SPEC CPU result comparison

15th October 2026 - Normalized report

SPECFP 2006: PGO+LTO -Ofast native - LLVM, ICX and GCC

-Ofast -native -flto -fprofile-instr-use vs -Ofast -march=native -flto -fprofile-use

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

Run-times

Benchmark LLVM ICX GCC ICX GCC
410.bwaves CE 100.00% 100.00%
416.gamess CE 100.00% 118.50%
433.milc 100.00% 83.97% 97.71% 100.00% 116.36%
434.zeusmp CE 100.00% 98.56%
435.gromacs CE CE CE
436.cactusADM CE CE CE
437.leslie3d CE 100.00% 100.00%
444.namd 100.00% 100.00% 109.58% 100.00% 109.58%
447.dealII 100.00% 99.12% 98.25% 100.00% 99.12%
450.soplex 100.00% 102.79% 102.79% 100.00% 100.00%
453.povray FE FE 100.00% FE
454.calculix CE CE CE
459.GemsFDTD CE 100.00% 103.55%
465.tonto CE 100.00% 67.43%
470.lbm 100.00% 97.04% 100.00% 100.00% 103.05%
481.wrf CE 100.00% 108.00%
482.sphinx3 100.00% 94.16% 101.33% 100.00% 107.61%
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)
ICX Intel(R) oneAPI DPC++/C++ Compiler 2026.1.1 (2026.1.1.20260724), ifx (IFX) 2026.1.1 20260724
GCC GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.

Non-debug executable sizes

Benchmark LLVM ICX GCC ICX GCC
410.bwaves CE 100.00% 4.13%
416.gamess CE 100.00% 98.23%
433.milc 100.00% 143.19% 71.77% 100.00% 50.12%
434.zeusmp CE 100.00% 99.88%
435.gromacs CE CE CE
436.cactusADM CE CE CE
437.leslie3d CE 100.00% 9.68%
444.namd 100.00% 225.27% 95.46% 100.00% 42.38%
447.dealII 100.00% 167.35% 67.66% 100.00% 40.43%
450.soplex 100.00% 132.53% 84.08% 100.00% 63.44%
453.povray FE FE 100.00% FE
454.calculix CE CE CE
459.GemsFDTD CE 100.00% 23.13%
465.tonto CE 100.00% 40.51%
470.lbm 100.00% 190.34% 91.97% 100.00% 48.32%
481.wrf CE 100.00% 66.66%
482.sphinx3 100.00% 251.81% 70.94% 100.00% 28.17%
Smaller binary sizes are always better.