The new fix for cloud-in-cell in #303 required changing the rtol in the spin_temp consistency test to 1e-4. Nominally, this should be machine-precision, though it's probably some rounding errors that are somehow different between the two consistency runs.
What we care about is whether there is any substantive difference between running run_coeval vs. spin_temp directly. We might be able to test this by just running spin_temp twice and seeing what the differences are. This would verify whether it's just random error, or whether its an actual difference between the two functions.
The new fix for cloud-in-cell in #303 required changing the rtol in the spin_temp consistency test to 1e-4. Nominally, this should be machine-precision, though it's probably some rounding errors that are somehow different between the two consistency runs.
What we care about is whether there is any substantive difference between running
run_coeval
vs.spin_temp
directly. We might be able to test this by just runningspin_temp
twice and seeing what the differences are. This would verify whether it's just random error, or whether its an actual difference between the two functions.