Craig Gidney

The simd bit table class *is* templated now, so smaller word sizes could be used for smaller tables, hidden behind the implementation details of a pybind tableau class. But ideally...

I would say it's illegal to call inplace_tramspose on a rectangular table, if there's no longer major axis padding, because it can increase the amount of space required and therefore...

Saw this again https://github.com/quantumlib/Stim/runs/6332576101?check_suite_focus=true ``` ============================= test session starts ============================== platform darwin -- Python 3.9.12, pytest-7.1.2, pluggy-1.0.0 rootdir: /Users/runner/work/Stim/Stim collected 491 items work/Stim/Stim/src/stim/circuit/circuit_gate_target_pybind_test.py ..... [ 1%] .... [ 1%] work/Stim/Stim/src/stim/circuit/circuit_instruction_pybind_test.py...

I think, both times, this failed on osx. This most recent one was OSX. If it happens again make sure to document. 

Saw one on windows this time: ``` ================================== FAILURES =================================== __ test_tableau_simulator_sampling_noisy_gates_agrees_with_cirq_data[gate36] __ gate = stimcirq.TwoQubitAsymmetricDepolarizingChannel((0.0033333333333333335, 0.006666666666666667, 0.01, 0.01333333333333333...66666666666667, 0.03, 0.03333333333333333, 0.03666666666666667, 0.04, 0.043333333333333335, 0.04666666666666667, 0.05)) @pytest.mark.parametrize("gate", ROUND_TRIP_NOISY_GATES) def test_tableau_simulator_sampling_noisy_gates_agrees_with_cirq_data(gate:...

@ChrisPattison No. I suspect it's just that the tests have tolerances that are too tight, but I'm not closing the issue until I actually do the math and work out...

Will fix

Okay. I suppose this will get fixed as a byproduct of rewriting to work natively on the detector graph instead of the detection event graph.

Fixed on dev branch

Fixed in dev branch