DPCT3001#

Message#

“<CMake common command>” is used to generate an output file by performing transformation on input file. You may need to update the name of the input and output file, as well as migrate the input file if the input file contains CUDA syntax.

Detailed Help#

“<CMake common command>” is used to generate an output file by performing transformation on input file. You may need to update the name of the input and output file, as well as migrate the input file if the input file contains CUDA syntax.

Suggestions to Fix#

For example, the CMake command configure_file(<input> <output> options) with the “@ONLY” option replaces variables in <input> file that are explicitly marked with “@VAR@” format (see “@FOO_VERSION@” in the example code) with the value of VAR defined in CMake script(see “FOO_VERSION” in the CMake script), and then save out as <output> file.

For example, this original CUDA* code:

1// src/bar.in.cu
2__global__ void hello_kernel() {}
3
4int main() {
5    std::cout << "FOO version: @FOO_VERSION@" << std::endl;
6    hello_kernel<<<1, 1>>>();
7    cudaDeviceSynchronize();
8    return 0;
9}

And the CMake script for the original CUDA* code:

1# src/CMakeLists.txt
2set(FOO_VERSION 1.00)
3
4configure_file(
5    ${CMAKE_SOURCE_DIR}/bar.in.cu
6    ${CMAKE_BINARY_DIR}/bar.cu
7    @ONLY
8)

result in the following migrated SYCL codes:

 1// build/bar.dp.cpp
 2void hello_kernel() {
 3}
 4
 5int main() {
 6    std::cout << "FOO version: 1.00" << std::endl;  // Example replacement
 7    dpct::get_in_order_queue().parallel_for(
 8        sycl::nd_range<3>(sycl::range<3>(1, 1, 1), sycl::range<3>(1, 1, 1)),
 9        [=](sycl::nd_item<3> item_ct1) {
10            hello_kernel();
11        });
12    dpct::get_current_device().queues_wait_and_throw();
13    return 0;
14}

 1# CMakeLists.txt
 2set(FOO_VERSION 1.00)
 3DPCT3001:1: "configure_file" is used to generate an output file by performing
 4transformation on input file. You may need to update the name of the input and
 5output file, as well as migrate the input file if the input file contains CUDA syntax
 6configure_file(
 7    ${CMAKE_SOURCE_DIR}/bar.in.dp.cpp
 8    ${CMAKE_BINARY_DIR}/bar.dp.cpp
 9    @ONLY
10)

which need to be rewritten to:

 1// src/bar.in.dp.cpp
 2void hello_kernel() {
 3}
 4
 5int main() {
 6    std::cout << "FOO version: @FOO_VERSION@" << std::endl;
 7    dpct::get_in_order_queue().parallel_for(
 8        sycl::nd_range<3>(sycl::range<3>(1, 1, 1), sycl::range<3>(1, 1, 1)),
 9        [=](sycl::nd_item<3> item_ct1) {
10            hello_kernel();
11        });
12    dpct::get_current_device().queues_wait_and_throw();
13    return 0;
14}

and:

1# src/CMakeLists.txt
2set(FOO_VERSION 1.00)
3configure_file(
4    ${CMAKE_SOURCE_DIR}/bar.in.dp.cpp
5    ${CMAKE_BINARY_DIR}/bar.dp.cpp
6    @ONLY
7)