I would like to compile a CUDA code with Fused Multiply-Add (FMA) operations disabled. I’m then using
Project -> Properties -> CUDA C/C++ -> Host -> Additional Compiler Options
and typing
-fdma=false
in the Additional Compiler Options line. Is that effective in disabling FMAs?
Thank you very much in advance.
Please see Options for Steering CUDA Compilation. You have a typo in your parameter but otherwise you are correct.