168 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py2; RUN: llc -mtriple=riscv32 -mattr=+v,+unaligned-vector-mem -target-abi=ilp32 \3; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV324; RUN: llc -mtriple=riscv64 -mattr=+v,+unaligned-vector-mem -target-abi=lp64 \5; RUN: -verify-machineinstrs < %s | FileCheck %s --check-prefixes=CHECK,RV646 7define void @constant_forward_stride(ptr %s, ptr %d) {8; CHECK-LABEL: constant_forward_stride:9; CHECK: # %bb.0:10; CHECK-NEXT: li a2, 1611; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma12; CHECK-NEXT: vlse16.v v8, (a0), a213; CHECK-NEXT: vse16.v v8, (a1)14; CHECK-NEXT: ret15 %1 = getelementptr inbounds i8, ptr %s, i64 1616 %2 = getelementptr inbounds i8, ptr %s, i64 3217 %3 = getelementptr inbounds i8, ptr %s, i64 4818 %4 = load <2 x i8>, ptr %s, align 119 %5 = load <2 x i8>, ptr %1, align 120 %6 = load <2 x i8>, ptr %2, align 121 %7 = load <2 x i8>, ptr %3, align 122 %8 = shufflevector <2 x i8> %4, <2 x i8> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>23 %9 = shufflevector <2 x i8> %6, <2 x i8> %7, <4 x i32> <i32 0, i32 1, i32 2, i32 3>24 %10 = shufflevector <4 x i8> %8, <4 x i8> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>25 store <8 x i8> %10, ptr %d, align 126 ret void27}28 29define void @constant_forward_stride2(ptr %s, ptr %d) {30; CHECK-LABEL: constant_forward_stride2:31; CHECK: # %bb.0:32; CHECK-NEXT: addi a0, a0, -4833; CHECK-NEXT: li a2, 1634; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma35; CHECK-NEXT: vlse16.v v8, (a0), a236; CHECK-NEXT: vse16.v v8, (a1)37; CHECK-NEXT: ret38 %1 = getelementptr inbounds i8, ptr %s, i64 -1639 %2 = getelementptr inbounds i8, ptr %s, i64 -3240 %3 = getelementptr inbounds i8, ptr %s, i64 -4841 %4 = load <2 x i8>, ptr %3, align 142 %5 = load <2 x i8>, ptr %2, align 143 %6 = load <2 x i8>, ptr %1, align 144 %7 = load <2 x i8>, ptr %s, align 145 %8 = shufflevector <2 x i8> %4, <2 x i8> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>46 %9 = shufflevector <2 x i8> %6, <2 x i8> %7, <4 x i32> <i32 0, i32 1, i32 2, i32 3>47 %10 = shufflevector <4 x i8> %8, <4 x i8> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>48 store <8 x i8> %10, ptr %d, align 149 ret void50}51 52define void @constant_forward_stride3(ptr %s, ptr %d) {53; CHECK-LABEL: constant_forward_stride3:54; CHECK: # %bb.0:55; CHECK-NEXT: li a2, 1656; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma57; CHECK-NEXT: vlse16.v v8, (a0), a258; CHECK-NEXT: vse16.v v8, (a1)59; CHECK-NEXT: ret60 %1 = getelementptr inbounds i8, ptr %s, i64 1661 %2 = getelementptr inbounds i8, ptr %s, i64 3262 %3 = getelementptr inbounds i8, ptr %s, i64 4863 %4 = getelementptr inbounds i8, ptr %1, i64 064 %5 = getelementptr inbounds i8, ptr %2, i64 065 %6 = getelementptr inbounds i8, ptr %3, i64 066 %7 = load <2 x i8>, ptr %s, align 167 %8 = load <2 x i8>, ptr %4, align 168 %9 = load <2 x i8>, ptr %5, align 169 %10 = load <2 x i8>, ptr %6, align 170 %11 = shufflevector <2 x i8> %7, <2 x i8> %8, <4 x i32> <i32 0, i32 1, i32 2, i32 3>71 %12 = shufflevector <2 x i8> %9, <2 x i8> %10, <4 x i32> <i32 0, i32 1, i32 2, i32 3>72 %13 = shufflevector <4 x i8> %11, <4 x i8> %12, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>73 store <8 x i8> %13, ptr %d, align 174 ret void75}76 77define void @constant_back_stride(ptr %s, ptr %d) {78; CHECK-LABEL: constant_back_stride:79; CHECK: # %bb.0:80; CHECK-NEXT: li a2, -1681; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma82; CHECK-NEXT: vlse16.v v8, (a0), a283; CHECK-NEXT: vse16.v v8, (a1)84; CHECK-NEXT: ret85 %1 = getelementptr inbounds i8, ptr %s, i64 -1686 %2 = getelementptr inbounds i8, ptr %s, i64 -3287 %3 = getelementptr inbounds i8, ptr %s, i64 -4888 %4 = load <2 x i8>, ptr %s, align 189 %5 = load <2 x i8>, ptr %1, align 190 %6 = load <2 x i8>, ptr %2, align 191 %7 = load <2 x i8>, ptr %3, align 192 %8 = shufflevector <2 x i8> %4, <2 x i8> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>93 %9 = shufflevector <2 x i8> %6, <2 x i8> %7, <4 x i32> <i32 0, i32 1, i32 2, i32 3>94 %10 = shufflevector <4 x i8> %8, <4 x i8> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>95 store <8 x i8> %10, ptr %d, align 196 ret void97}98 99define void @constant_back_stride2(ptr %s, ptr %d) {100; CHECK-LABEL: constant_back_stride2:101; CHECK: # %bb.0:102; CHECK-NEXT: addi a0, a0, 48103; CHECK-NEXT: li a2, -16104; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma105; CHECK-NEXT: vlse16.v v8, (a0), a2106; CHECK-NEXT: vse16.v v8, (a1)107; CHECK-NEXT: ret108 %1 = getelementptr inbounds i8, ptr %s, i64 16109 %2 = getelementptr inbounds i8, ptr %s, i64 32110 %3 = getelementptr inbounds i8, ptr %s, i64 48111 %4 = load <2 x i8>, ptr %3, align 1112 %5 = load <2 x i8>, ptr %2, align 1113 %6 = load <2 x i8>, ptr %1, align 1114 %7 = load <2 x i8>, ptr %s, align 1115 %8 = shufflevector <2 x i8> %4, <2 x i8> %5, <4 x i32> <i32 0, i32 1, i32 2, i32 3>116 %9 = shufflevector <2 x i8> %6, <2 x i8> %7, <4 x i32> <i32 0, i32 1, i32 2, i32 3>117 %10 = shufflevector <4 x i8> %8, <4 x i8> %9, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>118 store <8 x i8> %10, ptr %d, align 1119 ret void120}121 122define void @constant_back_stride3(ptr %s, ptr %d) {123; CHECK-LABEL: constant_back_stride3:124; CHECK: # %bb.0:125; CHECK-NEXT: li a2, -16126; CHECK-NEXT: vsetivli zero, 4, e16, mf2, ta, ma127; CHECK-NEXT: vlse16.v v8, (a0), a2128; CHECK-NEXT: vse16.v v8, (a1)129; CHECK-NEXT: ret130 %1 = getelementptr inbounds i8, ptr %s, i64 -16131 %2 = getelementptr inbounds i8, ptr %s, i64 -32132 %3 = getelementptr inbounds i8, ptr %s, i64 -48133 %4 = getelementptr inbounds i8, ptr %1, i64 0134 %5 = getelementptr inbounds i8, ptr %2, i64 0135 %6 = getelementptr inbounds i8, ptr %3, i64 0136 %7 = load <2 x i8>, ptr %s, align 1137 %8 = load <2 x i8>, ptr %4, align 1138 %9 = load <2 x i8>, ptr %5, align 1139 %10 = load <2 x i8>, ptr %6, align 1140 %11 = shufflevector <2 x i8> %7, <2 x i8> %8, <4 x i32> <i32 0, i32 1, i32 2, i32 3>141 %12 = shufflevector <2 x i8> %9, <2 x i8> %10, <4 x i32> <i32 0, i32 1, i32 2, i32 3>142 %13 = shufflevector <4 x i8> %11, <4 x i8> %12, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>143 store <8 x i8> %13, ptr %d, align 1144 ret void145}146 147define void @constant_zero_stride(ptr %s, ptr %d) {148; CHECK-LABEL: constant_zero_stride:149; CHECK: # %bb.0:150; CHECK-NEXT: vsetivli zero, 2, e8, mf8, ta, ma151; CHECK-NEXT: vle8.v v8, (a0)152; CHECK-NEXT: vmv1r.v v9, v8153; CHECK-NEXT: vsetivli zero, 4, e8, mf4, ta, ma154; CHECK-NEXT: vslideup.vi v9, v8, 2155; CHECK-NEXT: vse8.v v9, (a1)156; CHECK-NEXT: ret157 %1 = getelementptr inbounds i8, ptr %s, i64 0158 %2 = load <2 x i8>, ptr %s, align 1159 %3 = load <2 x i8>, ptr %1, align 1160 %4 = shufflevector <2 x i8> %2, <2 x i8> %3, <4 x i32> <i32 0, i32 1, i32 2, i32 3>161 store <4 x i8> %4, ptr %d, align 1162 ret void163}164 165;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:166; RV32: {{.*}}167; RV64: {{.*}}168