|
| 1 | +// REQUIRES: opencl, opencl_icd |
| 2 | +// |
| 3 | +// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple %s %opencl_lib -o %t.out |
| 4 | +// RUN: env SYCL_DEVICE_FILTER=opencl %CPU_RUN_PLACEHOLDER %t.out |
| 5 | +// RUN: env SYCL_DEVICE_FILTER=opencl %GPU_RUN_PLACEHOLDER %t.out |
| 6 | +// RUN: env SYCL_DEVICE_FILTER=opencl %ACC_RUN_PLACEHOLDER %t.out |
| 7 | + |
| 8 | +#include <sycl/sycl.hpp> |
| 9 | + |
| 10 | +using namespace sycl; |
| 11 | + |
| 12 | +const char *clSource = R"( |
| 13 | + kernel void test(global int *a, local float *b, int n) { |
| 14 | + if (get_local_id(0) == 0) { |
| 15 | + for (int i = 0; i < n; i++) |
| 16 | + b[i] = i; |
| 17 | + } |
| 18 | + barrier(CLK_LOCAL_MEM_FENCE); |
| 19 | +
|
| 20 | + bool ok = true; |
| 21 | + for (int i = 0; i < n; i++) |
| 22 | + ok &= (b[i] == i); |
| 23 | +
|
| 24 | + a[get_global_id(0)] = ok; |
| 25 | + } |
| 26 | +)"; |
| 27 | + |
| 28 | +kernel getKernel(const queue &Q) { |
| 29 | + cl_int status = CL_SUCCESS; |
| 30 | + |
| 31 | + auto device_cl = get_native<backend::opencl>(Q.get_device()); |
| 32 | + auto context_cl = get_native<backend::opencl>(Q.get_context()); |
| 33 | + auto program_cl = |
| 34 | + clCreateProgramWithSource(context_cl, 1, &clSource, nullptr, &status); |
| 35 | + assert(CL_SUCCESS == status); |
| 36 | + status = clBuildProgram(program_cl, 0, nullptr, "", nullptr, nullptr); |
| 37 | + assert(CL_SUCCESS == status); |
| 38 | + auto kernel_cl = clCreateKernel(program_cl, "test", &status); |
| 39 | + assert(CL_SUCCESS == status); |
| 40 | + |
| 41 | + return make_kernel<backend::opencl>(kernel_cl, Q.get_context()); |
| 42 | +} |
| 43 | + |
| 44 | +int main() { |
| 45 | + queue Q; |
| 46 | + auto K = getKernel(Q); |
| 47 | + |
| 48 | + constexpr cl_int N_slm = 256; |
| 49 | + constexpr int N_wg = 32; |
| 50 | + |
| 51 | + cl_int init[N_wg]; |
| 52 | + buffer<cl_int, 1> b(init, N_wg); |
| 53 | + |
| 54 | + Q.submit([&](handler &cgh) { |
| 55 | + auto acc_global = b.get_access<access::mode::write>(cgh); |
| 56 | + local_accessor<float, 1> acc_local(N_slm, cgh); |
| 57 | + |
| 58 | + cgh.set_arg(0, acc_global); |
| 59 | + cgh.set_arg(1, acc_local); |
| 60 | + cgh.set_arg(2, N_slm); |
| 61 | + |
| 62 | + cgh.parallel_for(nd_range<1>(N_wg, 1), K); |
| 63 | + }); |
| 64 | + |
| 65 | + auto acc_global = b.get_access<access::mode::read>(); |
| 66 | + for (int i = 0; i < N_wg; i++) { |
| 67 | + if (acc_global[i] != 1) { |
| 68 | + std::cerr << "Error in WG " << i << std::endl; |
| 69 | + exit(1); |
| 70 | + } |
| 71 | + } |
| 72 | + |
| 73 | + std::cout << "Success" << std::endl; |
| 74 | + return 0; |
| 75 | +} |
0 commit comments