Re: Floating point differences between GCC-7 and GCC-9?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Tue, 16 Mar 2021, David Aldrich via Gcc-help wrote:

> Hi
> 
> We are comparing results from our C++ app on Ubuntu 18.04 (gcc 7.3.0, glibc
> 2.27 ) and Ubuntu 20.04 (gcc 9.3.0, glibc 2.31 ).  We are seeing
> significant differences in floating point calculation results across these
> platforms.
> 
> Were there significant changes to the floating point library in gcc 9?

math library is part of Glibc, not GCC, and yes, it received updates between
2.27 and 2.31.

If you are using non-conservative optimization flags such as -Ofast or
-ffast-math, you can expect different results between GCC 7 and GCC 9 as well.

If you use an FMA-capable CPU and run GCC with -march=native or another flag
that allows use of FMA, you might see different compilers merging
a*b+c to fma(a, b, c) in different places. You can pass -ffp-contract=off
to disallow such contraction.

Alexander



[Index of Archives]     [Linux C Programming]     [Linux Kernel]     [eCos]     [Fedora Development]     [Fedora Announce]     [Autoconf]     [The DWARVES Debugging Tools]     [Yosemite Campsites]     [Yosemite News]     [Linux GCC]

  Powered by Linux