[haberman]

You are correct, I should have clarified, subsequent instructions that depend on the result of the cmov cannot execute until the cmov has executed. Whereas subsequent instructions that depend on the result of the branch instruction can be speculatively executed even before the branch conditional has been evaluated.

[bjourne]

True, but independently of whether "cmov rax, ..." or "jnz L; mov rax, ...; L:" is used, subsequent instructions that reads rax needs to stall until rax has been written to (or at least until cmov/mov has executed if bypasses are used).

[NohatCoder]

The difference is that in the case where the condition is false and predicted false the jump variant will not delay if the value being moved into rax is delayed, the cmov variant will. Effectively that value becomes a false dependency.

As best I can tell this case is rare enough that one shouldn't generally be afraid of cmov, and probably compiler authors should consider using it more frequently.

What one shouldn't do is to load values, that are likely in memory or L3, unnecessarily in order to be able to use cmov. It is the case that runs the greatest risk of degrading performance, and it puts extra load on resources that are shared between cores.

[haberman]

There is also the issue of the branch predicate itself. It is always a true dependency, but when is its value actually needed? For cmov, it is needed before dependent instructions can even be executed. For branch instructions, it is only needed before they can be retired. Speculative execution can keep the pipeline full in the meantime.

[bjourne]

Oh, right! I totally forgot about that. I guess it (at least theoretically) could make a big difference in code for the abs function if the noop is the common case and also easily predictable.