The hyperthreading of TSUBAME nodes is "ON", so it's available without customized settings.
For an MPI job submitted to the S or H queue,
t2sub -q S -l select=1:mpiprocs=12 ← not using hyperthreading
t2sub -q S -l select=1:mpiprocs=24 ← using hyperthreading
(Hyperthreading is used if over 12 processes/per is apecified)
For the V queue, the above selection is not possible.
The effect of hyperthreading depends concrete programs.
There is possiboly no effect for high density calculations such as matrix operartions.
(In this case the GPU might be a better choice.)
I think that it should be a valid method to check the situation of hyperthreading by running many short test cases.