Paper P0493 for WG21

Authors: Al Grant ([email protected]), Bronek Kozicki ([email protected]), Tim Northover ([email protected])

Audience: SG1, LEWG

Each rev* branch corresponds to a different revision of this paper. The most recet published revision is 4, which is in branch rev4.

Benchmark results may not be comparable between branches

• bench finds maximum value from a series generated by a PRNG
• D0493D5.md actual editable proposal text (unpublished draft)
• D0493D5.pdf generated with https://github.com/mpark/wg21 (thanks Michael Park !)
• fetch_max.hpp different implementations of atomic_fetch_max_explicit, as measured in benchmarks

Note, the default architecture for benchmarks is not Intel x86_64, as to our knowledge this architecture does not implement atomic fetch_max in hardware; instead it is ARMv8.1 ; to build the benchmarks on Intel you will need to set MARCH env. variable, e.g.

cd bench; MARCH=core2 make