Because of of the bug in GTPSA, the output map length must be nn and not nv. Therefore this temporary is unnecessarily big.
After the bug is fixed the affected functions inv!, prep_comp_inv_work_low and prep_inv_work_low should be corrected. Also instead of allocating a new array for the quaternion if nv<4 instead of reusing m1x_low or whatever it was, instead just make m1x_low big enough to reuse quaternion part
Because of of the bug in GTPSA, the output map length must be nn and not nv. Therefore this temporary is unnecessarily big.
After the bug is fixed the affected functions
inv!
,prep_comp_inv_work_low
andprep_inv_work_low
should be corrected. Also instead of allocating a new array for the quaternion ifnv
<4 instead of reusingm1x_low
or whatever it was, instead just make m1x_low big enough to reuse quaternion part