Closed yoshikikanemoto closed 5 months ago
tested on one actual system and AutoTester Pipeline #727062
One concern is that unlocking Gil for small operations might waste more CPU cycles. For example, are you sure Init function does enough computation?
One concern is that unlocking Gil for small operations might waste more CPU cycles. For example, are you sure Init function does enough computation?
I checked my profiling result again. KinBody::InitFromTriMesh
was indeed taking time so I am sure. But I think we can assume InitFromBoxes
/ InitFromSpheres
would finish fast. so i removed GIL unlock from there.
One concern is that unlocking Gil for small operations might waste more CPU cycles. For example, are you sure Init function does enough computation?
I checked my profiling result again.
KinBody::InitFromTriMesh
was indeed taking time so I am sure. But I think we can assumeInitFromBoxes
/InitFromSpheres
would finish fast. so i removed GIL unlock from there.
Actually, KinBody::InitFromTriMesh
was taking time on converting Python type to C++ type so unlocking GIL won't make a difference. I'll make another change.
Actually,
KinBody::InitFromTriMesh
was taking time on converting Python type to C++ type so unlocking GIL won't make a difference. I'll make another change.
done @rdiankov
thanks~
so that other python threads can make progresses