Computing aspects

The velocity-stack inversion has been fully automated, implying that a threshold is adaptively computed for each CMP gather independently. The input data are composed of 801 gathers with 60 traces each. In my implementation, I parallelized the computing using OpenMP (this concerns the operators only). On 10 processors of our SGI Power Challenge machine (notoriously slow), for 30 iterations, it took 30 hours to compute the 801 velocity panels. This process may be accelerated by faster machines.