I took Kian-Tat Lim's script that computes the expected number of TFLOPS from measurements taken a while back. I replace the parts that I could with new measurements taken in the process of completing OTT1 (
DM-3274). In particular, I took the timing for running SFM and DiffImTask on lsst-dev using HiTS data.
The result is 41.6 TFLOPS. This is an increase from the previous estimate of 38.5, but is still far below the max value of 231 TFLOPS.
I submit that this shows the current pipelines passing this KPM.