now works on 32 bit arm platforms, such as RPi

[zertyz]

No description provided.

[saierd]

Thank you for your changes, I included them in my copy of the code.

I had a small problem, though. Your check for 32 or 64 bit systems does not handle AArch64 correctly, which leads to segmentation faults on that platform. Since quite some people seem to copy the changes from this pull request, you might want to add this fix.

[saierd]

Should be:

#if __x86_64__ || __ppc64__ || __aarch64__

[pps83]

Suggested change

	#if _WIN32 || _WIN64
	#if SIZE_MAX == UINT64_MAX

same goes for every other place

[ya-solnyshko]

This code:

    int8_t next_size_over(size_t & size) const
    {
        size = std::max(size_t(2), detailv3::next_power_of_two(size));
        return 64 - detailv3::log2(size);
    }
    void commit(int8_t shift)
    {
        this->shift = shift;
    }
    void reset()
    {
        shift = 63;
    }
private:
    int8_t shift = 63;

also shall be adapted for 32 bits, because shift by more than 32 bits is UB