clang/test/CodeGen/SystemZ/builtins-systemz-zvector4.c

   1 // REQUIRES: systemz-registered-target
   2 // RUN: %clang_cc1 -target-cpu z16 -triple s390x-linux-gnu \
   3 // RUN: -O2 -fzvector -flax-vector-conversions=none \
   4 // RUN: -Wall -Wno-unused -Werror -emit-llvm %s -o - | FileCheck %s
   5 // RUN: %clang_cc1 -target-cpu z16 -triple s390x-linux-gnu \
   6 // RUN: -O2 -fzvector -flax-vector-conversions=none \
   7 // RUN: -Wall -Wno-unused -Werror -S %s -o - | FileCheck %s --check-prefix=CHECK-ASM
   8
   9 #include <vecintrin.h>
  10
  11 volatile vector signed char vsc;
  12 volatile vector signed short vss;
  13 volatile vector signed int vsi;
  14 volatile vector signed long long vsl;
  15 volatile vector unsigned char vuc;
  16 volatile vector unsigned short vus;
  17 volatile vector unsigned int vui;
  18 volatile vector unsigned long long vul;
  19 volatile vector bool char vbc;
  20 volatile vector bool short vbs;
  21 volatile vector bool int vbi;
  22 volatile vector bool long long vbl;
  23 volatile vector float vf;
  24 volatile vector double vd;
  25
  26 void test_nnp_assist(void) {
  27   // CHECK-ASM-LABEL: test_nnp_assist
  28
  29   vf = vec_extend_to_fp32_hi(vus, 0);
  30   // CHECK: call <4 x float> @llvm.s390.vclfnhs(<8 x i16> %{{.*}}, i32 0)
  31   // CHECK-ASM: vclfnh
  32   vf = vec_extend_to_fp32_hi(vus, 15);
  33   // CHECK: call <4 x float> @llvm.s390.vclfnhs(<8 x i16> %{{.*}}, i32 15)
  34   // CHECK-ASM: vclfnh
  35
  36   vf = vec_extend_to_fp32_lo(vus, 0);
  37   // CHECK: call <4 x float> @llvm.s390.vclfnls(<8 x i16> %{{.*}}, i32 0)
  38   // CHECK-ASM: vclfnl
  39   vf = vec_extend_to_fp32_lo(vus, 15);
  40   // CHECK: call <4 x float> @llvm.s390.vclfnls(<8 x i16> %{{.*}}, i32 15)
  41   // CHECK-ASM: vclfnl
  42
  43   vus = vec_round_from_fp32(vf, vf, 0);
  44   // CHECK: call <8 x i16> @llvm.s390.vcrnfs(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 0)
  45   // CHECK-ASM: vcrnf
  46   vus = vec_round_from_fp32(vf, vf, 15);
  47   // CHECK: call <8 x i16> @llvm.s390.vcrnfs(<4 x float> %{{.*}}, <4 x float> %{{.*}}, i32 15)
  48   // CHECK-ASM: vcrnf
  49
  50   vus = vec_convert_to_fp16(vus, 0);
  51   // CHECK: call <8 x i16> @llvm.s390.vcfn(<8 x i16> %{{.*}} i32 0)
  52   // CHECK-ASM: vcfn
  53   vus = vec_convert_to_fp16(vus, 15);
  54   // CHECK: call <8 x i16> @llvm.s390.vcfn(<8 x i16> %{{.*}} i32 15)
  55   // CHECK-ASM: vcfn
  56
  57   vus = vec_convert_from_fp16(vus, 0);
  58   // CHECK: call <8 x i16> @llvm.s390.vcnf(<8 x i16> %{{.*}} i32 0)
  59   // CHECK-ASM: vcnf
  60   vus = vec_convert_from_fp16(vus, 15);
  61   // CHECK: call <8 x i16> @llvm.s390.vcnf(<8 x i16> %{{.*}} i32 15)
  62   // CHECK-ASM: vcnf
  63 }