Files
LookingGlass/idd/LGCommon/WCCopy.cpp
Geoffrey McRae 4477ac0de1 [idd] clipboard: optimize write-combined reads
Add a reusable copy path for moving data from write-combined mappings
into normal cacheable memory. Dispatch to MOVNTDQA-based SSE4.1 or AVX2
loads when the processor and OS support them, with a fenced fallback.

Use the optimized path when staging inbound clipboard stream records from
the IVSHMEM mapping.
2026-08-15 16:09:28 +10:00

142 lines
3.9 KiB
C++

/**
* Looking Glass
* Copyright © 2017-2026 The Looking Glass Authors
* https://looking-glass.io
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the Free
* Software Foundation; either version 2 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 General Public License for
* more details.
*
* You should have received a copy of the GNU General Public License along
* with this program; if not, write to the Free Software Foundation, Inc., 59
* Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
#include "WCCopy.h"
#include <immintrin.h>
#include <intrin.h>
#include <stdint.h>
#include <string.h>
namespace WCCopy
{
namespace Detail
{
void CopyAVX2(void * destination, const void * source, size_t size);
}
namespace
{
using CopyFn = void (*)(void *, const void *, size_t);
static constexpr unsigned __int64 AVX_XSTATE_MASK = 0x6U;
void CopyFallback(void * destination, const void * source, size_t size)
{
_mm_mfence();
memcpy(destination, source, size);
_mm_mfence();
}
#if defined(_M_IX86) || defined(_M_X64)
__declspec(noinline) void CopySSE41(
void * destination, const void * source, size_t size)
{
uint8_t * dst = static_cast<uint8_t *>(destination);
const uint8_t * src = static_cast<const uint8_t *>(source);
_mm_mfence();
const size_t prefix =
(16U - (reinterpret_cast<uintptr_t>(src) & 15U)) & 15U;
if (prefix)
{
const size_t copy = prefix < size ? prefix : size;
memcpy(dst, src, copy);
dst += copy;
src += copy;
size -= copy;
}
while (size >= 64U)
{
// MSVC's SSE4.1 intrinsic omits const from its input pointer.
__m128i * input = reinterpret_cast<__m128i *>(
const_cast<uint8_t *>(src));
__m128i * output = reinterpret_cast<__m128i *>(dst);
const __m128i v0 = _mm_stream_load_si128(input + 0);
const __m128i v1 = _mm_stream_load_si128(input + 1);
const __m128i v2 = _mm_stream_load_si128(input + 2);
const __m128i v3 = _mm_stream_load_si128(input + 3);
_mm_storeu_si128(output + 0, v0);
_mm_storeu_si128(output + 1, v1);
_mm_storeu_si128(output + 2, v2);
_mm_storeu_si128(output + 3, v3);
dst += 64U;
src += 64U;
size -= 64U;
}
while (size >= 16U)
{
const __m128i value = _mm_stream_load_si128(
reinterpret_cast<__m128i *>(const_cast<uint8_t *>(src)));
_mm_storeu_si128(reinterpret_cast<__m128i *>(dst), value);
dst += 16U;
src += 16U;
size -= 16U;
}
if (size)
memcpy(dst, src, size);
_mm_mfence();
}
CopyFn SelectCopy()
{
int registers[4] = {};
__cpuid(registers, 0);
const int maximumLeaf = registers[0];
if (maximumLeaf < 1)
return &CopyFallback;
__cpuidex(registers, 1, 0);
const bool sse41 = (registers[2] & (1 << 19)) != 0;
const bool avx = (registers[2] & (1 << 28)) != 0;
const bool osxsave = (registers[2] & (1 << 27)) != 0;
if (maximumLeaf >= 7 && avx && osxsave &&
(_xgetbv(0) & AVX_XSTATE_MASK) == AVX_XSTATE_MASK)
{
__cpuidex(registers, 7, 0);
if ((registers[1] & (1 << 5)) != 0)
return &Detail::CopyAVX2;
}
return sse41 ? &CopySSE41 : &CopyFallback;
}
#else
CopyFn SelectCopy()
{
return &CopyFallback;
}
#endif
}
void Copy(void * destination, const void * source, size_t size)
{
if (!size)
return;
static const CopyFn copy = SelectCopy();
copy(destination, source, size);
}
}