I'm trying to make a memory pool class that allows me to return pointers from within a char* array, and cast them as various pointer. Very very simplified implementation:
class Mempool
{
char* mData=new char[1000];
int mCursor=0;
template <typename vtype>
vtype* New()
{
vtype* aResult=(vtype*)mData[mCursor];
mCursor+=sizeof(vtype);
return aResult;
}
};
(Remember, it's not intended to be dynamic, this is just random access memory that I only need for a short time and then it can all be tossed)
Here's my issue: I know that on some systems (iOS, at least) a float* has to be aligned. I assume the alignment is obtained via alignof(float)... but I can't ever be sure of the alignment of the initial char*'s allocation could be anywhere, because alignof(char) is 1.
What can I do to make sure that in my New() function, my mCursor is advanced to a proper alignment for the requested object?
alignof(std::max_align_t) will tell you the alignment required on your platform for standard functions like malloc(). You should use that if you don't know what alignment requirements the users of your allocator will have.
Simply round sizeof(vtype) up to the nearest multiple of alignof(std::max_align_t) when you increment mCursor.
Or for perhaps a little more efficiency, increment mCursor as you do now, then round it up as needed when the user requests an allocation. When allocating an object of size 2 for example, you know that 2-byte alignment is sufficient, you don't need the full alignof(std::max_align_t). You can use alignof(vtype) for this, to avoid wasted space if users allocate small objects next to each other.
As for this:
I can't ever be sure of the alignment of the initial char*'s allocation could be anywhere, because alignof(char) is 1
You can be sure: new char[N] will always be aligned to at least alignof(std::max_align_t). But my second approach (pad when New is called) does not rely on this assumption.
Related
Referencing this question and answer Memory alignment : how to use alignof / alignas? "Alignment is a restriction on which memory positions a value's first byte can be stored." Is there a portable way in C++ to find out the highest granularity of alignment that is required for instructions not to fault (for example in ARM)? Is alignof(intmax_t) sufficient since its the largest integer primitive type?
Further, how does something in malloc align data to a struct's boundaries? For example if I have a struct like so
alignas(16) struct SomethingElse { ... };
the SomethingElse struct is required to be aligned to a boundary that is a multiple of 16 bytes. Now if I request memory for a struct from malloc like so
SomethingElse* ptr = malloc(sizeof(SomethingElse));
Then what happens when malloc returns a pointer that points to an address like 40? Is that invalid since the pointer to SomethingElse objects must be a multiple of 16?
How to tell the maximum data alignment requirement in C++
There is std::max_align_t which
is a POD type whose alignment requirement is at least as strict (as large) as that of every scalar type.
Vector instructions may use array operands that require a higher alignment than any scalar. I don't know if there is a portable way to find out the the upper bound for that.
To find the alignment requirement for a particular type, you can use alignof.
Further, how does something in malloc align data to a struct's boundaries?
malloc aligns to some boundary that is enough for any scalar. In practice exactly the same as std::max_align_t.
It won't align correctly for "overaligned" types. By overaligned, I mean a type that has higher alignment requirement than std::max_align_t.
Then what happens when malloc returns a pointer that points to an address like 40?
Dereferencing such SomethingElse* would be have undefined behaviour.
Is that invalid since the pointer to SomethingElse objects must be a multiple of 16?
Yes.
how even would one go about fixing that? There is no way to tell malloc the alignment that you want right?
To allocate memory for overaligned types, you need std::aligned_alloc which is to be introduced in C++17.
Until then, you can use platform specific functions like posix_memalign or you can over allocate with std::malloc (allocate the alignment - 1 worth of extra bytes), then use std::align to find the correct boundary. But then you must keep track of both the start of the buffer, and the start of the memory block separately.
You might be looking for std::max_align_t
I know that there is no way in C++ to obtain the size of a dynamically created array, such as:
int* a;
a = new int[n];
What I would like to know is: Why? Did people just forget this in the specification of C++, or is there a technical reason for this?
Isn't the information stored somewhere? After all, the command
delete[] a;
seems to know how much memory it has to release, so it seems to me that delete[] has some way of knowing the size of a.
It's a follow on from the fundamental rule of "don't pay for what you don't need". In your example delete[] a; doesn't need to know the size of the array, because int doesn't have a destructor. If you had written:
std::string* a;
a = new std::string[n];
...
delete [] a;
Then the delete has to call destructors (and needs to know how many to call) - in which case the new has to save that count. However, given it doesn't need to be saved on all occasions, Bjarne decided not to give access to it.
(In hindsight, I think this was a mistake ...)
Even with int of course, something has to know about the size of the allocated memory, but:
Many allocators round up the size to some convenient multiple (say 64 bytes) for alignment and convenience reasons. The allocator knows that a block is 64 bytes long - but it doesn't know whether that is because n was 1 ... or 16.
The C++ run-time library may not have access to the size of the allocated block. If for example, new and delete are using malloc and free under the hood, then the C++ library has no way to know the size of a block returned by malloc. (Usually of course, new and malloc are both part of the same library - but not always.)
One fundamental reason is that there is no difference between a pointer to the first element of a dynamically allocated array of T and a pointer to any other T.
Consider a fictitious function that returns the number of elements a pointer points to.
Let's call it "size".
Sounds really nice, right?
If it weren't for the fact that all pointers are created equal:
char* p = new char[10];
size_t ps = size(p+1); // What?
char a[10] = {0};
size_t as = size(a); // Hmm...
size_t bs = size(a + 1); // Wut?
char i = 0;
size_t is = size(&i); // OK?
You could argue that the first should be 9, the second 10, the third 9, and the last 1, but to accomplish this you need to add a "size tag" on every single object.
A char will require 128 bits of storage (because of alignment) on a 64-bit machine. This is sixteen times more than what is necessary.
(Above, the ten-character array a would require at least 168 bytes.)
This may be convenient, but it's also unacceptably expensive.
You could of course envision a version that is only well-defined if the argument really is a pointer to the first element of a dynamic allocation by the default operator new, but this isn't nearly as useful as one might think.
You are right that some part of the system will have to know something about the size. But getting that information is probably not covered by the API of memory management system (think malloc/free), and the exact size that you requested may not be known, because it may have been rounded up.
You will often find that memory managers will only allocate space in a certain multiple, 64 bytes for example.
So, you may ask for new int[4], i.e. 16 bytes, but the memory manager will allocate 64 bytes for your request. To free this memory it doesn't need to know how much memory you asked for, only that it has allocated you one block of 64 bytes.
The next question may be, can it not store the requested size? This is an added overhead which not everybody is prepared to pay for. An Arduino Uno for example only has 2k of RAM, and in that context 4 bytes for each allocation suddenly becomes significant.
If you need that functionality then you have std::vector (or equivalent), or you have higher-level languages. C/C++ was designed to enable you to work with as little overhead as you choose to make use of, this being one example.
There is a curious case of overloading the operator delete that I found in the form of:
void operator delete[](void *p, size_t size);
The parameter size seems to default to the size (in bytes) of the block of memory to which void *p points. If this is true, it is reasonable to at least hope that it has a value passed by the invocation of operator new and, therefore, would merely need to be divided by sizeof(type) to deliver the number of elements stored in the array.
As for the "why" part of your question, Martin's rule of "don't pay for what you don't need" seems the most logical.
There's no way to know how you are going to use that array.
The allocation size does not necessarily match the element number so you cannot just use the allocation size (even if it was available).
This is a deep flaw in other languages not in C++.
You achieve the functionality you desire with std::vector yet still retain raw access to arrays. Retaining that raw access is critical for any code that actually has to do some work.
Many times you will perform operations on subsets of the array and when you have extra book-keeping built into the language you have to reallocate the sub-arrays and copy the data out to manipulate them with an API that expects a managed array.
Just consider the trite case of sorting the data elements.
If you have managed arrays then you can't use recursion without copying data to create new sub-arrays to pass recursively.
Another example is an FFT which recursively manipulates the data starting with 2x2 "butterflies" and works its way back to the whole array.
To fix the managed array you now need "something else" to patch over this defect and that "something else" is called 'iterators'. (You now have managed arrays but almost never pass them to any functions because you need iterators +90% of the time.)
The size of an array allocated with new[] is not visibly stored anywhere, so you can't access it. And new[] operator doesn't return an array, just a pointer to the array's first element. If you want to know the size of a dynamic array, you must store it manually or use classes from libraries such as std::vector
This is very similar to this question, but the answers don't really answer this, so I thought I'd ask again:
Sometimes I interact with functions that return variable-length structures; for example, FSCTL_GET_RETRIEVAL_POINTERS in Windows returns a variably-sized RETRIEVAL_POINTERS_BUFFER structure.
Using malloc/free is discouraged in C++, and so I was wondering:
What is the "proper" way to allocate variable-length buffers in standard C++ (i.e. no Boost, etc.)?
vector<char> is type-unsafe (and doesn't guarantee anything about alignment, if I understand correctly), new doesn't work with custom-sized allocations, and I can't think of a good substitute. Any ideas?
I would use std::vector<char> buffer(n). There's really no such thing as a variably sized structure in C++, so you have to fake it; throw type safety out the window.
If you like malloc()/free(), you can use
RETRIEVAL_POINTERS_BUFFER* ptr=new char [...appropriate size...];
... do stuff ...
delete[] ptr;
Quotation from the standard regarding alignment (expr.new/10):
For arrays of char and unsigned char, the difference between the
result of the new-expression and the address returned by the
allocation function shall be an integral multiple of the strictest
fundamental alignment requirement (3.11) of any object type whose size
is no greater than the size of the array being created. [ Note:
Because allocation functions are assumed to return pointers to storage
that is appropriately aligned for objects of any type with fundamental
alignment, this constraint on array allocation overhead permits the
common idiom of allocating character arrays into which objects of
other types will later be placed. — end note ]
I don't see any reason why you can't use std::vector<char>:
{
std::vector<char> raii(memory_size);
char* memory = &raii[0];
//Now use `memory` wherever you want
//Maybe, you want to use placement new as:
A *pA = new (memory) A(/*...*/); //assume memory_size >= sizeof(A);
pA->fun();
pA->~A(); //call the destructor, once done!
}//<--- just remember, memory is deallocated here, automatically!
Alright, I understand your alignment problem. It's not that complicated. You can do this:
A *pA = new (&memory[i]) A();
//choose `i` such that `&memory[i]` is multiple of four, or whatever alignment requires
//read the comments..
You may consider using a memory pool and, in the specific case of the RETRIEVAL_POINTERS_BUFFER structure, allocate pool memory amounts in accordance with its definition:
sizeof(DWORD) + sizeof(LARGE_INTEGER)
plus
ExtentCount * sizeof(Extents)
(I am sure you are more familiar with this data structure than I am -- the above is mostly for future readers of your question).
A memory pool boils down to "allocate a bunch of memory, then allocate that memory in small pieces using your own fast allocator".
You can build your own memory pool, but it may be worth looking at Boosts memory pool, which is a pure header (no DLLs!) library. Please note that I have not used the Boost memory pool library, but you did ask about Boost so I thought I'd mention it.
std::vector<char> is just fine. Typically you can call your low-level c-function with a zero-size argument, so you know how much is needed. Then you solve your alignment problem: just allocate more than you need, and offset the start pointer:
Say you want the buffer aligned to 4 bytes, allocate needed size + 4 and add 4 - ((&my_vect[0] - reinterpret_cast<char*>(0)) & 0x3).
Then call your c-function with the requested size and the offsetted pointer.
Ok, lets start from the beginning. Ideal way to return variable-length buffer would be:
MyStruct my_func(int a) { MyStruct s; /* magic here */ return s; }
Unfortunately, this does not work since sizeof(MyStruct) is calculated on compile-time. Anything variable-length just do not fit inside a buffer whose size is calculated on compile-time. The thing to notice that this happens with every variable or type supported by c++, since they all support sizeof. C++ has just one thing that can handle runtime sizes of buffers:
MyStruct *ptr = new MyStruct[count];
So anything that is going to solve this problem is necessarily going to use the array version of new. This includes std::vector and other solutions proposed earlier. Notice that tricks like the placement new to a char array has exactly the same problem with sizeof. Variable-length buffers just needs heap and arrays. There is no way around that restriction, if you want to stay within c++. Further it requires more than one object! This is important. You cannot make variable-length object with c++. It's just impossible.
The nearest one to variable-length object that the c++ provides is "jumping from type to type". Each and every object does not need to be of same type, and you can on runtime manipulate objects of different types. But each part and each complete object still supports sizeof and their sizes are determined on compile-time. Only thing left for programmer is to choose which type you use.
So what's our solution to the problem? How do you create variable-length objects? std::string provides the answer. It needs to have more than one character inside and use the array alternative for heap allocation. But this is all handled by the stdlib and programmer do not need to care. Then you'll have a class that manipulates those std::strings. std::string can do it because it's actually 2 separate memory areas. The sizeof(std::string) does return a memory block whose size can be calculated on compile-time. But the actual variable-length data is in separate memory block allocated by the array version of new.
The array version of new has some restrictions on it's own. sizeof(a[0])==sizeof(a[1]) etc. First allocating an array, and then doing placement new for several objects of different types will go around this limitation.
Usually data is aligned at power of two addresses depending on its size.
How should I align a struct or class with size of 20 bytes or another non-power-of-two size?
I'm creating a custom stack allocator so I guess that the compiler wont align data for me since I'm working with a continuous block of memory.
Some more context:
I have an Allocator class that uses malloc() to allocate a large amount of data.
Then I use void* allocate(U32 size_of_object) method to return the pointer that where I can store whether objects I need to store.
This way all objects are stored in the same region of memory and it will hopefully fit in the cache reducing cache misses.
C++11 has the alignof operator specifically for this purpose. Don't use any of the tricks mentioned in other posts, as they all have edge cases or may fail for certain compiler optimisations. The alignof operator is implemented by the compiler and knows the exact alignment being used.
See this description of c++11's new alignof operator
Although the compiler (or interpreter) normally allocates individual data items on aligned boundaries, data structures often have members with different alignment requirements. To maintain proper alignment the translator normally inserts additional unnamed data members so that each member is properly aligned. In addition the data structure as a whole may be padded with a final unnamed member. This allows each member of an array of structures to be properly aligned. http://en.wikipedia.org/wiki/Data_structure_alignment#Typical_alignment_of_C_structs_on_x86
This says that the compiler takes care of it for you, 99.9% of the time. As for how to force an object to align a specific way, that is compiler specific, and only works in certain circumstances.
MSVC: http://msdn.microsoft.com/en-us/library/83ythb65.aspx
__declspec(align(20))
struct S{ int a, b, c, d; };
//must be less than or equal to 20 bytes
GCC: http://gcc.gnu.org/onlinedocs/gcc-3.4.0/gcc/Type-Attributes.html
struct S{ int a, b, c, d; }
__attribute__ ((aligned (20)));
I don't know of a cross-platform way (including macros!) to do this, but there's probably neat macro somewhere.
Unless you want to access memory directly, or squeeze maximum data in a block of memory you don't worry about alignment -- the compiler takes case of that for you.
Due to the way processor data buses work, what you want to avoid is 'mis-aligned' access. Usually you can read a 32 bit value in a single access from addresses which are multiples of four; if you try to read it from an address that's not such a multiple, the CPU may have to grab it in two or more pieces. So if you're really worrying about things at this level of detail, what you need to be concerned about is not so much the overall struct, as the pieces within it. You'll find that compilers will frequently pad out structures with dummy bytes to ensure aligned access, unless you specifically force them not to with a pragma.
Since you've now added that you actually want to write your own allocator, the answer is straight-forward: Simply ensure that your allocator returns a pointer whose value is a multiple of the requested size. The object's size itself will already come suitably adjusted (via internal padding) so that all member objects themselves are properly aligned, so if you request sizeof(T) bytes, all your allocator needs to do is to return a pointer whose value is divisible by sizeof(T).
If your object does indeed have size 20 (as reported by sizeof), then you have nothing further to worry about. (On a 64-bit platform, the object would probably be padded to 24 bytes.)
Update: In fact, as I only now came to realize, strictly speaking you only need to ensure that the pointer is aligned, recursively, for the largest member of your type. That may be more efficient, but aligning to the size of the entire type is definitely not getting it wrong.
How should I align a struct or class with size of 20 bytes or another non-power-of-two size?
Alignment is CPU-specific, so there is no answer to this question without, at least, knowing the target CPU.
Generally speaking, alignment isn't something that you have to worry about; your compiler will have the rules implemented for you. It does come up once in a while, like when writing an allocator. The classic solution is discussed in The C Programming Language (K&R): use the worst possible alignment. malloc does this, although it's phrased as, "the pointer returned if the allocation succeeds shall be suitably aligned so that it may be assigned to a pointer to any type of object."
The way to do that is to use a union (the elements of a union are all allocated at the union's base address, and the union must therefore be aligned in such a way that each element could exist at that address; i.e., the union's alignment will be the same as the alignment of the element with the strictest rules):
typedef Align long;
union header {
// the inner struct has the important bookeeping info
struct {
unsigned size;
header* next;
} s;
// the align member only exists to make sure header_t's are always allocated
// using the alignment of a long, which is probably the worst alignment
// for the target architecture ("worst" == "strictest," something that meets
// the worst alignment will also meet all better alignment requirements)
Align align;
};
Memory is allocated by creating an array (using somthing like sbrk()) of headers large enough to satisfy the request, plus one additional header element that actually contains the bookkeeping information. If the array is called arry, the bookkeeping information is at arry[0], while the pointer returned points at arry[1] (the next element is meant for walking the free list).
This works, but can lead to wasted space ("In Sun's HotSpot JVM, object storage is aligned to the nearest 64-bit boundary"). I'm aware of a better approach that tries to get a type-specific alignment instead of "the alignment that will work for anything."
Compilers also often have compiler-specific commands. They aren't standard, and they require that you know the correct alignment requirements for the types in question. I would avoid them.
Is there any portable way to determine what the maximum possible alignment for any type is?
For example on x86, SSE instructions require 16-byte alignment, but as far as I'm aware, no instructions require more than that, so any type can be safely stored into a 16-byte aligned buffer.
I need to create a buffer (such as a char array) where I can write objects of arbitrary types, and so I need to be able to rely on the beginning of the buffer to be aligned.
If all else fails, I know that allocating a char array with new is guaranteed to have maximum alignment, but with the TR1/C++0x templates alignment_of and aligned_storage, I am wondering if it would be possible to create the buffer in-place in my buffer class, rather than requiring the extra pointer indirection of a dynamically allocated array.
Ideas?
I realize there are plenty of options for determining the max alignment for a bounded set of types: A union, or just alignment_of from TR1, but my problem is that the set of types is unbounded. I don't know in advance which objects must be stored into the buffer.
In C++11 std::max_align_t defined in header cstddef is a POD type whose alignment requirement is at least as strict (as large) as that of every scalar type.
Using the new alignof operator it would be as simple as alignof(std::max_align_t)
In C++0x, the Align template parameter of std::aligned_storage<Len, Align> has a default argument of "default-alignment," which is defined as (N3225 §20.7.6.6 Table 56):
The value of default-alignment shall be the most stringent alignment requirement for any C++ object type whose size is no greater than Len.
It isn't clear whether SSE types would be considered "C++ object types."
The default argument wasn't part of the TR1 aligned_storage; it was added for C++0x.
Unfortunately ensuring max alignment is a lot tougher than it should be, and there are no guaranteed solutions AFAIK. From the GotW blog (Fast Pimpl article):
union max_align {
short dummy0;
long dummy1;
double dummy2;
long double dummy3;
void* dummy4;
/*...and pointers to functions, pointers to
member functions, pointers to member data,
pointers to classes, eye of newt, ...*/
};
union {
max_align m;
char x_[sizeofx];
};
This isn't guaranteed to be fully
portable, but in practice it's close
enough because there are few or no
systems on which this won't work as
expected.
That's about the closest 'hack' I know for this.
There is another approach that I've used personally for super fast allocation. Note that it is evil, but I work in raytracing fields where speed is one of the greatest measures of quality and we profile code on a daily basis. It involves using a heap allocator with pre-allocated memory that works like the local stack (just increments a pointer on allocation and decrements one on deallocation).
I use it for Pimpls particularly. However, just having the allocator is not enough; for such an allocator to work, we have to assume that memory for a class, Foo, is allocated in a constructor, the same memory is likewise deallocated only in the destructor, and that Foo itself is created on the stack. To make it safe, I needed a function to see if the 'this' pointer of a class is on the local stack to determine if we can use our super fast heap-based stack allocator. For that we had to research OS-specific solutions: I used TIBs and TEBs for Win32/Win64, and my co-workers found solutions for Linux and Mac OS X.
The result, after a week of researching OS-specific methods to detect stack range, alignment requirements, and doing a lot of testing and profiling, was an allocator that could allocate memory in 4 clock cycles according to our tick counter benchmarks as opposed to about 400 cycles for malloc/operator new (our test involved thread contention so malloc is likely to be a bit faster than this in single-threaded cases, perhaps a couple of hundred cycles). We added a per-thread heap stack and detected which thread was being used which increased the time to about 12 cycles, though the client can keep track of the thread allocator to get the 4 cycle allocations. It wiped out memory allocation based hotspots off the map.
While you don't have to go through all that trouble, writing a fast allocator might be easier and more generally applicable (ex: allowing the amount of memory to allocate/deallocate to be determined at runtime) than something like max_align here. max_align is easy enough to use, but if you're after speed for memory allocations (and assuming you've already profiled your code and found hotspots in malloc/free/operator new/delete with major contributors being in code you have control over), writing your own allocator can really make the difference.
Short of some maximally_aligned_t type that all compilers promised faithfully to support for all architectures everywhere, I don't see how this could be solved at compile time. As you say, the set of potential types is unbounded. Is the extra pointer indirection really that big a deal?
Allocating aligned memory is trickier than it looks - see for example Implementation of aligned memory allocation
This is what I'm using. In addition to this, if you're allocating memory then a new()'d array of char with length greater than or equal to max_alignment will be aligned to max_alignment so you can then use indexes into that array to get aligned addresses.
enum {
max_alignment = boost::mpl::deref<
boost::mpl::max_element<
boost::mpl::vector<
boost::mpl::int_<boost::alignment_of<signed char>::value>::type,
boost::mpl::int_<boost::alignment_of<short int>::value>::type,
boost::mpl::int_<boost::alignment_of<int>::value>::type, boost::mpl::int_<boost::alignment_of<long int>::value>::type,
boost::mpl::int_<boost::alignment_of<float>::value>::type,
boost::mpl::int_<boost::alignment_of<double>::value>::type,
boost::mpl::int_<boost::alignment_of<long double>::value>::type,
boost::mpl::int_<boost::alignment_of<void*>::value>::type
>::type
>::type
>::type::value
};
}