SPEC CPU result comparison

15th October 2026 - Normalized report

766.femflow_r

The FemFlow program solves the compressible Navier-Stokes equations with high-order finite element methods. The equations are solved on locally refined meshes and use splitting methods, to treat the hyperbolic terms with a discontinuous Galerkin method on hexahedral elements of a 3D mesh (polynomial degree 4, over-integration on 6 points per direction) explicitly in time with Heun's method, while the parabolic terms are treated implicitly with a Crank-Nicolson scheme. The associated linear system is solved separately for velocity and temperature with a conjugate gradient method preconditioned by the matrix diagonal. The code relies on the deal.II finite element library, in particular its matrix-free infrastructure to quickly compute the underlying integrals with sum factorization.

Compilers:

Methods:

Levels:

Targets:


Other options:

 

AMD EPYC 8635P (zen5 based Sorano)

The values shown as "run-times" are not actual times but normalized times, the biggest obtained time was assigned to be 500 and the others were scaled appropriately



trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
gcc16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
gcc15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
gcc14 GCC 14.4.0 (revision 98859340e0916f004ac78d5fe20549fb100d41aa)
gcc13 GCC 13.5.0 (revision 0bba35ae26a574150f6d1ae55f8ab1f4c2314a56)
icx Intel(R) oneAPI DPC++/C++ Compiler 2026.1.1 (2026.1.1.20260724), ifx (IFX) 2026.1.1 20260724
llvm clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
aocc AMD clang version 17.0.6 (CLANG: AOCC_5.2.0-Build#2035 2026_04_10)
Smaller is better.

AMD EPYC 75F3 (zen3 based high-frequency Milan)

The values shown as "run-times" are not actual times but normalized times, the biggest obtained time was assigned to be 500 and the others were scaled appropriately



trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
gcc16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
gcc15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
icx Intel(R) oneAPI DPC++/C++ Compiler 2026.1.1 (2026.1.1.20260724), ifx (IFX) 2026.1.1 20260724
llvm clang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (https://github.com/llvm/llvm-project.git 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
Smaller is better.

Improvements

Time What Against Machine
20.00% trunk lto O3 x86_64_v3 gcc15 lto O3 x86_64_v3 AMD EPYC 8635P (zen5 based Sorano) Chart Table
14.55% trunk lto Ofast native gcc13 lto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart
14.13% trunk lto O3 x86_64_v3 gcc15 lto O3 x86_64_v3 AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
9.62% trunk lto Ofast native gcc15 lto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
8.82% trunk lto O2 generic gcc13 lto O2 generic AMD EPYC 8635P (zen5 based Sorano) Chart
8.27% trunk lto O2 generic gcc14 lto O2 generic AMD EPYC 8635P (zen5 based Sorano) Chart
8.08% trunk lto O2 generic gcc15 lto O2 generic AMD EPYC 8635P (zen5 based Sorano) Chart Table
6.25% trunk lto Ofast native gcc15 lto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
5.23% trunk lto O2 generic gcc15 lto O2 generic AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
5.00% trunk lto O3 native gcc15 lto O3 native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart
20.25% trunk pgolto O2 generic gcc15 pgolto O2 generic AMD EPYC 8635P (zen5 based Sorano) Chart Table
11.32% trunk pgolto Ofast native gcc15 pgolto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
8.43% trunk pgolto Ofast native gcc15 pgolto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
6.88% trunk pgo O2 generic gcc16 pgo O2 generic AMD EPYC 8635P (zen5 based Sorano) Chart Table
Improvements smaller than 5% are not shown.

Regressions

Time What Against Machine
8.22% trunk lto O3 x86_64_v3 gcc16 lto O3 x86_64_v3 AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
5.26% trunk lto O3 x86_64_v3 gcc16 lto O3 x86_64_v3 AMD EPYC 8635P (zen5 based Sorano) Chart Table
4.95% trunk lto O3 generic gcc16 lto O3 generic AMD EPYC 8635P (zen5 based Sorano) Chart
4.44% trunk lto Ofast native gcc14 lto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart
2.74% trunk lto Ofast native trunk std Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
2.17% trunk lto Ofast native trunk std Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
2.17% trunk lto O3 native gcc16 lto O3 native AMD EPYC 8635P (zen5 based Sorano) Chart
94.52% trunk pgo Ofast native trunk std Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
4.11% trunk pgolto Ofast native trunk std Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Tables: 1, 2
2.17% trunk pgolto Ofast native trunk std Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Tables: 1, 2
Regressions smaller than 2% are not shown.

Wins

Time What Against Machine
17.86% trunk std Ofast native llvm std Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Tables: 1, 2, 3
14.55% trunk lto Ofast native llvm thinlto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
14.55% trunk lto Ofast native aocc lto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
14.55% trunk lto O3 native llvm lto O3 native AMD EPYC 8635P (zen5 based Sorano) Chart
13.21% trunk std Ofast native icx std Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
14.55% trunk pgolto Ofast native aocc pgolto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Tables: 1, 2
14.55% trunk pgolto Ofast native llvm pgothinlto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
Wins smaller than 5% are not shown.
Please note that the optimization levels -O2 and -O3 have different meanings for different compilers despite being called the same.

Defeats

Time What Against Machine
17.19% trunk lto Ofast native icx lto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
17.19% trunk lto Ofast native llvm lto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Tables: 1, 2, 3, 4
15.22% trunk lto O3 generic llvm lto O3 generic AMD EPYC 8635P (zen5 based Sorano) Chart
14.49% trunk lto O3 x86_64_v3 llvm lto O3 x86_64_v3 AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
13.73% trunk lto Ofast x86_64_v3 llvm lto Ofast x86_64_v3 AMD EPYC 8635P (zen5 based Sorano) Chart Table
13.21% trunk lto O3 x86_64_v3 llvm lto O3 x86_64_v3 AMD EPYC 8635P (zen5 based Sorano) Chart Table
11.90% trunk lto Ofast native icx lto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
11.76% trunk lto O3 native llvm lto O3 native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart
10.29% trunk lto Ofast native llvm thinlto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
18.75% trunk pgolto Ofast native icx pgolto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
18.75% trunk pgolto Ofast native llvm pgolto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Tables: 1, 2, 3, 4
15.15% trunk pgolto Ofast native llvm pgothinlto Ofast native AMD EPYC 75F3 (zen3 based high-frequency Milan) Chart Table
11.90% trunk pgolto Ofast native icx pgolto Ofast native AMD EPYC 8635P (zen5 based Sorano) Chart Table
Defeats smaller than 5% are not shown.
Please note that the optimization levels -O2 and -O3 have different meanings for different compilers despite being called the same.