rk_engine/cpp/utils/cmp_memcpy.hpp

// Copyright (C) 2023 RozK
//
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
//
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
// GNU Affero General Public License for more details.
//
// You should have received a copy of the GNU Affero General Public License
// along with this program.  If not, see <http://www.gnu.org/licenses/>.

#ifndef RK_ENGINE_CMP_MEMCPY_H
#define RK_ENGINE_CMP_MEMCPY_H

#include "../types.hpp"

template<typename _small>
[[RK_FAST]]
inline bool _rk_cmp_memcpy_small(
    _small * __restrict dst,
    _small const * __restrict src,
    unsigned count) {
    _small cmp = 0;
    do {
        cmp |= *dst ^ *src;
        *dst++ = *src++;
    } while(--count > 0);
    return (cmp != 0);
}

template<typename _big, typename _small>
[[RK_FAST, RK_FLATTEN]]
inline bool _rk_cmp_memcpy_big(
    _small * const __restrict _dst,
    _small const * const __restrict _src,
    unsigned const _count) {
    unsigned const ratio = sizeof(_big) / sizeof(_small);
    unsigned big_count = _count / ratio;
    unsigned const small_count = _count % ratio;
    _big * dst = reinterpret_cast<_big *>(_dst);
    _big const * src = reinterpret_cast<_big const *>(_src);
    _big cmp = 0;
    do {
        cmp |= *dst ^ *src;
        *dst++ = *src++;
    } while(--big_count > 0);
    bool modified = (cmp != 0);
    if (small_count) {
        modified |= _rk_cmp_memcpy_small<_small>(
            reinterpret_cast<_small *>(dst), reinterpret_cast<_small const *>(src), small_count);
    }
    return modified;
}

#ifdef RK_CMP_MEMCPY_UNALIGNED
#define _rk_count_and_alignment(_t) (count >= (sizeof(_t) / sizeof(_small)))
#else
#define _rk_count_and_alignment(_t) ((count >= (sizeof(_t) / sizeof(_small))) && !(alignment % sizeof(_t)))
#endif

template<typename _small>
[[RK_HOT, RK_FAST, RK_FLATTEN]]
bool rk_cmp_memcpy(
    _small * const __restrict _dst,
    _small const * const __restrict _src,
    unsigned const count) {
#ifndef RK_CMP_MEMCPY_UNALIGNED
    unsigned const alignment = reinterpret_cast<uintptr_t>(_dst) | reinterpret_cast<uintptr_t const>(_src);
#endif
    if (sizeof(_small) < sizeof(rk_ullong)) {
        if (_rk_count_and_alignment(rk_ullong)) {
            return _rk_cmp_memcpy_big<rk_ullong, _small>(_dst, _src, count);
        }
    }
    if (sizeof(_small) < sizeof(rk_ulong)) {
        if (_rk_count_and_alignment(rk_ulong)) {
            return _rk_cmp_memcpy_big<rk_ulong, _small>(_dst, _src, count);
        }
    }
    if (sizeof(_small) < sizeof(rk_uint)) {
        if (_rk_count_and_alignment(rk_uint)) {
            return _rk_cmp_memcpy_big<rk_uint, _small>(_dst, _src, count);
        }
    }
    if (sizeof(_small) < sizeof(rk_ushort)) {
        if (_rk_count_and_alignment(rk_ushort)) {
            return _rk_cmp_memcpy_big<rk_ushort, _small>(_dst, _src, count);
        }
    }
    if (count) {
        return _rk_cmp_memcpy_small<_small>(_dst, _src, count);
    }
    return false;
}

#undef _rk_count_and_alignment

#endif // RK_ENGINE_CMP_MEMCPY_H