on.
On 10 October 2016 at 17:12, Jonathan Wakely wrote:
> On 09/10/16 16:23 +0100, Elliot Goodrich wrote:
>>
>> Hi,
>>
>> If we unroll the loop so that we iterate both forwards and backwards,
>> we can take advantage of memory-level parallelism when chasing
>>
30%-45%. Is it worth writing a new overload of
_M_reverse which takes the size of the list?
Tested on x86_64-linux-gnu.
Thanks!
commit 073c569818764c5b2e80cc09e89721f3386e1728
Author: Elliot Goodrich
Date: Sun Oct 9 14:34:59 2016 +0100
2016-10-09 Elliot Goodrich
* libstdc
