1 // REQUIRES: systemz-registered-target 2 // RUN: %clang_cc1 -target-cpu z13 -triple s390x-linux-gnu \ 3 // RUN: -O2 -fzvector -flax-vector-conversions=none \ 4 // RUN: -ffp-exception-behavior=strict \ 5 // RUN: -Wall -Wno-unused -Werror -emit-llvm %s -o - | FileCheck %s 6 // RUN: %clang_cc1 -target-cpu z13 -triple s390x-linux-gnu \ 7 // RUN: -O2 -fzvector -flax-vector-conversions=none \ 8 // RUN: -ffp-exception-behavior=strict \ 9 // RUN: -Wall -Wno-unused -Werror -S %s -o - | FileCheck %s --check-prefix=CHECK-ASM 10 11 #include <vecintrin.h> 12 13 volatile vector signed long long vsl; 14 volatile vector unsigned long long vul; 15 volatile vector bool long long vbl; 16 volatile vector double vd; 17 18 volatile double d; 19 20 const float * volatile cptrf; 21 const double * volatile cptrd; 22 23 float * volatile ptrf; 24 double * volatile ptrd; 25 26 volatile int idx; 27 28 void test_core(void) { 29 // CHECK-ASM-LABEL: test_core 30 31 d = vec_extract(vd, idx); 32 // CHECK: extractelement <2 x double> %{{.*}}, i32 %{{.*}} 33 // CHECK-ASM: vlgvg 34 35 vd = vec_insert(d, vd, idx); 36 // CHECK: insertelement <2 x double> %{{.*}}, double %{{.*}}, i32 %{{.*}} 37 // CHECK-ASM: vlvgg 38 39 vd = vec_promote(d, idx); 40 // CHECK: insertelement <2 x double> poison, double %{{.*}}, i32 %{{.*}} 41 // CHECK-ASM: vlvgg 42 43 vd = vec_insert_and_zero(cptrd); 44 // CHECK: [[ZVEC:%[^ ]+]] = insertelement <2 x double> <double poison, double 0.000000e+00>, double {{.*}}, i64 0 45 // CHECK-ASM: vllezg 46 47 vd = vec_revb(vd); 48 // CHECK-ASM: vperm 49 50 vd = vec_reve(vd); 51 // CHECK-ASM: {{vperm|vpdi}} 52 53 vd = vec_sel(vd, vd, vul); 54 // CHECK-ASM: vsel 55 vd = vec_sel(vd, vd, vbl); 56 // CHECK-ASM: vsel 57 58 vd = vec_gather_element(vd, vul, cptrd, 0); 59 // CHECK-ASM: vgeg %{{.*}}, 0(%{{.*}},%{{.*}}), 0 60 vd = vec_gather_element(vd, vul, cptrd, 1); 61 // CHECK-ASM: vgeg %{{.*}}, 0(%{{.*}},%{{.*}}), 1 62 63 vec_scatter_element(vd, vul, ptrd, 0); 64 // CHECK-ASM: vsceg %{{.*}}, 0(%{{.*}},%{{.*}}), 0 65 vec_scatter_element(vd, vul, ptrd, 1); 66 // CHECK-ASM: vsceg %{{.*}}, 0(%{{.*}},%{{.*}}), 1 67 68 vd = vec_xl(idx, cptrd); 69 // CHECK-ASM-NEXT: lgf %r5, 0(%r3) 70 // CHECK-ASM-NEXT: lg %r13, 0(%r4) 71 // CHECK-ASM-NEXT: vl %v0, 0(%r5,%r13){{$}} 72 // CHECK-ASM-NEXT: vst 73 74 vd = vec_xld2(idx, cptrd); 75 // CHECK-ASM: vst 76 77 vec_xst(vd, idx, ptrd); 78 79 vec_xstd2(vd, idx, ptrd); 80 81 vd = vec_splat(vd, 0); 82 // CHECK: shufflevector <2 x double> %{{.*}}, <2 x double> poison, <2 x i32> zeroinitializer 83 // CHECK-ASM: vrepg 84 vd = vec_splat(vd, 1); 85 // CHECK: shufflevector <2 x double> %{{.*}}, <2 x double> poison, <2 x i32> <i32 1, i32 1> 86 // CHECK-ASM: vrepg 87 88 vd = vec_splats(d); 89 // CHECK: shufflevector <2 x double> %{{.*}}, <2 x double> poison, <2 x i32> zeroinitializer 90 // CHECK-ASM: vlrepg 91 92 vd = vec_mergeh(vd, vd); 93 // shufflevector <2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x i32> <i32 0, i32 2> 94 // CHECK-ASM: vmrhg 95 96 vd = vec_mergel(vd, vd); 97 // shufflevector <2 x double> %{{.*}}, <2 x double> %{{.*}}, <i32 1, i32 3> 98 // CHECK-ASM: vmrlg 99 } 100 101 void test_compare(void) { 102 // CHECK-ASM-LABEL: test_compare 103 104 vbl = vec_cmpeq(vd, vd); 105 // CHECK: call <2 x i1> @llvm.experimental.constrained.fcmp.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !"oeq", metadata !{{.*}}) 106 // CHECK-ASM: vfcedb 107 108 vbl = vec_cmpge(vd, vd); 109 // CHECK: call <2 x i1> @llvm.experimental.constrained.fcmps.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !"oge", metadata !{{.*}}) 110 // CHECK-ASM: kdbr 111 // CHECK-ASM: kdbr 112 // CHECK-ASM: vst 113 114 vbl = vec_cmpgt(vd, vd); 115 // CHECK: call <2 x i1> @llvm.experimental.constrained.fcmps.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !"ogt", metadata !{{.*}}) 116 // CHECK-ASM: kdbr 117 // CHECK-ASM: kdbr 118 // CHECK-ASM: vst 119 120 vbl = vec_cmple(vd, vd); 121 // CHECK: call <2 x i1> @llvm.experimental.constrained.fcmps.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !"ole", metadata !{{.*}}) 122 // CHECK-ASM: kdbr 123 // CHECK-ASM: kdbr 124 // CHECK-ASM: vst 125 126 vbl = vec_cmplt(vd, vd); 127 // CHECK: call <2 x i1> @llvm.experimental.constrained.fcmps.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !"olt", metadata !{{.*}}) 128 // CHECK-ASM: kdbr 129 // CHECK-ASM: kdbr 130 // CHECK-ASM: vst 131 132 idx = vec_all_lt(vd, vd); 133 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 134 // CHECK-ASM: vfchdbs 135 136 idx = vec_all_nge(vd, vd); 137 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 138 // CHECK-ASM: vfchedbs 139 idx = vec_all_ngt(vd, vd); 140 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 141 // CHECK-ASM: vfchdbs 142 idx = vec_all_nle(vd, vd); 143 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 144 // CHECK-ASM: vfchedbs 145 idx = vec_all_nlt(vd, vd); 146 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 147 // CHECK-ASM: vfchdbs 148 149 idx = vec_all_nan(vd); 150 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vftcidb(<2 x double> %{{.*}}, i32 15) 151 // CHECK-ASM: vftcidb 152 idx = vec_all_numeric(vd); 153 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vftcidb(<2 x double> %{{.*}}, i32 15) 154 // CHECK-ASM: vftcidb 155 156 idx = vec_any_eq(vd, vd); 157 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfcedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 158 // CHECK-ASM: vfcedbs 159 160 idx = vec_any_ne(vd, vd); 161 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfcedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 162 // CHECK-ASM: vfcedbs 163 164 idx = vec_any_ge(vd, vd); 165 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 166 // CHECK-ASM: vfchedbs 167 168 idx = vec_any_gt(vd, vd); 169 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 170 // CHECK-ASM: vfchdbs 171 172 idx = vec_any_le(vd, vd); 173 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 174 // CHECK-ASM: vfchedbs 175 176 idx = vec_any_lt(vd, vd); 177 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 178 // CHECK-ASM: vfchdbs 179 180 idx = vec_any_nge(vd, vd); 181 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 182 // CHECK-ASM: vfchedbs 183 idx = vec_any_ngt(vd, vd); 184 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 185 // CHECK-ASM: vfchdbs 186 idx = vec_any_nle(vd, vd); 187 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchedbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 188 // CHECK-ASM: vfchedbs 189 idx = vec_any_nlt(vd, vd); 190 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vfchdbs(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 191 // CHECK-ASM: vfchdbs 192 193 idx = vec_any_nan(vd); 194 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vftcidb(<2 x double> %{{.*}}, i32 15) 195 // CHECK-ASM: vftcidb 196 idx = vec_any_numeric(vd); 197 // CHECK: call { <2 x i64>, i32 } @llvm.s390.vftcidb(<2 x double> %{{.*}}, i32 15) 198 // CHECK-ASM: vftcidb 199 } 200 201 void test_float(void) { 202 // CHECK-ASM-LABEL: test_float 203 204 vd = vec_abs(vd); 205 // CHECK: call <2 x double> @llvm.fabs.v2f64(<2 x double> %{{.*}}) 206 // CHECK-ASM: vflpdb 207 208 vd = vec_nabs(vd); 209 // CHECK: [[ABS:%[^ ]+]] = tail call <2 x double> @llvm.fabs.v2f64(<2 x double> %{{.*}}) 210 // CHECK-NEXT: fneg <2 x double> [[ABS]] 211 // CHECK-ASM: vflndb 212 213 vd = vec_madd(vd, vd, vd); 214 // CHECK: call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}}, metadata !{{.*}}) 215 // CHECK-ASM: vfmadb 216 vd = vec_msub(vd, vd, vd); 217 // CHECK: [[NEG:%[^ ]+]] = fneg <2 x double> %{{.*}} 218 // CHECK: call <2 x double> @llvm.experimental.constrained.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> [[NEG]], metadata !{{.*}}) 219 // CHECK-ASM: vfmsdb 220 vd = vec_sqrt(vd); 221 // CHECK: call <2 x double> @llvm.experimental.constrained.sqrt.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 222 // CHECK-ASM: vfsqdb 223 224 vd = vec_ld2f(cptrf); 225 // CHECK: [[VAL:%[^ ]+]] = load <2 x float>, ptr %{{.*}} 226 // CHECK: call <2 x double> @llvm.experimental.constrained.fpext.v2f64.v2f32(<2 x float> [[VAL]], metadata !{{.*}}) 227 // (emulated) 228 vec_st2f(vd, ptrf); 229 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x float> @llvm.experimental.constrained.fptrunc.v2f32.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 230 // CHECK: store <2 x float> [[VAL]], ptr %{{.*}} 231 // (emulated) 232 233 vd = vec_ctd(vsl, 0); 234 // CHECK: call <2 x double> @llvm.experimental.constrained.sitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 235 // (emulated) 236 vd = vec_ctd(vul, 0); 237 // CHECK: call <2 x double> @llvm.experimental.constrained.uitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 238 // (emulated) 239 vd = vec_ctd(vsl, 1); 240 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.sitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 241 // CHECK: call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> [[VAL]], <2 x double> splat (double 5.000000e-01), metadata !{{.*}}) 242 // (emulated) 243 vd = vec_ctd(vul, 1); 244 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.uitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 245 // CHECK: call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> [[VAL]], <2 x double> splat (double 5.000000e-01), metadata !{{.*}}) 246 // (emulated) 247 vd = vec_ctd(vsl, 31); 248 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.sitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 249 // CHECK: call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> [[VAL]], <2 x double> splat (double 0x3E00000000000000), metadata !{{.*}}) 250 // (emulated) 251 vd = vec_ctd(vul, 31); 252 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.uitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 253 // CHECK: call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> [[VAL]], <2 x double> splat (double 0x3E00000000000000), metadata !{{.*}}) 254 // (emulated) 255 256 vsl = vec_ctsl(vd, 0); 257 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptosi.v2i64.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 258 // (emulated) 259 vul = vec_ctul(vd, 0); 260 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptoui.v2i64.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 261 // (emulated) 262 vsl = vec_ctsl(vd, 1); 263 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> {{.*}}, <2 x double> splat (double 2.000000e+00), metadata !{{.*}}) 264 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptosi.v2i64.v2f64(<2 x double> [[VAL]], metadata !{{.*}}) 265 // (emulated) 266 vul = vec_ctul(vd, 1); 267 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %{{.*}}, <2 x double> splat (double 2.000000e+00), metadata !{{.*}}) 268 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptoui.v2i64.v2f64(<2 x double> [[VAL]], metadata !{{.*}}) 269 // (emulated) 270 vsl = vec_ctsl(vd, 31); 271 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %{{.*}}, <2 x double> splat (double 0x41E0000000000000), metadata !{{.*}}) 272 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptosi.v2i64.v2f64(<2 x double> [[VAL]], metadata !{{.*}}) 273 // (emulated) 274 vul = vec_ctul(vd, 31); 275 // CHECK: [[VAL:%[^ ]+]] = tail call <2 x double> @llvm.experimental.constrained.fmul.v2f64(<2 x double> %{{.*}}, <2 x double> splat (double 0x41E0000000000000), metadata !{{.*}}) 276 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptoui.v2i64.v2f64(<2 x double> [[VAL]], metadata !{{.*}}) 277 // (emulated) 278 279 vd = vec_double(vsl); 280 // CHECK: call <2 x double> @llvm.experimental.constrained.sitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 281 // CHECK-ASM: vcdgb 282 vd = vec_double(vul); 283 // CHECK: call <2 x double> @llvm.experimental.constrained.uitofp.v2f64.v2i64(<2 x i64> %{{.*}}, metadata !{{.*}}) 284 // CHECK-ASM: vcdlgb 285 286 vsl = vec_signed(vd); 287 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptosi.v2i64.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 288 // CHECK-ASM: vcgdb 289 vul = vec_unsigned(vd); 290 // CHECK: call <2 x i64> @llvm.experimental.constrained.fptoui.v2i64.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 291 // CHECK-ASM: vclgdb 292 293 vd = vec_roundp(vd); 294 // CHECK: call <2 x double> @llvm.experimental.constrained.ceil.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 295 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 6 296 vd = vec_ceil(vd); 297 // CHECK: call <2 x double> @llvm.experimental.constrained.ceil.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 298 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 6 299 vd = vec_roundm(vd); 300 // CHECK: call <2 x double> @llvm.experimental.constrained.floor.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 301 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 7 302 vd = vec_floor(vd); 303 // CHECK: call <2 x double> @llvm.experimental.constrained.floor.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 304 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 7 305 vd = vec_roundz(vd); 306 // CHECK: call <2 x double> @llvm.experimental.constrained.trunc.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 307 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 5 308 vd = vec_trunc(vd); 309 // CHECK: call <2 x double> @llvm.experimental.constrained.trunc.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 310 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 5 311 vd = vec_roundc(vd); 312 // CHECK: call <2 x double> @llvm.experimental.constrained.nearbyint.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 313 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 4, 0 314 vd = vec_rint(vd); 315 // CHECK: call <2 x double> @llvm.experimental.constrained.rint.v2f64(<2 x double> %{{.*}}, metadata !{{.*}}) 316 // CHECK-ASM: vfidb %{{.*}}, %{{.*}}, 0, 0 317 vd = vec_round(vd); 318 } 319