SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 FPrate: LTO -O2 generic - LLVM and GCC

-O2 -flto

Please note that the same "optimization levels" mean different things to different compilers. For example, LLVM and LLVM-based compiler tend to be much more aggressive at -O2, often with the corresponding consequences to size of the generated binaries and compilation time. GCC is sometimes more aggressive at -O3. LLVM is also often by default less strict when it comes to standard compliance by default and performs optimizations GCC would only do with a special compilation option or at -Ofast. Keep this is mind when comparing different compilers.

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark LLVM GCC
709.cactus_r 100.00% 108.92%
722.palm_r 100.00% 127.37%
731.astcenc_r 100.00% 156.59%
736.ocio_r 100.00% 108.41%
737.gmsh_r 100.00% 93.69%
748.flightdm_r 100.00% 99.11%
749.fotonik3d_r 100.00% 95.99%
765.roms_r 100.00% 90.89%
766.femflow_r 100.00% 505.56%
767.nest_r 100.00% 105.00%
772.marian_r 100.00% 431.34%
782.lbm_r 100.00% 97.01%
LLVM clang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc), flang version 23.1.3 (/labs/mjambor/git/llvm 0d261d1ca552c95a8f007e061c787ac7132fbcbc)
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM GCC
709.cactus_r 100.00% 94.69%
722.palm_r 100.00% 88.17%
731.astcenc_r 100.00% 120.54%
736.ocio_r 100.00% 88.05%
737.gmsh_r 100.00% 96.05%
748.flightdm_r 100.00% 87.31%
749.fotonik3d_r 100.00% 27.50%
765.roms_r 100.00% 73.48%
766.femflow_r 100.00% 51.84%
767.nest_r 100.00% 84.94%
772.marian_r 100.00% 89.95%
782.lbm_r 100.00% 82.05%
Smaller binary sizes are always better.

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

Run-times

Benchmark LLVM GCC
709.cactus_r 100.00% 94.32%
722.palm_r 100.00% 121.66%
731.astcenc_r 100.00% 150.61%
736.ocio_r 100.00% 108.03%
737.gmsh_r 100.00% 94.14%
748.flightdm_r 100.00% 100.45%
749.fotonik3d_r 100.00% 100.45%
765.roms_r 100.00% 90.96%
766.femflow_r 100.00% 420.54%
767.nest_r 100.00% 95.64%
772.marian_r 100.00% 431.76%
782.lbm_r 100.00% 96.98%
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)
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM GCC
709.cactus_r 100.00% 94.68%
722.palm_r 100.00% 88.17%
731.astcenc_r 100.00% 120.55%
736.ocio_r 100.00% 87.27%
737.gmsh_r 100.00% 96.12%
748.flightdm_r 100.00% 87.38%
749.fotonik3d_r 100.00% 27.51%
765.roms_r 100.00% 73.48%
766.femflow_r 100.00% 51.66%
767.nest_r 100.00% 85.07%
772.marian_r 100.00% 89.57%
782.lbm_r 100.00% 82.33%
Smaller binary sizes are always better.