BloombergLP::bdlma::ManagedAllocator

Name

Description

bslma::Allocator

This protocol class provides a pure abstract interface and contract for clients and suppliers of raw memory. If the requested memory cannot be returned, the contract requires that an std::bad_alloc exception be thrown. Note that memory is guaranteed to be sufficiently aligned for any object of the requested size on the current platform, which may be less than the maximal alignment guarantee afforded by global operator new.

Name

Description

size_type

Alias for an unsigned integral type capable of representing the number of bytes in this platform's virtual address space.

Name

Description

operator=

allocate

Return a newly allocated block of memory of the specified size bytes.

deallocate

Return the memory block at the specified address to this allocator.

deleteObject

deleteObject overloads

deleteObjectRaw

deleteObjectRaw overloads

is_equal

release [virtual]

Release all memory currently allocated through this allocator. The effect of using a pointer after this call that was obtained from this allocator before this call is undefined.

Name

Description

throwBadAlloc

Throw std::bad_alloc if exceptions are enabled or abort the program otherwise. Derived classes and helper functions will typically call this function when they are unable to satisfy an allocation request. This function never returns.

Name

Description

do_allocate [virtual]

Return a newly allocated block of memory of (at least) the specified positive bytes and having at least the specified alignment. Unless overriden in a derived class, the return value is this‐>allocate(bytes). If this allocator cannot return the requested number of bytes or cannot satisfy the alignment request, then it will throw a std::bad_alloc exception in an exception‐enabled build, or else will abort the program in a non‐exception build. Unless overriden in a derived class, this function will forward the allocation request to the allocate virtual function, padding bytes and adjusting the return value as necessary to ensure sufficient alignment. Note that if bytes is 0, the same non‐null value will be returned every time.

do_deallocate [virtual]

Return the memory block at the specified p address, having the specified bytes and specified alignment, back to this allocator. Unless overriden in a derived class, this function will forward the deallocation request to the deallocate virtual function, padding bytes and adjusting p as necessary to account for alignment values other than the natural alignment for an object of size bytes. The behavior is undefined unless address is a block allocated from this allocator object using the same bytes and alignment and not already deallocated.

do_is_equal [virtual]

Return true if this allocator is equal to the specified other allocator, meaning (at least) that a memory block allocated by one can be deallocated by the other; otherwise return false. Unless overriden, this method returns this == &other.

Name

Description

BufferedSequentialAllocator

This class implements the ManagedAllocator protocol to provide a fast allocator that dispenses heterogeneous blocks of memory (of varying, user‐specified sizes) from an external buffer whose address and size (in bytes) are supplied at construction. If an allocation request exceeds the remaining free memory space in the external buffer, memory will be supplied by an (optional) allocator also supplied at construction; if no allocator is supplied, the currently installed default allocator is used. This class is exception neutral: If memory cannot be allocated, the behavior is defined by the (optional) allocator supplied at construction. Note that in no case will the buffered sequential allocator attempt to deallocate the external buffer.

ConcurrentMultipoolAllocator

This class implements the bdlma::ManagedAllocator protocol to provide a thread‐safe allocator that maintains a configurable number of Pool objects, each dispensing memory blocks of a unique size. The Pool objects are placed in an array, with each successive pool managing memory blocks of size twice that of the previous pool. Each multipool allocation (deallocation) request allocates memory from (returns memory to) the internal pool having the smallest block size not less than the requested size, or, if no pool manages memory blocks of sufficient sized, from a separately managed list of memory blocks. Both the release method and the destructor of a ConcurrentMultipoolAllocator release all memory currently allocated via the object.

MultipoolAllocator

This class implements the bdlma::ManagedAllocator protocol to provide an allocator that maintains a configurable number of bdlma::Pool objects, each dispensing memory blocks of a unique size. The bdlma::Pool objects are placed in an array, with each successive pool managing memory blocks of size twice that of the previous pool. Each multipool allocation (deallocation) request allocates memory from (returns memory to) the internal pool having the smallest block size not less than the requested size, or, if no pool manages memory blocks of sufficient sized, from a separately managed list of memory blocks. Both the release method and the destructor of a bdlma::MultipoolAllocator release all memory currently allocated via the object.

SequentialAllocator

This class implements the ManagedAllocator protocol to provide a fast allocator that dispenses heterogeneous blocks of memory (of varying, user‐specified sizes) from a sequence of dynamically‐allocated buffers. Memory for the internal buffers is supplied by an (optional) allocator supplied at construction; if no allocator is supplied, the currently installed default allocator is used. If an allocation exceeds the remaining free memory space in the current buffer, the allocator replenishes its internal buffer with new memory to satisfy the request. This class is exception neutral: If memory cannot be allocated, the behavior is defined by the (optional) allocator specified at construction.

StackTraceTestAllocator

This class defines a concrete "test" allocator mechanism that implements the bdlma::ManagedAllocator protocol, and provides instrumentation to track the set of all blocks allocated by this allocator that have yet to be freed. At any time it can produce a report about such blocks, listing for each place that any unfreed blocks were allocated * the number of unfreed blocks allocated at that place * the stack trace at that place The allocator will also detect redundant frees of the same block, and frees by the wrong allocator. The client can choose whether such violations are handled by a core dump, or merely a report being written.