SPEC2017 INTrate: Trunk LTO native - O3 vs -Ofast
-O3 -flto -march=native vs -Ofast -flto -march=native
AMD EPYC 8635P (zen5 based Sorano)
Run-times
Benchmark
O3
Ofast
500.perlbench_r
100.00%
99.34%
502.gcc_r
100.00%
100.00%
505.mcf_r
100.00%
99.45%
520.omnetpp_r
100.00%
102.29%
523.xalancbmk_r
100.00%
100.00%
525.x264_r
100.00%
100.00%
531.deepsjeng_r
100.00%
100.00%
541.leela_r
100.00%
98.82%
548.exchange2_r
100.00%
101.50%
557.xz_r
100.00%
100.00%
| Benchmark | O3 | Ofast |
|---|---|---|
| 500.perlbench_r | 100.00% | 99.34% |
| 502.gcc_r | 100.00% | 100.00% |
| 505.mcf_r | 100.00% | 99.45% |
| 520.omnetpp_r | 100.00% | 102.29% |
| 523.xalancbmk_r | 100.00% | 100.00% |
| 525.x264_r | 100.00% | 100.00% |
| 531.deepsjeng_r | 100.00% | 100.00% |
| 541.leela_r | 100.00% | 98.82% |
| 548.exchange2_r | 100.00% | 101.50% |
| 557.xz_r | 100.00% | 100.00% |
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
500.perlbench_r
100.00%
99.93%
502.gcc_r
100.00%
100.04%
505.mcf_r
100.00%
100.34%
520.omnetpp_r
100.00%
100.13%
523.xalancbmk_r
100.00%
100.57%
525.x264_r
100.00%
100.07%
531.deepsjeng_r
100.00%
100.00%
541.leela_r
100.00%
99.86%
548.exchange2_r
100.00%
98.80%
557.xz_r
100.00%
100.02%
Smaller binary sizes are always better.
| Benchmark | O3 | Ofast |
|---|---|---|
| 500.perlbench_r | 100.00% | 99.93% |
| 502.gcc_r | 100.00% | 100.04% |
| 505.mcf_r | 100.00% | 100.34% |
| 520.omnetpp_r | 100.00% | 100.13% |
| 523.xalancbmk_r | 100.00% | 100.57% |
| 525.x264_r | 100.00% | 100.07% |
| 531.deepsjeng_r | 100.00% | 100.00% |
| 541.leela_r | 100.00% | 99.86% |
| 548.exchange2_r | 100.00% | 98.80% |
| 557.xz_r | 100.00% | 100.02% |
AMD EPYC 75F3 (zen3 based high-frequency Milan)
Run-times
Benchmark
O3
Ofast
500.perlbench_r
100.00%
100.91%
502.gcc_r
100.00%
100.00%
505.mcf_r
100.00%
100.00%
520.omnetpp_r
100.00%
99.53%
523.xalancbmk_r
100.00%
99.50%
525.x264_r
100.00%
100.00%
531.deepsjeng_r
100.00%
100.00%
541.leela_r
100.00%
100.00%
548.exchange2_r
100.00%
99.12%
557.xz_r
100.00%
100.00%
| Benchmark | O3 | Ofast |
|---|---|---|
| 500.perlbench_r | 100.00% | 100.91% |
| 502.gcc_r | 100.00% | 100.00% |
| 505.mcf_r | 100.00% | 100.00% |
| 520.omnetpp_r | 100.00% | 99.53% |
| 523.xalancbmk_r | 100.00% | 99.50% |
| 525.x264_r | 100.00% | 100.00% |
| 531.deepsjeng_r | 100.00% | 100.00% |
| 541.leela_r | 100.00% | 100.00% |
| 548.exchange2_r | 100.00% | 99.12% |
| 557.xz_r | 100.00% | 100.00% |
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
500.perlbench_r
100.00%
99.86%
502.gcc_r
100.00%
100.11%
505.mcf_r
100.00%
100.36%
520.omnetpp_r
100.00%
100.12%
523.xalancbmk_r
100.00%
100.62%
525.x264_r
100.00%
100.05%
531.deepsjeng_r
100.00%
100.00%
541.leela_r
100.00%
99.90%
548.exchange2_r
100.00%
99.16%
557.xz_r
100.00%
99.98%
Smaller binary sizes are always better.
| Benchmark | O3 | Ofast |
|---|---|---|
| 500.perlbench_r | 100.00% | 99.86% |
| 502.gcc_r | 100.00% | 100.11% |
| 505.mcf_r | 100.00% | 100.36% |
| 520.omnetpp_r | 100.00% | 100.12% |
| 523.xalancbmk_r | 100.00% | 100.62% |
| 525.x264_r | 100.00% | 100.05% |
| 531.deepsjeng_r | 100.00% | 100.00% |
| 541.leela_r | 100.00% | 99.90% |
| 548.exchange2_r | 100.00% | 99.16% |
| 557.xz_r | 100.00% | 99.98% |