SPEC CPU result comparison

15th October 2026 - Normalized report

SPECINT 2006: LTO -Ofast native - LLVM, ICX and GCC

-Ofast -march=native -flto

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

Run-times

Benchmark LLVM ICX GCC ICX GCC
400.perlbench 100.00% 95.85% 94.61% 100.00% 98.70%
401.bzip2 100.00% 98.19% 96.57% 100.00% 98.36%
403.gcc 100.00% 99.16% 101.68% 100.00% 102.55%
429.mcf 100.00% 93.52% 92.71% 100.00% 99.13%
445.gobmk RE RE RE
456.hmmer 100.00% 58.35% 40.71% 100.00% 69.76%
458.sjeng 100.00% 99.79% 100.41% 100.00% 100.62%
462.libquantum 100.00% 81.98% 89.27% 100.00% 108.89%
464.h264ref 100.00% 91.83% 100.43% 100.00% 109.37%
471.omnetpp 100.00% 100.00% 93.24% 100.00% 93.24%
473.astar 100.00% 96.17% 92.98% 100.00% 96.68%
483.xalancbmk 100.00% 86.75% 72.22% 100.00% 83.25%
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
400.perlbench 100.00% 118.38% 116.35% 100.00% 98.29%
401.bzip2 100.00% 199.31% 99.67% 100.00% 50.01%
403.gcc 100.00% 103.45% 103.99% 100.00% 100.52%
429.mcf 100.00% 203.31% 74.91% 100.00% 36.84%
445.gobmk RE RE RE
456.hmmer 100.00% 161.99% 98.46% 100.00% 60.78%
458.sjeng 100.00% 103.07% 102.25% 100.00% 99.21%
462.libquantum 100.00% 521.03% 81.59% 100.00% 15.66%
464.h264ref 100.00% 138.69% 96.56% 100.00% 69.62%
471.omnetpp 100.00% 132.99% 92.24% 100.00% 69.36%
473.astar 100.00% 287.77% 77.68% 100.00% 26.99%
483.xalancbmk 100.00% 100.71% 115.23% 100.00% 114.42%
Smaller binary sizes are always better.