sample

sample.cpp

        {
          Halide::Var x("x"), xo("xo"), xi("xi");
          Halide::Func f("f");

          f(x) = x + 777;
          f.bound(x, 0, 16);
          f.split(x, xo, xi, 8).gpu_threads(xi).gpu_blocks(xo);
          f.output_buffer().dim(0).set_bounds(0, 16);

          Halide::Target target  = Halide::get_host_target();
          target.set_feature(Halide::Target::OpenCL);
          // target.set_feature(Halide::Target::UserContext);
          target.set_feature(Halide::Target::NoBoundsQuery);
          target.set_feature(Halide::Target::NoAsserts);
          // target.set_feature(Halide::Target::NoRuntime);
          // f.compile_jit(target);

          Halide::Module m = f.compile_to_module({}, "myfunc", target).resolve_submodules();
          Halide::Internal::LoweredFunc lf = m.get_function_by_name("myfunc");


          Halide::Internal::CodeGen_OpenCL_Dev codegen(target);
          codegen.init_module();

          Halide::Internal::DeviceArgument devArg("_f_buffer", true, Halide::Int(32), 1, 16);

          codegen.add_kernel(lf.body, m.functions()[0].name, {devArg});

          std::vector<char> src = codegen.compile_to_src();
          std::cout << std::string(src.begin(), src.end()) << '\n';
          CV_Assert(false);
        }