[optimization](simd) optimize count_zero_num for ColumnNullable #19124

This commit is contained in:
zclllyybb
2023-04-29 14:50:39 +08:00
committed by GitHub
parent f2b15c03ca
commit d383f1f3d7

View File

@ -87,6 +87,36 @@ inline size_t count_zero_num(const int8_t* __restrict data, const uint8_t* __res
size_t size) {
size_t num = 0;
const int8_t* end = data + size;
#if defined(__SSE2__) && defined(__POPCNT__)
const __m128i zero16 = _mm_setzero_si128();
const int8_t* end64 = data + (size / 64 * 64);
for (; data < end64; data += 64) {
num += __builtin_popcountll(
static_cast<uint64_t>(_mm_movemask_epi8(_mm_or_si128(
_mm_cmpeq_epi8(_mm_loadu_si128(reinterpret_cast<const __m128i*>(data)),
zero16),
_mm_loadu_si128(reinterpret_cast<const __m128i*>(null_map))))) |
(static_cast<uint64_t>(_mm_movemask_epi8(_mm_or_si128(
_mm_cmpeq_epi8(
_mm_loadu_si128(reinterpret_cast<const __m128i*>(data + 16)),
zero16),
_mm_loadu_si128(reinterpret_cast<const __m128i*>(null_map + 16)))))
<< 16u) |
(static_cast<uint64_t>(_mm_movemask_epi8(_mm_or_si128(
_mm_cmpeq_epi8(
_mm_loadu_si128(reinterpret_cast<const __m128i*>(data + 32)),
zero16),
_mm_loadu_si128(reinterpret_cast<const __m128i*>(null_map + 32)))))
<< 32u) |
(static_cast<uint64_t>(_mm_movemask_epi8(_mm_or_si128(
_mm_cmpeq_epi8(
_mm_loadu_si128(reinterpret_cast<const __m128i*>(data + 48)),
zero16),
_mm_loadu_si128(reinterpret_cast<const __m128i*>(null_map + 48)))))
<< 48u));
}
#endif
for (; data < end; ++data, ++null_map) {
num += ((*data == 0) | *null_map);
}