SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 INTrate: PGO -Ofast native - LLVM and GCC

-Ofast -march=native -mtune=native and -fprofile-instr-use vs. -profile-use

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 LLVM GCC
706.stockfish_r 100.00% 39.64%
707.ntest_r 100.00% 97.70%
708.sqlite_r 100.00% 92.27%
710.omnetpp_r 100.00% 87.19%
714.cpython_r 100.00% 92.31%
721.gcc_r 100.00% 97.17%
723.llvm_r 100.00% 100.00%
727.cppcheck_r 100.00% 108.63%
729.abc_r 100.00% 101.75%
734.vpr_r 100.00% 96.18%
735.gem5_r 100.00% 93.48%
750.sealcrypto_r 100.00% 96.30%
753.ns3_r 100.00% 97.89%
777.zstd_r 100.00% 104.96%
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
706.stockfish_r 100.00% 101.33%
707.ntest_r 100.00% 95.58%
708.sqlite_r 100.00% 102.60%
710.omnetpp_r 100.00% 82.78%
714.cpython_r 100.00% 87.69%
721.gcc_r 100.00% 93.77%
723.llvm_r 100.00% 88.65%
727.cppcheck_r 100.00% 102.29%
729.abc_r 100.00% 82.83%
734.vpr_r 100.00% 89.69%
735.gem5_r 100.00% 89.90%
750.sealcrypto_r 100.00% 86.22%
753.ns3_r 100.00% 86.41%
777.zstd_r 100.00% 82.08%
Smaller binary sizes are always better.