Two branch runs with same configurations produce different results

[Zi-Wen Han]

Hello everyone.
I recently conducted some experiments using "branch runs" (RUN_TYPE = branch) based on a set of restart files.
I guess that "branch runs" can reproduce the CTRL bit-by-bit, just like the experiment hadn't stopped. To test this, I conducted two "branch runs" with absolutely same configurations for comparisons. However, the two runs did not show exactly the same results (bit-by-bit), which is beyond my expectations.
Does anyone do why? Thanks in advance!

Han​

 

[slevis]

If everything was the same, including the pe-layout, I would also have expected bit-for-bit same answers.