Question about AMD vs INTEL CPU using IFC

Question about AMD vs INTEL CPU using IFC

Brian A.'s picture
Brian A.
Tue, 07/26/2011 - 08:40

If we compile a program that generates mathematical results on an INTEL CPU and a end user runs it on a AMD CPU, could they get different results? Is the floating point calculations different? We have a end user who is getting different results than us. THey are using RedHat Linux Enterprise 6.1 on a AMD Opteron CPU. We have tried openSuse, RedHat Linux Enterprise 6, and Windows 7 on a INTEL CPU, and all three get the same resluts but different than the end user.

RSS Top
2 posts / 0 new
Last post
For more complete information about compiler optimizations, see our Optimization Notice.
TimP (Intel)'s picture
TimP (Intel)
Tue, 07/26/2011 - 08:47

The imf-arch-consistency=true option, first implemented in ifort 12.0.4, is intended to produce more accurate results uniformly across various CPU models. It reduces performance of svml (vectorized math library).
-prec-div -prec-sqrt also avoid generation of code which gives slightly different results on AMD.

Top

Login to leave a comment.