Skip to content

Commit

Permalink
reduce memory allocations
Browse files Browse the repository at this point in the history
  • Loading branch information
@maxbachmann
maxbachmann committed Oct 31, 2023
1 parent bed2d6c commit 163fc13
Show file tree
Hide file tree
Showing 2 changed files with 54 additions and 33 deletions.
44 changes: 27 additions & 17 deletions extras/rapidfuzz_amalgamated.hpp
View file
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
// Licensed under the MIT License <http://opensource.org/licenses/MIT>.
// SPDX-License-Identifier: MIT
// RapidFuzz v1.0.2
// Generated: 2023-10-31 11:09:46.332642
// Generated: 2023-10-31 11:32:43.133377
// ----------------------------------------------------------
// This file is an amalgamation of multiple different files.
// You probably shouldn't edit it directly.
Expand Down Expand Up @@ -5822,29 +5822,43 @@ jaro_similarity_simd_long_s2(Range<double*> scores, const detail::BlockPatternMa
assert(block.size() % vecs == 0);
assert(static_cast<size_t>(s2.size()) > sizeof(VecType) * 8);

struct AlignedAlloc {
AlignedAlloc(size_t size)
{
// work around compilation failure in msvc
memory = operator new[](size, std::align_val_t(native_simd<VecType>::alignment));
}

~AlignedAlloc()
{
::operator delete[](memory, std::align_val_t(native_simd<VecType>::alignment));
}

void* memory = nullptr;
};

native_simd<VecType> zero(VecType(0));
native_simd<VecType> one(1);
size_t result_index = 0;

size_t s2_block_count = static_cast<size_t>(detail::ceil_div(s2.size(), sizeof(VecType) * 8));
std::vector<native_simd<VecType>> T_flag;
T_flag.resize(s2_block_count);

std::vector<native_simd<VecType>> counter;
counter.resize(s2_block_count);
AlignedAlloc memory(2 * s2_block_count * sizeof(native_simd<VecType>));

std::vector<std::array<VecType, vec_width>> T_flags;
T_flags.resize(s2_block_count);
native_simd<VecType>* T_flag = static_cast<native_simd<VecType>*>(memory.memory);
// reuse the same memory since counter is only required in the first half of the algorithm while
// T_flags is required in the second half
native_simd<VecType>* counter = static_cast<native_simd<VecType>*>(memory.memory) + s2_block_count;
VecType* T_flags = static_cast<VecType*>(memory.memory) + s2_block_count * vec_width;

for (size_t cur_vec = 0; cur_vec < block.size(); cur_vec += vecs) {
auto s2_cur = s2;
auto bounds = jaro_similarity_prepare_bound_long_s2(s1_lengths + result_index, s2_cur);

native_simd<VecType> P_flag(VecType(0));

std::fill(T_flag.begin(), T_flag.begin() + detail::ceil_div(s2_cur.size(), sizeof(VecType) * 8),
std::fill(T_flag, T_flag + detail::ceil_div(s2_cur.size(), sizeof(VecType) * 8),
native_simd<VecType>(VecType(0)));
std::fill(counter.begin(), counter.begin() + detail::ceil_div(s2_cur.size(), sizeof(VecType) * 8),
std::fill(counter, counter + detail::ceil_div(s2_cur.size(), sizeof(VecType) * 8),
native_simd<VecType>(VecType(1)));

// In case s2 is longer than all of the elements in s1_lengths boundMaskSize
Expand Down Expand Up @@ -5886,11 +5900,7 @@ jaro_similarity_simd_long_s2(Range<double*> scores, const detail::BlockPatternMa
P_flag.store(P_flags.data());

for (size_t i = 0; i < static_cast<size_t>(detail::ceil_div(s2_cur.size(), sizeof(VecType) * 8)); ++i)
{
alignas(alignment) std::array<VecType, vec_width> T_flags_;
T_flag[i].store(T_flags_.data());
T_flags[i] = T_flags_;
}
T_flag[i].store(T_flags + i * vec_width);

for (size_t i = 0; i < vec_width; ++i) {
VecType CommonChars = counts[i];
Expand All @@ -5911,11 +5921,11 @@ jaro_similarity_simd_long_s2(Range<double*> scores, const detail::BlockPatternMa

{
size_t T_word_index = 0;
VecType T_flag_cur = T_flags[T_word_index][i];
VecType T_flag_cur = T_flags[T_word_index * vec_width + i];
while (P_flag_cur) {
while (!T_flag_cur) {
++T_word_index;
T_flag_cur = T_flags[T_word_index][i];
T_flag_cur = T_flags[T_word_index * vec_width + i];
}

VecType PatternFlagMask = blsi(P_flag_cur);
Expand Down
43 changes: 27 additions & 16 deletions rapidfuzz/distance/Jaro_impl.hpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -585,28 +585,43 @@ static inline void jaro_similarity_simd_long_s2(Range<double*> scores, const det
assert(block.size() % vecs == 0);
assert(static_cast<size_t>(s2.size()) > sizeof(VecType)*8);

struct AlignedAlloc
{
AlignedAlloc(size_t size)
{
// work around compilation failure in msvc
memory = operator new[](size, std::align_val_t(native_simd<VecType>::alignment));
}

~AlignedAlloc()
{
::operator delete[] (memory, std::align_val_t(native_simd<VecType>::alignment));
}

void* memory = nullptr;
};

native_simd<VecType> zero(VecType(0));
native_simd<VecType> one(1);
size_t result_index = 0;

size_t s2_block_count = static_cast<size_t>(detail::ceil_div(s2.size(), sizeof(VecType)*8));
std::vector<native_simd<VecType>> T_flag;
T_flag.resize(s2_block_count);

std::vector<native_simd<VecType>> counter;
counter.resize(s2_block_count);
AlignedAlloc memory(2 * s2_block_count * sizeof(native_simd<VecType>));

std::vector<std::array<VecType, vec_width>> T_flags;
T_flags.resize(s2_block_count);
native_simd<VecType>* T_flag = static_cast<native_simd<VecType>*>(memory.memory);
// reuse the same memory since counter is only required in the first half of the algorithm while
// T_flags is required in the second half
native_simd<VecType>* counter = static_cast<native_simd<VecType>*>(memory.memory) + s2_block_count;
VecType* T_flags = static_cast<VecType*>(memory.memory) + s2_block_count * vec_width;

for (size_t cur_vec = 0; cur_vec < block.size(); cur_vec += vecs) {
auto s2_cur = s2;
auto bounds = jaro_similarity_prepare_bound_long_s2(s1_lengths + result_index, s2_cur);

native_simd<VecType> P_flag(VecType(0));

std::fill(T_flag.begin(), T_flag.begin() + detail::ceil_div(s2_cur.size(), sizeof(VecType)*8), native_simd<VecType>(VecType(0)));
std::fill(counter.begin(), counter.begin() + detail::ceil_div(s2_cur.size(), sizeof(VecType)*8), native_simd<VecType>(VecType(1)));
std::fill(T_flag, T_flag + detail::ceil_div(s2_cur.size(), sizeof(VecType)*8), native_simd<VecType>(VecType(0)));
std::fill(counter, counter + detail::ceil_div(s2_cur.size(), sizeof(VecType)*8), native_simd<VecType>(VecType(1)));

// In case s2 is longer than all of the elements in s1_lengths boundMaskSize
// might have all bits set and therefor the condition ((boundMask <= boundMaskSize) & one)
Expand Down Expand Up @@ -649,11 +664,7 @@ static inline void jaro_similarity_simd_long_s2(Range<double*> scores, const det
P_flag.store(P_flags.data());

for(size_t i = 0; i < static_cast<size_t>(detail::ceil_div(s2_cur.size(), sizeof(VecType)*8)); ++i)
{
alignas(alignment) std::array<VecType, vec_width> T_flags_;
T_flag[i].store(T_flags_.data());
T_flags[i] = T_flags_;
}
T_flag[i].store(T_flags + i * vec_width);

for (size_t i = 0; i < vec_width; ++i) {
VecType CommonChars = counts[i];
Expand All @@ -672,13 +683,13 @@ static inline void jaro_similarity_simd_long_s2(Range<double*> scores, const det

{
size_t T_word_index = 0;
VecType T_flag_cur = T_flags[T_word_index][i];
VecType T_flag_cur = T_flags[T_word_index * vec_width + i];
while(P_flag_cur)
{
while(!T_flag_cur)
{
++T_word_index;
T_flag_cur = T_flags[T_word_index][i];
T_flag_cur = T_flags[T_word_index * vec_width + i];
}

VecType PatternFlagMask = blsi(P_flag_cur);
Expand Down

0 comments on commit 163fc13

Please sign in to comment.