gh-159186: Skip branch-history recording in builds without Tier 2

[lgeiger]

As far as I can tell, branch history is consumed only if Tier 2 interpreter is enabled. This disables branch history recording for builds without Tier 2 which saves a few instructions for the POP_JUMP_IF_{TRUE,FALSE} op codes.

A toy benchmarks on macOS with ./configure --enable-optimizations --with-lto=full --with-tail-call-interp shows a tiny improvement but its very small so this might be noise. In any case, I don't see any downsides to this, so opening the PR even if performance improvement will be minimal.

Benchmark	Baseline	Candidate	pyperf result
Alternating true/false	4.37 ns	4.15 ns	1.05× faster
Shuffled true/false	7.42 ns	7.64 ns	1.03× slower
Always true	4.81 ns	4.61 ns	1.04× faster
Always false	4.73 ns	4.64 ns	1.02× faster
No-branch control	3.18 ns	3.25 ns	1.02× slower

def branch_loop(values):
    result = 0
    for value in values:
        if value:
            result = 1
        else:
            result = 2
    return result


def control_loop(values):
    result = 0
    for value in values:
        result = 1
    return result


if __name__ == "__main__":
    import random
    import pyperf

    runner = pyperf.Runner()
    size = 10000
    mixed_shuffled = [True, False] * (size // 2)
    random.Random(0).shuffle(mixed_shuffled)
    runner.bench_func("branch_mixed", branch_loop, [True, False] * (size // 2), inner_loops=size)
    runner.bench_func("branch_mixed_shuffled", branch_loop, mixed_shuffled, inner_loops=size)
    runner.bench_func("branch_all_true", branch_loop, [True] * size, inner_loops=size)
    runner.bench_func("branch_all_false", branch_loop, [False] * size, inner_loops=size)
    runner.bench_func("no_branch_control", control_loop, mixed_shuffled, inner_loops=size)

• Issue: Skip branch-history recording in builds without Tier 2 #159186