pinam45 / dynamic_bitset Goto Github PK

Could you probably add an own namespace? Otherwise it may be a bit difficult to use in larger projects.

To simplify serialization with json I use on my side a push_back_block:

constexpr void dynamic_bitset<Block, Allocator>::push_back_block(Block blockValue)
{
	m_blocks.push_back(blockValue);
}

Binary serialization used .data();
for json 'for each block' save an uint32 or uint64 based on internal blocktype

The following is a comparison of different bitset implementations: https://cs.up.ac.za/cs/vpieterse/pub/PieterseEtAl_SAICSIT2010.pdf.
It would be interesting to see a similar comparison with this library.

Very cool,

I'm using it on my side with minor modification:

constexpr std::byte* data()
{
	return reinterpret_cast<std::byte*>(&m_blocks[0]);
}
constexpr std::byte const* data() const
{
	return reinterpret_cast<std::byte const*>(&m_blocks[0]);
}

Which is on my side convenient for binary serialization.

I didn't find an api that converts to long long, similar to the standard library's bitset

I wonder if we can make sul::dynamic_bitset compatible with fmt (https://fmt.dev/latest/index.html, https://hackingcpp.com/cpp/libs/fmt.html).
To do so, I think the missing is the begin/end accessing directly one bit with an iterator class (not for blocks) and an constructor with std::initializer_list< bool >.
https://hackingcpp.com/cpp/libs/fmt.html#panel-fold68a

Any thought on that?

Hi, I want to use this library in an openmp program. Since different threads can set bits in a same block, I wonder how I can set a bit atomically without using a mutex?

Binary (assignment) operations require the same size() of their operands (example). While this makes the implementation easy it enforces additional boilerplate code if the user wants to use differently-sized sets.

The implementation will still be easy as the resulting bit set will either have the smaller or the larger of both size()s:

In case of assignment operators I would expect the LHS operand to expand its size() to fit the RHS if required. Of course, the loops must only iterate to the minimum of both sizes.

In case of binary operators it should be sufficient to swap LHS and RHS in order to avoid additional reallocations.

When debugging code that uses dynamic_bitset with gdb I get a segmentation fault when I try to print a bitset.

(gdb) print bitset[0]
$1 = {m_block = @0xd, m_mask = 93825020145982}
(gdb) print bitset
$2 = {Segmentation fault 

Also, how should I read the output of the first print?