Fix:: replace deprecated CUDA_VERSION and handle CUDA 13+ architecture requirements

[Flying-dragon-boxing]

<Copied from Flying-dragon-boxing#24 >

Reminder

• Have you linked an issue with this pull request?
• Have you added adequate unit tests and/or case tests for your pull request?
• Have you noticed possible changes of behavior below or in the linked issue?
• Have you explained the changes of codes in core modules of ESolver, HSolver, ElecState, Hamilt, Operator or Psi? (ignore if not applicable)

Linked Issue

Addresses the CMake CUDA configuration issues discussed in the repository.

Unit Tests and/or Case Tests for my changes

• Existing CUDA CI workflow will validate these changes when CUDA builds are triggered.

What's changed?

Problem: CMake CUDA configuration had three critical bugs:

1. Used undefined CUDA_VERSION variable (deprecated since CMake 3.10)
2. find_package(CUDAToolkit) only called when architectures not manually defined, breaking version detection
3. Default architectures included 60/70 which CUDA 13+ no longer supports, causing build failures

Changes:

• CMakeLists.txt (lines 332-377):

  • Moved find_package(CUDAToolkit REQUIRED) to execute unconditionally when USE_CUDA=ON
  • Replaced CUDA_VERSION with CUDAToolkit_VERSION from FindCUDAToolkit module
  • Added conditional architecture selection:
    • CUDA < 13.0: includes architectures 60, 70, 75, 80, 86, 89, 90 (version-gated)
    • CUDA ≥ 13.0: excludes architectures < 75 (starts empty, appends 75+)
  • Fixed typo: "architechures" → "architectures"

• docs/advanced/acceleration/cuda.md:

  • Updated to reflect version-dependent architecture selection behavior

Behavior changes:

• CUDA 13+ builds will no longer fail due to unsupported architectures 60/70
• C++ standard properly set to 17 for CUDA 13+ (was broken due to undefined variable)
• Users can still override via CMAKE_CUDA_ARCHITECTURES or CUDAARCHS environment variable

Any changes of core modules? (ignore if not applicable)

No changes to core solver modules.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Potential CMake logic bugs with CUDA</issue_title>
<issue_description>### Describe the Code Quality Issue

Implementation in #6777 of CMakeLists.txt doesn't really detect the CUDA version. The CUDA version is a manually-defined variable, making this CXX 17 issue not automatically solved.

  if(CUDA_VERSION VERSION_GREATER_EQUAL "13.0")
    message(STATUS "CUDA ${CUDA_VERSION} detected. Setting CMAKE_CUDA_STANDARD to 17.")
    set_if_higher(CMAKE_CXX_STANDARD 17)
  endif()

Actually when executing find_package(CUDAToolkit REQUIRED), CMake automatically introduces a variable CUDAToolkit_VERSION. Is there any difficulty introducing this variable in your situation?

Also, CUDA 13 has dropped the support for older architectures lower than 75, which is not properly dealt with in our CMake configuration, causing ABACUS with CUDA 13 to refuse to build.

set(CMAKE_CUDA_ARCHITECTURES
        60 # P100
        70 # V100
        # Add your CUDA arch here Check the Compute Capability version of your
        # GPU at: https://en.wikipedia.org/wiki/CUDA#GPUs_supported
    )

Besides, when defining CMAKE_CUDA_ARCHITECTURES, our CMake configuration does NOT find CUDAToolkit anymore. Why?

  if(NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
    find_package(CUDAToolkit REQUIRED)

Additional Context

No response

Task list for Issue attackers (only for developers)

• Identify the specific code file or section with the code quality issue.
• Investigate the issue and determine the root cause.
• Research best practices and potential solutions for the identified issue.
• Refactor the code to improve code quality, following the suggested solution.
• Ensure the refactored code adheres to the project's coding standards.
• Test the refactored code to ensure it functions as expected.
• Update any relevant documentation, if necessary.
• Submit a pull request with the refactored code and a description of the changes made.</issue_description>

Comments on the Issue (you are @copilot in this section)

@Flying-dragon-boxing `CUDA_VERSION` is **deprecated** together with `FindCUDA` since CMake 3.10, we have already ported to `FindCUDAToolkit`, so we should use `CUDAToolkit_VERSION` instead. https://cmake.org/cmake/help/latest/module/FindCUDA.html

• Fixes Potential CMake logic bugs with CUDA #6807

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.