godot/core/templates/paged_array.h
reduz 37c6434e16 Add a paged array template
Used for reusable stacks, or filling arrays from multiple threads efficiently.
2020-12-21 15:06:50 -03:00

350 lines
11 KiB
C++

/*************************************************************************/
/* paged_array.h */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/*************************************************************************/
/* Copyright (c) 2007-2020 Juan Linietsky, Ariel Manzur. */
/* Copyright (c) 2014-2020 Godot Engine contributors (cf. AUTHORS.md). */
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/*************************************************************************/
#ifndef PAGED_ARRAY_H
#define PAGED_ARRAY_H
#include "core/os/memory.h"
#include "core/os/spin_lock.h"
#include "core/typedefs.h"
// PagedArray is used mainly for filling a very large array from multiple threads efficiently and without causing major fragmentation
// PageArrayPool manages central page allocation in a thread safe matter
template <class T>
class PagedArrayPool {
T **page_pool = nullptr;
uint32_t pages_allocated = 0;
uint32_t *available_page_pool = nullptr;
uint32_t pages_available = 0;
uint32_t page_size = 0;
SpinLock spin_lock;
public:
uint32_t alloc_page() {
spin_lock.lock();
if (unlikely(pages_available == 0)) {
uint32_t pages_used = pages_allocated;
pages_allocated++;
page_pool = (T **)memrealloc(page_pool, sizeof(T *) * pages_allocated);
available_page_pool = (uint32_t *)memrealloc(available_page_pool, sizeof(uint32_t) * pages_allocated);
page_pool[pages_used] = (T *)memalloc(sizeof(T) * page_size);
available_page_pool[0] = pages_used;
pages_available++;
}
pages_available--;
uint32_t page = available_page_pool[pages_available];
spin_lock.unlock();
return page;
}
T *get_page(uint32_t p_page_id) {
return page_pool[p_page_id];
}
void free_page(uint32_t p_page_id) {
spin_lock.lock();
available_page_pool[pages_available] = p_page_id;
pages_available++;
spin_lock.unlock();
}
uint32_t get_page_size_shift() const {
return get_shift_from_power_of_2(page_size);
}
uint32_t get_page_size_mask() const {
return page_size - 1;
}
void reset() {
ERR_FAIL_COND(pages_available < pages_allocated);
if (pages_allocated) {
for (uint32_t i = 0; i < pages_allocated; i++) {
memfree(page_pool[i]);
}
memfree(page_pool);
memfree(available_page_pool);
page_pool = nullptr;
available_page_pool = nullptr;
pages_allocated = 0;
pages_available = 0;
}
}
bool is_configured() const {
return page_size > 0;
}
void configure(uint32_t p_page_size) {
ERR_FAIL_COND(page_pool != nullptr); //sanity check
ERR_FAIL_COND(p_page_size == 0);
page_size = nearest_power_of_2_templated(p_page_size);
}
PagedArrayPool(uint32_t p_page_size = 4096) { // power of 2 recommended because of alignment with OS page sizes. Even if element is bigger, its still a multiple and get rounded amount of pages
configure(p_page_size);
}
~PagedArrayPool() {
ERR_FAIL_COND_MSG(pages_available < pages_allocated, "Pages in use exist at exit in PagedArrayPool");
reset();
}
};
// PageArray is a local array that is optimized to grow in place, then be cleared often.
// It does so by allocating pages from a PagedArrayPool.
// It is safe to use multiple PagedArrays from different threads, sharing a single PagedArrayPool
template <class T>
class PagedArray {
PagedArrayPool<T> *page_pool = nullptr;
T **page_data = nullptr;
uint32_t *page_ids = nullptr;
uint32_t max_pages_used = 0;
uint32_t page_size_shift = 0;
uint32_t page_size_mask = 0;
uint64_t count = 0;
_FORCE_INLINE_ uint32_t _get_pages_in_use() const {
if (count == 0) {
return 0;
} else {
return ((count - 1) >> page_size_shift) + 1;
}
}
void _grow_page_array() {
//no more room in the page array to put the new page, make room
if (max_pages_used == 0) {
max_pages_used = 1;
} else {
max_pages_used *= 2; // increase in powers of 2 to keep allocations to minimum
}
page_data = (T **)memrealloc(page_data, sizeof(T *) * max_pages_used);
page_ids = (uint32_t *)memrealloc(page_ids, sizeof(uint32_t) * max_pages_used);
}
public:
_FORCE_INLINE_ const T &operator[](uint64_t p_index) const {
CRASH_BAD_UNSIGNED_INDEX(p_index, count);
uint32_t page = p_index >> page_size_shift;
uint32_t offset = p_index & page_size_mask;
return page_data[page][offset];
}
_FORCE_INLINE_ T &operator[](uint64_t p_index) {
CRASH_BAD_UNSIGNED_INDEX(p_index, count);
uint32_t page = p_index >> page_size_shift;
uint32_t offset = p_index & page_size_mask;
return page_data[page][offset];
}
_FORCE_INLINE_ void push_back(const T &p_value) {
uint32_t remainder = count & page_size_mask;
if (unlikely(remainder == 0)) {
// at 0, so time to request a new page
uint32_t page_count = _get_pages_in_use();
uint32_t new_page_count = page_count + 1;
if (unlikely(new_page_count > max_pages_used)) {
ERR_FAIL_COND(page_pool == nullptr); //sanity check
_grow_page_array(); //keep out of inline
}
uint32_t page_id = page_pool->alloc_page();
page_data[page_count] = page_pool->get_page(page_id);
page_ids[page_count] = page_id;
}
// place the new value
uint32_t page = count >> page_size_shift;
uint32_t offset = count & page_size_mask;
if (!__has_trivial_constructor(T)) {
memnew_placement(&page_data[page][offset], T(p_value));
} else {
page_data[page][offset] = p_value;
}
count++;
}
void clear() {
//destruct if needed
if (!__has_trivial_destructor(T)) {
for (uint64_t i = 0; i < count; i++) {
uint32_t page = i >> page_size_shift;
uint32_t offset = i & page_size_mask;
page_data[page][offset].~T();
}
}
//return the pages to the pagepool, so they can be used by another array eventually
uint32_t pages_used = _get_pages_in_use();
for (uint32_t i = 0; i < pages_used; i++) {
page_pool->free_page(page_ids[i]);
}
count = 0;
//note we leave page_data and page_indices intact for next use. If you really want to clear them call reset()
}
void reset() {
clear();
if (page_data) {
memfree(page_data);
memfree(page_ids);
page_data = nullptr;
page_ids = nullptr;
max_pages_used = 0;
}
}
// This takes the pages from a source array and merges them to this one
// resulting order is undefined, but content is merged very efficiently,
// making it ideal to fill content on several threads to later join it.
void merge_unordered(PagedArray<T> &p_array) {
ERR_FAIL_COND(page_pool != p_array.page_pool);
uint32_t remainder = count & page_size_mask;
T *remainder_page = nullptr;
uint32_t remainder_page_id;
if (remainder > 0) {
uint32_t last_page = _get_pages_in_use() - 1;
remainder_page = page_data[last_page];
remainder_page_id = page_ids[last_page];
}
count -= remainder;
uint32_t src_pages = p_array._get_pages_in_use();
uint32_t page_size = page_size_mask + 1;
for (uint32_t i = 0; i < src_pages; i++) {
uint32_t page_count = _get_pages_in_use();
uint32_t new_page_count = page_count + 1;
if (unlikely(new_page_count > max_pages_used)) {
_grow_page_array(); //keep out of inline
}
page_data[page_count] = p_array.page_data[i];
page_ids[page_count] = p_array.page_ids[i];
if (i == src_pages - 1) {
//last page, only increment with remainder
count += p_array.count & page_size_mask;
} else {
count += page_size;
}
}
p_array.count = 0; //take away the other array pages
//handle the remainder page if exists
if (remainder_page) {
uint32_t new_remainder = count & page_size_mask;
if (new_remainder > 0) {
//must merge old remainder with new remainder
T *dst_page = page_data[_get_pages_in_use() - 1];
uint32_t to_copy = MIN(page_size - new_remainder, remainder);
for (uint32_t i = 0; i < to_copy; i++) {
if (!__has_trivial_constructor(T)) {
memnew_placement(&dst_page[i + new_remainder], T(remainder_page[i + remainder - to_copy]));
} else {
dst_page[i + new_remainder] = remainder_page[i + remainder - to_copy];
}
if (!__has_trivial_destructor(T)) {
remainder_page[i + remainder - to_copy].~T();
}
}
remainder -= to_copy; //subtract what was copied from remainder
count += to_copy; //add what was copied to the count
if (remainder == 0) {
//entire remainder copied, let go of remainder page
page_pool->free_page(remainder_page_id);
remainder_page = nullptr;
}
}
if (remainder > 0) {
//there is still remainder, append it
uint32_t page_count = _get_pages_in_use();
uint32_t new_page_count = page_count + 1;
if (unlikely(new_page_count > max_pages_used)) {
_grow_page_array(); //keep out of inline
}
page_data[page_count] = remainder_page;
page_ids[page_count] = remainder_page_id;
count += remainder;
}
}
}
uint64_t size() const {
return count;
}
void set_page_pool(PagedArrayPool<T> *p_page_pool) {
ERR_FAIL_COND(max_pages_used > 0); //sanity check
page_pool = p_page_pool;
page_size_mask = page_pool->get_page_size_mask();
page_size_shift = page_pool->get_page_size_shift();
}
~PagedArray() {
reset();
}
};
#endif // PAGED_ARRAY_H