SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 INTrate: Trunk LTO native - O3 vs -Ofast

-O3 -flto -march=native vs -Ofast -flto -march=native

Please note that even though the header of this comparison says -Ofast, benchmarks 710.omnetpp_r, 734.vpr_r, 735.gem5_r and 753.ns3_r were additionally compiled with -fno-finite-math-only.

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark O3 Ofast
706.stockfish_r 100.00% 100.00%
707.ntest_r 100.00% 100.00%
708.sqlite_r 100.00% 100.64%
710.omnetpp_r 100.00% 98.87%
714.cpython_r 100.00% 100.00%
721.gcc_r 100.00% 100.64%
723.llvm_r 100.00% 98.73%
727.cppcheck_r 100.00% 102.23%
729.abc_r 100.00% 100.00%
734.vpr_r 100.00% 100.00%
735.gem5_r 100.00% 101.26%
750.sealcrypto_r 100.00% 100.00%
753.ns3_r 100.00% 98.81%
777.zstd_r 100.00% 100.00%
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 O3 Ofast
706.stockfish_r 100.00% 100.29%
707.ntest_r 100.00% 99.96%
708.sqlite_r 100.00% 100.05%
710.omnetpp_r 100.00% 99.93%
714.cpython_r 100.00% 99.85%
721.gcc_r 100.00% 100.02%
723.llvm_r 100.00% 100.01%
727.cppcheck_r 100.00% 99.98%
729.abc_r 100.00% 100.28%
734.vpr_r 100.00% 100.15%
735.gem5_r 100.00% 100.00%
750.sealcrypto_r 100.00% 100.04%
753.ns3_r 100.00% 100.12%
777.zstd_r 100.00% 100.01%
Smaller binary sizes are always better.

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

Run-times

Benchmark O3 Ofast
706.stockfish_r 100.00% 100.51%
707.ntest_r 100.00% 99.52%
708.sqlite_r 100.00% 98.86%
710.omnetpp_r 100.00% 101.88%
714.cpython_r 100.00% 100.00%
721.gcc_r 100.00% 100.43%
723.llvm_r 100.00% 100.00%
727.cppcheck_r 100.00% 96.88%
729.abc_r 100.00% 100.00%
734.vpr_r 100.00% 100.00%
735.gem5_r 100.00% 102.30%
750.sealcrypto_r 100.00% 100.00%
753.ns3_r 100.00% 99.49%
777.zstd_r 100.00% 100.00%
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 O3 Ofast
706.stockfish_r 100.00% 100.05%
707.ntest_r 100.00% 100.01%
708.sqlite_r 100.00% 99.75%
710.omnetpp_r 100.00% 100.08%
714.cpython_r 100.00% 99.85%
721.gcc_r 100.00% 100.03%
723.llvm_r 100.00% 100.02%
727.cppcheck_r 100.00% 100.01%
729.abc_r 100.00% 100.20%
734.vpr_r 100.00% 100.19%
735.gem5_r 100.00% 100.01%
750.sealcrypto_r 100.00% 100.08%
753.ns3_r 100.00% 100.11%
777.zstd_r 100.00% 100.00%
Smaller binary sizes are always better.