mirror of
https://github.com/FEX-Emu/FEX.git
synced 2026-10-14 08:00:50 +02:00
When the result of an atomic fetch operation is unused then we can safely convert it to a non-fetching version of the operation. This happens hundreds of times per process as far as I can tell. No idea if this actually helps any hardware, but theoretically it can allow CPUs to not stall waiting for writeback of the atomic operation. Couldn't detect any performance improvements in the various little things I was poking at least. Very trivial to support so add it. Unaligned variants are already handled in our unaligned fault handler since the only difference is the acquire semantic is dropped and the destination register is the zero register.