SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: LTO generic -O2 - Released GCCs vs Trunk

-O2 -flto

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
709.cactus_r 100.00% 108.26% 118.88% 100.00% 109.81%
722.palm_r 100.00% 101.54% 100.00% 100.00% 98.49%
731.astcenc_r 100.00% 100.82% 100.41% 100.00% 99.59%
736.ocio_r 100.00% 100.82% 100.82% 100.00% 100.00%
737.gmsh_r 100.00% 99.13% 100.00% 100.00% 100.88%
748.flightdm_r 100.00% 98.21% 100.00% 100.00% 101.82%
749.fotonik3d_r 100.00% 99.30% 100.47% 100.00% 101.17%
765.roms_r 100.00% 98.05% 97.08% 100.00% 99.01%
766.femflow_r 100.00% 92.53% 91.92% 100.00% 99.34%
767.nest_r 100.00% 121.24% 124.19% 100.00% 102.44%
772.marian_r 100.00% 57.43% 58.03% 100.00% 101.05%
782.lbm_r 100.00% 100.89% 100.89% 100.00% 100.00%
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 15 GCC 16 Trunk GCC 16 Trunk
709.cactus_r 100.00% 101.35% 102.32% 100.00% 100.95%
722.palm_r 100.00% 101.85% 101.82% 100.00% 99.97%
731.astcenc_r 100.00% 95.35% 101.02% 100.00% 105.95%
736.ocio_r 100.00% 101.92% 101.52% 100.00% 99.61%
737.gmsh_r 100.00% 104.24% 104.46% 100.00% 100.21%
748.flightdm_r 100.00% 103.07% 99.30% 100.00% 96.34%
749.fotonik3d_r 100.00% 102.03% 99.53% 100.00% 97.54%
765.roms_r 100.00% 101.71% 101.82% 100.00% 100.12%
766.femflow_r 100.00% 101.80% 102.85% 100.00% 101.03%
767.nest_r 100.00% 101.39% 101.97% 100.00% 100.58%
772.marian_r 100.00% 102.50% 102.14% 100.00% 99.64%
782.lbm_r 100.00% 100.00% 99.76% 100.00% 99.76%
Smaller binary sizes are always better.

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

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
709.cactus_r 100.00% 111.36% 114.81% 100.00% 103.10%
722.palm_r 100.00% 101.43% 98.77% 100.00% 97.38%
731.astcenc_r 100.00% 101.02% 101.43% 100.00% 100.40%
736.ocio_r 100.00% 100.00% 101.01% 100.00% 101.01%
737.gmsh_r 100.00% 99.36% 100.00% 100.00% 100.65%
748.flightdm_r 100.00% 101.10% 98.23% 100.00% 97.16%
749.fotonik3d_r 100.00% 98.45% 98.45% 100.00% 100.00%
765.roms_r 100.00% 95.01% 94.18% 100.00% 99.12%
766.femflow_r 100.00% 93.76% 94.77% 100.00% 101.07%
767.nest_r 100.00% 118.52% 121.96% 100.00% 102.90%
772.marian_r 100.00% 73.49% 73.69% 100.00% 100.27%
782.lbm_r 100.00% 101.12% 101.12% 100.00% 100.00%
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 15 GCC 16 Trunk GCC 16 Trunk
709.cactus_r 100.00% 101.35% 102.32% 100.00% 100.95%
722.palm_r 100.00% 101.85% 101.82% 100.00% 99.97%
731.astcenc_r 100.00% 95.35% 101.02% 100.00% 105.95%
736.ocio_r 100.00% 101.92% 101.52% 100.00% 99.61%
737.gmsh_r 100.00% 104.23% 104.46% 100.00% 100.22%
748.flightdm_r 100.00% 103.07% 99.29% 100.00% 96.34%
749.fotonik3d_r 100.00% 102.03% 99.53% 100.00% 97.55%
765.roms_r 100.00% 101.71% 101.82% 100.00% 100.12%
766.femflow_r 100.00% 101.80% 102.85% 100.00% 101.03%
767.nest_r 100.00% 101.39% 101.97% 100.00% 100.58%
772.marian_r 100.00% 102.50% 102.14% 100.00% 99.64%
782.lbm_r 100.00% 100.00% 99.76% 100.00% 99.76%
Smaller binary sizes are always better.