Stephen Hines | 176edba | 2014-12-01 14:53:08 -0800 | [diff] [blame] | 1 | // RUN: %clang_cc1 -verify -fopenmp=libiomp5 -ast-print %s | FileCheck %s |
| 2 | // RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -emit-pch -o %t %s |
| 3 | // RUN: %clang_cc1 -fopenmp=libiomp5 -std=c++11 -include-pch %t -fsyntax-only -verify %s -ast-print | FileCheck %s |
| 4 | // expected-no-diagnostics |
| 5 | |
| 6 | #ifndef HEADER |
| 7 | #define HEADER |
| 8 | |
| 9 | void foo() {} |
| 10 | int g_ind = 1; |
| 11 | template<class T, class N> T reduct(T* arr, N num) { |
| 12 | N i; |
| 13 | N ind; |
| 14 | N myind; |
| 15 | T sum = (T)0; |
| 16 | // CHECK: T sum = (T)0; |
| 17 | #pragma omp parallel for simd private(myind, g_ind), linear(ind), aligned(arr) |
| 18 | // CHECK-NEXT: #pragma omp parallel for simd private(myind,g_ind) linear(ind) aligned(arr) |
| 19 | for (i = 0; i < num; ++i) { |
| 20 | myind = ind; |
| 21 | T cur = arr[myind]; |
| 22 | ind += g_ind; |
| 23 | sum += cur; |
| 24 | } |
| 25 | } |
| 26 | |
| 27 | template<class T> struct S { |
| 28 | S(const T &a) |
| 29 | :m_a(a) |
| 30 | {} |
| 31 | T result(T *v) const { |
| 32 | T res; |
| 33 | T val; |
| 34 | T lin = 0; |
| 35 | // CHECK: T res; |
| 36 | // CHECK: T val; |
| 37 | // CHECK: T lin = 0; |
| 38 | #pragma omp parallel for simd private(val) safelen(7) linear(lin : -5) lastprivate(res) |
| 39 | // CHECK-NEXT: #pragma omp parallel for simd private(val) safelen(7) linear(lin: -5) lastprivate(res) |
| 40 | for (T i = 7; i < m_a; ++i) { |
| 41 | val = v[i-7] + m_a; |
| 42 | res = val; |
| 43 | lin -= 5; |
| 44 | } |
| 45 | const T clen = 3; |
| 46 | // CHECK: T clen = 3; |
| 47 | #pragma omp parallel for simd safelen(clen-1) |
| 48 | // CHECK-NEXT: #pragma omp parallel for simd safelen(clen - 1) |
| 49 | for(T i = clen+2; i < 20; ++i) { |
| 50 | // CHECK-NEXT: for (T i = clen + 2; i < 20; ++i) { |
| 51 | v[i] = v[v-clen] + 1; |
| 52 | // CHECK-NEXT: v[i] = v[v - clen] + 1; |
| 53 | } |
| 54 | // CHECK-NEXT: } |
| 55 | return res; |
| 56 | } |
| 57 | ~S() |
| 58 | {} |
| 59 | T m_a; |
| 60 | }; |
| 61 | |
| 62 | template<int LEN> struct S2 { |
| 63 | static void func(int n, float *a, float *b, float *c) { |
| 64 | int k1 = 0, k2 = 0; |
| 65 | #pragma omp parallel for simd safelen(LEN) linear(k1,k2:LEN) aligned(a:LEN) |
| 66 | for(int i = 0; i < n; i++) { |
| 67 | c[i] = a[i] + b[i]; |
| 68 | c[k1] = a[k1] + b[k1]; |
| 69 | c[k2] = a[k2] + b[k2]; |
| 70 | k1 = k1 + LEN; |
| 71 | k2 = k2 + LEN; |
| 72 | } |
| 73 | } |
| 74 | }; |
| 75 | |
| 76 | // S2<4>::func is called below in main. |
| 77 | // CHECK: template <int LEN = 4> struct S2 { |
| 78 | // CHECK-NEXT: static void func(int n, float *a, float *b, float *c) { |
| 79 | // CHECK-NEXT: int k1 = 0, k2 = 0; |
| 80 | // CHECK-NEXT: #pragma omp parallel for simd safelen(4) linear(k1,k2: 4) aligned(a: 4) |
| 81 | // CHECK-NEXT: for (int i = 0; i < n; i++) { |
| 82 | // CHECK-NEXT: c[i] = a[i] + b[i]; |
| 83 | // CHECK-NEXT: c[k1] = a[k1] + b[k1]; |
| 84 | // CHECK-NEXT: c[k2] = a[k2] + b[k2]; |
| 85 | // CHECK-NEXT: k1 = k1 + 4; |
| 86 | // CHECK-NEXT: k2 = k2 + 4; |
| 87 | // CHECK-NEXT: } |
| 88 | // CHECK-NEXT: } |
| 89 | |
| 90 | int main (int argc, char **argv) { |
| 91 | int b = argc, c, d, e, f, g; |
| 92 | int k1=0,k2=0; |
| 93 | static int *a; |
| 94 | // CHECK: static int *a; |
| 95 | #pragma omp parallel for simd |
| 96 | // CHECK-NEXT: #pragma omp parallel for simd |
| 97 | for (int i=0; i < 2; ++i)*a=2; |
| 98 | // CHECK-NEXT: for (int i = 0; i < 2; ++i) |
| 99 | // CHECK-NEXT: *a = 2; |
| 100 | #pragma omp parallel |
| 101 | #pragma omp parallel for simd private(argc, b),lastprivate(d,f) collapse(2) aligned(a : 4) ,firstprivate( g ) |
| 102 | for (int i = 0; i < 10; ++i) |
| 103 | for (int j = 0; j < 10; ++j) {foo(); k1 += 8; k2 += 8;} |
| 104 | // CHECK-NEXT: #pragma omp parallel |
| 105 | // CHECK-NEXT: #pragma omp parallel for simd private(argc,b) lastprivate(d,f) collapse(2) aligned(a: 4) firstprivate(g) |
| 106 | // CHECK-NEXT: for (int i = 0; i < 10; ++i) |
| 107 | // CHECK-NEXT: for (int j = 0; j < 10; ++j) { |
| 108 | // CHECK-NEXT: foo(); |
| 109 | // CHECK-NEXT: k1 += 8; |
| 110 | // CHECK-NEXT: k2 += 8; |
| 111 | // CHECK-NEXT: } |
| 112 | for (int i = 0; i < 10; ++i)foo(); |
| 113 | // CHECK-NEXT: for (int i = 0; i < 10; ++i) |
| 114 | // CHECK-NEXT: foo(); |
| 115 | const int CLEN = 4; |
| 116 | // CHECK-NEXT: const int CLEN = 4; |
| 117 | #pragma omp parallel for simd aligned(a:CLEN) linear(a:CLEN) safelen(CLEN) collapse( 1 ) |
| 118 | // CHECK-NEXT: #pragma omp parallel for simd aligned(a: CLEN) linear(a: CLEN) safelen(CLEN) collapse(1) |
| 119 | for (int i = 0; i < 10; ++i)foo(); |
| 120 | // CHECK-NEXT: for (int i = 0; i < 10; ++i) |
| 121 | // CHECK-NEXT: foo(); |
| 122 | |
| 123 | float arr[16]; |
| 124 | S2<4>::func(0,arr,arr,arr); |
| 125 | return (0); |
| 126 | } |
| 127 | |
| 128 | #endif |