1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=loop-vectorize -mtriple=riscv64 -mattr=+v -S | FileCheck %s 3; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=off -mtriple=riscv64 -mattr=+v -S | FileCheck %s --check-prefix=FIXED 4; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -mtriple=riscv64 -mattr=+v -S | FileCheck %s --check-prefix=SCALABLE 5 6define void @load_store_factor2_i32(ptr %p) { 7; CHECK-LABEL: @load_store_factor2_i32( 8; CHECK-NEXT: entry: 9; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 10; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 4 11; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 12; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 13; CHECK: vector.ph: 14; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 15; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 4 16; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 17; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 18; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 19; CHECK-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 20; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 21; CHECK: vector.body: 22; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 23; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 24; CHECK-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 25; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP7]] 26; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP8]], align 4 27; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]]) 28; CHECK-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0 29; CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1 30; CHECK-NEXT: [[TMP12:%.*]] = add <vscale x 4 x i32> [[TMP10]], splat (i32 1) 31; CHECK-NEXT: [[TMP15:%.*]] = add <vscale x 4 x i32> [[TMP11]], splat (i32 2) 32; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 8 x i32> @llvm.vector.interleave2.nxv8i32(<vscale x 4 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP15]]) 33; CHECK-NEXT: store <vscale x 8 x i32> [[INTERLEAVED_VEC]], ptr [[TMP8]], align 4 34; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 35; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 36; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 37; CHECK: middle.block: 38; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 39; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 40; CHECK: scalar.ph: 41; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 42; CHECK-NEXT: br label [[LOOP:%.*]] 43; CHECK: loop: 44; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 45; CHECK-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 46; CHECK-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 47; CHECK-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 48; CHECK-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 49; CHECK-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 50; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 51; CHECK-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 52; CHECK-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 53; CHECK-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 54; CHECK-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 55; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 56; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 57; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] 58; CHECK: exit: 59; CHECK-NEXT: ret void 60; 61; FIXED-LABEL: @load_store_factor2_i32( 62; FIXED-NEXT: entry: 63; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 64; FIXED: vector.ph: 65; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 66; FIXED: vector.body: 67; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 68; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 69; FIXED-NEXT: [[TMP1:%.*]] = shl i64 [[TMP0]], 1 70; FIXED-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP1]] 71; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, ptr [[TMP2]], align 4 72; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 73; FIXED-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 74; FIXED-NEXT: [[TMP4:%.*]] = add <8 x i32> [[STRIDED_VEC]], splat (i32 1) 75; FIXED-NEXT: [[TMP7:%.*]] = add <8 x i32> [[STRIDED_VEC1]], splat (i32 2) 76; FIXED-NEXT: [[TMP9:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> [[TMP7]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 77; FIXED-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i32> [[TMP9]], <16 x i32> poison, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15> 78; FIXED-NEXT: store <16 x i32> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 4 79; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 80; FIXED-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 81; FIXED-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 82; FIXED: middle.block: 83; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 84; FIXED: scalar.ph: 85; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 86; FIXED-NEXT: br label [[LOOP:%.*]] 87; FIXED: loop: 88; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 89; FIXED-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 90; FIXED-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 91; FIXED-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 92; FIXED-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 93; FIXED-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 94; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 95; FIXED-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 96; FIXED-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 97; FIXED-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 98; FIXED-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 99; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 100; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 101; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] 102; FIXED: exit: 103; FIXED-NEXT: ret void 104; 105; SCALABLE-LABEL: @load_store_factor2_i32( 106; SCALABLE-NEXT: entry: 107; SCALABLE-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 108; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 4 109; SCALABLE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 110; SCALABLE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 111; SCALABLE: vector.ph: 112; SCALABLE-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 113; SCALABLE-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 4 114; SCALABLE-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 115; SCALABLE-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 116; SCALABLE-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 117; SCALABLE-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 118; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 119; SCALABLE: vector.body: 120; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 121; SCALABLE-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 122; SCALABLE-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 123; SCALABLE-NEXT: [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP7]] 124; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP8]], align 4 125; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]]) 126; SCALABLE-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0 127; SCALABLE-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1 128; SCALABLE-NEXT: [[TMP12:%.*]] = add <vscale x 4 x i32> [[TMP10]], splat (i32 1) 129; SCALABLE-NEXT: [[TMP15:%.*]] = add <vscale x 4 x i32> [[TMP11]], splat (i32 2) 130; SCALABLE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 8 x i32> @llvm.vector.interleave2.nxv8i32(<vscale x 4 x i32> [[TMP12]], <vscale x 4 x i32> [[TMP15]]) 131; SCALABLE-NEXT: store <vscale x 8 x i32> [[INTERLEAVED_VEC]], ptr [[TMP8]], align 4 132; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 133; SCALABLE-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 134; SCALABLE-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] 135; SCALABLE: middle.block: 136; SCALABLE-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 137; SCALABLE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 138; SCALABLE: scalar.ph: 139; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 140; SCALABLE-NEXT: br label [[LOOP:%.*]] 141; SCALABLE: loop: 142; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 143; SCALABLE-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 144; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 145; SCALABLE-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 146; SCALABLE-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 147; SCALABLE-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 148; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 149; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 150; SCALABLE-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 151; SCALABLE-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 152; SCALABLE-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 153; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 154; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 155; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]] 156; SCALABLE: exit: 157; SCALABLE-NEXT: ret void 158; 159entry: 160 br label %loop 161loop: 162 %i = phi i64 [0, %entry], [%nexti, %loop] 163 164 %offset0 = shl i64 %i, 1 165 %q0 = getelementptr i32, ptr %p, i64 %offset0 166 %x0 = load i32, ptr %q0 167 %y0 = add i32 %x0, 1 168 store i32 %y0, ptr %q0 169 170 %offset1 = add i64 %offset0, 1 171 %q1 = getelementptr i32, ptr %p, i64 %offset1 172 %x1 = load i32, ptr %q1 173 %y1 = add i32 %x1, 2 174 store i32 %y1, ptr %q1 175 176 %nexti = add i64 %i, 1 177 %done = icmp eq i64 %nexti, 1024 178 br i1 %done, label %exit, label %loop 179exit: 180 ret void 181} 182 183define void @load_store_factor2_i64(ptr %p) { 184; CHECK-LABEL: @load_store_factor2_i64( 185; CHECK-NEXT: entry: 186; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 187; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2 188; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 189; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 190; CHECK: vector.ph: 191; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 192; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 2 193; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 194; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 195; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 196; CHECK-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 2 197; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 198; CHECK: vector.body: 199; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 200; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 201; CHECK-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 202; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP7]] 203; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 4 x i64>, ptr [[TMP8]], align 8 204; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.vector.deinterleave2.nxv4i64(<vscale x 4 x i64> [[WIDE_VEC]]) 205; CHECK-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 0 206; CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1 207; CHECK-NEXT: [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP10]], splat (i64 1) 208; CHECK-NEXT: [[TMP15:%.*]] = add <vscale x 2 x i64> [[TMP11]], splat (i64 2) 209; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 4 x i64> @llvm.vector.interleave2.nxv4i64(<vscale x 2 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP15]]) 210; CHECK-NEXT: store <vscale x 4 x i64> [[INTERLEAVED_VEC]], ptr [[TMP8]], align 8 211; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 212; CHECK-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 213; CHECK-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 214; CHECK: middle.block: 215; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 216; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 217; CHECK: scalar.ph: 218; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 219; CHECK-NEXT: br label [[LOOP:%.*]] 220; CHECK: loop: 221; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 222; CHECK-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 223; CHECK-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 224; CHECK-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 225; CHECK-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 226; CHECK-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 227; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 228; CHECK-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 229; CHECK-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 230; CHECK-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 231; CHECK-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 232; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 233; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 234; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] 235; CHECK: exit: 236; CHECK-NEXT: ret void 237; 238; FIXED-LABEL: @load_store_factor2_i64( 239; FIXED-NEXT: entry: 240; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 241; FIXED: vector.ph: 242; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 243; FIXED: vector.body: 244; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 245; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 246; FIXED-NEXT: [[TMP1:%.*]] = shl i64 [[TMP0]], 1 247; FIXED-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 248; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP2]], align 8 249; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 250; FIXED-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 251; FIXED-NEXT: [[TMP4:%.*]] = add <4 x i64> [[STRIDED_VEC]], splat (i64 1) 252; FIXED-NEXT: [[TMP7:%.*]] = add <4 x i64> [[STRIDED_VEC1]], splat (i64 2) 253; FIXED-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP4]], <4 x i64> [[TMP7]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 254; FIXED-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i64> [[TMP9]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7> 255; FIXED-NEXT: store <8 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 256; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 257; FIXED-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 258; FIXED-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 259; FIXED: middle.block: 260; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 261; FIXED: scalar.ph: 262; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 263; FIXED-NEXT: br label [[LOOP:%.*]] 264; FIXED: loop: 265; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 266; FIXED-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 267; FIXED-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 268; FIXED-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 269; FIXED-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 270; FIXED-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 271; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 272; FIXED-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 273; FIXED-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 274; FIXED-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 275; FIXED-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 276; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 277; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 278; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] 279; FIXED: exit: 280; FIXED-NEXT: ret void 281; 282; SCALABLE-LABEL: @load_store_factor2_i64( 283; SCALABLE-NEXT: entry: 284; SCALABLE-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 285; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2 286; SCALABLE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 287; SCALABLE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 288; SCALABLE: vector.ph: 289; SCALABLE-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 290; SCALABLE-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 2 291; SCALABLE-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 292; SCALABLE-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 293; SCALABLE-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 294; SCALABLE-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 2 295; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 296; SCALABLE: vector.body: 297; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 298; SCALABLE-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 299; SCALABLE-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 300; SCALABLE-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP7]] 301; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 4 x i64>, ptr [[TMP8]], align 8 302; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.vector.deinterleave2.nxv4i64(<vscale x 4 x i64> [[WIDE_VEC]]) 303; SCALABLE-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 0 304; SCALABLE-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1 305; SCALABLE-NEXT: [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP10]], splat (i64 1) 306; SCALABLE-NEXT: [[TMP15:%.*]] = add <vscale x 2 x i64> [[TMP11]], splat (i64 2) 307; SCALABLE-NEXT: [[INTERLEAVED_VEC:%.*]] = call <vscale x 4 x i64> @llvm.vector.interleave2.nxv4i64(<vscale x 2 x i64> [[TMP12]], <vscale x 2 x i64> [[TMP15]]) 308; SCALABLE-NEXT: store <vscale x 4 x i64> [[INTERLEAVED_VEC]], ptr [[TMP8]], align 8 309; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 310; SCALABLE-NEXT: [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 311; SCALABLE-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]] 312; SCALABLE: middle.block: 313; SCALABLE-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 314; SCALABLE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 315; SCALABLE: scalar.ph: 316; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 317; SCALABLE-NEXT: br label [[LOOP:%.*]] 318; SCALABLE: loop: 319; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 320; SCALABLE-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 321; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 322; SCALABLE-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 323; SCALABLE-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 324; SCALABLE-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 325; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 326; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 327; SCALABLE-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 328; SCALABLE-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 329; SCALABLE-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 330; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 331; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 332; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]] 333; SCALABLE: exit: 334; SCALABLE-NEXT: ret void 335; 336entry: 337 br label %loop 338loop: 339 %i = phi i64 [0, %entry], [%nexti, %loop] 340 341 %offset0 = shl i64 %i, 1 342 %q0 = getelementptr i64, ptr %p, i64 %offset0 343 %x0 = load i64, ptr %q0 344 %y0 = add i64 %x0, 1 345 store i64 %y0, ptr %q0 346 347 %offset1 = add i64 %offset0, 1 348 %q1 = getelementptr i64, ptr %p, i64 %offset1 349 %x1 = load i64, ptr %q1 350 %y1 = add i64 %x1, 2 351 store i64 %y1, ptr %q1 352 353 %nexti = add i64 %i, 1 354 %done = icmp eq i64 %nexti, 1024 355 br i1 %done, label %exit, label %loop 356exit: 357 ret void 358} 359 360define void @load_store_factor3_i32(ptr %p) { 361; CHECK-LABEL: @load_store_factor3_i32( 362; CHECK-NEXT: entry: 363; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 364; CHECK: vector.ph: 365; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 366; CHECK: vector.body: 367; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 368; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 369; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 370; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP1]] 371; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <24 x i32>, ptr [[TMP2]], align 4 372; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 373; CHECK-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 374; CHECK-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 375; CHECK-NEXT: [[TMP4:%.*]] = add <8 x i32> [[STRIDED_VEC]], splat (i32 1) 376; CHECK-NEXT: [[TMP6:%.*]] = add <8 x i32> [[STRIDED_VEC1]], splat (i32 2) 377; CHECK-NEXT: [[TMP9:%.*]] = add <8 x i32> [[STRIDED_VEC2]], splat (i32 3) 378; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> [[TMP6]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 379; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <8 x i32> [[TMP9]], <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison> 380; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <16 x i32> [[TMP11]], <16 x i32> [[TMP12]], <24 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23> 381; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <24 x i32> [[TMP13]], <24 x i32> poison, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 382; CHECK-NEXT: store <24 x i32> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 4 383; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 384; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 385; CHECK-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 386; CHECK: middle.block: 387; CHECK-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 388; CHECK: scalar.ph: 389; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 390; CHECK-NEXT: br label [[LOOP:%.*]] 391; CHECK: loop: 392; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 393; CHECK-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 394; CHECK-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 395; CHECK-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 396; CHECK-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 397; CHECK-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 398; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 399; CHECK-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 400; CHECK-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 401; CHECK-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 402; CHECK-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 403; CHECK-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 404; CHECK-NEXT: [[Q2:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET2]] 405; CHECK-NEXT: [[X2:%.*]] = load i32, ptr [[Q2]], align 4 406; CHECK-NEXT: [[Y2:%.*]] = add i32 [[X2]], 3 407; CHECK-NEXT: store i32 [[Y2]], ptr [[Q2]], align 4 408; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 409; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 410; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] 411; CHECK: exit: 412; CHECK-NEXT: ret void 413; 414; FIXED-LABEL: @load_store_factor3_i32( 415; FIXED-NEXT: entry: 416; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 417; FIXED: vector.ph: 418; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 419; FIXED: vector.body: 420; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 421; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 422; FIXED-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 423; FIXED-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP1]] 424; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <24 x i32>, ptr [[TMP2]], align 4 425; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 426; FIXED-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 427; FIXED-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 428; FIXED-NEXT: [[TMP4:%.*]] = add <8 x i32> [[STRIDED_VEC]], splat (i32 1) 429; FIXED-NEXT: [[TMP6:%.*]] = add <8 x i32> [[STRIDED_VEC1]], splat (i32 2) 430; FIXED-NEXT: [[TMP9:%.*]] = add <8 x i32> [[STRIDED_VEC2]], splat (i32 3) 431; FIXED-NEXT: [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> [[TMP6]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 432; FIXED-NEXT: [[TMP12:%.*]] = shufflevector <8 x i32> [[TMP9]], <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison> 433; FIXED-NEXT: [[TMP13:%.*]] = shufflevector <16 x i32> [[TMP11]], <16 x i32> [[TMP12]], <24 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23> 434; FIXED-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <24 x i32> [[TMP13]], <24 x i32> poison, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 435; FIXED-NEXT: store <24 x i32> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 4 436; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 437; FIXED-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 438; FIXED-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 439; FIXED: middle.block: 440; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 441; FIXED: scalar.ph: 442; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 443; FIXED-NEXT: br label [[LOOP:%.*]] 444; FIXED: loop: 445; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 446; FIXED-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 447; FIXED-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 448; FIXED-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 449; FIXED-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 450; FIXED-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 451; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 452; FIXED-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 453; FIXED-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 454; FIXED-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 455; FIXED-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 456; FIXED-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 457; FIXED-NEXT: [[Q2:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET2]] 458; FIXED-NEXT: [[X2:%.*]] = load i32, ptr [[Q2]], align 4 459; FIXED-NEXT: [[Y2:%.*]] = add i32 [[X2]], 3 460; FIXED-NEXT: store i32 [[Y2]], ptr [[Q2]], align 4 461; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 462; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 463; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] 464; FIXED: exit: 465; FIXED-NEXT: ret void 466; 467; SCALABLE-LABEL: @load_store_factor3_i32( 468; SCALABLE-NEXT: entry: 469; SCALABLE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 470; SCALABLE: vector.ph: 471; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 472; SCALABLE: vector.body: 473; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 474; SCALABLE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 475; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 476; SCALABLE-NEXT: [[TMP2:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP1]] 477; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <24 x i32>, ptr [[TMP2]], align 4 478; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21> 479; SCALABLE-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 1, i32 4, i32 7, i32 10, i32 13, i32 16, i32 19, i32 22> 480; SCALABLE-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <24 x i32> [[WIDE_VEC]], <24 x i32> poison, <8 x i32> <i32 2, i32 5, i32 8, i32 11, i32 14, i32 17, i32 20, i32 23> 481; SCALABLE-NEXT: [[TMP4:%.*]] = add <8 x i32> [[STRIDED_VEC]], splat (i32 1) 482; SCALABLE-NEXT: [[TMP6:%.*]] = add <8 x i32> [[STRIDED_VEC1]], splat (i32 2) 483; SCALABLE-NEXT: [[TMP9:%.*]] = add <8 x i32> [[STRIDED_VEC2]], splat (i32 3) 484; SCALABLE-NEXT: [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP4]], <8 x i32> [[TMP6]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 485; SCALABLE-NEXT: [[TMP12:%.*]] = shufflevector <8 x i32> [[TMP9]], <8 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison> 486; SCALABLE-NEXT: [[TMP13:%.*]] = shufflevector <16 x i32> [[TMP11]], <16 x i32> [[TMP12]], <24 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23> 487; SCALABLE-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <24 x i32> [[TMP13]], <24 x i32> poison, <24 x i32> <i32 0, i32 8, i32 16, i32 1, i32 9, i32 17, i32 2, i32 10, i32 18, i32 3, i32 11, i32 19, i32 4, i32 12, i32 20, i32 5, i32 13, i32 21, i32 6, i32 14, i32 22, i32 7, i32 15, i32 23> 488; SCALABLE-NEXT: store <24 x i32> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 4 489; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 490; SCALABLE-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 491; SCALABLE-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]] 492; SCALABLE: middle.block: 493; SCALABLE-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 494; SCALABLE: scalar.ph: 495; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 496; SCALABLE-NEXT: br label [[LOOP:%.*]] 497; SCALABLE: loop: 498; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 499; SCALABLE-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 500; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 501; SCALABLE-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 502; SCALABLE-NEXT: [[Y0:%.*]] = add i32 [[X0]], 1 503; SCALABLE-NEXT: store i32 [[Y0]], ptr [[Q0]], align 4 504; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 505; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 506; SCALABLE-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 507; SCALABLE-NEXT: [[Y1:%.*]] = add i32 [[X1]], 2 508; SCALABLE-NEXT: store i32 [[Y1]], ptr [[Q1]], align 4 509; SCALABLE-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 510; SCALABLE-NEXT: [[Q2:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET2]] 511; SCALABLE-NEXT: [[X2:%.*]] = load i32, ptr [[Q2]], align 4 512; SCALABLE-NEXT: [[Y2:%.*]] = add i32 [[X2]], 3 513; SCALABLE-NEXT: store i32 [[Y2]], ptr [[Q2]], align 4 514; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 515; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 516; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP7:![0-9]+]] 517; SCALABLE: exit: 518; SCALABLE-NEXT: ret void 519; 520entry: 521 br label %loop 522loop: 523 %i = phi i64 [0, %entry], [%nexti, %loop] 524 525 %offset0 = mul i64 %i, 3 526 %q0 = getelementptr i32, ptr %p, i64 %offset0 527 %x0 = load i32, ptr %q0 528 %y0 = add i32 %x0, 1 529 store i32 %y0, ptr %q0 530 531 %offset1 = add i64 %offset0, 1 532 %q1 = getelementptr i32, ptr %p, i64 %offset1 533 %x1 = load i32, ptr %q1 534 %y1 = add i32 %x1, 2 535 store i32 %y1, ptr %q1 536 537 %offset2 = add i64 %offset1, 1 538 %q2 = getelementptr i32, ptr %p, i64 %offset2 539 %x2 = load i32, ptr %q2 540 %y2 = add i32 %x2, 3 541 store i32 %y2, ptr %q2 542 543 %nexti = add i64 %i, 1 544 %done = icmp eq i64 %nexti, 1024 545 br i1 %done, label %exit, label %loop 546exit: 547 ret void 548} 549 550define void @load_store_factor3_i64(ptr %p) { 551; CHECK-LABEL: @load_store_factor3_i64( 552; CHECK-NEXT: entry: 553; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 554; CHECK: vector.ph: 555; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 556; CHECK: vector.body: 557; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 558; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 559; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 560; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 561; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <12 x i64>, ptr [[TMP2]], align 8 562; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9> 563; CHECK-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10> 564; CHECK-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11> 565; CHECK-NEXT: [[TMP4:%.*]] = add <4 x i64> [[STRIDED_VEC]], splat (i64 1) 566; CHECK-NEXT: [[TMP6:%.*]] = add <4 x i64> [[STRIDED_VEC1]], splat (i64 2) 567; CHECK-NEXT: [[TMP9:%.*]] = add <4 x i64> [[STRIDED_VEC2]], splat (i64 3) 568; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP4]], <4 x i64> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 569; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison> 570; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <8 x i64> [[TMP11]], <8 x i64> [[TMP12]], <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11> 571; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <12 x i64> [[TMP13]], <12 x i64> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 572; CHECK-NEXT: store <12 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 573; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 574; CHECK-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 575; CHECK-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] 576; CHECK: middle.block: 577; CHECK-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 578; CHECK: scalar.ph: 579; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 580; CHECK-NEXT: br label [[LOOP:%.*]] 581; CHECK: loop: 582; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 583; CHECK-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 584; CHECK-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 585; CHECK-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 586; CHECK-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 587; CHECK-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 588; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 589; CHECK-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 590; CHECK-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 591; CHECK-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 592; CHECK-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 593; CHECK-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 594; CHECK-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 595; CHECK-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 596; CHECK-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 597; CHECK-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 598; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 599; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 600; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] 601; CHECK: exit: 602; CHECK-NEXT: ret void 603; 604; FIXED-LABEL: @load_store_factor3_i64( 605; FIXED-NEXT: entry: 606; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 607; FIXED: vector.ph: 608; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 609; FIXED: vector.body: 610; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 611; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 612; FIXED-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 613; FIXED-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 614; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <12 x i64>, ptr [[TMP2]], align 8 615; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9> 616; FIXED-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10> 617; FIXED-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11> 618; FIXED-NEXT: [[TMP4:%.*]] = add <4 x i64> [[STRIDED_VEC]], splat (i64 1) 619; FIXED-NEXT: [[TMP6:%.*]] = add <4 x i64> [[STRIDED_VEC1]], splat (i64 2) 620; FIXED-NEXT: [[TMP9:%.*]] = add <4 x i64> [[STRIDED_VEC2]], splat (i64 3) 621; FIXED-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP4]], <4 x i64> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 622; FIXED-NEXT: [[TMP12:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison> 623; FIXED-NEXT: [[TMP13:%.*]] = shufflevector <8 x i64> [[TMP11]], <8 x i64> [[TMP12]], <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11> 624; FIXED-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <12 x i64> [[TMP13]], <12 x i64> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 625; FIXED-NEXT: store <12 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 626; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 627; FIXED-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 628; FIXED-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] 629; FIXED: middle.block: 630; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 631; FIXED: scalar.ph: 632; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 633; FIXED-NEXT: br label [[LOOP:%.*]] 634; FIXED: loop: 635; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 636; FIXED-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 637; FIXED-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 638; FIXED-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 639; FIXED-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 640; FIXED-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 641; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 642; FIXED-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 643; FIXED-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 644; FIXED-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 645; FIXED-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 646; FIXED-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 647; FIXED-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 648; FIXED-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 649; FIXED-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 650; FIXED-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 651; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 652; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 653; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] 654; FIXED: exit: 655; FIXED-NEXT: ret void 656; 657; SCALABLE-LABEL: @load_store_factor3_i64( 658; SCALABLE-NEXT: entry: 659; SCALABLE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 660; SCALABLE: vector.ph: 661; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 662; SCALABLE: vector.body: 663; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 664; SCALABLE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 665; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 3 666; SCALABLE-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 667; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <12 x i64>, ptr [[TMP2]], align 8 668; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9> 669; SCALABLE-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 1, i32 4, i32 7, i32 10> 670; SCALABLE-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 2, i32 5, i32 8, i32 11> 671; SCALABLE-NEXT: [[TMP4:%.*]] = add <4 x i64> [[STRIDED_VEC]], splat (i64 1) 672; SCALABLE-NEXT: [[TMP6:%.*]] = add <4 x i64> [[STRIDED_VEC1]], splat (i64 2) 673; SCALABLE-NEXT: [[TMP9:%.*]] = add <4 x i64> [[STRIDED_VEC2]], splat (i64 3) 674; SCALABLE-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP4]], <4 x i64> [[TMP6]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 675; SCALABLE-NEXT: [[TMP12:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison> 676; SCALABLE-NEXT: [[TMP13:%.*]] = shufflevector <8 x i64> [[TMP11]], <8 x i64> [[TMP12]], <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11> 677; SCALABLE-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <12 x i64> [[TMP13]], <12 x i64> poison, <12 x i32> <i32 0, i32 4, i32 8, i32 1, i32 5, i32 9, i32 2, i32 6, i32 10, i32 3, i32 7, i32 11> 678; SCALABLE-NEXT: store <12 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 679; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 680; SCALABLE-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 681; SCALABLE-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]] 682; SCALABLE: middle.block: 683; SCALABLE-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 684; SCALABLE: scalar.ph: 685; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 686; SCALABLE-NEXT: br label [[LOOP:%.*]] 687; SCALABLE: loop: 688; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 689; SCALABLE-NEXT: [[OFFSET0:%.*]] = mul i64 [[I]], 3 690; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 691; SCALABLE-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 692; SCALABLE-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 693; SCALABLE-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 694; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 695; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 696; SCALABLE-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 697; SCALABLE-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 698; SCALABLE-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 699; SCALABLE-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 700; SCALABLE-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 701; SCALABLE-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 702; SCALABLE-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 703; SCALABLE-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 704; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 705; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 706; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]] 707; SCALABLE: exit: 708; SCALABLE-NEXT: ret void 709; 710entry: 711 br label %loop 712loop: 713 %i = phi i64 [0, %entry], [%nexti, %loop] 714 715 %offset0 = mul i64 %i, 3 716 %q0 = getelementptr i64, ptr %p, i64 %offset0 717 %x0 = load i64, ptr %q0 718 %y0 = add i64 %x0, 1 719 store i64 %y0, ptr %q0 720 721 %offset1 = add i64 %offset0, 1 722 %q1 = getelementptr i64, ptr %p, i64 %offset1 723 %x1 = load i64, ptr %q1 724 %y1 = add i64 %x1, 2 725 store i64 %y1, ptr %q1 726 727 %offset2 = add i64 %offset1, 1 728 %q2 = getelementptr i64, ptr %p, i64 %offset2 729 %x2 = load i64, ptr %q2 730 %y2 = add i64 %x2, 3 731 store i64 %y2, ptr %q2 732 733 %nexti = add i64 %i, 1 734 %done = icmp eq i64 %nexti, 1024 735 br i1 %done, label %exit, label %loop 736exit: 737 ret void 738} 739 740define void @load_store_factor8(ptr %p) { 741; CHECK-LABEL: @load_store_factor8( 742; CHECK-NEXT: entry: 743; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 744; CHECK: vector.ph: 745; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 746; CHECK: vector.body: 747; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 748; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 749; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[TMP0]], 3 750; CHECK-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 751; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <16 x i64>, ptr [[TMP2]], align 8 752; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 0, i32 8> 753; CHECK-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 1, i32 9> 754; CHECK-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 2, i32 10> 755; CHECK-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 3, i32 11> 756; CHECK-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 4, i32 12> 757; CHECK-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 5, i32 13> 758; CHECK-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 6, i32 14> 759; CHECK-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 7, i32 15> 760; CHECK-NEXT: [[TMP4:%.*]] = add <2 x i64> [[STRIDED_VEC]], splat (i64 1) 761; CHECK-NEXT: [[TMP6:%.*]] = add <2 x i64> [[STRIDED_VEC1]], splat (i64 2) 762; CHECK-NEXT: [[TMP8:%.*]] = add <2 x i64> [[STRIDED_VEC2]], splat (i64 3) 763; CHECK-NEXT: [[TMP10:%.*]] = add <2 x i64> [[STRIDED_VEC3]], splat (i64 4) 764; CHECK-NEXT: [[TMP12:%.*]] = add <2 x i64> [[STRIDED_VEC4]], splat (i64 5) 765; CHECK-NEXT: [[TMP14:%.*]] = add <2 x i64> [[STRIDED_VEC5]], splat (i64 6) 766; CHECK-NEXT: [[TMP16:%.*]] = add <2 x i64> [[STRIDED_VEC6]], splat (i64 7) 767; CHECK-NEXT: [[TMP19:%.*]] = add <2 x i64> [[STRIDED_VEC7]], splat (i64 8) 768; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <2 x i64> [[TMP4]], <2 x i64> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 769; CHECK-NEXT: [[TMP22:%.*]] = shufflevector <2 x i64> [[TMP8]], <2 x i64> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 770; CHECK-NEXT: [[TMP23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> [[TMP14]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 771; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <2 x i64> [[TMP16]], <2 x i64> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 772; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <4 x i64> [[TMP21]], <4 x i64> [[TMP22]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 773; CHECK-NEXT: [[TMP26:%.*]] = shufflevector <4 x i64> [[TMP23]], <4 x i64> [[TMP24]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 774; CHECK-NEXT: [[TMP27:%.*]] = shufflevector <8 x i64> [[TMP25]], <8 x i64> [[TMP26]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 775; CHECK-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i64> [[TMP27]], <16 x i64> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 776; CHECK-NEXT: store <16 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 777; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 778; CHECK-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 779; CHECK-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] 780; CHECK: middle.block: 781; CHECK-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 782; CHECK: scalar.ph: 783; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 784; CHECK-NEXT: br label [[LOOP:%.*]] 785; CHECK: loop: 786; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 787; CHECK-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 3 788; CHECK-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 789; CHECK-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 790; CHECK-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 791; CHECK-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 792; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 793; CHECK-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 794; CHECK-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 795; CHECK-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 796; CHECK-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 797; CHECK-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 798; CHECK-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 799; CHECK-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 800; CHECK-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 801; CHECK-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 802; CHECK-NEXT: [[OFFSET3:%.*]] = add i64 [[OFFSET2]], 1 803; CHECK-NEXT: [[Q3:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET3]] 804; CHECK-NEXT: [[X3:%.*]] = load i64, ptr [[Q3]], align 8 805; CHECK-NEXT: [[Y3:%.*]] = add i64 [[X3]], 4 806; CHECK-NEXT: store i64 [[Y3]], ptr [[Q3]], align 8 807; CHECK-NEXT: [[OFFSET4:%.*]] = add i64 [[OFFSET3]], 1 808; CHECK-NEXT: [[Q4:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET4]] 809; CHECK-NEXT: [[X4:%.*]] = load i64, ptr [[Q4]], align 8 810; CHECK-NEXT: [[Y4:%.*]] = add i64 [[X4]], 5 811; CHECK-NEXT: store i64 [[Y4]], ptr [[Q4]], align 8 812; CHECK-NEXT: [[OFFSET5:%.*]] = add i64 [[OFFSET4]], 1 813; CHECK-NEXT: [[Q5:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET5]] 814; CHECK-NEXT: [[X5:%.*]] = load i64, ptr [[Q5]], align 8 815; CHECK-NEXT: [[Y5:%.*]] = add i64 [[X5]], 6 816; CHECK-NEXT: store i64 [[Y5]], ptr [[Q5]], align 8 817; CHECK-NEXT: [[OFFSET6:%.*]] = add i64 [[OFFSET5]], 1 818; CHECK-NEXT: [[Q6:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET6]] 819; CHECK-NEXT: [[X6:%.*]] = load i64, ptr [[Q6]], align 8 820; CHECK-NEXT: [[Y6:%.*]] = add i64 [[X6]], 7 821; CHECK-NEXT: store i64 [[Y6]], ptr [[Q6]], align 8 822; CHECK-NEXT: [[OFFSET7:%.*]] = add i64 [[OFFSET6]], 1 823; CHECK-NEXT: [[Q7:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET7]] 824; CHECK-NEXT: [[X7:%.*]] = load i64, ptr [[Q7]], align 8 825; CHECK-NEXT: [[Y7:%.*]] = add i64 [[X7]], 8 826; CHECK-NEXT: store i64 [[Y7]], ptr [[Q7]], align 8 827; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 828; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 829; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]] 830; CHECK: exit: 831; CHECK-NEXT: ret void 832; 833; FIXED-LABEL: @load_store_factor8( 834; FIXED-NEXT: entry: 835; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 836; FIXED: vector.ph: 837; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 838; FIXED: vector.body: 839; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 840; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 841; FIXED-NEXT: [[TMP1:%.*]] = shl i64 [[TMP0]], 3 842; FIXED-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 843; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <16 x i64>, ptr [[TMP2]], align 8 844; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 0, i32 8> 845; FIXED-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 1, i32 9> 846; FIXED-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 2, i32 10> 847; FIXED-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 3, i32 11> 848; FIXED-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 4, i32 12> 849; FIXED-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 5, i32 13> 850; FIXED-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 6, i32 14> 851; FIXED-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 7, i32 15> 852; FIXED-NEXT: [[TMP4:%.*]] = add <2 x i64> [[STRIDED_VEC]], splat (i64 1) 853; FIXED-NEXT: [[TMP6:%.*]] = add <2 x i64> [[STRIDED_VEC1]], splat (i64 2) 854; FIXED-NEXT: [[TMP8:%.*]] = add <2 x i64> [[STRIDED_VEC2]], splat (i64 3) 855; FIXED-NEXT: [[TMP10:%.*]] = add <2 x i64> [[STRIDED_VEC3]], splat (i64 4) 856; FIXED-NEXT: [[TMP12:%.*]] = add <2 x i64> [[STRIDED_VEC4]], splat (i64 5) 857; FIXED-NEXT: [[TMP14:%.*]] = add <2 x i64> [[STRIDED_VEC5]], splat (i64 6) 858; FIXED-NEXT: [[TMP16:%.*]] = add <2 x i64> [[STRIDED_VEC6]], splat (i64 7) 859; FIXED-NEXT: [[TMP19:%.*]] = add <2 x i64> [[STRIDED_VEC7]], splat (i64 8) 860; FIXED-NEXT: [[TMP21:%.*]] = shufflevector <2 x i64> [[TMP4]], <2 x i64> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 861; FIXED-NEXT: [[TMP22:%.*]] = shufflevector <2 x i64> [[TMP8]], <2 x i64> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 862; FIXED-NEXT: [[TMP23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> [[TMP14]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 863; FIXED-NEXT: [[TMP24:%.*]] = shufflevector <2 x i64> [[TMP16]], <2 x i64> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 864; FIXED-NEXT: [[TMP25:%.*]] = shufflevector <4 x i64> [[TMP21]], <4 x i64> [[TMP22]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 865; FIXED-NEXT: [[TMP26:%.*]] = shufflevector <4 x i64> [[TMP23]], <4 x i64> [[TMP24]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 866; FIXED-NEXT: [[TMP27:%.*]] = shufflevector <8 x i64> [[TMP25]], <8 x i64> [[TMP26]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 867; FIXED-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i64> [[TMP27]], <16 x i64> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 868; FIXED-NEXT: store <16 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 869; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 870; FIXED-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 871; FIXED-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] 872; FIXED: middle.block: 873; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 874; FIXED: scalar.ph: 875; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 876; FIXED-NEXT: br label [[LOOP:%.*]] 877; FIXED: loop: 878; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 879; FIXED-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 3 880; FIXED-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 881; FIXED-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 882; FIXED-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 883; FIXED-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 884; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 885; FIXED-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 886; FIXED-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 887; FIXED-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 888; FIXED-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 889; FIXED-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 890; FIXED-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 891; FIXED-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 892; FIXED-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 893; FIXED-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 894; FIXED-NEXT: [[OFFSET3:%.*]] = add i64 [[OFFSET2]], 1 895; FIXED-NEXT: [[Q3:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET3]] 896; FIXED-NEXT: [[X3:%.*]] = load i64, ptr [[Q3]], align 8 897; FIXED-NEXT: [[Y3:%.*]] = add i64 [[X3]], 4 898; FIXED-NEXT: store i64 [[Y3]], ptr [[Q3]], align 8 899; FIXED-NEXT: [[OFFSET4:%.*]] = add i64 [[OFFSET3]], 1 900; FIXED-NEXT: [[Q4:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET4]] 901; FIXED-NEXT: [[X4:%.*]] = load i64, ptr [[Q4]], align 8 902; FIXED-NEXT: [[Y4:%.*]] = add i64 [[X4]], 5 903; FIXED-NEXT: store i64 [[Y4]], ptr [[Q4]], align 8 904; FIXED-NEXT: [[OFFSET5:%.*]] = add i64 [[OFFSET4]], 1 905; FIXED-NEXT: [[Q5:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET5]] 906; FIXED-NEXT: [[X5:%.*]] = load i64, ptr [[Q5]], align 8 907; FIXED-NEXT: [[Y5:%.*]] = add i64 [[X5]], 6 908; FIXED-NEXT: store i64 [[Y5]], ptr [[Q5]], align 8 909; FIXED-NEXT: [[OFFSET6:%.*]] = add i64 [[OFFSET5]], 1 910; FIXED-NEXT: [[Q6:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET6]] 911; FIXED-NEXT: [[X6:%.*]] = load i64, ptr [[Q6]], align 8 912; FIXED-NEXT: [[Y6:%.*]] = add i64 [[X6]], 7 913; FIXED-NEXT: store i64 [[Y6]], ptr [[Q6]], align 8 914; FIXED-NEXT: [[OFFSET7:%.*]] = add i64 [[OFFSET6]], 1 915; FIXED-NEXT: [[Q7:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET7]] 916; FIXED-NEXT: [[X7:%.*]] = load i64, ptr [[Q7]], align 8 917; FIXED-NEXT: [[Y7:%.*]] = add i64 [[X7]], 8 918; FIXED-NEXT: store i64 [[Y7]], ptr [[Q7]], align 8 919; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 920; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 921; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]] 922; FIXED: exit: 923; FIXED-NEXT: ret void 924; 925; SCALABLE-LABEL: @load_store_factor8( 926; SCALABLE-NEXT: entry: 927; SCALABLE-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 928; SCALABLE: vector.ph: 929; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 930; SCALABLE: vector.body: 931; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 932; SCALABLE-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 933; SCALABLE-NEXT: [[TMP1:%.*]] = shl i64 [[TMP0]], 3 934; SCALABLE-NEXT: [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP1]] 935; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <16 x i64>, ptr [[TMP2]], align 8 936; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 0, i32 8> 937; SCALABLE-NEXT: [[STRIDED_VEC1:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 1, i32 9> 938; SCALABLE-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 2, i32 10> 939; SCALABLE-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 3, i32 11> 940; SCALABLE-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 4, i32 12> 941; SCALABLE-NEXT: [[STRIDED_VEC5:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 5, i32 13> 942; SCALABLE-NEXT: [[STRIDED_VEC6:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 6, i32 14> 943; SCALABLE-NEXT: [[STRIDED_VEC7:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <2 x i32> <i32 7, i32 15> 944; SCALABLE-NEXT: [[TMP4:%.*]] = add <2 x i64> [[STRIDED_VEC]], splat (i64 1) 945; SCALABLE-NEXT: [[TMP6:%.*]] = add <2 x i64> [[STRIDED_VEC1]], splat (i64 2) 946; SCALABLE-NEXT: [[TMP8:%.*]] = add <2 x i64> [[STRIDED_VEC2]], splat (i64 3) 947; SCALABLE-NEXT: [[TMP10:%.*]] = add <2 x i64> [[STRIDED_VEC3]], splat (i64 4) 948; SCALABLE-NEXT: [[TMP12:%.*]] = add <2 x i64> [[STRIDED_VEC4]], splat (i64 5) 949; SCALABLE-NEXT: [[TMP14:%.*]] = add <2 x i64> [[STRIDED_VEC5]], splat (i64 6) 950; SCALABLE-NEXT: [[TMP16:%.*]] = add <2 x i64> [[STRIDED_VEC6]], splat (i64 7) 951; SCALABLE-NEXT: [[TMP19:%.*]] = add <2 x i64> [[STRIDED_VEC7]], splat (i64 8) 952; SCALABLE-NEXT: [[TMP21:%.*]] = shufflevector <2 x i64> [[TMP4]], <2 x i64> [[TMP6]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 953; SCALABLE-NEXT: [[TMP22:%.*]] = shufflevector <2 x i64> [[TMP8]], <2 x i64> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 954; SCALABLE-NEXT: [[TMP23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> [[TMP14]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 955; SCALABLE-NEXT: [[TMP24:%.*]] = shufflevector <2 x i64> [[TMP16]], <2 x i64> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> 956; SCALABLE-NEXT: [[TMP25:%.*]] = shufflevector <4 x i64> [[TMP21]], <4 x i64> [[TMP22]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 957; SCALABLE-NEXT: [[TMP26:%.*]] = shufflevector <4 x i64> [[TMP23]], <4 x i64> [[TMP24]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 958; SCALABLE-NEXT: [[TMP27:%.*]] = shufflevector <8 x i64> [[TMP25]], <8 x i64> [[TMP26]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> 959; SCALABLE-NEXT: [[INTERLEAVED_VEC:%.*]] = shufflevector <16 x i64> [[TMP27]], <16 x i64> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 960; SCALABLE-NEXT: store <16 x i64> [[INTERLEAVED_VEC]], ptr [[TMP2]], align 8 961; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2 962; SCALABLE-NEXT: [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 963; SCALABLE-NEXT: br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]] 964; SCALABLE: middle.block: 965; SCALABLE-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 966; SCALABLE: scalar.ph: 967; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 968; SCALABLE-NEXT: br label [[LOOP:%.*]] 969; SCALABLE: loop: 970; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 971; SCALABLE-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 3 972; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 973; SCALABLE-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 974; SCALABLE-NEXT: [[Y0:%.*]] = add i64 [[X0]], 1 975; SCALABLE-NEXT: store i64 [[Y0]], ptr [[Q0]], align 8 976; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 977; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 978; SCALABLE-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 979; SCALABLE-NEXT: [[Y1:%.*]] = add i64 [[X1]], 2 980; SCALABLE-NEXT: store i64 [[Y1]], ptr [[Q1]], align 8 981; SCALABLE-NEXT: [[OFFSET2:%.*]] = add i64 [[OFFSET1]], 1 982; SCALABLE-NEXT: [[Q2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET2]] 983; SCALABLE-NEXT: [[X2:%.*]] = load i64, ptr [[Q2]], align 8 984; SCALABLE-NEXT: [[Y2:%.*]] = add i64 [[X2]], 3 985; SCALABLE-NEXT: store i64 [[Y2]], ptr [[Q2]], align 8 986; SCALABLE-NEXT: [[OFFSET3:%.*]] = add i64 [[OFFSET2]], 1 987; SCALABLE-NEXT: [[Q3:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET3]] 988; SCALABLE-NEXT: [[X3:%.*]] = load i64, ptr [[Q3]], align 8 989; SCALABLE-NEXT: [[Y3:%.*]] = add i64 [[X3]], 4 990; SCALABLE-NEXT: store i64 [[Y3]], ptr [[Q3]], align 8 991; SCALABLE-NEXT: [[OFFSET4:%.*]] = add i64 [[OFFSET3]], 1 992; SCALABLE-NEXT: [[Q4:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET4]] 993; SCALABLE-NEXT: [[X4:%.*]] = load i64, ptr [[Q4]], align 8 994; SCALABLE-NEXT: [[Y4:%.*]] = add i64 [[X4]], 5 995; SCALABLE-NEXT: store i64 [[Y4]], ptr [[Q4]], align 8 996; SCALABLE-NEXT: [[OFFSET5:%.*]] = add i64 [[OFFSET4]], 1 997; SCALABLE-NEXT: [[Q5:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET5]] 998; SCALABLE-NEXT: [[X5:%.*]] = load i64, ptr [[Q5]], align 8 999; SCALABLE-NEXT: [[Y5:%.*]] = add i64 [[X5]], 6 1000; SCALABLE-NEXT: store i64 [[Y5]], ptr [[Q5]], align 8 1001; SCALABLE-NEXT: [[OFFSET6:%.*]] = add i64 [[OFFSET5]], 1 1002; SCALABLE-NEXT: [[Q6:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET6]] 1003; SCALABLE-NEXT: [[X6:%.*]] = load i64, ptr [[Q6]], align 8 1004; SCALABLE-NEXT: [[Y6:%.*]] = add i64 [[X6]], 7 1005; SCALABLE-NEXT: store i64 [[Y6]], ptr [[Q6]], align 8 1006; SCALABLE-NEXT: [[OFFSET7:%.*]] = add i64 [[OFFSET6]], 1 1007; SCALABLE-NEXT: [[Q7:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET7]] 1008; SCALABLE-NEXT: [[X7:%.*]] = load i64, ptr [[Q7]], align 8 1009; SCALABLE-NEXT: [[Y7:%.*]] = add i64 [[X7]], 8 1010; SCALABLE-NEXT: store i64 [[Y7]], ptr [[Q7]], align 8 1011; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 1012; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1013; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]] 1014; SCALABLE: exit: 1015; SCALABLE-NEXT: ret void 1016; 1017entry: 1018 br label %loop 1019loop: 1020 %i = phi i64 [0, %entry], [%nexti, %loop] 1021 1022 %offset0 = shl i64 %i, 3 1023 %q0 = getelementptr i64, ptr %p, i64 %offset0 1024 %x0 = load i64, ptr %q0 1025 %y0 = add i64 %x0, 1 1026 store i64 %y0, ptr %q0 1027 1028 %offset1 = add i64 %offset0, 1 1029 %q1 = getelementptr i64, ptr %p, i64 %offset1 1030 %x1 = load i64, ptr %q1 1031 %y1 = add i64 %x1, 2 1032 store i64 %y1, ptr %q1 1033 1034 %offset2 = add i64 %offset1, 1 1035 %q2 = getelementptr i64, ptr %p, i64 %offset2 1036 %x2 = load i64, ptr %q2 1037 %y2 = add i64 %x2, 3 1038 store i64 %y2, ptr %q2 1039 1040 %offset3 = add i64 %offset2, 1 1041 %q3 = getelementptr i64, ptr %p, i64 %offset3 1042 %x3 = load i64, ptr %q3 1043 %y3 = add i64 %x3, 4 1044 store i64 %y3, ptr %q3 1045 1046 %offset4 = add i64 %offset3, 1 1047 %q4 = getelementptr i64, ptr %p, i64 %offset4 1048 %x4 = load i64, ptr %q4 1049 %y4 = add i64 %x4, 5 1050 store i64 %y4, ptr %q4 1051 1052 %offset5 = add i64 %offset4, 1 1053 %q5 = getelementptr i64, ptr %p, i64 %offset5 1054 %x5 = load i64, ptr %q5 1055 %y5 = add i64 %x5, 6 1056 store i64 %y5, ptr %q5 1057 1058 %offset6 = add i64 %offset5, 1 1059 %q6 = getelementptr i64, ptr %p, i64 %offset6 1060 %x6 = load i64, ptr %q6 1061 %y6 = add i64 %x6, 7 1062 store i64 %y6, ptr %q6 1063 1064 %offset7 = add i64 %offset6, 1 1065 %q7 = getelementptr i64, ptr %p, i64 %offset7 1066 %x7 = load i64, ptr %q7 1067 %y7 = add i64 %x7, 8 1068 store i64 %y7, ptr %q7 1069 1070 %nexti = add i64 %i, 1 1071 %done = icmp eq i64 %nexti, 1024 1072 br i1 %done, label %exit, label %loop 1073exit: 1074 ret void 1075} 1076 1077define void @combine_load_factor2_i32(ptr noalias %p, ptr noalias %q) { 1078; CHECK-LABEL: @combine_load_factor2_i32( 1079; CHECK-NEXT: entry: 1080; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 1081; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 4 1082; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 1083; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1084; CHECK: vector.ph: 1085; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 1086; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 4 1087; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 1088; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 1089; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 1090; CHECK-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 1091; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 1092; CHECK: vector.body: 1093; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1094; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 1095; CHECK-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 1096; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP7]] 1097; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP8]], align 4 1098; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]]) 1099; CHECK-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0 1100; CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1 1101; CHECK-NEXT: [[TMP12:%.*]] = add <vscale x 4 x i32> [[TMP10]], [[TMP11]] 1102; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[TMP6]] 1103; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0 1104; CHECK-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP14]], align 4 1105; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 1106; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 1107; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] 1108; CHECK: middle.block: 1109; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 1110; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 1111; CHECK: scalar.ph: 1112; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1113; CHECK-NEXT: br label [[LOOP:%.*]] 1114; CHECK: loop: 1115; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1116; CHECK-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1117; CHECK-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 1118; CHECK-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 1119; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1120; CHECK-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 1121; CHECK-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 1122; CHECK-NEXT: [[RES:%.*]] = add i32 [[X0]], [[X1]] 1123; CHECK-NEXT: [[DST:%.*]] = getelementptr i32, ptr [[Q]], i64 [[I]] 1124; CHECK-NEXT: store i32 [[RES]], ptr [[DST]], align 4 1125; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 1126; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1127; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]] 1128; CHECK: exit: 1129; CHECK-NEXT: ret void 1130; 1131; FIXED-LABEL: @combine_load_factor2_i32( 1132; FIXED-NEXT: entry: 1133; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1134; FIXED: vector.ph: 1135; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 1136; FIXED: vector.body: 1137; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1138; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 1139; FIXED-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 8 1140; FIXED-NEXT: [[TMP2:%.*]] = shl i64 [[TMP0]], 1 1141; FIXED-NEXT: [[TMP3:%.*]] = shl i64 [[TMP1]], 1 1142; FIXED-NEXT: [[TMP4:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP2]] 1143; FIXED-NEXT: [[TMP5:%.*]] = getelementptr i32, ptr [[P]], i64 [[TMP3]] 1144; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <16 x i32>, ptr [[TMP4]], align 4 1145; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 1146; FIXED-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <16 x i32> [[WIDE_VEC]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 1147; FIXED-NEXT: [[WIDE_VEC1:%.*]] = load <16 x i32>, ptr [[TMP5]], align 4 1148; FIXED-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <16 x i32> [[WIDE_VEC1]], <16 x i32> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14> 1149; FIXED-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <16 x i32> [[WIDE_VEC1]], <16 x i32> poison, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15> 1150; FIXED-NEXT: [[TMP8:%.*]] = add <8 x i32> [[STRIDED_VEC]], [[STRIDED_VEC3]] 1151; FIXED-NEXT: [[TMP9:%.*]] = add <8 x i32> [[STRIDED_VEC2]], [[STRIDED_VEC4]] 1152; FIXED-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[TMP0]] 1153; FIXED-NEXT: [[TMP12:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0 1154; FIXED-NEXT: [[TMP13:%.*]] = getelementptr i32, ptr [[TMP10]], i32 8 1155; FIXED-NEXT: store <8 x i32> [[TMP8]], ptr [[TMP12]], align 4 1156; FIXED-NEXT: store <8 x i32> [[TMP9]], ptr [[TMP13]], align 4 1157; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16 1158; FIXED-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 1159; FIXED-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] 1160; FIXED: middle.block: 1161; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 1162; FIXED: scalar.ph: 1163; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1164; FIXED-NEXT: br label [[LOOP:%.*]] 1165; FIXED: loop: 1166; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1167; FIXED-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1168; FIXED-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 1169; FIXED-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 1170; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1171; FIXED-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 1172; FIXED-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 1173; FIXED-NEXT: [[RES:%.*]] = add i32 [[X0]], [[X1]] 1174; FIXED-NEXT: [[DST:%.*]] = getelementptr i32, ptr [[Q]], i64 [[I]] 1175; FIXED-NEXT: store i32 [[RES]], ptr [[DST]], align 4 1176; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 1177; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1178; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]] 1179; FIXED: exit: 1180; FIXED-NEXT: ret void 1181; 1182; SCALABLE-LABEL: @combine_load_factor2_i32( 1183; SCALABLE-NEXT: entry: 1184; SCALABLE-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 1185; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 4 1186; SCALABLE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 1187; SCALABLE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1188; SCALABLE: vector.ph: 1189; SCALABLE-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 1190; SCALABLE-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 4 1191; SCALABLE-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 1192; SCALABLE-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 1193; SCALABLE-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 1194; SCALABLE-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 4 1195; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 1196; SCALABLE: vector.body: 1197; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1198; SCALABLE-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 1199; SCALABLE-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 1200; SCALABLE-NEXT: [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[TMP7]] 1201; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 8 x i32>, ptr [[TMP8]], align 4 1202; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_VEC]]) 1203; SCALABLE-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0 1204; SCALABLE-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1 1205; SCALABLE-NEXT: [[TMP12:%.*]] = add <vscale x 4 x i32> [[TMP10]], [[TMP11]] 1206; SCALABLE-NEXT: [[TMP13:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[TMP6]] 1207; SCALABLE-NEXT: [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0 1208; SCALABLE-NEXT: store <vscale x 4 x i32> [[TMP12]], ptr [[TMP14]], align 4 1209; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 1210; SCALABLE-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 1211; SCALABLE-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] 1212; SCALABLE: middle.block: 1213; SCALABLE-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 1214; SCALABLE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 1215; SCALABLE: scalar.ph: 1216; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1217; SCALABLE-NEXT: br label [[LOOP:%.*]] 1218; SCALABLE: loop: 1219; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1220; SCALABLE-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1221; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET0]] 1222; SCALABLE-NEXT: [[X0:%.*]] = load i32, ptr [[Q0]], align 4 1223; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1224; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i32, ptr [[P]], i64 [[OFFSET1]] 1225; SCALABLE-NEXT: [[X1:%.*]] = load i32, ptr [[Q1]], align 4 1226; SCALABLE-NEXT: [[RES:%.*]] = add i32 [[X0]], [[X1]] 1227; SCALABLE-NEXT: [[DST:%.*]] = getelementptr i32, ptr [[Q]], i64 [[I]] 1228; SCALABLE-NEXT: store i32 [[RES]], ptr [[DST]], align 4 1229; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 1230; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1231; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]] 1232; SCALABLE: exit: 1233; SCALABLE-NEXT: ret void 1234; 1235entry: 1236 br label %loop 1237loop: 1238 %i = phi i64 [0, %entry], [%nexti, %loop] 1239 1240 %offset0 = shl i64 %i, 1 1241 %q0 = getelementptr i32, ptr %p, i64 %offset0 1242 %x0 = load i32, ptr %q0 1243 1244 %offset1 = add i64 %offset0, 1 1245 %q1 = getelementptr i32, ptr %p, i64 %offset1 1246 %x1 = load i32, ptr %q1 1247 1248 %res = add i32 %x0, %x1 1249 1250 %dst = getelementptr i32, ptr %q, i64 %i 1251 store i32 %res, ptr %dst 1252 1253 %nexti = add i64 %i, 1 1254 %done = icmp eq i64 %nexti, 1024 1255 br i1 %done, label %exit, label %loop 1256exit: 1257 ret void 1258} 1259 1260define void @combine_load_factor2_i64(ptr noalias %p, ptr noalias %q) { 1261; CHECK-LABEL: @combine_load_factor2_i64( 1262; CHECK-NEXT: entry: 1263; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 1264; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2 1265; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 1266; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1267; CHECK: vector.ph: 1268; CHECK-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 1269; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 2 1270; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 1271; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 1272; CHECK-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 1273; CHECK-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 2 1274; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 1275; CHECK: vector.body: 1276; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1277; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 1278; CHECK-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 1279; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP7]] 1280; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 4 x i64>, ptr [[TMP8]], align 8 1281; CHECK-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.vector.deinterleave2.nxv4i64(<vscale x 4 x i64> [[WIDE_VEC]]) 1282; CHECK-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 0 1283; CHECK-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1 1284; CHECK-NEXT: [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP10]], [[TMP11]] 1285; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[TMP6]] 1286; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i64, ptr [[TMP13]], i32 0 1287; CHECK-NEXT: store <vscale x 2 x i64> [[TMP12]], ptr [[TMP14]], align 8 1288; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 1289; CHECK-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 1290; CHECK-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] 1291; CHECK: middle.block: 1292; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 1293; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 1294; CHECK: scalar.ph: 1295; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1296; CHECK-NEXT: br label [[LOOP:%.*]] 1297; CHECK: loop: 1298; CHECK-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1299; CHECK-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1300; CHECK-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 1301; CHECK-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 1302; CHECK-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1303; CHECK-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 1304; CHECK-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 1305; CHECK-NEXT: [[RES:%.*]] = add i64 [[X0]], [[X1]] 1306; CHECK-NEXT: [[DST:%.*]] = getelementptr i64, ptr [[Q]], i64 [[I]] 1307; CHECK-NEXT: store i64 [[RES]], ptr [[DST]], align 8 1308; CHECK-NEXT: [[NEXTI]] = add i64 [[I]], 1 1309; CHECK-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1310; CHECK-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] 1311; CHECK: exit: 1312; CHECK-NEXT: ret void 1313; 1314; FIXED-LABEL: @combine_load_factor2_i64( 1315; FIXED-NEXT: entry: 1316; FIXED-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1317; FIXED: vector.ph: 1318; FIXED-NEXT: br label [[VECTOR_BODY:%.*]] 1319; FIXED: vector.body: 1320; FIXED-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1321; FIXED-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0 1322; FIXED-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 4 1323; FIXED-NEXT: [[TMP2:%.*]] = shl i64 [[TMP0]], 1 1324; FIXED-NEXT: [[TMP3:%.*]] = shl i64 [[TMP1]], 1 1325; FIXED-NEXT: [[TMP4:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP2]] 1326; FIXED-NEXT: [[TMP5:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]] 1327; FIXED-NEXT: [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8 1328; FIXED-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 1329; FIXED-NEXT: [[STRIDED_VEC3:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 1330; FIXED-NEXT: [[WIDE_VEC1:%.*]] = load <8 x i64>, ptr [[TMP5]], align 8 1331; FIXED-NEXT: [[STRIDED_VEC2:%.*]] = shufflevector <8 x i64> [[WIDE_VEC1]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6> 1332; FIXED-NEXT: [[STRIDED_VEC4:%.*]] = shufflevector <8 x i64> [[WIDE_VEC1]], <8 x i64> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7> 1333; FIXED-NEXT: [[TMP8:%.*]] = add <4 x i64> [[STRIDED_VEC]], [[STRIDED_VEC3]] 1334; FIXED-NEXT: [[TMP9:%.*]] = add <4 x i64> [[STRIDED_VEC2]], [[STRIDED_VEC4]] 1335; FIXED-NEXT: [[TMP10:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[TMP0]] 1336; FIXED-NEXT: [[TMP12:%.*]] = getelementptr i64, ptr [[TMP10]], i32 0 1337; FIXED-NEXT: [[TMP13:%.*]] = getelementptr i64, ptr [[TMP10]], i32 4 1338; FIXED-NEXT: store <4 x i64> [[TMP8]], ptr [[TMP12]], align 8 1339; FIXED-NEXT: store <4 x i64> [[TMP9]], ptr [[TMP13]], align 8 1340; FIXED-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8 1341; FIXED-NEXT: [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024 1342; FIXED-NEXT: br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] 1343; FIXED: middle.block: 1344; FIXED-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]] 1345; FIXED: scalar.ph: 1346; FIXED-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1347; FIXED-NEXT: br label [[LOOP:%.*]] 1348; FIXED: loop: 1349; FIXED-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1350; FIXED-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1351; FIXED-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 1352; FIXED-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 1353; FIXED-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1354; FIXED-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 1355; FIXED-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 1356; FIXED-NEXT: [[RES:%.*]] = add i64 [[X0]], [[X1]] 1357; FIXED-NEXT: [[DST:%.*]] = getelementptr i64, ptr [[Q]], i64 [[I]] 1358; FIXED-NEXT: store i64 [[RES]], ptr [[DST]], align 8 1359; FIXED-NEXT: [[NEXTI]] = add i64 [[I]], 1 1360; FIXED-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1361; FIXED-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] 1362; FIXED: exit: 1363; FIXED-NEXT: ret void 1364; 1365; SCALABLE-LABEL: @combine_load_factor2_i64( 1366; SCALABLE-NEXT: entry: 1367; SCALABLE-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64() 1368; SCALABLE-NEXT: [[TMP1:%.*]] = mul i64 [[TMP0]], 2 1369; SCALABLE-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]] 1370; SCALABLE-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]] 1371; SCALABLE: vector.ph: 1372; SCALABLE-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64() 1373; SCALABLE-NEXT: [[TMP3:%.*]] = mul i64 [[TMP2]], 2 1374; SCALABLE-NEXT: [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]] 1375; SCALABLE-NEXT: [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]] 1376; SCALABLE-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64() 1377; SCALABLE-NEXT: [[TMP5:%.*]] = mul i64 [[TMP4]], 2 1378; SCALABLE-NEXT: br label [[VECTOR_BODY:%.*]] 1379; SCALABLE: vector.body: 1380; SCALABLE-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 1381; SCALABLE-NEXT: [[TMP6:%.*]] = add i64 [[INDEX]], 0 1382; SCALABLE-NEXT: [[TMP7:%.*]] = shl i64 [[TMP6]], 1 1383; SCALABLE-NEXT: [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[TMP7]] 1384; SCALABLE-NEXT: [[WIDE_VEC:%.*]] = load <vscale x 4 x i64>, ptr [[TMP8]], align 8 1385; SCALABLE-NEXT: [[STRIDED_VEC:%.*]] = call { <vscale x 2 x i64>, <vscale x 2 x i64> } @llvm.vector.deinterleave2.nxv4i64(<vscale x 4 x i64> [[WIDE_VEC]]) 1386; SCALABLE-NEXT: [[TMP10:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 0 1387; SCALABLE-NEXT: [[TMP11:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1 1388; SCALABLE-NEXT: [[TMP12:%.*]] = add <vscale x 2 x i64> [[TMP10]], [[TMP11]] 1389; SCALABLE-NEXT: [[TMP13:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[TMP6]] 1390; SCALABLE-NEXT: [[TMP14:%.*]] = getelementptr i64, ptr [[TMP13]], i32 0 1391; SCALABLE-NEXT: store <vscale x 2 x i64> [[TMP12]], ptr [[TMP14]], align 8 1392; SCALABLE-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]] 1393; SCALABLE-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 1394; SCALABLE-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] 1395; SCALABLE: middle.block: 1396; SCALABLE-NEXT: [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]] 1397; SCALABLE-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]] 1398; SCALABLE: scalar.ph: 1399; SCALABLE-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ] 1400; SCALABLE-NEXT: br label [[LOOP:%.*]] 1401; SCALABLE: loop: 1402; SCALABLE-NEXT: [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ] 1403; SCALABLE-NEXT: [[OFFSET0:%.*]] = shl i64 [[I]], 1 1404; SCALABLE-NEXT: [[Q0:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET0]] 1405; SCALABLE-NEXT: [[X0:%.*]] = load i64, ptr [[Q0]], align 8 1406; SCALABLE-NEXT: [[OFFSET1:%.*]] = add i64 [[OFFSET0]], 1 1407; SCALABLE-NEXT: [[Q1:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET1]] 1408; SCALABLE-NEXT: [[X1:%.*]] = load i64, ptr [[Q1]], align 8 1409; SCALABLE-NEXT: [[RES:%.*]] = add i64 [[X0]], [[X1]] 1410; SCALABLE-NEXT: [[DST:%.*]] = getelementptr i64, ptr [[Q]], i64 [[I]] 1411; SCALABLE-NEXT: store i64 [[RES]], ptr [[DST]], align 8 1412; SCALABLE-NEXT: [[NEXTI]] = add i64 [[I]], 1 1413; SCALABLE-NEXT: [[DONE:%.*]] = icmp eq i64 [[NEXTI]], 1024 1414; SCALABLE-NEXT: br i1 [[DONE]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP15:![0-9]+]] 1415; SCALABLE: exit: 1416; SCALABLE-NEXT: ret void 1417; 1418entry: 1419 br label %loop 1420loop: 1421 %i = phi i64 [0, %entry], [%nexti, %loop] 1422 1423 %offset0 = shl i64 %i, 1 1424 %q0 = getelementptr i64, ptr %p, i64 %offset0 1425 %x0 = load i64, ptr %q0 1426 1427 %offset1 = add i64 %offset0, 1 1428 %q1 = getelementptr i64, ptr %p, i64 %offset1 1429 %x1 = load i64, ptr %q1 1430 1431 %res = add i64 %x0, %x1 1432 1433 %dst = getelementptr i64, ptr %q, i64 %i 1434 store i64 %res, ptr %dst 1435 1436 %nexti = add i64 %i, 1 1437 %done = icmp eq i64 %nexti, 1024 1438 br i1 %done, label %exit, label %loop 1439exit: 1440 ret void 1441} 1442