clang/test/CodeGen/attr-arm-sve-vector-bits-codegen.c

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
   2 // RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +bf16 -mvscale-min=4 -mvscale-max=4 -S -disable-llvm-passes -emit-llvm -o - %s | FileCheck %s
   3
   4 // REQUIRES: aarch64-registered-target
   5
   6 #include <arm_sve.h>
   7
   8 #define N __ARM_FEATURE_SVE_BITS
   9
  10 typedef svint32_t fixed_int32_t __attribute__((arm_sve_vector_bits(N)));
  11 typedef svbool_t fixed_bool_t __attribute__((arm_sve_vector_bits(N)));
  12 typedef uint8_t uint8_vec_t __attribute__((vector_size(N / 64)));
  13
  14 fixed_bool_t global_pred;
  15 fixed_int32_t global_vec;
  16
  17 // CHECK-LABEL: @foo(
  18 // CHECK-NEXT:  entry:
  19 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca <16 x i32>, align 16
  20 // CHECK-NEXT:    [[PRED_ADDR:%.*]] = alloca <vscale x 16 x i1>, align 2
  21 // CHECK-NEXT:    [[VEC_ADDR:%.*]] = alloca <vscale x 4 x i32>, align 16
  22 // CHECK-NEXT:    [[PG:%.*]] = alloca <vscale x 16 x i1>, align 2
  23 // CHECK-NEXT:    store <vscale x 16 x i1> [[PRED:%.*]], ptr [[PRED_ADDR]], align 2
  24 // CHECK-NEXT:    store <vscale x 4 x i32> [[VEC:%.*]], ptr [[VEC_ADDR]], align 16
  25 // CHECK-NEXT:    [[TMP0:%.*]] = load <vscale x 16 x i1>, ptr [[PRED_ADDR]], align 2
  26 // CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i8>, ptr @global_pred, align 2
  27 // CHECK-NEXT:    [[CASTFIXEDSVE:%.*]] = call <vscale x 2 x i8> @llvm.vector.insert.nxv2i8.v8i8(<vscale x 2 x i8> undef, <8 x i8> [[TMP1]], i64 0)
  28 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <vscale x 2 x i8> [[CASTFIXEDSVE]] to <vscale x 16 x i1>
  29 // CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, ptr @global_pred, align 2
  30 // CHECK-NEXT:    [[CASTFIXEDSVE2:%.*]] = call <vscale x 2 x i8> @llvm.vector.insert.nxv2i8.v8i8(<vscale x 2 x i8> undef, <8 x i8> [[TMP3]], i64 0)
  31 // CHECK-NEXT:    [[TMP4:%.*]] = bitcast <vscale x 2 x i8> [[CASTFIXEDSVE2]] to <vscale x 16 x i1>
  32 // CHECK-NEXT:    [[TMP5:%.*]] = call <vscale x 16 x i1> @llvm.aarch64.sve.and.z.nxv16i1(<vscale x 16 x i1> [[TMP0]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP4]])
  33 // CHECK-NEXT:    store <vscale x 16 x i1> [[TMP5]], ptr [[PG]], align 2
  34 // CHECK-NEXT:    [[TMP6:%.*]] = load <vscale x 16 x i1>, ptr [[PG]], align 2
  35 // CHECK-NEXT:    [[TMP7:%.*]] = load <16 x i32>, ptr @global_vec, align 16
  36 // CHECK-NEXT:    [[CASTSCALABLESVE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP7]], i64 0)
  37 // CHECK-NEXT:    [[TMP8:%.*]] = load <vscale x 4 x i32>, ptr [[VEC_ADDR]], align 16
  38 // CHECK-NEXT:    [[TMP9:%.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[TMP6]])
  39 // CHECK-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.add.nxv4i32(<vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[CASTSCALABLESVE]], <vscale x 4 x i32> [[TMP8]])
  40 // CHECK-NEXT:    [[CASTFIXEDSVE3:%.*]] = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> [[TMP10]], i64 0)
  41 // CHECK-NEXT:    store <16 x i32> [[CASTFIXEDSVE3]], ptr [[RETVAL]], align 16
  42 // CHECK-NEXT:    [[TMP11:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 16
  43 // CHECK-NEXT:    [[CASTSCALABLESVE4:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP11]], i64 0)
  44 // CHECK-NEXT:    ret <vscale x 4 x i32> [[CASTSCALABLESVE4]]
  45 //
  46 fixed_int32_t foo(svbool_t pred, svint32_t vec) {
  47   svbool_t pg = svand_z(pred, global_pred, global_pred);
  48   return svadd_m(pg, global_vec, vec);
  49 }
  50
  51 // CHECK-LABEL: @test_ptr_to_global(
  52 // CHECK-NEXT:  entry:
  53 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca <16 x i32>, align 16
  54 // CHECK-NEXT:    [[GLOBAL_VEC_PTR:%.*]] = alloca ptr, align 8
  55 // CHECK-NEXT:    store ptr @global_vec, ptr [[GLOBAL_VEC_PTR]], align 8
  56 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[GLOBAL_VEC_PTR]], align 8
  57 // CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i32>, ptr [[TMP0]], align 16
  58 // CHECK-NEXT:    store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 16
  59 // CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 16
  60 // CHECK-NEXT:    [[CASTSCALABLESVE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP2]], i64 0)
  61 // CHECK-NEXT:    ret <vscale x 4 x i32> [[CASTSCALABLESVE]]
  62 //
  63 fixed_int32_t test_ptr_to_global() {
  64   fixed_int32_t *global_vec_ptr;
  65   global_vec_ptr = &global_vec;
  66   return *global_vec_ptr;
  67 }
  68
  69 //
  70 // Test casting pointer from fixed-length array to scalable vector.
  71 // CHECK-LABEL: @array_arg(
  72 // CHECK-NEXT:  entry:
  73 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca <16 x i32>, align 16
  74 // CHECK-NEXT:    [[ARR_ADDR:%.*]] = alloca ptr, align 8
  75 // CHECK-NEXT:    store ptr [[ARR:%.*]], ptr [[ARR_ADDR]], align 8
  76 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[ARR_ADDR]], align 8
  77 // CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds <16 x i32>, ptr [[TMP0]], i64 0
  78 // CHECK-NEXT:    [[TMP1:%.*]] = load <16 x i32>, ptr [[ARRAYIDX]], align 16
  79 // CHECK-NEXT:    store <16 x i32> [[TMP1]], ptr [[RETVAL]], align 16
  80 // CHECK-NEXT:    [[TMP2:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 16
  81 // CHECK-NEXT:    [[CASTSCALABLESVE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP2]], i64 0)
  82 // CHECK-NEXT:    ret <vscale x 4 x i32> [[CASTSCALABLESVE]]
  83 //
  84 fixed_int32_t array_arg(fixed_int32_t arr[]) {
  85   return arr[0];
  86 }
  87
  88 // CHECK-LABEL: @address_of_array_idx(
  89 // CHECK-NEXT:  entry:
  90 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca <8 x i8>, align 2
  91 // CHECK-NEXT:    [[ARR:%.*]] = alloca [3 x <8 x i8>], align 2
  92 // CHECK-NEXT:    [[PARR:%.*]] = alloca ptr, align 8
  93 // CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [3 x <8 x i8>], ptr [[ARR]], i64 0, i64 0
  94 // CHECK-NEXT:    store ptr [[ARRAYIDX]], ptr [[PARR]], align 8
  95 // CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[PARR]], align 8
  96 // CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i8>, ptr [[TMP0]], align 2
  97 // CHECK-NEXT:    store <8 x i8> [[TMP1]], ptr [[RETVAL]], align 2
  98 // CHECK-NEXT:    [[TMP2:%.*]] = load <8 x i8>, ptr [[RETVAL]], align 2
  99 // CHECK-NEXT:    [[CASTFIXEDSVE:%.*]] = call <vscale x 2 x i8> @llvm.vector.insert.nxv2i8.v8i8(<vscale x 2 x i8> undef, <8 x i8> [[TMP2]], i64 0)
 100 // CHECK-NEXT:    [[TMP3:%.*]] = bitcast <vscale x 2 x i8> [[CASTFIXEDSVE]] to <vscale x 16 x i1>
 101 // CHECK-NEXT:    ret <vscale x 16 x i1> [[TMP3]]
 102 //
 103 fixed_bool_t address_of_array_idx() {
 104   fixed_bool_t arr[3];
 105   fixed_bool_t *parr;
 106   parr = &arr[0];
 107   return *parr;
 108 }
 109
 110 // CHECK-LABEL: @test_cast(
 111 // CHECK-NEXT:  entry:
 112 // CHECK-NEXT:    [[RETVAL:%.*]] = alloca <16 x i32>, align 16
 113 // CHECK-NEXT:    [[PRED_ADDR:%.*]] = alloca <vscale x 16 x i1>, align 2
 114 // CHECK-NEXT:    [[VEC_ADDR:%.*]] = alloca <vscale x 4 x i32>, align 16
 115 // CHECK-NEXT:    [[XX:%.*]] = alloca <8 x i8>, align 8
 116 // CHECK-NEXT:    [[YY:%.*]] = alloca <8 x i8>, align 8
 117 // CHECK-NEXT:    [[PG:%.*]] = alloca <vscale x 16 x i1>, align 2
 118 // CHECK-NEXT:    store <vscale x 16 x i1> [[PRED:%.*]], ptr [[PRED_ADDR]], align 2
 119 // CHECK-NEXT:    store <vscale x 4 x i32> [[VEC:%.*]], ptr [[VEC_ADDR]], align 16
 120 // CHECK-NEXT:    store <8 x i8> <i8 1, i8 2, i8 3, i8 4, i8 0, i8 0, i8 0, i8 0>, ptr [[XX]], align 8
 121 // CHECK-NEXT:    store <8 x i8> <i8 2, i8 5, i8 4, i8 6, i8 0, i8 0, i8 0, i8 0>, ptr [[YY]], align 8
 122 // CHECK-NEXT:    [[TMP0:%.*]] = load <vscale x 16 x i1>, ptr [[PRED_ADDR]], align 2
 123 // CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i8>, ptr @global_pred, align 2
 124 // CHECK-NEXT:    [[CASTFIXEDSVE:%.*]] = call <vscale x 2 x i8> @llvm.vector.insert.nxv2i8.v8i8(<vscale x 2 x i8> undef, <8 x i8> [[TMP1]], i64 0)
 125 // CHECK-NEXT:    [[TMP2:%.*]] = bitcast <vscale x 2 x i8> [[CASTFIXEDSVE]] to <vscale x 16 x i1>
 126 // CHECK-NEXT:    [[TMP3:%.*]] = load <8 x i8>, ptr [[XX]], align 8
 127 // CHECK-NEXT:    [[TMP4:%.*]] = load <8 x i8>, ptr [[YY]], align 8
 128 // CHECK-NEXT:    [[ADD:%.*]] = add <8 x i8> [[TMP3]], [[TMP4]]
 129 // CHECK-NEXT:    [[CASTFIXEDSVE2:%.*]] = call <vscale x 2 x i8> @llvm.vector.insert.nxv2i8.v8i8(<vscale x 2 x i8> undef, <8 x i8> [[ADD]], i64 0)
 130 // CHECK-NEXT:    [[TMP5:%.*]] = bitcast <vscale x 2 x i8> [[CASTFIXEDSVE2]] to <vscale x 16 x i1>
 131 // CHECK-NEXT:    [[TMP6:%.*]] = call <vscale x 16 x i1> @llvm.aarch64.sve.and.z.nxv16i1(<vscale x 16 x i1> [[TMP0]], <vscale x 16 x i1> [[TMP2]], <vscale x 16 x i1> [[TMP5]])
 132 // CHECK-NEXT:    store <vscale x 16 x i1> [[TMP6]], ptr [[PG]], align 2
 133 // CHECK-NEXT:    [[TMP7:%.*]] = load <vscale x 16 x i1>, ptr [[PG]], align 2
 134 // CHECK-NEXT:    [[TMP8:%.*]] = load <16 x i32>, ptr @global_vec, align 16
 135 // CHECK-NEXT:    [[CASTSCALABLESVE:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP8]], i64 0)
 136 // CHECK-NEXT:    [[TMP9:%.*]] = load <vscale x 4 x i32>, ptr [[VEC_ADDR]], align 16
 137 // CHECK-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[TMP7]])
 138 // CHECK-NEXT:    [[TMP11:%.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.add.nxv4i32(<vscale x 4 x i1> [[TMP10]], <vscale x 4 x i32> [[CASTSCALABLESVE]], <vscale x 4 x i32> [[TMP9]])
 139 // CHECK-NEXT:    [[CASTFIXEDSVE3:%.*]] = call <16 x i32> @llvm.vector.extract.v16i32.nxv4i32(<vscale x 4 x i32> [[TMP11]], i64 0)
 140 // CHECK-NEXT:    store <16 x i32> [[CASTFIXEDSVE3]], ptr [[RETVAL]], align 16
 141 // CHECK-NEXT:    [[TMP12:%.*]] = load <16 x i32>, ptr [[RETVAL]], align 16
 142 // CHECK-NEXT:    [[CASTSCALABLESVE4:%.*]] = call <vscale x 4 x i32> @llvm.vector.insert.nxv4i32.v16i32(<vscale x 4 x i32> undef, <16 x i32> [[TMP12]], i64 0)
 143 // CHECK-NEXT:    ret <vscale x 4 x i32> [[CASTSCALABLESVE4]]
 144 //
 145 fixed_int32_t test_cast(svbool_t pred, svint32_t vec) {
 146   uint8_vec_t xx = {1, 2, 3, 4};
 147   uint8_vec_t yy = {2, 5, 4, 6};
 148   svbool_t pg = svand_z(pred, global_pred, xx + yy);
 149   return svadd_m(pg, global_vec, vec);
 150 }