[PATCH] Remove redundant loop in unsynchronized_pool_resource code

[Jonathan Wakely]

On 13/11/18 23:19 +0000, Jonathan Wakely wrote:
>On 13/11/18 22:59 +0000, Jonathan Wakely wrote:
>>	* src/c++17/memory_resource.cc (bitset::find_first_unset()): Remove
>>	unused function.
>>	(bitset::get_first_unset()): Remove loop, if there's are unset bits
>>	then _M_next_word refers to the first one and there's no need to loop.
>>	(_Pool::_Pool(size_t, size_t), _Pool::block_size()): Remove dead code.
>
>
>>    size_type get_first_unset() noexcept
>>    {
>>-      for (size_type i = _M_next_word; i < nwords(); ++i)
>>+      if (_M_next_word < nwords())
>>	{
>>-	  const size_type n = std::__countr_one(_M_words[i]);
>>+	  const size_type n = std::__countr_one(_M_words[_M_next_word]);
>>	  if (n < bits_per_word)
>>	    {
>>	      const word bit = word(1) << n;
>>-	      _M_words[i] |= bit;
>>-	      if (i == _M_next_word)
>>+	      _M_words[_M_next_word] |= bit;
>>+	      const size_t res = (_M_next_word * bits_per_word) + n;
>>+	      if (n == (bits_per_word - 1))
>>		update_next_word();
>>-	      return (i * bits_per_word) + n;
>>+	      return res;
>>	    }
>>	}
>>      return size_type(-1);
>
>I'm not sure why, but this version seems to perform measurably worse.
>I'll investigate, and maybe revert the change.

The attached patch restores the previous performance. I'll finish
testing it tomorrow.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: patch.txt
Type: text/x-patch
Size: 948 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/libstdc++/attachments/20181114/0ff91397/attachment.bin>