Currently, when doing SuperPMI asm diffs, a % improvement/regressed number is calculated by jit-analyze using only the functions that contained diffs as the total amount of code. This gives skewed data, as we only generate the asm for functions that have any changes. We should calculate the diff using the total amount of code generated in the replay, whether it contained a diff or not.
category:eng-sys
theme:super-pmi
skill-level:beginner
cost:small
Currently, when doing SuperPMI asm diffs, a % improvement/regressed number is calculated by jit-analyze using only the functions that contained diffs as the total amount of code. This gives skewed data, as we only generate the asm for functions that have any changes. We should calculate the diff using the total amount of code generated in the replay, whether it contained a diff or not.
category:eng-sys
theme:super-pmi
skill-level:beginner
cost:small