SPEC CPU result comparison

15th October 2026 - Normalized report

SPEC2017 INTrate: PGO+LTO native -Ofast - Released GCCs vs Trunk

-Ofast -flto -march=native -mtune=native -fprofile-use

AMD EPYC 8635P (zen5 based Sorano)

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
500.perlbench_r 100.00% 100.00% 99.26% 100.00% 99.26%
502.gcc_r 100.00% 100.00% 99.07% 100.00% 99.07%
505.mcf_r 100.00% 101.70% 103.41% 100.00% 101.68%
520.omnetpp_r 100.00% 97.52% 99.45% 100.00% 101.98%
523.xalancbmk_r 100.00% 105.30% 104.34% 100.00% 99.08%
525.x264_r 100.00% 89.73% 94.18% 100.00% 104.96%
531.deepsjeng_r 100.00% 103.00% 103.00% 100.00% 100.00%
541.leela_r 100.00% 98.87% 98.42% 100.00% 99.54%
548.exchange2_r 100.00% 95.21% 95.21% 100.00% 100.00%
557.xz_r 100.00% 99.13% 99.57% 100.00% 100.44%
GCC 15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
GCC 16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
Trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.
See also continuous gcc branch benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/branch.

Non-debug executable sizes

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
500.perlbench_r 100.00% 107.11% 104.67% 100.00% 97.72%
502.gcc_r 100.00% 111.06% 104.19% 100.00% 93.81%
505.mcf_r 100.00% 104.91% 104.56% 100.00% 99.67%
520.omnetpp_r 100.00% 102.54% 101.40% 100.00% 98.89%
523.xalancbmk_r 100.00% 102.42% 101.07% 100.00% 98.68%
525.x264_r 100.00% 115.79% 102.91% 100.00% 88.87%
531.deepsjeng_r 100.00% 100.20% 100.22% 100.00% 100.02%
541.leela_r 100.00% 99.30% 99.17% 100.00% 99.87%
548.exchange2_r 100.00% 91.91% 89.92% 100.00% 97.84%
557.xz_r 100.00% 103.36% 106.64% 100.00% 103.17%
Smaller binary sizes are always better.

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

Run-times

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
500.perlbench_r 100.00% 97.89% 97.89% 100.00% 100.00%
502.gcc_r 100.00% 99.27% 98.54% 100.00% 99.27%
505.mcf_r 100.00% 101.12% 102.25% 100.00% 101.11%
520.omnetpp_r 100.00% 103.74% 100.50% 100.00% 96.88%
523.xalancbmk_r 100.00% 102.82% 102.35% 100.00% 99.54%
525.x264_r 100.00% 84.86% 88.29% 100.00% 104.04%
531.deepsjeng_r 100.00% 101.00% 101.00% 100.00% 100.00%
541.leela_r 100.00% 98.16% 97.93% 100.00% 99.77%
548.exchange2_r 100.00% 93.00% 93.00% 100.00% 100.00%
557.xz_r 100.00% 98.68% 100.88% 100.00% 102.23%
GCC 15 GCC 15.3.0 (revision 4db0e8df15bef836558857c291c323add11d035c)
GCC 16 GCC 16.2.0 (revision 78d4ac73dd391005b895a6148cd9831e28e1208b)
Trunk GCC 17.0.0 20261015 (experimental) (revision 84348add236100e5f69a3e9ec4f7b1038bf2b83d)
Naturally, the smaller the run-times, the better.
See also continuous gcc branch benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/branch.

Non-debug executable sizes

Benchmark GCC 15 GCC 16 Trunk GCC 16 Trunk
500.perlbench_r 100.00% 106.87% 104.68% 100.00% 97.95%
502.gcc_r 100.00% 111.26% 104.37% 100.00% 93.81%
505.mcf_r 100.00% 103.37% 103.99% 100.00% 100.60%
520.omnetpp_r 100.00% 102.62% 101.45% 100.00% 98.86%
523.xalancbmk_r 100.00% 102.58% 101.22% 100.00% 98.68%
525.x264_r 100.00% 115.05% 103.82% 100.00% 90.23%
531.deepsjeng_r 100.00% 100.20% 100.21% 100.00% 100.01%
541.leela_r 100.00% 101.24% 101.05% 100.00% 99.81%
548.exchange2_r 100.00% 92.79% 92.24% 100.00% 99.40%
557.xz_r 100.00% 103.51% 106.55% 100.00% 102.94%
Smaller binary sizes are always better.