[TMP16]] +; CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4 +; CHECK-NEXT: [[CMP3:%.*]] = icmp sgt i32 [[TMP16]], [[TMP17]] ; CHECK-NEXT: br i1 [[CMP3]], label [[IF_THEN:%.*]], label [[IF_END14]] ; CHECK: if.then: -; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[TMP15]], 19 +; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[TMP16]], 19 ; CHECK-NEXT: br i1 [[CMP6]], label [[IF_END14]], label [[IF_ELSE:%.*]] ; CHECK: if.else: -; CHECK-NEXT: [[CMP10:%.*]] = icmp slt i32 [[TMP16]], 4 +; CHECK-NEXT: [[CMP10:%.*]] = icmp slt i32 [[TMP17]], 4 ; CHECK-NEXT: [[DOT:%.*]] = select i1 [[CMP10]], i32 4, i32 5 ; CHECK-NEXT: br label [[IF_END14]] ; CHECK: if.end14: diff --git a/llvm/test/Transforms/LoopVectorize/induction.ll b/llvm/test/Transforms/LoopVectorize/induction.ll --- a/llvm/test/Transforms/LoopVectorize/induction.ll +++ b/llvm/test/Transforms/LoopVectorize/induction.ll @@ -331,12 +331,12 @@ ; IND-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2 ; IND-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]] ; IND-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2 -; IND-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]] -; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]] +; IND-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]] ; IND-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2 ; IND-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] -; IND-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]] -; IND-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]] +; IND-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; IND-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]] ; IND-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]] ; IND-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]] ; IND-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -348,11 +348,11 @@ ; IND-NEXT: br label [[VECTOR_BODY:%.*]] ; IND: vector.body: ; IND-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; IND-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET]] -; IND-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]] +; IND-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; IND-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i64 [[OFFSET]] ; IND-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !alias.scope !4, !noalias !7 -; IND-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]] -; IND-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP7]] +; IND-NEXT: [[TMP7:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; IND-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i64 [[OFFSET2]] ; IND-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP8]], align 4, !alias.scope !7 ; IND-NEXT: [[TMP9:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]] ; IND-NEXT: [[TMP10:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP9]] @@ -368,11 +368,11 @@ ; IND-NEXT: br label [[FOR_BODY:%.*]] ; IND: for.body: ; IND-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] -; IND-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]] -; IND-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]] +; IND-NEXT: [[TMP12:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; IND-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i64 [[OFFSET]] ; IND-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4 -; IND-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]] -; IND-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]] +; IND-NEXT: [[TMP13:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; IND-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i64 [[OFFSET2]] ; IND-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4 ; IND-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]] ; IND-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]] @@ -391,12 +391,12 @@ ; UNROLL-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2 ; UNROLL-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]] ; UNROLL-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2 -; UNROLL-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]] -; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]] +; UNROLL-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]] ; UNROLL-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2 ; UNROLL-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] -; UNROLL-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]] -; UNROLL-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]] +; UNROLL-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; UNROLL-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]] ; UNROLL-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]] ; UNROLL-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]] ; UNROLL-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -408,13 +408,13 @@ ; UNROLL-NEXT: br label [[VECTOR_BODY:%.*]] ; UNROLL: vector.body: ; UNROLL-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; UNROLL-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET]] -; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]] +; UNROLL-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i64 [[OFFSET]] ; UNROLL-NEXT: [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP6]], align 4, !alias.scope !4, !noalias !7 ; UNROLL-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 2 ; UNROLL-NEXT: [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope !4, !noalias !7 -; UNROLL-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], [[OFFSET2]] -; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP8]] +; UNROLL-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; UNROLL-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 [[OFFSET2]] ; UNROLL-NEXT: [[WIDE_LOAD5:%.*]] = load <2 x float>, ptr [[TMP9]], align 4, !alias.scope !7 ; UNROLL-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i64 2 ; UNROLL-NEXT: [[WIDE_LOAD6:%.*]] = load <2 x float>, ptr [[TMP10]], align 4, !alias.scope !7 @@ -435,11 +435,11 @@ ; UNROLL-NEXT: br label [[FOR_BODY:%.*]] ; UNROLL: for.body: ; UNROLL-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] -; UNROLL-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]] -; UNROLL-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]] +; UNROLL-NEXT: [[TMP16:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; UNROLL-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[OFFSET]] ; UNROLL-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4 -; UNROLL-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]] -; UNROLL-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]] +; UNROLL-NEXT: [[TMP17:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; UNROLL-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[OFFSET2]] ; UNROLL-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4 ; UNROLL-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]] ; UNROLL-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]] @@ -534,12 +534,12 @@ ; INTERLEAVE-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2 ; INTERLEAVE-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]] ; INTERLEAVE-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2 -; INTERLEAVE-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]] -; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]] +; INTERLEAVE-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]] ; INTERLEAVE-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2 ; INTERLEAVE-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] -; INTERLEAVE-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]] -; INTERLEAVE-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]] +; INTERLEAVE-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; INTERLEAVE-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]] ; INTERLEAVE-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]] ; INTERLEAVE-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]] ; INTERLEAVE-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -551,13 +551,13 @@ ; INTERLEAVE-NEXT: br label [[VECTOR_BODY:%.*]] ; INTERLEAVE: vector.body: ; INTERLEAVE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; INTERLEAVE-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET]] -; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]] +; INTERLEAVE-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i64 [[OFFSET]] ; INTERLEAVE-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP6]], align 4, !alias.scope !4, !noalias !7 ; INTERLEAVE-NEXT: [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 4 ; INTERLEAVE-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP7]], align 4, !alias.scope !4, !noalias !7 -; INTERLEAVE-NEXT: [[TMP8:%.*]] = add i64 [[INDEX]], [[OFFSET2]] -; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP8]] +; INTERLEAVE-NEXT: [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; INTERLEAVE-NEXT: [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i64 [[OFFSET2]] ; INTERLEAVE-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP9]], align 4, !alias.scope !7 ; INTERLEAVE-NEXT: [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i64 4 ; INTERLEAVE-NEXT: [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP10]], align 4, !alias.scope !7 @@ -578,11 +578,11 @@ ; INTERLEAVE-NEXT: br label [[FOR_BODY:%.*]] ; INTERLEAVE: for.body: ; INTERLEAVE-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] -; INTERLEAVE-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]] -; INTERLEAVE-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]] +; INTERLEAVE-NEXT: [[TMP16:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; INTERLEAVE-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[OFFSET]] ; INTERLEAVE-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4 -; INTERLEAVE-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]] -; INTERLEAVE-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]] +; INTERLEAVE-NEXT: [[TMP17:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; INTERLEAVE-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[OFFSET2]] ; INTERLEAVE-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4 ; INTERLEAVE-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]] ; INTERLEAVE-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]] @@ -1632,8 +1632,8 @@ ; IND-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1 ; IND-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64 ; IND-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3 -; IND-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 8 -; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]] +; IND-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]] +; IND-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 8 ; IND-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 4 ; IND-NEXT: [[TMP8:%.*]] = or i64 [[TMP7]], 4 ; IND-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP8]] @@ -1695,8 +1695,8 @@ ; UNROLL-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1 ; UNROLL-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64 ; UNROLL-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3 -; UNROLL-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 8 -; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]] +; UNROLL-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]] +; UNROLL-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 8 ; UNROLL-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 4 ; UNROLL-NEXT: [[TMP8:%.*]] = or i64 [[TMP7]], 4 ; UNROLL-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP8]] @@ -1851,8 +1851,8 @@ ; INTERLEAVE-NEXT: [[TMP3:%.*]] = add i32 [[N]], -1 ; INTERLEAVE-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64 ; INTERLEAVE-NEXT: [[TMP5:%.*]] = shl nuw nsw i64 [[TMP4]], 3 -; INTERLEAVE-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 8 -; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP6]] +; INTERLEAVE-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP5]] +; INTERLEAVE-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP6]], i64 8 ; INTERLEAVE-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP4]], 4 ; INTERLEAVE-NEXT: [[TMP8:%.*]] = or i64 [[TMP7]], 4 ; INTERLEAVE-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP8]] diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll --- a/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll +++ b/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll @@ -1,4 +1,4 @@ -; NOTE: Assertions have been autogenerated by utils/update_test_checks.py +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --prefix-filecheck-ir-name T ; RUN: opt -S -passes=loop-vectorize,instcombine -force-vector-width=4 -force-vector-interleave=1 -enable-interleaved-mem-accesses=true -runtime-memory-check-threshold=24 < %s | FileCheck %s target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128" @@ -441,8 +441,8 @@ ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]] ; CHECK-NEXT: [[TMP:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; CHECK-NEXT: [[MUL:%.*]] = shl nsw i32 [[TMP]], 1 -; CHECK-NEXT: [[TMP1:%.*]] = lshr exact i64 [[INDVARS_IV]], 1 -; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP1]] +; CHECK-NEXT: [[TTMP1:%.*]] = lshr exact i64 [[INDVARS_IV]], 1 +; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TTMP1]] ; CHECK-NEXT: store i32 [[MUL]], ptr [[ARRAYIDX2]], align 4 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 2 ; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV]], 1022 @@ -518,8 +518,8 @@ ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDVARS_IV]] ; CHECK-NEXT: [[TMP:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 ; CHECK-NEXT: [[MUL:%.*]] = shl nsw i32 [[TMP]], 1 -; CHECK-NEXT: [[TMP1:%.*]] = lshr exact i64 [[INDVARS_IV]], 1 -; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP1]] +; CHECK-NEXT: [[TTMP1:%.*]] = lshr exact i64 [[INDVARS_IV]], 1 +; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TTMP1]] ; CHECK-NEXT: store i32 [[MUL]], ptr [[ARRAYIDX2]], align 4 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 2 ; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], [[N]] @@ -1377,28 +1377,28 @@ ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = or i64 [[TMP4]], 3 ; CHECK-NEXT: [[TMP5:%.*]] = or i64 [[TMP4]], 5 ; CHECK-NEXT: [[TMP6:%.*]] = or i64 [[TMP4]], 7 -; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[TMP4]], 9 -; CHECK-NEXT: [[TMP8:%.*]] = add <4 x i64> [[VEC_IND]], -; CHECK-NEXT: [[TMP9:%.*]] = add <4 x i64> [[VEC_IND]], -; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]] -; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]] -; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]] -; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP7]] -; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i64> [[TMP8]], i64 0 +; CHECK-NEXT: [[TMP7:%.*]] = add <4 x i64> [[VEC_IND]], +; CHECK-NEXT: [[TMP8:%.*]] = add <4 x i64> [[VEC_IND]], +; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[OFFSET_IDX]] +; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP5]] +; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]] +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i64 9 +; CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i64> [[TMP7]], i64 0 ; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]] -; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i64> [[TMP8]], i64 1 +; CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i64> [[TMP7]], i64 1 ; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP16]] -; CHECK-NEXT: [[TMP18:%.*]] = extractelement <4 x i64> [[TMP8]], i64 2 +; CHECK-NEXT: [[TMP18:%.*]] = extractelement <4 x i64> [[TMP7]], i64 2 ; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP18]] -; CHECK-NEXT: [[TMP20:%.*]] = extractelement <4 x i64> [[TMP8]], i64 3 +; CHECK-NEXT: [[TMP20:%.*]] = extractelement <4 x i64> [[TMP7]], i64 3 ; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP20]] -; CHECK-NEXT: [[TMP22:%.*]] = extractelement <4 x i64> [[TMP9]], i64 0 +; CHECK-NEXT: [[TMP22:%.*]] = extractelement <4 x i64> [[TMP8]], i64 0 ; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP22]] -; CHECK-NEXT: [[TMP24:%.*]] = extractelement <4 x i64> [[TMP9]], i64 1 +; CHECK-NEXT: [[TMP24:%.*]] = extractelement <4 x i64> [[TMP8]], i64 1 ; CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP24]] -; CHECK-NEXT: [[TMP26:%.*]] = extractelement <4 x i64> [[TMP9]], i64 2 +; CHECK-NEXT: [[TMP26:%.*]] = extractelement <4 x i64> [[TMP8]], i64 2 ; CHECK-NEXT: [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP26]] -; CHECK-NEXT: [[TMP28:%.*]] = extractelement <4 x i64> [[TMP9]], i64 3 +; CHECK-NEXT: [[TMP28:%.*]] = extractelement <4 x i64> [[TMP8]], i64 3 ; CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP28]] ; CHECK-NEXT: store i32 [[X:%.*]], ptr [[TMP15]], align 4 ; CHECK-NEXT: store i32 [[X]], ptr [[TMP17]], align 4 @@ -1408,9 +1408,9 @@ ; CHECK-NEXT: store i32 [[Y]], ptr [[TMP25]], align 4 ; CHECK-NEXT: store i32 [[Y]], ptr [[TMP27]], align 4 ; CHECK-NEXT: store i32 [[Y]], ptr [[TMP29]], align 4 -; CHECK-NEXT: store i32 [[Z:%.*]], ptr [[TMP10]], align 4 +; CHECK-NEXT: store i32 [[Z:%.*]], ptr [[TMP9]], align 4 +; CHECK-NEXT: store i32 [[Z]], ptr [[TMP10]], align 4 ; CHECK-NEXT: store i32 [[Z]], ptr [[TMP11]], align 4 -; CHECK-NEXT: store i32 [[Z]], ptr [[TMP12]], align 4 ; CHECK-NEXT: store i32 [[Z]], ptr [[TMP13]], align 4 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 ; CHECK-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], @@ -1424,11 +1424,11 @@ ; CHECK-NEXT: br label [[FOR_BODY:%.*]] ; CHECK: for.body: ; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ] -; CHECK-NEXT: [[I_MINUS_1:%.*]] = add i64 [[I]], -1 -; CHECK-NEXT: [[I_MINUS_3:%.*]] = add i64 [[I]], -3 ; CHECK-NEXT: [[A_I:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I]] -; CHECK-NEXT: [[A_I_MINUS_1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_MINUS_1]] -; CHECK-NEXT: [[A_I_MINUS_3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_MINUS_3]] +; CHECK-NEXT: [[TMP31:%.*]] = getelementptr i32, ptr [[A]], i64 [[I]] +; CHECK-NEXT: [[A_I_MINUS_1:%.*]] = getelementptr inbounds i32, ptr [[TMP31]], i64 -1 +; CHECK-NEXT: [[TMP32:%.*]] = getelementptr i32, ptr [[A]], i64 [[I]] +; CHECK-NEXT: [[A_I_MINUS_3:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i64 -3 ; CHECK-NEXT: store i32 [[X]], ptr [[A_I_MINUS_1]], align 4 ; CHECK-NEXT: store i32 [[Y]], ptr [[A_I_MINUS_3]], align 4 ; CHECK-NEXT: store i32 [[Z]], ptr [[A_I]], align 4 @@ -1480,11 +1480,11 @@ ; CHECK: vector.memcheck: ; CHECK-NEXT: [[TMP2:%.*]] = shl i64 [[N]], 1 ; CHECK-NEXT: [[TMP3:%.*]] = and i64 [[TMP2]], -4 -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], 4 -; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[TMP4]], i64 4 ; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 2 -; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[TMP3]], 6 -; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP5]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] +; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[TMP5]], i64 6 ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ugt ptr [[SCEVGEP2]], [[B]] ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP1]], [[SCEVGEP]] ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -1534,7 +1534,8 @@ ; CHECK-NEXT: [[GEP1:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[IV1]] ; CHECK-NEXT: [[LOAD1:%.*]] = load i16, ptr [[GEP1]], align 4 ; CHECK-NEXT: [[CONV1:%.*]] = sext i16 [[LOAD1]] to i32 -; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[IV2]] +; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i16, ptr [[A]], i64 [[IV]] +; CHECK-NEXT: [[GEP2:%.*]] = getelementptr inbounds i16, ptr [[TMP16]], i64 2 ; CHECK-NEXT: [[LOAD2]] = load i16, ptr [[GEP2]], align 4 ; CHECK-NEXT: [[CONV2:%.*]] = sext i16 [[LOAD2]] to i32 ; CHECK-NEXT: [[MUL01:%.*]] = mul nsw i32 [[CONV]], [[CONV1]] diff --git a/llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll b/llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll --- a/llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll +++ b/llvm/test/Transforms/LoopVectorize/invariant-store-vectorization.ll @@ -347,73 +347,73 @@ ; CHECK-NEXT: br i1 [[CMP20]], label [[FOR_END10:%.*]], label [[FOR_COND1_PREHEADER_PREHEADER:%.*]] ; CHECK: for.cond1.preheader.preheader: ; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[VAR2:%.*]], i64 4 +; CHECK-NEXT: [[INVARIANT_GEP5:%.*]] = getelementptr i8, ptr [[VAR1:%.*]], i64 4 ; CHECK-NEXT: br label [[FOR_COND1_PREHEADER:%.*]] ; CHECK: for.cond1.preheader: ; CHECK-NEXT: [[INDVARS_IV23:%.*]] = phi i64 [ [[INDVARS_IV_NEXT24:%.*]], [[FOR_INC8:%.*]] ], [ 0, [[FOR_COND1_PREHEADER_PREHEADER]] ] ; CHECK-NEXT: [[J_022:%.*]] = phi i32 [ [[J_1_LCSSA:%.*]], [[FOR_INC8]] ], [ 0, [[FOR_COND1_PREHEADER_PREHEADER]] ] ; CHECK-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[INDVARS_IV23]], 2 -; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[VAR1:%.*]], i64 [[TMP0]] -; CHECK-NEXT: [[TMP1:%.*]] = add nuw i64 [[TMP0]], 4 -; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[VAR1]], i64 [[TMP1]] +; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[VAR1]], i64 [[TMP0]] +; CHECK-NEXT: [[GEP6:%.*]] = getelementptr i8, ptr [[INVARIANT_GEP5]], i64 [[TMP0]] ; CHECK-NEXT: [[CMP218:%.*]] = icmp ult i32 [[J_022]], [[ITR]] ; CHECK-NEXT: br i1 [[CMP218]], label [[FOR_BODY3_LR_PH:%.*]], label [[FOR_INC8]] ; CHECK: for.body3.lr.ph: ; CHECK-NEXT: [[ARRAYIDX5:%.*]] = getelementptr inbounds i32, ptr [[VAR1]], i64 [[INDVARS_IV23]] -; CHECK-NEXT: [[TMP2:%.*]] = zext i32 [[J_022]] to i64 +; CHECK-NEXT: [[TMP1:%.*]] = zext i32 [[J_022]] to i64 ; CHECK-NEXT: [[ARRAYIDX5_PROMOTED:%.*]] = load i32, ptr [[ARRAYIDX5]], align 4 -; CHECK-NEXT: [[TMP3:%.*]] = xor i32 [[J_022]], -1 -; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[TMP3]], [[ITR]] -; CHECK-NEXT: [[TMP5:%.*]] = zext i32 [[TMP4]] to i64 -; CHECK-NEXT: [[TMP6:%.*]] = add nuw nsw i64 [[TMP5]], 1 -; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP4]], 3 +; CHECK-NEXT: [[TMP2:%.*]] = xor i32 [[J_022]], -1 +; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], [[ITR]] +; CHECK-NEXT: [[TMP4:%.*]] = zext i32 [[TMP3]] to i64 +; CHECK-NEXT: [[TMP5:%.*]] = add nuw nsw i64 [[TMP4]], 1 +; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP3]], 3 ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] ; CHECK: vector.memcheck: -; CHECK-NEXT: [[TMP7:%.*]] = shl nuw nsw i64 [[TMP2]], 2 -; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[VAR2]], i64 [[TMP7]] -; CHECK-NEXT: [[TMP8:%.*]] = xor i32 [[J_022]], -1 -; CHECK-NEXT: [[TMP9:%.*]] = add i32 [[TMP8]], [[ITR]] -; CHECK-NEXT: [[TMP10:%.*]] = zext i32 [[TMP9]] to i64 -; CHECK-NEXT: [[TMP11:%.*]] = add nuw nsw i64 [[TMP2]], [[TMP10]] -; CHECK-NEXT: [[TMP12:%.*]] = shl nuw nsw i64 [[TMP11]], 2 -; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i8, ptr [[SCEVGEP3]], i64 [[TMP12]] +; CHECK-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[TMP1]], 2 +; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[VAR2]], i64 [[TMP6]] +; CHECK-NEXT: [[TMP7:%.*]] = xor i32 [[J_022]], -1 +; CHECK-NEXT: [[TMP8:%.*]] = add i32 [[TMP7]], [[ITR]] +; CHECK-NEXT: [[TMP9:%.*]] = zext i32 [[TMP8]] to i64 +; CHECK-NEXT: [[TMP10:%.*]] = add nuw nsw i64 [[TMP1]], [[TMP9]] +; CHECK-NEXT: [[TMP11:%.*]] = shl nuw nsw i64 [[TMP10]], 2 +; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i8, ptr [[SCEVGEP3]], i64 [[TMP11]] ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP4]] -; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]] +; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[GEP6]] ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] ; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] ; CHECK: vector.ph: -; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP6]], 8589934588 -; CHECK-NEXT: [[IND_END:%.*]] = add nuw nsw i64 [[N_VEC]], [[TMP2]] -; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x i32> , i32 [[ARRAYIDX5_PROMOTED]], i64 0 +; CHECK-NEXT: [[N_VEC:%.*]] = and i64 [[TMP5]], 8589934588 +; CHECK-NEXT: [[IND_END:%.*]] = add nuw nsw i64 [[N_VEC]], [[TMP1]] +; CHECK-NEXT: [[TMP12:%.*]] = insertelement <4 x i32> , i32 [[ARRAYIDX5_PROMOTED]], i64 0 +; CHECK-NEXT: [[INVARIANT_GEP:%.*]] = getelementptr i32, ptr [[VAR2]], i64 [[TMP1]] ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP13]], [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ] -; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 [[INDEX]], [[TMP2]] -; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[VAR2]], i64 [[OFFSET_IDX]] -; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP14]], align 4, !alias.scope !23 -; CHECK-NEXT: [[TMP15:%.*]] = add <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]] -; CHECK-NEXT: [[TMP16]] = add <4 x i32> [[TMP15]], +; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP12]], [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[GEP:%.*]] = getelementptr i32, ptr [[INVARIANT_GEP]], i64 [[INDEX]] +; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[GEP]], align 4, !alias.scope !23 +; CHECK-NEXT: [[TMP13:%.*]] = add <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]] +; CHECK-NEXT: [[TMP14]] = add <4 x i32> [[TMP13]], ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 -; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] -; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]] +; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] +; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]] ; CHECK: middle.block: -; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi <4 x i32> [ [[TMP16]], [[VECTOR_BODY]] ] -; CHECK-NEXT: [[TMP18:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[DOTLCSSA]]) -; CHECK-NEXT: store i32 [[TMP18]], ptr [[ARRAYIDX5]], align 4 -; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP6]], [[N_VEC]] +; CHECK-NEXT: [[DOTLCSSA:%.*]] = phi <4 x i32> [ [[TMP14]], [[VECTOR_BODY]] ] +; CHECK-NEXT: [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[DOTLCSSA]]) +; CHECK-NEXT: store i32 [[TMP16]], ptr [[ARRAYIDX5]], align 4 +; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP5]], [[N_VEC]] ; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_INC8_LOOPEXIT:%.*]], label [[SCALAR_PH]] ; CHECK: scalar.ph: -; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[TMP2]], [[FOR_BODY3_LR_PH]] ], [ [[TMP2]], [[VECTOR_MEMCHECK]] ] -; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP18]], [[MIDDLE_BLOCK]] ], [ [[ARRAYIDX5_PROMOTED]], [[FOR_BODY3_LR_PH]] ], [ [[ARRAYIDX5_PROMOTED]], [[VECTOR_MEMCHECK]] ] +; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[TMP1]], [[FOR_BODY3_LR_PH]] ], [ [[TMP1]], [[VECTOR_MEMCHECK]] ] +; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP16]], [[MIDDLE_BLOCK]] ], [ [[ARRAYIDX5_PROMOTED]], [[FOR_BODY3_LR_PH]] ], [ [[ARRAYIDX5_PROMOTED]], [[VECTOR_MEMCHECK]] ] ; CHECK-NEXT: br label [[FOR_BODY3:%.*]] ; CHECK: for.body3: -; CHECK-NEXT: [[TMP19:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP21:%.*]], [[FOR_BODY3]] ] +; CHECK-NEXT: [[TMP17:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[TMP19:%.*]], [[FOR_BODY3]] ] ; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY3]] ] ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[VAR2]], i64 [[INDVARS_IV]] -; CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 -; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP19]], [[TMP20]] -; CHECK-NEXT: [[TMP21]] = add nsw i32 [[ADD]], 1 -; CHECK-NEXT: store i32 [[TMP21]], ptr [[ARRAYIDX5]], align 4 +; CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX]], align 4 +; CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP18]] +; CHECK-NEXT: [[TMP19]] = add nsw i32 [[ADD]], 1 +; CHECK-NEXT: store i32 [[TMP19]], ptr [[ARRAYIDX5]], align 4 ; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 ; CHECK-NEXT: [[LFTR_WIDEIV:%.*]] = trunc i64 [[INDVARS_IV_NEXT]] to i32 ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[LFTR_WIDEIV]], [[ITR]] diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check.ll b/llvm/test/Transforms/LoopVectorize/runtime-check.ll --- a/llvm/test/Transforms/LoopVectorize/runtime-check.ll +++ b/llvm/test/Transforms/LoopVectorize/runtime-check.ll @@ -114,12 +114,12 @@ ; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[OFFSET:%.*]], 2 ; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[TMP0]] ; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[N]], 2 -; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]] -; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[TMP2]], i64 [[TMP0]] ; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[OFFSET2:%.*]], 2 ; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP1]], [[TMP3]] -; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP1]] +; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[TMP4]], i64 [[TMP3]] ; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP3]] ; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[SCEVGEP2]], [[SCEVGEP1]] ; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] @@ -131,11 +131,11 @@ ; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] ; CHECK: vector.body: ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] -; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[INDEX]], [[OFFSET]] -; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP5]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i64 [[OFFSET]] ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP6]], align 4, !alias.scope !15, !noalias !18 -; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[INDEX]], [[OFFSET2]] -; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP7]] +; CHECK-NEXT: [[TMP7:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]] +; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i64 [[OFFSET2]] ; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP8]], align 4, !alias.scope !18 ; CHECK-NEXT: [[TMP9:%.*]] = fmul fast <4 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]] ; CHECK-NEXT: [[TMP10:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[TMP9]] @@ -151,11 +151,11 @@ ; CHECK-NEXT: br label [[FOR_BODY:%.*]] ; CHECK: for.body: ; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] -; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET]] -; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM]] +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i64 [[OFFSET]] ; CHECK-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4 -; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2]] -; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]] +; CHECK-NEXT: [[TMP13:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i64 [[OFFSET2]] ; CHECK-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4 ; CHECK-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B]] ; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]] @@ -225,17 +225,17 @@ ; CHECK-NEXT: br label [[FOR_BODY:%.*]] ; CHECK: for.body: ; CHECK-NEXT: [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ] -; CHECK-NEXT: [[IND_SUM:%.*]] = add i64 [[IV]], [[OFFSET:%.*]] -; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IND_SUM]] +; CHECK-NEXT: [[TMP0:%.*]] = getelementptr float, ptr [[A:%.*]], i64 [[IV]] +; CHECK-NEXT: [[ARR_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i64 [[OFFSET:%.*]] ; CHECK-NEXT: [[L1:%.*]] = load float, ptr [[ARR_IDX]], align 4 -; CHECK-NEXT: [[IND_SUM2:%.*]] = add i64 [[IV]], [[OFFSET2:%.*]] -; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IND_SUM2]] +; CHECK-NEXT: [[TMP1:%.*]] = getelementptr float, ptr [[A]], i64 [[IV]] +; CHECK-NEXT: [[ARR_IDX2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i64 [[OFFSET2:%.*]] ; CHECK-NEXT: [[L2:%.*]] = load float, ptr [[ARR_IDX2]], align 4 ; CHECK-NEXT: [[M:%.*]] = fmul fast float [[L2]], [[B:%.*]] ; CHECK-NEXT: [[AD:%.*]] = fadd fast float [[L1]], [[M]] ; CHECK-NEXT: store float [[AD]], ptr [[ARR_IDX]], align 4 -; CHECK-NEXT: [[C_IND:%.*]] = add nsw i64 [[IV]], -1 -; CHECK-NEXT: [[C_IDX:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i64 [[C_IND]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr float, ptr [[C:%.*]], i64 [[IV]] +; CHECK-NEXT: [[C_IDX:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i64 -1 ; CHECK-NEXT: [[LC:%.*]] = load float, ptr [[C_IDX]], align 4 ; CHECK-NEXT: [[VC:%.*]] = fadd float [[LC]], 1.000000e+00 ; CHECK-NEXT: [[C_IDX2:%.*]] = getelementptr inbounds float, ptr [[C]], i64 [[IV]] diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll --- a/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll +++ b/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused-loops.ll @@ -27,14 +27,14 @@ ; CHECK-NEXT: br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT: [[TMP1:%.*]] = getelementptr double, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr double, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8 ; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP2]], i64 4 ; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT: [[TMP5:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr double, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr double, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT: [[COL_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8 ; CHECK-NEXT: [[VEC_GEP3:%.*]] = getelementptr double, ptr [[TMP5]], i64 4 ; CHECK-NEXT: [[COL_LOAD4:%.*]] = load <2 x double>, ptr [[VEC_GEP3]], align 8 @@ -55,8 +55,8 @@ ; CHECK-NEXT: [[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT: [[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT: [[TMP10:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP10]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP12:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP11]] +; CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[C:%.*]], i64 [[TMP10]] +; CHECK-NEXT: [[TMP12:%.*]] = getelementptr double, ptr [[TMP11]], i64 [[ROWS_IV]] ; CHECK-NEXT: store <2 x double> [[TMP7]], ptr [[TMP12]], align 8 ; CHECK-NEXT: [[VEC_GEP16:%.*]] = getelementptr double, ptr [[TMP12]], i64 4 ; CHECK-NEXT: store <2 x double> [[TMP9]], ptr [[VEC_GEP16]], align 8 @@ -103,14 +103,14 @@ ; CHECK-NEXT: br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[INNER_IV]], 1 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 2 ; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT: [[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT: [[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 4 ; CHECK-NEXT: [[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -135,8 +135,8 @@ ; CHECK-NEXT: [[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT: [[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT: [[TMP14:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[C:%.*]], i64 [[TMP15]] +; CHECK-NEXT: [[TMP15:%.*]] = getelementptr i64, ptr [[C:%.*]], i64 [[TMP14]] +; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP15]], i64 [[ROWS_IV]] ; CHECK-NEXT: store <2 x i64> [[TMP9]], ptr [[TMP16]], align 8 ; CHECK-NEXT: [[VEC_GEP16:%.*]] = getelementptr i64, ptr [[TMP16]], i64 2 ; CHECK-NEXT: store <2 x i64> [[TMP13]], ptr [[VEC_GEP16]], align 8 @@ -189,14 +189,14 @@ ; CHECK-NEXT: br label [[INNER_BODY:%.*]] ; CHECK: inner.body: ; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[INNER_IV]], 2 -; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP1]] +; CHECK-NEXT: [[TMP1:%.*]] = getelementptr i64, ptr [[A:%.*]], i64 [[TMP0]] +; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i64 [[ROWS_IV]] ; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8 ; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr i64, ptr [[TMP2]], i64 4 ; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x i64>, ptr [[VEC_GEP]], align 8 ; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], [[INNER_IV]] -; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP4]] +; CHECK-NEXT: [[TMP4:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[TMP3]] +; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i64 [[INNER_IV]] ; CHECK-NEXT: [[COL_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8 ; CHECK-NEXT: [[VEC_GEP3:%.*]] = getelementptr i64, ptr [[TMP5]], i64 2 ; CHECK-NEXT: [[COL_LOAD4:%.*]] = load <2 x i64>, ptr [[VEC_GEP3]], align 8 @@ -221,8 +221,8 @@ ; CHECK-NEXT: [[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT: [[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_STEP]], 4 ; CHECK-NEXT: [[TMP14:%.*]] = shl i64 [[COLS_IV]], 2 -; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP14]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[C:%.*]], i64 [[TMP15]] +; CHECK-NEXT: [[TMP15:%.*]] = getelementptr i64, ptr [[C:%.*]], i64 [[TMP14]] +; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP15]], i64 [[ROWS_IV]] ; CHECK-NEXT: store <2 x i64> [[TMP9]], ptr [[TMP16]], align 8 ; CHECK-NEXT: [[VEC_GEP16:%.*]] = getelementptr i64, ptr [[TMP16]], i64 4 ; CHECK-NEXT: store <2 x i64> [[TMP13]], ptr [[VEC_GEP16]], align 8 @@ -334,8 +334,8 @@ ; CHECK-NEXT: [[ROWS_STEP]] = add i64 [[ROWS_IV]], 2 ; CHECK-NEXT: [[ROWS_COND_NOT:%.*]] = icmp eq i64 [[ROWS_IV]], 0 ; CHECK-NEXT: [[TMP18:%.*]] = shl i64 [[COLS_IV]], 1 -; CHECK-NEXT: [[TMP19:%.*]] = add i64 [[TMP18]], [[ROWS_IV]] -; CHECK-NEXT: [[TMP20:%.*]] = getelementptr float, ptr [[C]], i64 [[TMP19]] +; CHECK-NEXT: [[TMP19:%.*]] = getelementptr float, ptr [[C]], i64 [[TMP18]] +; CHECK-NEXT: [[TMP20:%.*]] = getelementptr float, ptr [[TMP19]], i64 [[ROWS_IV]] ; CHECK-NEXT: store <2 x float> [[TMP15]], ptr [[TMP20]], align 8 ; CHECK-NEXT: [[VEC_GEP23:%.*]] = getelementptr float, ptr [[TMP20]], i64 2 ; CHECK-NEXT: store <2 x float> [[TMP17]], ptr [[VEC_GEP23]], align 8