SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: Trunk native -Ofast - standard vs LTO vs LTO+PGO

-Ofast -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 Normal LTO LTO+PGO LTO LTO+PGO
709.cactus_r 100.00% 108.30% 104.53% 100.00% 96.52%
722.palm_r 100.00% 100.00% 96.85% 100.00% 96.85%
731.astcenc_r 100.00% 100.99% 114.80% 100.00% 113.68%
736.ocio_r 100.00% 100.00% 99.14% 100.00% 99.14%
737.gmsh_r 100.00% 98.23% 98.23% 100.00% 100.00%
748.flightdm_r 100.00% 91.95% 80.54% 100.00% 87.59%
749.fotonik3d_r 100.00% 101.32% 112.70% 100.00% 111.23%
765.roms_r 100.00% 99.10% 101.35% 100.00% 102.26%
766.femflow_r 100.00% 102.17% 102.17% 100.00% 100.00%
767.nest_r 100.00% 101.99% 103.13% 100.00% 101.12%
772.marian_r 100.00% 100.00% 122.22% 100.00% 122.22%
782.lbm_r 100.00% 99.04% 96.63% 100.00% 97.57%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc options benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/options.

Non-debug executable sizes

Benchmark Normal LTO LTO+PGO LTO LTO+PGO
709.cactus_r 100.00% 86.73% 61.15% 100.00% 70.51%
722.palm_r 100.00% 89.72% 71.28% 100.00% 79.44%
731.astcenc_r 100.00% 94.16% 49.62% 100.00% 52.69%
736.ocio_r 100.00% 51.12% 33.25% 100.00% 65.04%
737.gmsh_r 100.00% 61.98% 36.24% 100.00% 58.47%
748.flightdm_r 100.00% 69.18% 61.07% 100.00% 88.27%
749.fotonik3d_r 100.00% 97.69% 60.10% 100.00% 61.52%
765.roms_r 100.00% 84.68% 68.29% 100.00% 80.64%
766.femflow_r 100.00% 6.68% 3.80% 100.00% 56.80%
767.nest_r 100.00% 78.51% 50.82% 100.00% 64.73%
772.marian_r 100.00% 68.56% 47.55% 100.00% 69.35%
782.lbm_r 100.00% 76.11% 74.10% 100.00% 97.36%
Smaller binary sizes are always better.

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

Run-times

Benchmark Normal LTO LTO+PGO LTO LTO+PGO
709.cactus_r 100.00% 100.28% 99.43% 100.00% 99.15%
722.palm_r 100.00% 99.48% 96.10% 100.00% 96.61%
731.astcenc_r 100.00% 99.41% 109.09% 100.00% 109.73%
736.ocio_r 100.00% 98.90% 99.45% 100.00% 100.56%
737.gmsh_r 100.00% 97.83% 97.17% 100.00% 99.33%
748.flightdm_r 100.00% 89.52% 79.26% 100.00% 88.54%
749.fotonik3d_r 100.00% 100.78% 109.04% 100.00% 108.21%
765.roms_r 100.00% 99.64% 100.72% 100.00% 101.09%
766.femflow_r 100.00% 102.74% 104.11% 100.00% 101.33%
767.nest_r 100.00% 98.19% 102.58% 100.00% 104.47%
772.marian_r 100.00% 100.00% 106.90% 100.00% 106.90%
782.lbm_r 100.00% 104.93% 100.00% 100.00% 95.30%
The compiler used was GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d).
Naturally, the smaller the run-times, the better.
See also continuous gcc options benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/options.

Non-debug executable sizes

Benchmark Normal LTO LTO+PGO LTO LTO+PGO
709.cactus_r 100.00% 87.17% 61.72% 100.00% 70.81%
722.palm_r 100.00% 89.52% 71.35% 100.00% 79.70%
731.astcenc_r 100.00% 95.30% 46.33% 100.00% 48.62%
736.ocio_r 100.00% 50.50% 33.12% 100.00% 65.58%
737.gmsh_r 100.00% 61.91% 36.59% 100.00% 59.10%
748.flightdm_r 100.00% 69.08% 61.14% 100.00% 88.50%
749.fotonik3d_r 100.00% 97.59% 58.38% 100.00% 59.81%
765.roms_r 100.00% 84.72% 68.11% 100.00% 80.40%
766.femflow_r 100.00% 6.76% 3.92% 100.00% 58.04%
767.nest_r 100.00% 78.80% 51.15% 100.00% 64.91%
772.marian_r 100.00% 68.54% 47.72% 100.00% 69.63%
782.lbm_r 100.00% 76.94% 74.47% 100.00% 96.79%
Smaller binary sizes are always better.