I've been writing ring buffers wrong all these years

Re:I've been writing ring buffers wrong all these years by smallnamespace on Wednesday December 14, 2016 @09:54PM (#13181989)

Can't the CPU just continue execute out-of-order while waiting on the cmov data dependency to finish though?

In which case cmov would be relatively cheap since it isn't blocking execution of other instructions.

[ Parent | on HN ]

Re:I've been writing ring buffers wrong all these years by gpderetta on Thursday December 15, 2016 @05:06AM (#13183563)

It does, but it may run out of non-dependent instructions to execute while waiting for the long pole of the cmov dependency chain to finish.

This used to be a problem on Pentium4 where cmov had high latency (4 cycles or more), but today, IIRC, a register-to-register cmov is only one cycle, so it is safe to use whenever a branch could have a non-trivial misprediction rate.

[ Parent | on HN ]

Re:I've been writing ring buffers wrong all these years by jnordwick on Saturday December 17, 2016 @07:21PM (#13203568)

Just looked it up. Yes, from Broadwell forward a reg to reg cmov has latency 1. On Atom processors it is still 6.

If you decrement your array index you don't even need the cmp instruction. The compiler could probably gen so good code.

[ Parent | on HN ]