Inconsistent result in release/debug build

I found the result for large floating point computation is different in release/debug build on Linux/gcc. Please help how to fix it. My goal is to make it consistent. On Windows/MSVC, the result in release/debug matches well. And Linux's debug build matches with results in Windows/MSVC. But only the release build on Linux is different. Thx

Added by: tomek, 2019 VII 31

Could you show a sample code please?