1 // REQUIRES: systemz-registered-target 2 // RUN: %clang_cc1 -target-cpu z14 -triple s390x-ibm-linux -flax-vector-conversions=none \ 3 // RUN: -Wall -Wno-unused -Werror -emit-llvm %s -o - | FileCheck %s 4 5 typedef __attribute__((vector_size(16))) signed char vec_schar; 6 typedef __attribute__((vector_size(16))) signed short vec_sshort; 7 typedef __attribute__((vector_size(16))) signed int vec_sint; 8 typedef __attribute__((vector_size(16))) signed long long vec_slong; 9 typedef __attribute__((vector_size(16))) unsigned char vec_uchar; 10 typedef __attribute__((vector_size(16))) unsigned short vec_ushort; 11 typedef __attribute__((vector_size(16))) unsigned int vec_uint; 12 typedef __attribute__((vector_size(16))) unsigned long long vec_ulong; 13 typedef __attribute__((vector_size(16))) double vec_double; 14 typedef __attribute__((vector_size(16))) float vec_float; 15 16 volatile vec_schar vsc; 17 volatile vec_sshort vss; 18 volatile vec_sint vsi; 19 volatile vec_slong vsl; 20 volatile vec_uchar vuc; 21 volatile vec_ushort vus; 22 volatile vec_uint vui; 23 volatile vec_ulong vul; 24 volatile vec_double vd; 25 volatile vec_float vf; 26 27 volatile unsigned int len; 28 const void * volatile cptr; 29 void * volatile ptr; 30 int cc; 31 32 void test_core(void) { 33 vul = __builtin_s390_vbperm(vuc, vuc); 34 // CHECK: call <2 x i64> @llvm.s390.vbperm(<16 x i8> %{{.*}}, <16 x i8> %{{.*}}) 35 36 vsc = __builtin_s390_vlrl(len, cptr); 37 // CHECK: call <16 x i8> @llvm.s390.vlrl(i32 %{{.*}}, i8* %{{.*}}) 38 39 __builtin_s390_vstrl(vsc, len, ptr); 40 // CHECK: call void @llvm.s390.vstrl(<16 x i8> %{{.*}}, i32 %{{.*}}, i8* %{{.*}}) 41 } 42 43 void test_integer(void) { 44 vuc = __builtin_s390_vmslg(vul, vul, vuc, 0); 45 // CHECK: call <16 x i8> @llvm.s390.vmslg(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}, <16 x i8> %{{.*}}, i32 0) 46 vuc = __builtin_s390_vmslg(vul, vul, vuc, 15); 47 // CHECK: call <16 x i8> @llvm.s390.vmslg(<2 x i64> %{{.*}}, <2 x i64> %{{.*}}, <16 x i8> %{{.*}}, i32 15) 48 } 49 50 void test_float(void) { 51 vd = __builtin_s390_vfmaxdb(vd, vd, 4); 52 // CHECK: call <2 x double> @llvm.maxnum.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 53 vd = __builtin_s390_vfmaxdb(vd, vd, 0); 54 // CHECK: call <2 x double> @llvm.s390.vfmaxdb(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 0) 55 vd = __builtin_s390_vfmaxdb(vd, vd, 15); 56 // CHECK: call <2 x double> @llvm.s390.vfmaxdb(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 15) 57 58 vd = __builtin_s390_vfmindb(vd, vd, 4); 59 // CHECK: call <2 x double> @llvm.minnum.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}) 60 vd = __builtin_s390_vfmindb(vd, vd, 0); 61 // CHECK: call <2 x double> @llvm.s390.vfmindb(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 0) 62 vd = __builtin_s390_vfmindb(vd, vd, 15); 63 // CHECK: call <2 x double> @llvm.s390.vfmindb(<2 x double> %{{.*}}, <2 x double> %{{.*}}, i32 15) 64 65 vd = __builtin_s390_vfnmadb(vd, vd, vd); 66 // CHECK: [[RES:%[^ ]+]] = call <2 x double> @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> %{{.*}}) 67 // CHECK: fneg <2 x double> [[RES]] 68 vd = __builtin_s390_vfnmsdb(vd, vd, vd); 69 // CHECK: [[NEG:%[^ ]+]] = fneg <2 x double> %{{.*}} 70 // CHECK: [[RES:%[^ ]+]] = call <2 x double> @llvm.fma.v2f64(<2 x double> %{{.*}}, <2 x double> %{{.*}}, <2 x double> [[NEG]]) 71 // CHECK: fneg <2 x double> [[RES]] 72 73 vsi = __builtin_s390_vfcesbs(vf, vf, &cc); 74 // CHECK: call { <4 x i32>, i32 } @llvm.s390.vfcesbs(<4 x float> %{{.*}}, <4 x float> %{{.*}}) 75 vsi = __builtin_s390_vfchsbs(vf, vf, &cc); 76 // CHECK: call { <4 x i32>, i32 } @llvm.s390.vfchsbs(<4 x float> %{{.*}}, <4 x float> %{{.*}}) 77 vsi = __builtin_s390_vfchesbs(vf, vf, &cc); 78 // CHECK: call { <4 x i32>, i32 } @llvm.s390.vfchesbs(<4 x float> %{{.*}}, <4 x float> %{{.*}}) 79 80 vsi = __builtin_s390_vftcisb(vf, 0, &cc); 81 // CHECK: call { <4 x i32>, i32 } @llvm.s390.vftcisb(<4 x float> %{{.*}}, i32 0) 82 vsi = __builtin_s390_vftcisb(vf, 4095, &cc); 83 // CHECK: call { <4 x i32>, i32 } @llvm.s390.vftcisb(<4 x float> %{{.*}}, i32 4095) 84 85 vf = __builtin_s390_vfmaxsb(vf, vf, 4); 86 // CHECK: call <4 x float> @llvm.maxnum.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}) 87 vf = __builtin_s390_vfmaxsb(vf, vf, 0); 88 // CHECK: call <4 x float> @llvm.s390.vfmaxsb(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 0) 89 vf = __builtin_s390_vfmaxsb(vf, vf, 15); 90 // CHECK: call <4 x float> @llvm.s390.vfmaxsb(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 15) 91 92 vf = __builtin_s390_vfminsb(vf, vf, 4); 93 // CHECK: call <4 x float> @llvm.minnum.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}) 94 vf = __builtin_s390_vfminsb(vf, vf, 0); 95 // CHECK: call <4 x float> @llvm.s390.vfminsb(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 0) 96 vf = __builtin_s390_vfminsb(vf, vf, 15); 97 // CHECK: call <4 x float> @llvm.s390.vfminsb(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 15) 98 99 vf = __builtin_s390_vfsqsb(vf); 100 // CHECK: call <4 x float> @llvm.sqrt.v4f32(<4 x float> %{{.*}}) 101 102 vf = __builtin_s390_vfmasb(vf, vf, vf); 103 // CHECK: call <4 x float> @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> %{{.*}}) 104 vf = __builtin_s390_vfmssb(vf, vf, vf); 105 // CHECK: [[NEG:%[^ ]+]] = fneg <4 x float> %{{.*}} 106 // CHECK: call <4 x float> @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> [[NEG]]) 107 vf = __builtin_s390_vfnmasb(vf, vf, vf); 108 // CHECK: [[RES:%[^ ]+]] = call <4 x float> @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> %{{.*}}) 109 // CHECK: fneg <4 x float> [[RES]] 110 vf = __builtin_s390_vfnmssb(vf, vf, vf); 111 // CHECK: [[NEG:%[^ ]+]] = fneg <4 x float> %{{.*}} 112 // CHECK: [[RES:%[^ ]+]] = call <4 x float> @llvm.fma.v4f32(<4 x float> %{{.*}}, <4 x float> %{{.*}}, <4 x float> [[NEG]]) 113 // CHECK: fneg <4 x float> [[RES]] 114 115 vf = __builtin_s390_vflpsb(vf); 116 // CHECK: call <4 x float> @llvm.fabs.v4f32(<4 x float> %{{.*}}) 117 vf = __builtin_s390_vflnsb(vf); 118 // CHECK: [[ABS:%[^ ]+]] = call <4 x float> @llvm.fabs.v4f32(<4 x float> %{{.*}}) 119 // CHECK: fneg <4 x float> [[ABS]] 120 121 vf = __builtin_s390_vfisb(vf, 0, 0); 122 // CHECK: call <4 x float> @llvm.rint.v4f32(<4 x float> %{{.*}}) 123 vf = __builtin_s390_vfisb(vf, 4, 0); 124 // CHECK: call <4 x float> @llvm.nearbyint.v4f32(<4 x float> %{{.*}}) 125 vf = __builtin_s390_vfisb(vf, 4, 1); 126 // CHECK: call <4 x float> @llvm.round.v4f32(<4 x float> %{{.*}}) 127 vf = __builtin_s390_vfisb(vf, 4, 5); 128 // CHECK: call <4 x float> @llvm.trunc.v4f32(<4 x float> %{{.*}}) 129 vf = __builtin_s390_vfisb(vf, 4, 6); 130 // CHECK: call <4 x float> @llvm.ceil.v4f32(<4 x float> %{{.*}}) 131 vf = __builtin_s390_vfisb(vf, 4, 7); 132 // CHECK: call <4 x float> @llvm.floor.v4f32(<4 x float> %{{.*}}) 133 vf = __builtin_s390_vfisb(vf, 4, 4); 134 // CHECK: call <4 x float> @llvm.s390.vfisb(<4 x float> %{{.*}}, i32 4, i32 4) 135 } 136 137