SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2026 INTrate: Trunk native -Ofast - standard vs LTO vs LTO+PGO

-Ofast -march=native -mtune=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 Normal LTO LTO+PGO LTO LTO+PGO
706.stockfish_r 100.00% 99.05% 99.05% 100.00% 100.00%
707.ntest_r 100.00% 98.58% 94.58% 100.00% 95.93%
708.sqlite_r 100.00% 102.18% 93.68% 100.00% 91.68%
710.omnetpp_r 100.00% 80.05% 68.12% 100.00% 85.10%
714.cpython_r 100.00% 93.26% 74.13% 100.00% 79.49%
721.gcc_r 100.00% 97.30% 86.93% 100.00% 89.34%
723.llvm_r 100.00% 99.36% 84.08% 100.00% 84.62%
727.cppcheck_r 100.00% 101.10% 84.11% 100.00% 83.19%
729.abc_r 100.00% 99.09% 105.48% 100.00% 106.45%
734.vpr_r 100.00% 87.98% 84.13% 100.00% 95.62%
735.gem5_r 100.00% 93.50% 103.25% 100.00% 110.42%
750.sealcrypto_r 100.00% 100.00% 116.67% 100.00% 116.67%
753.ns3_r 100.00% 86.61% 83.47% 100.00% 96.38%
777.zstd_r 100.00% 103.62% 101.45% 100.00% 97.90%
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 Normal LTO LTO+PGO LTO LTO+PGO
706.stockfish_r 100.00% 89.90% 83.69% 100.00% 93.09%
707.ntest_r 100.00% 70.09% 63.91% 100.00% 91.18%
708.sqlite_r 100.00% 96.90% 65.67% 100.00% 67.77%
710.omnetpp_r 100.00% 82.64% 58.46% 100.00% 70.74%
714.cpython_r 100.00% 78.97% 59.79% 100.00% 75.71%
721.gcc_r 100.00% 102.48% 82.47% 100.00% 80.47%
723.llvm_r 100.00% 67.82% 44.13% 100.00% 65.07%
727.cppcheck_r 100.00% 81.54% 64.87% 100.00% 79.56%
729.abc_r 100.00% 62.83% 35.85% 100.00% 57.07%
734.vpr_r 100.00% 78.24% 58.25% 100.00% 74.45%
735.gem5_r 100.00% 69.34% 48.40% 100.00% 69.80%
750.sealcrypto_r 100.00% 42.78% 26.88% 100.00% 62.82%
753.ns3_r 100.00% 73.57% 54.12% 100.00% 73.56%
777.zstd_r 100.00% 78.30% 41.35% 100.00% 52.81%
Smaller binary sizes are always better.

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

Run-times

Benchmark Normal LTO LTO+PGO LTO LTO+PGO
706.stockfish_r 100.00% 98.99% 102.01% 100.00% 103.05%
707.ntest_r 100.00% 96.76% 91.20% 100.00% 94.26%
708.sqlite_r 100.00% 100.46% 90.26% 100.00% 89.84%
710.omnetpp_r 100.00% 75.81% 64.42% 100.00% 84.97%
714.cpython_r 100.00% 89.76% 71.68% 100.00% 79.85%
721.gcc_r 100.00% 97.27% 84.70% 100.00% 87.07%
723.llvm_r 100.00% 100.63% 76.10% 100.00% 75.62%
727.cppcheck_r 100.00% 99.15% 88.70% 100.00% 89.46%
729.abc_r 100.00% 98.62% 105.05% 100.00% 106.51%
734.vpr_r 100.00% 88.36% 83.84% 100.00% 94.88%
735.gem5_r 100.00% 95.02% 104.50% 100.00% 109.98%
750.sealcrypto_r 100.00% 100.00% 103.49% 100.00% 103.49%
753.ns3_r 100.00% 81.12% 69.50% 100.00% 85.68%
777.zstd_r 100.00% 103.37% 95.06% 100.00% 91.96%
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 Normal LTO LTO+PGO LTO LTO+PGO
706.stockfish_r 100.00% 90.01% 83.68% 100.00% 92.97%
707.ntest_r 100.00% 69.80% 64.48% 100.00% 92.38%
708.sqlite_r 100.00% 96.49% 63.68% 100.00% 66.00%
710.omnetpp_r 100.00% 82.69% 58.63% 100.00% 70.91%
714.cpython_r 100.00% 77.87% 59.67% 100.00% 76.62%
721.gcc_r 100.00% 101.65% 82.75% 100.00% 81.41%
723.llvm_r 100.00% 67.73% 45.16% 100.00% 66.67%
727.cppcheck_r 100.00% 81.68% 65.15% 100.00% 79.77%
729.abc_r 100.00% 62.78% 36.07% 100.00% 57.45%
734.vpr_r 100.00% 78.19% 58.46% 100.00% 74.78%
735.gem5_r 100.00% 69.92% 48.81% 100.00% 69.80%
750.sealcrypto_r 100.00% 41.72% 27.25% 100.00% 65.30%
753.ns3_r 100.00% 73.77% 54.49% 100.00% 73.86%
777.zstd_r 100.00% 79.84% 42.79% 100.00% 53.59%
Smaller binary sizes are always better.