SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: PGO native -Ofast - Released GCCs vs Trunk

-Ofast -fprofile-use -march=native -mtune=native

Please note that event though the header of this comparison says -Ofast, benchmark 767.nest_r was additionally compiled with -fno-finite-math-only. Furthermore, benchmarks 737.gmsh_r and 748.flightdm_r were compiled with -fno-fast-math, so a big chunk of -Ofast was in fact disabled. (This might put GCC at a disadvantage compared to LLVM-based compilers because the latter defaults to -fno-math-errno and -fno-trapping-math, in future measurements, we will try to enable these flags.)

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark GCC 16 Trunk
709.cactus_r 100.00% 100.00%
722.palm_r 100.00% 94.44%
731.astcenc_r 100.00% 100.00%
736.ocio_r 100.00% 98.02%
737.gmsh_r 100.00% 99.12%
748.flightdm_r 100.00% 99.05%
749.fotonik3d_r 100.00% 98.83%
765.roms_r 100.00% 99.55%
766.femflow_r 100.00% CE
767.nest_r 100.00% 98.96%
772.marian_r 100.00% 76.92%
782.lbm_r 100.00% 101.01%
GCC 15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
GCC 16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
Trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.
See also continuous gcc branch benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/branch.

Non-debug executable sizes

Benchmark GCC 16 Trunk
709.cactus_r 100.00% 100.33%
722.palm_r 100.00% 99.48%
731.astcenc_r 100.00% 99.45%
736.ocio_r 100.00% 100.74%
737.gmsh_r 100.00% 99.74%
748.flightdm_r 100.00% 100.15%
749.fotonik3d_r 100.00% 98.06%
765.roms_r 100.00% 97.44%
766.femflow_r 100.00% CE
767.nest_r 100.00% 100.04%
772.marian_r 100.00% 101.99%
782.lbm_r 100.00% 99.41%
Smaller binary sizes are always better.