DPCT Error on matching constructors to variables

[FARSHAD_A_Intel]

Hello,

I wonder the case I am seeing is a DPCT limitation as there is no issue with NVCC. Essentially, I see that DPCT is having issue with complex data types.  In below code snipped, I get an error described as below:

: error: no matching constructor for initialization of 'hpmc::hpmc_counters_t'
hpmc_counters_t result;

Any guidance as how to get this code through DPCT?

================================================

here is the code causing above error:

struct hpmc_counters_t

    {

    unsigned long long int translate_accept_count;      //!< Count of accepted translation moves

    unsigned long long int translate_reject_count;      //!< Count of rejected translation moves

    unsigned long long int rotate_accept_count;         //!< Count of accepted rotation moves

    unsigned long long int rotate_reject_count;         //!< Count of rejected rotation moves

    unsigned long long int overlap_checks;              //!< Count of the number of overlap checks

    unsigned int overlap_err_count;                     //!< Count of the number of times overlap checks encounter errors

    //! Construct a zero set of counters

    hpmc_counters_t()

        {

        translate_accept_count = 0;

        translate_reject_count = 0;

        rotate_accept_count = 0;

        rotate_reject_count = 0;

        overlap_checks = 0;

        overlap_err_count = 0;

        }

    //! Get the translate acceptance

    /*! \returns The ratio of translation moves that are accepted, or 0 if there are no translation moves

    */

    DEVICE double getTranslateAcceptance()

        {

        if (translate_reject_count + translate_accept_count == 0)

            return 0.0;

        else

            return double(translate_accept_count) / double(translate_reject_count + translate_accept_count);

        }

    //! Get the rotate acceptance

    /*! \returns The ratio of rotation moves that are accepted, or 0 if there are no rotation moves

    */

    DEVICE double getRotateAcceptance()

        {

        if (rotate_reject_count + rotate_accept_count == 0)

            return 0.0;

        else

            return double(rotate_accept_count) / double(rotate_reject_count + rotate_accept_count);

        }

    //! Get the number of moves

    /*! \return The total number of moves

    */

    DEVICE unsigned long long int getNMoves()

        {

        return translate_accept_count + translate_reject_count + rotate_accept_count + rotate_reject_count;

        }

    };

//! Take the difference of two sets of counters

DEVICE inline hpmc_counters_t operator-(const hpmc_counters_t& a, const hpmc_counters_t& b)

    {

    hpmc_counters_t result;

    result.translate_accept_count = a.translate_accept_count - b.translate_accept_count;

    result.rotate_accept_count = a.rotate_accept_count - b.rotate_accept_count;

    result.translate_reject_count = a.translate_reject_count - b.translate_reject_count;

    result.rotate_reject_count = a.rotate_reject_count - b.rotate_reject_count;

    result.overlap_checks = a.overlap_checks - b.overlap_checks;

    result.overlap_err_count = a.overlap_err_count - b.overlap_err_count;

    return result;

    }

[RahulV_intel]

Hi,

Could you attach the complete source code of the files that you are trying to migrate?

I will try to reproduce this issue from my end and get back to you.

Thanks,

Rahul

[FARSHAD_A_Intel]

It may not be an easy thing to do due to nested dependencies spreading through multiple libraries. I will try to isolate the issue. But one question: I read somewhere that DPCT is not supporting inline functions. Is that true? Because this is an in-line function here.

[RahulV_intel]

Hi,

Yes, that would be helpful.

For inline functions, can you give it a try by providing this extra flag during DPCT migration: --extra-arg="-std=c++17"

Thanks,

Rahul

[RahulV_intel]

Hi,

Just a quick reminder to attach a small reproducer for your issue.

Thanks,

Rahul

[FARSHAD_A_Intel]

Hi Rahul,

this turned out not to be an easy task. There are more than 20,000 errors I had to navigate through to create a test case. I have tabled this and will come back to it later. you can unfollow this issue for now.

thanks

Farshad.

[RahulV_intel]

Hi,

As per your request, I'll go ahead and close this thread from my end. Post a new question once you have the reproducer ready.

Thanks,

Rahul