Missing float conversions in codegen-kernels from lbmpy
When using codgen-kernels in waLBerla (e.g. at commit f0a00fb0), some generated kernels contain the number literal
0.5. It is neither surrounded by a conversion (
real_c(0.5)) nor appended with a conversion to float (
0.5f) when building waLBerla with single precision.
This leads to implicit-conversion warnings.
In !491 (closed), this problem is fixed in src/python/walberla_lbm_generation.py. However, the proposed fixed is only valid for the number literal
0.5. While we did not observe any other number literal to be problematic here, a more flexible way of solving this problem might be required in the future.