cacay / memorypool Goto Github PK

I understand that this repo is no longer being maintained. I was searching through GitHub for C++ memory pool implementations. I believe that this is effectively a copy of yours, but with the license headers removed.

I created an issue on his repo as well.

Hi,
I want to use the pool allocator for thousands of small arrays:

typedef uint16_t elementType;
typedef elementType* arrayP;

typedef elementType array_2[2];
MemoryPool<array_2> mMemPool_2;

arrayP foo;
foo = reinterpret_cast<elementType(*)>(mMemPool_2->newElement());
...
mMemPool_2->deleteElement(reinterpret_cast<array_2(*)>(foo));

(the recast I do because I need always foo-like elements because I put the pointers in a hashmap, I take care of the size of the array myself, i.e. I store the length at[0])

With that I always get an compile error in the deleteElement function. When I comment out the call to the destructor (placement new destructor?) it works.

template <typename T, size_t BlockSize>
inline void MemoryPool<T, BlockSize>::deleteElement(pointer p) {
if (p != nullptr) {
// p->~value_type();
deallocate(p);
}
}

Why is that and is it correct in my special case?
Thanks!
steffen

the compile error (gcc 4.8.1):

MemoryPool.tcc: In instantiation of ‘void MemoryPool<T, BlockSize>::deleteElement(MemoryPool<T, BlockSize>::pointer) [with T = short unsigned int [2]; long unsigned int BlockSize = 1966080ul; MemoryPool<T, BlockSize>::pointer = short unsigned int ()[2]]’:
MinHashEncoder.h:310:84: required from here
MemoryPool.tcc:228:4: error: request for member ‘~MemoryPool<short unsigned int [2], 1966080ul>::value_type’ in ‘ p’, which is of non-class type ‘short unsigned int [2]’
p->~value_type();

When compiling with the highest warning level -Wall -Werror -Wextra I receive these error messages regarding unused variables.

May be in the next release you could add this code in order it does not fail when using the very aggressive compiler flags:

(void) variable-name;

The author apparently does not appreciate any response because old issue entries do not have any comment. I wrote my own memory pool which can be at least compiled in a professional environment...

Hi,

I was reading the testcases and something that piqued my curiosity greatly was a mention that MemoryPool would be really useful with multiple linked lists. I'd really love to see an example; could you write one up? My own knowledge of implementing algorithms would really benefit. :)

Thanks for your time,
Colin

I might be wrong but shouldn't

be

operator delete(reinterpret_cast<void*>(curr), BlockSize);

to delete the whole block?

when i use the padding function from this code,i found that the summary of address and padding mod align bytes cant be zero sometimes.Should it be a problem? I google the padding problem,someone say the padding function should be like this: inline size_t padding(int* p, size_t align) noexcept
{
uintptr_t result = reinterpret_cast<uintptr_t>(p);
return ((align - (result%align)) % align);
}`
I make a few tests,it works

C-11/MemoryPool.h should include type_traits and utility as it depend on them.

hello There's a question I'd like to ask.
template <typename T, size_t BlockSize>
MemoryPool<T, BlockSize>::~MemoryPool()
noexcept
{
slot_pointer_ curr = currentBlock_;
while (curr != nullptr) {
slot_pointer_ prev = curr->next;
operator delete(reinterpret_cast<void*>(curr));
curr = prev;
}
}

currentBlock_ is not operator new(BlockSize) return address. If there are multiple blocks, MemoryPool destroy Unable to release memory.
I think we need to add a linked list record operator new(BlockSize) address