clang/test/CodeGen/RISCV/rvv-vls-shift-ops.c

   1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
   2 // RUN: %clang_cc1 -triple riscv64-none-linux-gnu -target-feature +zve64d \
   3 // RUN: -target-feature +f -target-feature +d -disable-O0-optnone \
   4 // RUN: -mvscale-min=4 -mvscale-max=4 -emit-llvm -o - %s | \
   5 // RUN: opt -S -passes=sroa | FileCheck %s
   6
   7 // REQUIRES: riscv-registered-target
   8
   9 #include <stdint.h>
  10
  11 typedef __rvv_int8m1_t vint8m1_t;
  12 typedef __rvv_uint8m1_t vuint8m1_t;
  13 typedef __rvv_int16m1_t vint16m1_t;
  14 typedef __rvv_uint16m1_t vuint16m1_t;
  15 typedef __rvv_int32m1_t vint32m1_t;
  16 typedef __rvv_uint32m1_t vuint32m1_t;
  17 typedef __rvv_int64m1_t vint64m1_t;
  18 typedef __rvv_uint64m1_t vuint64m1_t;
  19 typedef __rvv_float32m1_t vfloat32m1_t;
  20 typedef __rvv_float64m1_t vfloat64m1_t;
  21
  22 typedef vint8m1_t fixed_int8m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  23 typedef vint16m1_t fixed_int16m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  24 typedef vint32m1_t fixed_int32m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  25 typedef vint64m1_t fixed_int64m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  26
  27 typedef vuint8m1_t fixed_uint8m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  28 typedef vuint16m1_t fixed_uint16m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  29 typedef vuint32m1_t fixed_uint32m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  30 typedef vuint64m1_t fixed_uint64m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  31
  32 typedef vfloat32m1_t fixed_float32m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  33 typedef vfloat64m1_t fixed_float64m1_t __attribute__((riscv_rvv_vector_bits(__riscv_v_fixed_vlen)));
  34
  35 // CHECK-LABEL: @lshift_i8(
  36 // CHECK-NEXT:  entry:
  37 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
  38 // CHECK-NEXT:    [[B:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[B_COERCE:%.*]], i64 0)
  39 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[A]], [[B]]
  40 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
  41 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
  42 //
  43 fixed_int8m1_t lshift_i8(fixed_int8m1_t a, fixed_int8m1_t b) {
  44   return a << b;
  45 }
  46
  47 // CHECK-LABEL: @rshift_i8(
  48 // CHECK-NEXT:  entry:
  49 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
  50 // CHECK-NEXT:    [[B:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[B_COERCE:%.*]], i64 0)
  51 // CHECK-NEXT:    [[SHR:%.*]] = ashr <32 x i8> [[A]], [[B]]
  52 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
  53 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
  54 //
  55 fixed_int8m1_t rshift_i8(fixed_int8m1_t a, fixed_int8m1_t b) {
  56   return a >> b;
  57 }
  58
  59 // CHECK-LABEL: @lshift_u8(
  60 // CHECK-NEXT:  entry:
  61 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
  62 // CHECK-NEXT:    [[B:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[B_COERCE:%.*]], i64 0)
  63 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[A]], [[B]]
  64 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
  65 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
  66 //
  67 fixed_uint8m1_t lshift_u8(fixed_uint8m1_t a, fixed_uint8m1_t b) {
  68   return a << b;
  69 }
  70
  71 // CHECK-LABEL: @rshift_u8(
  72 // CHECK-NEXT:  entry:
  73 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
  74 // CHECK-NEXT:    [[B:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[B_COERCE:%.*]], i64 0)
  75 // CHECK-NEXT:    [[SHR:%.*]] = lshr <32 x i8> [[A]], [[B]]
  76 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
  77 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
  78 //
  79 fixed_uint8m1_t rshift_u8(fixed_uint8m1_t a, fixed_uint8m1_t b) {
  80   return a >> b;
  81 }
  82
  83 // CHECK-LABEL: @lshift_i16(
  84 // CHECK-NEXT:  entry:
  85 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
  86 // CHECK-NEXT:    [[B:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[B_COERCE:%.*]], i64 0)
  87 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[A]], [[B]]
  88 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
  89 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
  90 //
  91 fixed_int16m1_t lshift_i16(fixed_int16m1_t a, fixed_int16m1_t b) {
  92   return a << b;
  93 }
  94
  95 // CHECK-LABEL: @rshift_i16(
  96 // CHECK-NEXT:  entry:
  97 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
  98 // CHECK-NEXT:    [[B:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[B_COERCE:%.*]], i64 0)
  99 // CHECK-NEXT:    [[SHR:%.*]] = ashr <16 x i16> [[A]], [[B]]
 100 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 101 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 102 //
 103 fixed_int16m1_t rshift_i16(fixed_int16m1_t a, fixed_int16m1_t b) {
 104   return a >> b;
 105 }
 106
 107 // CHECK-LABEL: @lshift_u16(
 108 // CHECK-NEXT:  entry:
 109 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 110 // CHECK-NEXT:    [[B:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[B_COERCE:%.*]], i64 0)
 111 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[A]], [[B]]
 112 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
 113 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 114 //
 115 fixed_uint16m1_t lshift_u16(fixed_uint16m1_t a, fixed_uint16m1_t b) {
 116   return a << b;
 117 }
 118
 119 // CHECK-LABEL: @rshift_u16(
 120 // CHECK-NEXT:  entry:
 121 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 122 // CHECK-NEXT:    [[B:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[B_COERCE:%.*]], i64 0)
 123 // CHECK-NEXT:    [[SHR:%.*]] = lshr <16 x i16> [[A]], [[B]]
 124 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 125 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 126 //
 127 fixed_uint16m1_t rshift_u16(fixed_uint16m1_t a, fixed_uint16m1_t b) {
 128   return a >> b;
 129 }
 130
 131 // CHECK-LABEL: @lshift_i32(
 132 // CHECK-NEXT:  entry:
 133 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 134 // CHECK-NEXT:    [[B:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[B_COERCE:%.*]], i64 0)
 135 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[A]], [[B]]
 136 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 137 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 138 //
 139 fixed_int32m1_t lshift_i32(fixed_int32m1_t a, fixed_int32m1_t b) {
 140   return a << b;
 141 }
 142
 143 // CHECK-LABEL: @rshift_i32(
 144 // CHECK-NEXT:  entry:
 145 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 146 // CHECK-NEXT:    [[B:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[B_COERCE:%.*]], i64 0)
 147 // CHECK-NEXT:    [[SHR:%.*]] = ashr <8 x i32> [[A]], [[B]]
 148 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 149 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 150 //
 151 fixed_int32m1_t rshift_i32(fixed_int32m1_t a, fixed_int32m1_t b) {
 152   return a >> b;
 153 }
 154
 155 // CHECK-LABEL: @lshift_u32(
 156 // CHECK-NEXT:  entry:
 157 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 158 // CHECK-NEXT:    [[B:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[B_COERCE:%.*]], i64 0)
 159 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[A]], [[B]]
 160 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 161 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 162 //
 163 fixed_uint32m1_t lshift_u32(fixed_uint32m1_t a, fixed_uint32m1_t b) {
 164   return a << b;
 165 }
 166
 167 // CHECK-LABEL: @rshift_u32(
 168 // CHECK-NEXT:  entry:
 169 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 170 // CHECK-NEXT:    [[B:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[B_COERCE:%.*]], i64 0)
 171 // CHECK-NEXT:    [[SHR:%.*]] = lshr <8 x i32> [[A]], [[B]]
 172 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 173 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 174 //
 175 fixed_uint32m1_t rshift_u32(fixed_uint32m1_t a, fixed_uint32m1_t b) {
 176   return a >> b;
 177 }
 178
 179 // CHECK-LABEL: @lshift_i64(
 180 // CHECK-NEXT:  entry:
 181 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 182 // CHECK-NEXT:    [[B:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[B_COERCE:%.*]], i64 0)
 183 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[A]], [[B]]
 184 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 185 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 186 //
 187 fixed_int64m1_t lshift_i64(fixed_int64m1_t a, fixed_int64m1_t b) {
 188   return a << b;
 189 }
 190
 191 // CHECK-LABEL: @rshift_i64(
 192 // CHECK-NEXT:  entry:
 193 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 194 // CHECK-NEXT:    [[B:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[B_COERCE:%.*]], i64 0)
 195 // CHECK-NEXT:    [[SHR:%.*]] = ashr <4 x i64> [[A]], [[B]]
 196 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 197 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 198 //
 199 fixed_int64m1_t rshift_i64(fixed_int64m1_t a, fixed_int64m1_t b) {
 200   return a >> b;
 201 }
 202
 203 // CHECK-LABEL: @lshift_u64(
 204 // CHECK-NEXT:  entry:
 205 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 206 // CHECK-NEXT:    [[B:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[B_COERCE:%.*]], i64 0)
 207 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[A]], [[B]]
 208 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 209 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 210 //
 211 fixed_uint64m1_t lshift_u64(fixed_uint64m1_t a, fixed_uint64m1_t b) {
 212   return a << b;
 213 }
 214
 215 // CHECK-LABEL: @rshift_u64(
 216 // CHECK-NEXT:  entry:
 217 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 218 // CHECK-NEXT:    [[B:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[B_COERCE:%.*]], i64 0)
 219 // CHECK-NEXT:    [[SHR:%.*]] = lshr <4 x i64> [[A]], [[B]]
 220 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 221 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 222 //
 223 fixed_uint64m1_t rshift_u64(fixed_uint64m1_t a, fixed_uint64m1_t b) {
 224   return a >> b;
 225 }
 226
 227 // CHECK-LABEL: @lshift_i8_rsplat(
 228 // CHECK-NEXT:  entry:
 229 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 230 // CHECK-NEXT:    [[CONV:%.*]] = sext i8 [[B:%.*]] to i32
 231 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[CONV]], i64 0
 232 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i32> [[SPLAT_SPLATINSERT]], <32 x i32> poison, <32 x i32> zeroinitializer
 233 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <32 x i32> [[SPLAT_SPLAT]] to <32 x i8>
 234 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[A]], [[SH_PROM]]
 235 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
 236 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 237 //
 238 fixed_int8m1_t lshift_i8_rsplat(fixed_int8m1_t a, int8_t b) {
 239   return a << b;
 240 }
 241
 242 // CHECK-LABEL: @lshift_i8_lsplat(
 243 // CHECK-NEXT:  entry:
 244 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 245 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i8> poison, i8 [[B:%.*]], i64 0
 246 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i8> [[SPLAT_SPLATINSERT]], <32 x i8> poison, <32 x i32> zeroinitializer
 247 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[SPLAT_SPLAT]], [[A]]
 248 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
 249 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 250 //
 251 fixed_int8m1_t lshift_i8_lsplat(fixed_int8m1_t a, int8_t b) {
 252   return b << a;
 253 }
 254
 255 // CHECK-LABEL: @rshift_i8_rsplat(
 256 // CHECK-NEXT:  entry:
 257 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 258 // CHECK-NEXT:    [[CONV:%.*]] = sext i8 [[B:%.*]] to i32
 259 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[CONV]], i64 0
 260 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i32> [[SPLAT_SPLATINSERT]], <32 x i32> poison, <32 x i32> zeroinitializer
 261 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <32 x i32> [[SPLAT_SPLAT]] to <32 x i8>
 262 // CHECK-NEXT:    [[SHR:%.*]] = ashr <32 x i8> [[A]], [[SH_PROM]]
 263 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
 264 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 265 //
 266 fixed_int8m1_t rshift_i8_rsplat(fixed_int8m1_t a, int8_t b) {
 267   return a >> b;
 268 }
 269
 270 // CHECK-LABEL: @rshift_i8_lsplat(
 271 // CHECK-NEXT:  entry:
 272 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 273 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i8> poison, i8 [[B:%.*]], i64 0
 274 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i8> [[SPLAT_SPLATINSERT]], <32 x i8> poison, <32 x i32> zeroinitializer
 275 // CHECK-NEXT:    [[SHR:%.*]] = ashr <32 x i8> [[SPLAT_SPLAT]], [[A]]
 276 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
 277 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 278 //
 279 fixed_int8m1_t rshift_i8_lsplat(fixed_int8m1_t a, int8_t b) {
 280   return b >> a;
 281 }
 282
 283 // CHECK-LABEL: @lshift_u8_rsplat(
 284 // CHECK-NEXT:  entry:
 285 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 286 // CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[B:%.*]] to i32
 287 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[CONV]], i64 0
 288 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i32> [[SPLAT_SPLATINSERT]], <32 x i32> poison, <32 x i32> zeroinitializer
 289 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <32 x i32> [[SPLAT_SPLAT]] to <32 x i8>
 290 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[A]], [[SH_PROM]]
 291 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
 292 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 293 //
 294 fixed_uint8m1_t lshift_u8_rsplat(fixed_uint8m1_t a, uint8_t b) {
 295   return a << b;
 296 }
 297
 298 // CHECK-LABEL: @lshift_u8_lsplat(
 299 // CHECK-NEXT:  entry:
 300 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 301 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i8> poison, i8 [[B:%.*]], i64 0
 302 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i8> [[SPLAT_SPLATINSERT]], <32 x i8> poison, <32 x i32> zeroinitializer
 303 // CHECK-NEXT:    [[SHL:%.*]] = shl <32 x i8> [[SPLAT_SPLAT]], [[A]]
 304 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHL]], i64 0)
 305 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 306 //
 307 fixed_uint8m1_t lshift_u8_lsplat(fixed_uint8m1_t a, uint8_t b) {
 308   return b << a;
 309 }
 310
 311 // CHECK-LABEL: @rshift_u8_rsplat(
 312 // CHECK-NEXT:  entry:
 313 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 314 // CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[B:%.*]] to i32
 315 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i32> poison, i32 [[CONV]], i64 0
 316 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i32> [[SPLAT_SPLATINSERT]], <32 x i32> poison, <32 x i32> zeroinitializer
 317 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <32 x i32> [[SPLAT_SPLAT]] to <32 x i8>
 318 // CHECK-NEXT:    [[SHR:%.*]] = lshr <32 x i8> [[A]], [[SH_PROM]]
 319 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
 320 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 321 //
 322 fixed_uint8m1_t rshift_u8_rsplat(fixed_uint8m1_t a, uint8_t b) {
 323   return a >> b;
 324 }
 325
 326 // CHECK-LABEL: @rshift_u8_lsplat(
 327 // CHECK-NEXT:  entry:
 328 // CHECK-NEXT:    [[A:%.*]] = call <32 x i8> @llvm.vector.extract.v32i8.nxv8i8(<vscale x 8 x i8> [[A_COERCE:%.*]], i64 0)
 329 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <32 x i8> poison, i8 [[B:%.*]], i64 0
 330 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <32 x i8> [[SPLAT_SPLATINSERT]], <32 x i8> poison, <32 x i32> zeroinitializer
 331 // CHECK-NEXT:    [[SHR:%.*]] = lshr <32 x i8> [[SPLAT_SPLAT]], [[A]]
 332 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 8 x i8> @llvm.vector.insert.nxv8i8.v32i8(<vscale x 8 x i8> poison, <32 x i8> [[SHR]], i64 0)
 333 // CHECK-NEXT:    ret <vscale x 8 x i8> [[CAST_SCALABLE]]
 334 //
 335 fixed_uint8m1_t rshift_u8_lsplat(fixed_uint8m1_t a, uint8_t b) {
 336   return b >> a;
 337 }
 338
 339 // CHECK-LABEL: @lshift_i16_rsplat(
 340 // CHECK-NEXT:  entry:
 341 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 342 // CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[B:%.*]] to i32
 343 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[CONV]], i64 0
 344 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i32> [[SPLAT_SPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
 345 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <16 x i32> [[SPLAT_SPLAT]] to <16 x i16>
 346 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[A]], [[SH_PROM]]
 347 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
 348 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 349 //
 350 fixed_int16m1_t lshift_i16_rsplat(fixed_int16m1_t a, int16_t b) {
 351   return a << b;
 352 }
 353
 354 // CHECK-LABEL: @lshift_i16_lsplat(
 355 // CHECK-NEXT:  entry:
 356 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 357 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i16> poison, i16 [[B:%.*]], i64 0
 358 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i16> [[SPLAT_SPLATINSERT]], <16 x i16> poison, <16 x i32> zeroinitializer
 359 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[SPLAT_SPLAT]], [[A]]
 360 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
 361 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 362 //
 363 fixed_int16m1_t lshift_i16_lsplat(fixed_int16m1_t a, int16_t b) {
 364   return b << a;
 365 }
 366
 367 // CHECK-LABEL: @rshift_i16_rsplat(
 368 // CHECK-NEXT:  entry:
 369 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 370 // CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[B:%.*]] to i32
 371 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[CONV]], i64 0
 372 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i32> [[SPLAT_SPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
 373 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <16 x i32> [[SPLAT_SPLAT]] to <16 x i16>
 374 // CHECK-NEXT:    [[SHR:%.*]] = ashr <16 x i16> [[A]], [[SH_PROM]]
 375 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 376 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 377 //
 378 fixed_int16m1_t rshift_i16_rsplat(fixed_int16m1_t a, int16_t b) {
 379   return a >> b;
 380 }
 381
 382 // CHECK-LABEL: @rshift_i16_lsplat(
 383 // CHECK-NEXT:  entry:
 384 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 385 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i16> poison, i16 [[B:%.*]], i64 0
 386 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i16> [[SPLAT_SPLATINSERT]], <16 x i16> poison, <16 x i32> zeroinitializer
 387 // CHECK-NEXT:    [[SHR:%.*]] = ashr <16 x i16> [[SPLAT_SPLAT]], [[A]]
 388 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 389 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 390 //
 391 fixed_int16m1_t rshift_i16_lsplat(fixed_int16m1_t a, int16_t b) {
 392   return b >> a;
 393 }
 394
 395 // CHECK-LABEL: @lshift_u16_rsplat(
 396 // CHECK-NEXT:  entry:
 397 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 398 // CHECK-NEXT:    [[CONV:%.*]] = zext i16 [[B:%.*]] to i32
 399 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[CONV]], i64 0
 400 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i32> [[SPLAT_SPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
 401 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <16 x i32> [[SPLAT_SPLAT]] to <16 x i16>
 402 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[A]], [[SH_PROM]]
 403 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
 404 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 405 //
 406 fixed_uint16m1_t lshift_u16_rsplat(fixed_uint16m1_t a, uint16_t b) {
 407   return a << b;
 408 }
 409
 410 // CHECK-LABEL: @lshift_u16_lsplat(
 411 // CHECK-NEXT:  entry:
 412 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 413 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i16> poison, i16 [[B:%.*]], i64 0
 414 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i16> [[SPLAT_SPLATINSERT]], <16 x i16> poison, <16 x i32> zeroinitializer
 415 // CHECK-NEXT:    [[SHL:%.*]] = shl <16 x i16> [[SPLAT_SPLAT]], [[A]]
 416 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHL]], i64 0)
 417 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 418 //
 419 fixed_uint16m1_t lshift_u16_lsplat(fixed_uint16m1_t a, uint16_t b) {
 420   return b << a;
 421 }
 422
 423 // CHECK-LABEL: @rshift_u16_rsplat(
 424 // CHECK-NEXT:  entry:
 425 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 426 // CHECK-NEXT:    [[CONV:%.*]] = zext i16 [[B:%.*]] to i32
 427 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[CONV]], i64 0
 428 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i32> [[SPLAT_SPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
 429 // CHECK-NEXT:    [[SH_PROM:%.*]] = trunc <16 x i32> [[SPLAT_SPLAT]] to <16 x i16>
 430 // CHECK-NEXT:    [[SHR:%.*]] = lshr <16 x i16> [[A]], [[SH_PROM]]
 431 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 432 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 433 //
 434 fixed_uint16m1_t rshift_u16_rsplat(fixed_uint16m1_t a, uint16_t b) {
 435   return a >> b;
 436 }
 437
 438 // CHECK-LABEL: @rshift_u16_lsplat(
 439 // CHECK-NEXT:  entry:
 440 // CHECK-NEXT:    [[A:%.*]] = call <16 x i16> @llvm.vector.extract.v16i16.nxv4i16(<vscale x 4 x i16> [[A_COERCE:%.*]], i64 0)
 441 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <16 x i16> poison, i16 [[B:%.*]], i64 0
 442 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <16 x i16> [[SPLAT_SPLATINSERT]], <16 x i16> poison, <16 x i32> zeroinitializer
 443 // CHECK-NEXT:    [[SHR:%.*]] = lshr <16 x i16> [[SPLAT_SPLAT]], [[A]]
 444 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 4 x i16> @llvm.vector.insert.nxv4i16.v16i16(<vscale x 4 x i16> poison, <16 x i16> [[SHR]], i64 0)
 445 // CHECK-NEXT:    ret <vscale x 4 x i16> [[CAST_SCALABLE]]
 446 //
 447 fixed_uint16m1_t rshift_u16_lsplat(fixed_uint16m1_t a, uint16_t b) {
 448   return b >> a;
 449 }
 450
 451 // CHECK-LABEL: @lshift_i32_rsplat(
 452 // CHECK-NEXT:  entry:
 453 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 454 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 455 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 456 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[A]], [[SPLAT_SPLAT]]
 457 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 458 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 459 //
 460 fixed_int32m1_t lshift_i32_rsplat(fixed_int32m1_t a, int32_t b) {
 461   return a << b;
 462 }
 463
 464 // CHECK-LABEL: @lshift_i32_lsplat(
 465 // CHECK-NEXT:  entry:
 466 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 467 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 468 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 469 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[SPLAT_SPLAT]], [[A]]
 470 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 471 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 472 //
 473 fixed_int32m1_t lshift_i32_lsplat(fixed_int32m1_t a, int32_t b) {
 474   return b << a;
 475 }
 476
 477 // CHECK-LABEL: @rshift_i32_rsplat(
 478 // CHECK-NEXT:  entry:
 479 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 480 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 481 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 482 // CHECK-NEXT:    [[SHR:%.*]] = ashr <8 x i32> [[A]], [[SPLAT_SPLAT]]
 483 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 484 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 485 //
 486 fixed_int32m1_t rshift_i32_rsplat(fixed_int32m1_t a, int32_t b) {
 487   return a >> b;
 488 }
 489
 490 // CHECK-LABEL: @rshift_i32_lsplat(
 491 // CHECK-NEXT:  entry:
 492 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 493 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 494 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 495 // CHECK-NEXT:    [[SHR:%.*]] = ashr <8 x i32> [[SPLAT_SPLAT]], [[A]]
 496 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 497 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 498 //
 499 fixed_int32m1_t rshift_i32_lsplat(fixed_int32m1_t a, int32_t b) {
 500   return b >> a;
 501 }
 502
 503 // CHECK-LABEL: @lshift_u32_rsplat(
 504 // CHECK-NEXT:  entry:
 505 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 506 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 507 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 508 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[A]], [[SPLAT_SPLAT]]
 509 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 510 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 511 //
 512 fixed_uint32m1_t lshift_u32_rsplat(fixed_uint32m1_t a, uint32_t b) {
 513   return a << b;
 514 }
 515
 516 // CHECK-LABEL: @lshift_u32_lsplat(
 517 // CHECK-NEXT:  entry:
 518 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 519 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 520 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 521 // CHECK-NEXT:    [[SHL:%.*]] = shl <8 x i32> [[SPLAT_SPLAT]], [[A]]
 522 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHL]], i64 0)
 523 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 524 //
 525 fixed_uint32m1_t lshift_u32_lsplat(fixed_uint32m1_t a, uint32_t b) {
 526   return b << a;
 527 }
 528
 529 // CHECK-LABEL: @rshift_u32_rsplat(
 530 // CHECK-NEXT:  entry:
 531 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 532 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 533 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 534 // CHECK-NEXT:    [[SHR:%.*]] = lshr <8 x i32> [[A]], [[SPLAT_SPLAT]]
 535 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 536 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 537 //
 538 fixed_uint32m1_t rshift_u32_rsplat(fixed_uint32m1_t a, uint32_t b) {
 539   return a >> b;
 540 }
 541
 542 // CHECK-LABEL: @rshift_u32_lsplat(
 543 // CHECK-NEXT:  entry:
 544 // CHECK-NEXT:    [[A:%.*]] = call <8 x i32> @llvm.vector.extract.v8i32.nxv2i32(<vscale x 2 x i32> [[A_COERCE:%.*]], i64 0)
 545 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <8 x i32> poison, i32 [[B:%.*]], i64 0
 546 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <8 x i32> [[SPLAT_SPLATINSERT]], <8 x i32> poison, <8 x i32> zeroinitializer
 547 // CHECK-NEXT:    [[SHR:%.*]] = lshr <8 x i32> [[SPLAT_SPLAT]], [[A]]
 548 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 2 x i32> @llvm.vector.insert.nxv2i32.v8i32(<vscale x 2 x i32> poison, <8 x i32> [[SHR]], i64 0)
 549 // CHECK-NEXT:    ret <vscale x 2 x i32> [[CAST_SCALABLE]]
 550 //
 551 fixed_uint32m1_t rshift_u32_lsplat(fixed_uint32m1_t a, uint32_t b) {
 552   return b >> a;
 553 }
 554
 555 // CHECK-LABEL: @lshift_i64_rsplat(
 556 // CHECK-NEXT:  entry:
 557 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 558 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 559 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 560 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[A]], [[SPLAT_SPLAT]]
 561 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 562 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 563 //
 564 fixed_int64m1_t lshift_i64_rsplat(fixed_int64m1_t a, int64_t b) {
 565   return a << b;
 566 }
 567
 568 // CHECK-LABEL: @lshift_i64_lsplat(
 569 // CHECK-NEXT:  entry:
 570 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 571 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 572 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 573 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[SPLAT_SPLAT]], [[A]]
 574 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 575 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 576 //
 577 fixed_int64m1_t lshift_i64_lsplat(fixed_int64m1_t a, int64_t b) {
 578   return b << a;
 579 }
 580
 581 // CHECK-LABEL: @rshift_i64_rsplat(
 582 // CHECK-NEXT:  entry:
 583 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 584 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 585 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 586 // CHECK-NEXT:    [[SHR:%.*]] = ashr <4 x i64> [[A]], [[SPLAT_SPLAT]]
 587 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 588 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 589 //
 590 fixed_int64m1_t rshift_i64_rsplat(fixed_int64m1_t a, int64_t b) {
 591   return a >> b;
 592 }
 593
 594 // CHECK-LABEL: @rshift_i64_lsplat(
 595 // CHECK-NEXT:  entry:
 596 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 597 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 598 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 599 // CHECK-NEXT:    [[SHR:%.*]] = ashr <4 x i64> [[SPLAT_SPLAT]], [[A]]
 600 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 601 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 602 //
 603 fixed_int64m1_t rshift_i64_lsplat(fixed_int64m1_t a, int64_t b) {
 604   return b >> a;
 605 }
 606
 607 // CHECK-LABEL: @lshift_u64_rsplat(
 608 // CHECK-NEXT:  entry:
 609 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 610 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 611 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 612 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[A]], [[SPLAT_SPLAT]]
 613 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 614 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 615 //
 616 fixed_uint64m1_t lshift_u64_rsplat(fixed_uint64m1_t a, uint64_t b) {
 617   return a << b;
 618 }
 619
 620 // CHECK-LABEL: @lshift_u64_lsplat(
 621 // CHECK-NEXT:  entry:
 622 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 623 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 624 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 625 // CHECK-NEXT:    [[SHL:%.*]] = shl <4 x i64> [[SPLAT_SPLAT]], [[A]]
 626 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHL]], i64 0)
 627 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 628 //
 629 fixed_uint64m1_t lshift_u64_lsplat(fixed_uint64m1_t a, uint64_t b) {
 630   return b << a;
 631 }
 632
 633 // CHECK-LABEL: @rshift_u64_rsplat(
 634 // CHECK-NEXT:  entry:
 635 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 636 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 637 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 638 // CHECK-NEXT:    [[SHR:%.*]] = lshr <4 x i64> [[A]], [[SPLAT_SPLAT]]
 639 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 640 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 641 //
 642 fixed_uint64m1_t rshift_u64_rsplat(fixed_uint64m1_t a, uint64_t b) {
 643   return a >> b;
 644 }
 645
 646 // CHECK-LABEL: @rshift_u64_lsplat(
 647 // CHECK-NEXT:  entry:
 648 // CHECK-NEXT:    [[A:%.*]] = call <4 x i64> @llvm.vector.extract.v4i64.nxv1i64(<vscale x 1 x i64> [[A_COERCE:%.*]], i64 0)
 649 // CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[B:%.*]], i64 0
 650 // CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <4 x i64> [[SPLAT_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 651 // CHECK-NEXT:    [[SHR:%.*]] = lshr <4 x i64> [[SPLAT_SPLAT]], [[A]]
 652 // CHECK-NEXT:    [[CAST_SCALABLE:%.*]] = call <vscale x 1 x i64> @llvm.vector.insert.nxv1i64.v4i64(<vscale x 1 x i64> poison, <4 x i64> [[SHR]], i64 0)
 653 // CHECK-NEXT:    ret <vscale x 1 x i64> [[CAST_SCALABLE]]
 654 //
 655 fixed_uint64m1_t rshift_u64_lsplat(fixed_uint64m1_t a, uint64_t b) {
 656   return b >> a;
 657 }