Function TEST_F

dnn/test/cuda/relayout.cpp:20–38

}  // namespace

TEST_F(CUDA, RELAYOUT_TRANSPOSE) {
    Checker<Relayout> checker(handle_cuda());
    auto run = [&](size_t batch, size_t m, size_t n, size_t c, DType dtype) {
        checker.set_dtype(0, dtype).set_dtype(1, dtype);
        TensorLayout src = {{batch, m, n, c}, dtype};
        src.init_contiguous_stride();
        TensorLayout dst = {{batch, m, n, c}, dtype};
        dst.stride[0] = m * n * c;
        dst.stride[1] = c;
        dst.stride[2] = m * c;
        dst.stride[3] = 1;
        checker.execl({src, dst});
    };
    run(16, 30, 40, 4, dtype::Int8());
    run(16, 20, 10, 4, dtype::Int8());
    run(1, 30, 20, 1, dtype::Int32());
    run(1, 20, 30, 1, dtype::Int32());
    run(1, 11, 21, 1, dtype::Float32());
}

#if MEGDNN_WITH_BENCHMARK
TEST_F(CUDA, BENCHMARK_RELAYOUT_TRANSPOSE) {