normal code (no --vector_size) does not support mc_grouped_subproc = True anymore

oliviermattelaer commented 2 years ago

generate p p > mu+ m- output launch

is an example of the issue

oliviermattelaer commented 2 years ago

generate p p > t t~
output
launch

now compiles ... but returns wrong cross-section (334 instead of the normal 504.7 that one gets when setting --vector_size=X (X>1)

valassi commented 2 years ago

Thanks... hm I will stay with th eversion before then for the moment ;-)

oliviermattelaer commented 2 years ago

(quick) Update on the "p p > t t~" that fails with "--vector_size=1" (or no --vector_size) If i do run with "set fixed_scale 100" then I got

Cross-section :   692.6 +- 0.8957 pb

while the LTS version returns

Cross-section :   692.8 +- 1.126 pb

So the issue is (seems to be) related to the alphas running.

dynamical scale with HT/2 is back to normal:

 Cross-section :   505.5 +- 0.6627 pb (versus in the LTS: 505.8 +- 0.8033 pb)

However when using MLM scale this is still off

LO_VECT:  249.2 +- 0.3562 pb
LTS: 505.8 +- 0.8033 pb

Update 3: The issue here is with the pdf (likely the scale used to evaluate the pdf), when setting those PDF to one:

(VECT): Results Last   3 iters: Integral =   0.7582E-02
(LTS): Results Last   3 iters: Integral =   0.7685E-02

PDF evaluate at fixed scale (100)

(VECT):  Results Last   3 iters: Integral =   0.1599E+02
(LTS):  Results Last   3 iters: Integral =   0.1616E+02

Update 4:

Isssue witheCKM scale is fixed but MLM merging is off...

oliviermattelaer commented 2 years ago