On 5/25/2011 1:39 PM, Frank Mori Hess wrote:
What version of boost are you referring to?
boost_1_46_1
make_shared used to be slow due to it doing a lot of unnecessary copying of the storage area for the pointee, but that should have been fixed probably a couple years back now. Copying a shared_ptr is fairly slow due to the atomic reference counting, but I would expect a compiler to be able to elide the copy of the make_shared return value.
That optimization doesn't exist and it's worse with the Intel Compiler: $ bin/benchmark.test/intel-linux/release/benchmark make: auto_ptr: 9406481 ns shared_ptr: 24266241 ns block_ptr: 126199145 ns new: auto_ptr: 4637891 ns shared_ptr: 11943452 ns block_ptr: 73166351 ns $ bin/benchmark.test/gcc-4.1.2/release/benchmark make: auto_ptr: 10980716 ns shared_ptr: 15384949 ns block_ptr: 118651908 ns new: auto_ptr: 4608425 ns shared_ptr: 10733654 ns block_ptr: 69710242 ns -Phil