Intel Fortran Compiler Results Difference??? I have an explicit transient dynamics Fortran90/95 program Fma-3D with 204,506 lines of code (count includes comment lines). The program has 151+ verification/regression test cases. PROBLEM: Quite different simualtion results from two different INTEL FORTRAN compilers based on exactly the same FORTRAN code. ====================================================== CASE: Good Results. ====================================================== With Intel's -- Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 18.0.3.210 Build 20180410 Copyright (C) 1985-2018 Intel Corporation. All rights reserved. Problem VP80 gets the results it always has, namely, GOOD Output results for problem VP80 ------------------------------------------------------ FMA-3D Verification Log Date Of Execution: Wed, Feb 20, 2019 10:27:38 AM Problems Requested: short #PASS# Equals Success *FAIL* Equals Failure 80 Beam with a 90 Degree Spiral Twist (Paired 4-Node Plates, KG Form.) File: reference/vp80.cro Line: 963 Value: 1.0588E-02 File: vp80.cro Line: 2009 Value: 1.0588E-02 #PASS#" ------------------------------------------------------ GOOD Computed Results 'snap-shoot.' ------------------------------------------------------ Time T = 2.00010E-02 Time Step = 19464 0 NP X-DISPLACE Y-DISPLACE Z-DISPLACE X-VELOCITY Y-VELOCITY Z-VELOCITY X-ACCELERATE Y-ACCELERATE Z-ACCELERATE NP 37 3.5090E-03 -6.4205E-04 1.0588E-02 -2.5293E-01 6.5876E-02 -6.6740E-01 -1.3692E+03 -1.1300E+03 1.6509E+03 37 38 3.5090E-03 -6.0086E-06 1.0588E-02 -2.5376E-01 9.3550E-04 -6.6778E-01 -1.3620E+03 -4.1982E+00 1.6682E+03 38 39 3.5088E-03 6.3003E-04 1.0589E-02 -2.5510E-01 -6.4011E-02 -6.6756E-01 -1.3304E+03 1.1218E+03 1.6535E+03 39 ------------------------------------------------------ ====================================================== ====================================================== CASE: Bad Results. ====================================================== However, with Intel's new Whiz-Bang Intel(R) Visual Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 19.0.2.190 Build 20190117 Copyright (C) 1985-2019 Intel Corporation. All rights reserved. Problem VP80 gets results it never has before, namely, BAD Output results for problem VP80 ------------------------------------------------------ FMA-3D Verification Log Date Of Execution: Wed, Feb 20, 2019 11:35:07 AM Problems Requested: short #PASS# Equals Success *FAIL* Equals Failure 80 Beam with a 90 Degree Spiral Twist (Paired 4-Node Plates, KG Form.) File: reference/vp80.cro Line: 963 Value: 1.0588E-02 *FAIL* ------------------------------------------------------ BAD Computed Results 'snap-shoot.' ------------------------------------------------------ Time T = 2.00010E-02 Time Step = 19474 0 NP X-DISPLACE Y-DISPLACE Z-DISPLACE X-VELOCITY Y-VELOCITY Z-VELOCITY X-ACCELERATE Y-ACCELERATE Z-ACCELERATE NP 37 3.1320E-01 -1.2398E-01 9.2295E-01 5.5841E+03 -1.7370E+03 3.2168E+03 2.4400E+08 -1.1449E+08 7.6017E+07 37 38 4.4459E-01 -9.2594E-02 9.3223E-01 1.5940E+03 -4.6621E+01 1.5206E+03 4.9748E+07 -1.9775E+07 1.6839E+07 38 39 5.8434E-01 -5.9188E-02 9.4708E-01 -3.9777E+03 1.4123E+02 -8.1344E+02 -2.9865E+08 -2.0920E+07 -1.2048E+08 39 ------------------------------------------------------ ====================================================== Notes: (1) These two calculations are based on the exact same Fortran Coding!! (2) These two calculations are based on the exaxt same input file! (3) These two results evaluations use the exact same awk-script! (4) Start by comparing the Uz displancement column; header "Z-DISPLACE" (5) Compile line for likely routine where difference occurs ==> ifort -nologo -fpp -D_NETLIB -D_CVFNT -compile-only -QxHost -O3 -inline:speed -align:dcommons -check:bounds -extend_source -traceback zzz.f -object:platq.o