Closed atalon-lip6 closed 4 months ago
preliminary experiments:
ctBones (upsampled to 512³)
[DiscreteGradient] Built discrete gradient ..................
24 cores
averages over 5 runs
c++-14: 36.927s
c++-17: 38.7794
So 5% more time :-( I experienced the 10% faster on scalarFieldSmoother, I wonder if it still persists with ctBones.
PR #1006 requires migrating to C++17. I've run again the performance tests and couldn't reproduce the 5% perf loss (on discrete gradient computation, the performances between C++14 and C++17 were pretty much on par). I've checked visually the example data base and everything's OK. let's go!
This changes is prompted by the use of if constexpr expressions in the near future, unsuported with std=c++14.