diff options
Diffstat (limited to 'llvm/test/Analysis')
6 files changed, 238 insertions, 49 deletions
diff --git a/llvm/test/Analysis/BasicAA/intrinsics.ll b/llvm/test/Analysis/BasicAA/intrinsics.ll index f8b30df..56d762b 100644 --- a/llvm/test/Analysis/BasicAA/intrinsics.ll +++ b/llvm/test/Analysis/BasicAA/intrinsics.ll @@ -1,3 +1,4 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -aa-pipeline=basic-aa -passes=gvn -S < %s | FileCheck %s target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:32:64-v128:32:128-a0:0:32-n32" @@ -5,12 +6,15 @@ target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32- ; BasicAA should prove that these calls don't interfere, since they are ; IntrArgReadMem and have noalias pointers. -; CHECK: define <8 x i16> @test0(ptr noalias %p, ptr noalias %q, <8 x i16> %y, <8 x i1> %m, <8 x i16> %pt) { -; CHECK-NEXT: entry: -; CHECK-NEXT: %a = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr %p, i32 16, <8 x i1> %m, <8 x i16> %pt) [[ATTR:#[0-9]+]] -; CHECK-NEXT: call void @llvm.masked.store.v8i16.p0(<8 x i16> %y, ptr %q, i32 16, <8 x i1> %m) -; CHECK-NEXT: %c = add <8 x i16> %a, %a define <8 x i16> @test0(ptr noalias %p, ptr noalias %q, <8 x i16> %y, <8 x i1> %m, <8 x i16> %pt) { +; CHECK-LABEL: define <8 x i16> @test0( +; CHECK-SAME: ptr noalias [[P:%.*]], ptr noalias [[Q:%.*]], <8 x i16> [[Y:%.*]], <8 x i1> [[M:%.*]], <8 x i16> [[PT:%.*]]) { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[A:%.*]] = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr [[P]], i32 16, <8 x i1> [[M]], <8 x i16> [[PT]]) #[[ATTR2:[0-9]+]] +; CHECK-NEXT: call void @llvm.masked.store.v8i16.p0(<8 x i16> [[Y]], ptr [[Q]], i32 16, <8 x i1> [[M]]) +; CHECK-NEXT: [[C:%.*]] = add <8 x i16> [[A]], [[A]] +; CHECK-NEXT: ret <8 x i16> [[C]] +; entry: %a = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr %p, i32 16, <8 x i1> %m, <8 x i16> %pt) nounwind call void @llvm.masked.store.v8i16.p0(<8 x i16> %y, ptr %q, i32 16, <8 x i1> %m) @@ -24,4 +28,3 @@ declare void @llvm.masked.store.v8i16.p0(<8 x i16>, ptr, i32, <8 x i1>) nounwind ; CHECK: attributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: read) } ; CHECK: attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: write) } -; CHECK: attributes [[ATTR]] = { nounwind } diff --git a/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll b/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll index 7e980c9..ffd8259 100644 --- a/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll +++ b/llvm/test/Analysis/BasicAA/scalable-dse-aa.ll @@ -1,10 +1,23 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt < %s -aa-pipeline=basic-aa -passes=dse -S | FileCheck %s define <vscale x 4 x float> @dead_scalable_store(ptr %0) { ; CHECK-LABEL: define <vscale x 4 x float> @dead_scalable_store( -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask) -; CHECK-NOT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.32, ptr nonnull %gep.arr.32, i32 1, <vscale x 4 x i1> %mask) -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <vscale x 4 x i1> %mask) +; CHECK-SAME: ptr [[TMP0:%.*]]) { +; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4 +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) +; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16 +; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48 +; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16 +; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48 +; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]] +; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]] ; %arr = alloca [64 x i32], align 4 %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) @@ -34,9 +47,21 @@ define <vscale x 4 x float> @dead_scalable_store(ptr %0) { define <4 x float> @dead_scalable_store_fixed(ptr %0) { ; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed( -; CHECK: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <4 x i1> %mask) -; CHECK-NOT: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.32, ptr nonnull %gep.arr.36, i32 1, <4 x i1> %mask2) -; CHECK: call void @llvm.masked.store.v4f32.p0(<4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <4 x i1> %mask) +; CHECK-SAME: ptr [[TMP0:%.*]]) { +; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4 +; CHECK-NEXT: [[MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 4) +; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16 +; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48 +; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16 +; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48 +; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <4 x i1> [[MASK]]) +; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <4 x i1> [[MASK]]) +; CHECK-NEXT: [[FADDOP0:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: [[FADDOP1:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: [[FADD:%.*]] = fadd <4 x float> [[FADDOP0]], [[FADDOP1]] +; CHECK-NEXT: ret <4 x float> [[FADD]] ; %arr = alloca [64 x i32], align 4 %mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 4) @@ -67,9 +92,25 @@ define <4 x float> @dead_scalable_store_fixed(ptr %0) { define <vscale x 4 x float> @scalable_store_partial_overwrite(ptr %0) { ; CHECK-LABEL: define <vscale x 4 x float> @scalable_store_partial_overwrite( -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask) -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.30, ptr nonnull %gep.arr.30, i32 1, <vscale x 4 x i1> %mask) -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.48, ptr nonnull %gep.arr.48, i32 1, <vscale x 4 x i1> %mask) +; CHECK-SAME: ptr [[TMP0:%.*]]) { +; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4 +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) +; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16 +; CHECK-NEXT: [[GEP_0_30:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 30 +; CHECK-NEXT: [[GEP_0_48:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 48 +; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16 +; CHECK-NEXT: [[GEP_ARR_30:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 30 +; CHECK-NEXT: [[GEP_ARR_48:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 48 +; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[LOAD_0_30:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_30]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_30]], ptr nonnull [[GEP_ARR_30]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[LOAD_0_48:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_48]], ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_48]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]] +; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]] ; %arr = alloca [64 x i32], align 4 %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) @@ -99,9 +140,23 @@ define <vscale x 4 x float> @scalable_store_partial_overwrite(ptr %0) { define <vscale x 4 x float> @dead_scalable_store_small_mask(ptr %0) { ; CHECK-LABEL: define <vscale x 4 x float> @dead_scalable_store_small_mask( -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.16, ptr nonnull %gep.arr.16, i32 1, <vscale x 4 x i1> %mask) -; CHECK-NOT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.30, ptr nonnull %gep.arr.30, i32 1, <vscale x 4 x i1> %mask) -; CHECK: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> %load.0.46, ptr nonnull %gep.arr.46, i32 1, <vscale x 4 x i1> %mask) +; CHECK-SAME: ptr [[TMP0:%.*]]) { +; CHECK-NEXT: [[ARR:%.*]] = alloca [64 x i32], align 4 +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) +; CHECK-NEXT: [[GEP_0_16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 16 +; CHECK-NEXT: [[GEP_0_46:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP0]], i64 46 +; CHECK-NEXT: [[GEP_ARR_16:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 16 +; CHECK-NEXT: [[GEP_ARR_46:%.*]] = getelementptr inbounds nuw i8, ptr [[ARR]], i64 46 +; CHECK-NEXT: [[LOAD_0_16:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_16]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_16]], ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[LOAD_0_46:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_0_46]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0_46]], ptr nonnull [[GEP_ARR_46]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[SMALLMASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 2) +; CHECK-NEXT: [[FADDOP0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_16]], i32 1, <vscale x 4 x i1> [[SMALLMASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADDOP1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[GEP_ARR_46]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: [[FADD:%.*]] = fadd <vscale x 4 x float> [[FADDOP0]], [[FADDOP1]] +; CHECK-NEXT: ret <vscale x 4 x float> [[FADD]] +; %arr = alloca [64 x i32], align 4 %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 4) @@ -131,7 +186,12 @@ define <vscale x 4 x float> @dead_scalable_store_small_mask(ptr %0) { define <vscale x 4 x float> @dead_scalar_store(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 4 x float> @dead_scalar_store( -; CHECK-NOT: store i32 20, ptr %gep.1.12 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i128(i128 0, i128 4) +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]] ; %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i128(i128 0, i128 4) %gep.1.12 = getelementptr inbounds nuw i8, ptr %1, i64 12 @@ -144,10 +204,17 @@ define <vscale x 4 x float> @dead_scalar_store(ptr noalias %0, ptr %1) { } -; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed_large_mask( -; CHECK-NOT: store i32 20, ptr %1 -; CHECK: store i32 50, ptr %gep.5 define <4 x float> @dead_scalable_store_fixed_large_mask(ptr noalias %0, ptr %1) { +; CHECK-LABEL: define <4 x float> @dead_scalable_store_fixed_large_mask( +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 7) +; CHECK-NEXT: [[GEP_5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i64 5 +; CHECK-NEXT: store i32 50, ptr [[GEP_5]], align 4 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[TMP0]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.v4f32.p0(<4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr nonnull [[TMP1]], i32 1, <4 x i1> [[MASK]], <4 x float> zeroinitializer) +; CHECK-NEXT: ret <4 x float> [[RETVAL]] +; %mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 0, i32 7) store i32 20, ptr %1 @@ -164,8 +231,16 @@ define <4 x float> @dead_scalable_store_fixed_large_mask(ptr noalias %0, ptr %1) ; This get active lane mask may cover 4 or 8 integers define <vscale x 4 x float> @mask_gt_minimum_num_elts(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 4 x float> @mask_gt_minimum_num_elts( -; CHECK: store i32 10, ptr %gep.1.12 -; CHECK: store i32 20, ptr %gep.1.28 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 8) +; CHECK-NEXT: [[GEP_1_12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 12 +; CHECK-NEXT: store i32 10, ptr [[GEP_1_12]], align 4 +; CHECK-NEXT: [[GEP_1_28:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 28 +; CHECK-NEXT: store i32 20, ptr [[GEP_1_28]], align 4 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]] ; %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 0, i32 8) %gep.1.12 = getelementptr inbounds nuw i8, ptr %1, i64 12 @@ -182,7 +257,13 @@ define <vscale x 4 x float> @mask_gt_minimum_num_elts(ptr noalias %0, ptr %1) { ; Don't do anything if the mask's Op1 < Op0 define <vscale x 4 x float> @active_lane_mask_lt(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 4 x float> @active_lane_mask_lt( -; CHECK: store i32 20, ptr %1 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 4, i32 2) +; CHECK-NEXT: store i32 20, ptr [[TMP1]], align 4 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]] ; %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 4, i32 2) store i32 20, ptr %1 @@ -196,7 +277,13 @@ define <vscale x 4 x float> @active_lane_mask_lt(ptr noalias %0, ptr %1) { ; Don't do anything if the mask's Op1 == Op0 define <vscale x 4 x float> @active_lane_mask_eq(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 4 x float> @active_lane_mask_eq( -; CHECK: store i32 20, ptr %1 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 2, i32 2) +; CHECK-NEXT: store i32 20, ptr [[TMP1]], align 4 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]] ; %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 2, i32 2) store i32 20, ptr %1 @@ -209,8 +296,14 @@ define <vscale x 4 x float> @active_lane_mask_eq(ptr noalias %0, ptr %1) { define <vscale x 16 x i8> @scalar_stores_small_mask(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 16 x i8> @scalar_stores_small_mask( -; CHECK-NOT: store i8 60, ptr %gep.1.6 -; CHECK: store i8 120, ptr %gep.1.8 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i8(i8 0, i8 7) +; CHECK-NEXT: [[GEP_1_8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 8 +; CHECK-NEXT: store i8 120, ptr [[GEP_1_8]], align 1 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 16 x i1> [[MASK]], <vscale x 16 x i8> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[LOAD_0]], ptr [[TMP1]], i32 1, <vscale x 16 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP1]], i32 1, <vscale x 16 x i1> [[MASK]], <vscale x 16 x i8> zeroinitializer) +; CHECK-NEXT: ret <vscale x 16 x i8> [[RETVAL]] ; %mask = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i8.i8(i8 0, i8 7) %gep.1.6 = getelementptr inbounds nuw i8, ptr %1, i64 6 @@ -226,10 +319,14 @@ define <vscale x 16 x i8> @scalar_stores_small_mask(ptr noalias %0, ptr %1) { define <vscale x 4 x float> @dead_scalar_store_offset(ptr noalias %0, ptr %1) { ; CHECK-LABEL: define <vscale x 4 x float> @dead_scalar_store_offset( -; CHECK-NOT: store i32 10, ptr %gep.1.0 -; CHECK-NOT: store i32 20, ptr %gep.1.4 -; CHECK-NOT: store i32 30, ptr %gep.1.8 -; CHECK: store i32 40, ptr %gep.1.12 +; CHECK-SAME: ptr noalias [[TMP0:%.*]], ptr [[TMP1:%.*]]) { +; CHECK-NEXT: [[MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 1, i32 4) +; CHECK-NEXT: [[GEP_1_12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i64 12 +; CHECK-NEXT: store i32 40, ptr [[GEP_1_12]], align 4 +; CHECK-NEXT: [[LOAD_0:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP0]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[LOAD_0]], ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]]) +; CHECK-NEXT: [[RETVAL:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr nonnull [[TMP1]], i32 1, <vscale x 4 x i1> [[MASK]], <vscale x 4 x float> zeroinitializer) +; CHECK-NEXT: ret <vscale x 4 x float> [[RETVAL]] ; %mask = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i32(i32 1, i32 4) %gep.1.0 = getelementptr inbounds nuw i8, ptr %1, i64 0 diff --git a/llvm/test/Analysis/CostModel/AArch64/sve-illegal-types.ll b/llvm/test/Analysis/CostModel/AArch64/sve-illegal-types.ll index 3e85760..2c838e2 100644 --- a/llvm/test/Analysis/CostModel/AArch64/sve-illegal-types.ll +++ b/llvm/test/Analysis/CostModel/AArch64/sve-illegal-types.ll @@ -1,14 +1,17 @@ +; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=aarch64--linux-gnu -mattr=+sve < %s | FileCheck %s target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128" define void @load_store(ptr %ptrs) { ; CHECK-LABEL: 'load_store' -; CHECK-NEXT: Invalid cost for instruction: %load1 = load <vscale x 1 x i128>, ptr undef -; CHECK-NEXT: Invalid cost for instruction: %load2 = load <vscale x 2 x i128>, ptr undef -; CHECK-NEXT: Invalid cost for instruction: %load3 = load <vscale x 1 x fp128>, ptr undef -; CHECK-NEXT: Invalid cost for instruction: %load4 = load <vscale x 2 x fp128>, ptr undef -; CHECK-NEXT: Invalid cost for instruction: store <vscale x 1 x i128> %load1, ptr %ptrs +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %load1 = load <vscale x 1 x i128>, ptr undef, align 16 +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %load2 = load <vscale x 2 x i128>, ptr undef, align 32 +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %load3 = load <vscale x 1 x fp128>, ptr undef, align 16 +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %load4 = load <vscale x 2 x fp128>, ptr undef, align 32 +; CHECK-NEXT: Cost Model: Invalid cost for instruction: store <vscale x 1 x i128> %load1, ptr %ptrs, align 16 +; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void +; %load1 = load <vscale x 1 x i128>, ptr undef %load2 = load <vscale x 2 x i128>, ptr undef %load3 = load <vscale x 1 x fp128>, ptr undef @@ -19,8 +22,10 @@ define void @load_store(ptr %ptrs) { define void @masked_load_store(ptr %ptrs, ptr %val, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) { ; CHECK-LABEL: 'masked_load_store' -; CHECK-NEXT: Invalid cost for instruction: %mload = call <vscale x 1 x i128> @llvm.masked.load.nxv1i128.p0(ptr %val, i32 8, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) -; CHECK-NEXT: Invalid cost for instruction: call void @llvm.masked.store.nxv1i128.p0(<vscale x 1 x i128> %mload, ptr %ptrs, i32 8, <vscale x 1 x i1> %mask) +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %mload = call <vscale x 1 x i128> @llvm.masked.load.nxv1i128.p0(ptr %val, i32 8, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) +; CHECK-NEXT: Cost Model: Invalid cost for instruction: call void @llvm.masked.store.nxv1i128.p0(<vscale x 1 x i128> %mload, ptr %ptrs, i32 8, <vscale x 1 x i1> %mask) +; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void +; %mload = call <vscale x 1 x i128> @llvm.masked.load.nxv1i128(ptr %val, i32 8, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) call void @llvm.masked.store.nxv1i128(<vscale x 1 x i128> %mload, ptr %ptrs, i32 8, <vscale x 1 x i1> %mask) ret void @@ -28,8 +33,10 @@ define void @masked_load_store(ptr %ptrs, ptr %val, <vscale x 1 x i1> %mask, <vs define void @masked_gather_scatter(<vscale x 1 x ptr> %ptrs, <vscale x 1 x ptr> %val, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) { ; CHECK-LABEL: 'masked_gather_scatter' -; CHECK-NEXT: Invalid cost for instruction: %mgather = call <vscale x 1 x i128> @llvm.masked.gather.nxv1i128.nxv1p0(<vscale x 1 x ptr> %val, i32 0, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) -; CHECK-NEXT: Invalid cost for instruction: call void @llvm.masked.scatter.nxv1i128.nxv1p0(<vscale x 1 x i128> %mgather, <vscale x 1 x ptr> %ptrs, i32 0, <vscale x 1 x i1> %mask) +; CHECK-NEXT: Cost Model: Invalid cost for instruction: %mgather = call <vscale x 1 x i128> @llvm.masked.gather.nxv1i128.nxv1p0(<vscale x 1 x ptr> %val, i32 0, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) +; CHECK-NEXT: Cost Model: Invalid cost for instruction: call void @llvm.masked.scatter.nxv1i128.nxv1p0(<vscale x 1 x i128> %mgather, <vscale x 1 x ptr> %ptrs, i32 0, <vscale x 1 x i1> %mask) +; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret void +; %mgather = call <vscale x 1 x i128> @llvm.masked.gather.nxv1i128(<vscale x 1 x ptr> %val, i32 0, <vscale x 1 x i1> %mask, <vscale x 1 x i128> %passthru) call void @llvm.masked.scatter.nxv1i128(<vscale x 1 x i128> %mgather, <vscale x 1 x ptr> %ptrs, i32 0, <vscale x 1 x i1> %mask) ret void diff --git a/llvm/test/Analysis/DXILResource/buffer-frombinding.ll b/llvm/test/Analysis/DXILResource/buffer-frombinding.ll index aeeb21e..ab1945d 100644 --- a/llvm/test/Analysis/DXILResource/buffer-frombinding.ll +++ b/llvm/test/Analysis/DXILResource/buffer-frombinding.ll @@ -150,7 +150,7 @@ define void @test_typedbuffer() { ; CHECK: Kind: CBuffer ; CHECK: CBuffer size: 4 - %cb1 = call target("dx.CBuffer", target("dx.Layout", {float}, 4, 0)) + %cb1 = call target("dx.CBuffer", <{ [2 x <{ float, target("dx.Padding", 12) }>], float }>) @llvm.dx.resource.handlefrombinding(i32 1, i32 8, i32 1, i32 0, ptr @Constants.str) ; CHECK: Resource [[CB1:[0-9]+]]: ; CHECK: Name: Constants @@ -161,7 +161,7 @@ define void @test_typedbuffer() { ; CHECK: Size: 1 ; CHECK: Class: CBV ; CHECK: Kind: CBuffer - ; CHECK: CBuffer size: 4 + ; CHECK: CBuffer size: 36 ; CHECK-NOT: Resource {{[0-9]+}}: diff --git a/llvm/test/Analysis/ScalarEvolution/ne-guard-multiple-trip-count.ll b/llvm/test/Analysis/ScalarEvolution/ne-guard-multiple-trip-count.ll new file mode 100644 index 0000000..220c5a1 --- /dev/null +++ b/llvm/test/Analysis/ScalarEvolution/ne-guard-multiple-trip-count.ll @@ -0,0 +1,72 @@ +; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5 +; RUN: opt -passes='print<scalar-evolution>' -disable-output %s 2>&1 | FileCheck %s + +declare void @foo() + +; Tests with multiple guards for the same value and different values. + +define void @test_guard_order_b_then_c_and_d(ptr %a, ptr %b, ptr %c, ptr %d) { +; CHECK-LABEL: 'test_guard_order_b_then_c_and_d' +; CHECK-NEXT: Classifying expressions for: @test_guard_order_b_then_c_and_d +; CHECK-NEXT: %iv = phi ptr [ %a, %entry ], [ %iv.next, %loop ] +; CHECK-NEXT: --> {%a,+,1}<%loop> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64) + %a) LoopDispositions: { %loop: Computable } +; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 1 +; CHECK-NEXT: --> {(1 + %a),+,1}<%loop> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64) + %a) LoopDispositions: { %loop: Computable } +; CHECK-NEXT: Determining loop execution counts for: @test_guard_order_b_then_c_and_d +; CHECK-NEXT: Loop %loop: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64)) +; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 -2 +; CHECK-NEXT: Loop %loop: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64)) +; CHECK-NEXT: Loop %loop: Trip multiple is 1 +; +entry: + %cmp.eq.b = icmp ne ptr %a, %b + %cmp.eq.c = icmp ne ptr %a, %c + %cmp.eq.d = icmp ne ptr %b, %d + call void @llvm.assume(i1 %cmp.eq.b) + call void @llvm.assume(i1 %cmp.eq.c) + call void @llvm.assume(i1 %cmp.eq.d) + br label %loop + +loop: + %iv = phi ptr [ %a, %entry ], [ %iv.next, %loop ] + %iv.next = getelementptr i8, ptr %iv, i64 1 + call void @foo() + %ec = icmp eq ptr %iv.next, %b + br i1 %ec, label %exit, label %loop + +exit: + ret void +} + +define void @test_guard_order_d_then_c_and_b(ptr %a, ptr %b, ptr %c, ptr %d) { +; CHECK-LABEL: 'test_guard_order_d_then_c_and_b' +; CHECK-NEXT: Classifying expressions for: @test_guard_order_d_then_c_and_b +; CHECK-NEXT: %iv = phi ptr [ %a, %entry ], [ %iv.next, %loop ] +; CHECK-NEXT: --> {%a,+,1}<%loop> U: full-set S: full-set Exits: (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64) + %a) LoopDispositions: { %loop: Computable } +; CHECK-NEXT: %iv.next = getelementptr i8, ptr %iv, i64 1 +; CHECK-NEXT: --> {(1 + %a),+,1}<%loop> U: full-set S: full-set Exits: ((-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64) + %a) LoopDispositions: { %loop: Computable } +; CHECK-NEXT: Determining loop execution counts for: @test_guard_order_d_then_c_and_b +; CHECK-NEXT: Loop %loop: backedge-taken count is (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64)) +; CHECK-NEXT: Loop %loop: constant max backedge-taken count is i64 -2 +; CHECK-NEXT: Loop %loop: symbolic max backedge-taken count is (-1 + (-1 * (ptrtoint ptr %a to i64)) + (ptrtoint ptr %b to i64)) +; CHECK-NEXT: Loop %loop: Trip multiple is 1 +; +entry: + %cmp.eq.b = icmp ne ptr %a, %b + %cmp.eq.c = icmp ne ptr %a, %c + %cmp.eq.d = icmp ne ptr %b, %d + call void @llvm.assume(i1 %cmp.eq.d) + call void @llvm.assume(i1 %cmp.eq.c) + call void @llvm.assume(i1 %cmp.eq.b) + br label %loop + +loop: + %iv = phi ptr [ %a, %entry ], [ %iv.next, %loop ] + %iv.next = getelementptr i8, ptr %iv, i64 1 + call void @foo() + %ec = icmp eq ptr %iv.next, %b + br i1 %ec, label %exit, label %loop + +exit: + ret void +} diff --git a/llvm/test/Analysis/TypeBasedAliasAnalysis/intrinsics.ll b/llvm/test/Analysis/TypeBasedAliasAnalysis/intrinsics.ll index 7ec674a..dc4a72e 100644 --- a/llvm/test/Analysis/TypeBasedAliasAnalysis/intrinsics.ll +++ b/llvm/test/Analysis/TypeBasedAliasAnalysis/intrinsics.ll @@ -1,3 +1,4 @@ +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6 ; RUN: opt -aa-pipeline=tbaa,basic-aa -passes=gvn -S < %s | FileCheck %s target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:32:64-v128:32:128-a0:0:32-n32" @@ -5,12 +6,15 @@ target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32- ; TBAA should prove that these calls don't interfere, since they are ; IntrArgReadMem and have TBAA metadata. -; CHECK: define <8 x i16> @test0(ptr %p, ptr %q, <8 x i16> %y, <8 x i1> %m, <8 x i16> %pt) { -; CHECK-NEXT: entry: -; CHECK-NEXT: %a = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr %p, i32 16, <8 x i1> %m, <8 x i16> %pt) [[NUW:#[0-9]+]] -; CHECK-NEXT: call void @llvm.masked.store.v8i16.p0(<8 x i16> %y, ptr %q, i32 16, <8 x i1> %m) -; CHECK-NEXT: %c = add <8 x i16> %a, %a define <8 x i16> @test0(ptr %p, ptr %q, <8 x i16> %y, <8 x i1> %m, <8 x i16> %pt) { +; CHECK-LABEL: define <8 x i16> @test0( +; CHECK-SAME: ptr [[P:%.*]], ptr [[Q:%.*]], <8 x i16> [[Y:%.*]], <8 x i1> [[M:%.*]], <8 x i16> [[PT:%.*]]) { +; CHECK-NEXT: [[ENTRY:.*:]] +; CHECK-NEXT: [[A:%.*]] = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr [[P]], i32 16, <8 x i1> [[M]], <8 x i16> [[PT]]) #[[ATTR2:[0-9]+]], !tbaa [[B_TBAA0:![0-9]+]] +; CHECK-NEXT: call void @llvm.masked.store.v8i16.p0(<8 x i16> [[Y]], ptr [[Q]], i32 16, <8 x i1> [[M]]), !tbaa [[A_TBAA3:![0-9]+]] +; CHECK-NEXT: [[C:%.*]] = add <8 x i16> [[A]], [[A]] +; CHECK-NEXT: ret <8 x i16> [[C]] +; entry: %a = call <8 x i16> @llvm.masked.load.v8i16.p0(ptr %p, i32 16, <8 x i1> %m, <8 x i16> %pt) nounwind, !tbaa !2 call void @llvm.masked.store.v8i16.p0(<8 x i16> %y, ptr %q, i32 16, <8 x i1> %m), !tbaa !1 @@ -24,10 +28,16 @@ declare void @llvm.masked.store.v8i16.p0(<8 x i16>, ptr, i32, <8 x i1>) nounwind ; CHECK: attributes #0 = { nocallback nofree nosync nounwind willreturn memory(argmem: read) } ; CHECK: attributes #1 = { nocallback nofree nosync nounwind willreturn memory(argmem: write) } -; CHECK: attributes [[NUW]] = { nounwind } !0 = !{!"tbaa root"} !1 = !{!3, !3, i64 0} !2 = !{!4, !4, i64 0} !3 = !{!"A", !0} !4 = !{!"B", !0} +;. +; CHECK: [[B_TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0} +; CHECK: [[META1]] = !{!"B", [[META2:![0-9]+]]} +; CHECK: [[META2]] = !{!"tbaa root"} +; CHECK: [[A_TBAA3]] = !{[[META4:![0-9]+]], [[META4]], i64 0} +; CHECK: [[META4]] = !{!"A", [[META2]]} +;. |