SPEC2017 INTrate: Trunk standard -Ofast - generic vs native
-Ofast vs -Ofast -march=native -mtune=native
AMD EPYC 8635P (zen5 based Sorano)
Run-times
Benchmark
Generic
Native
500.perlbench_r
100.00%
102.87%
502.gcc_r
100.00%
100.00%
505.mcf_r
100.00%
99.53%
520.omnetpp_r
100.00%
103.10%
523.xalancbmk_r
100.00%
92.45%
525.x264_r
100.00%
92.91%
531.deepsjeng_r
100.00%
95.24%
541.leela_r
100.00%
98.05%
548.exchange2_r
100.00%
101.08%
557.xz_r
100.00%
101.91%
| Benchmark | Generic | Native |
|---|---|---|
| 500.perlbench_r | 100.00% | 102.87% |
| 502.gcc_r | 100.00% | 100.00% |
| 505.mcf_r | 100.00% | 99.53% |
| 520.omnetpp_r | 100.00% | 103.10% |
| 523.xalancbmk_r | 100.00% | 92.45% |
| 525.x264_r | 100.00% | 92.91% |
| 531.deepsjeng_r | 100.00% | 95.24% |
| 541.leela_r | 100.00% | 98.05% |
| 548.exchange2_r | 100.00% | 101.08% |
| 557.xz_r | 100.00% | 101.91% |
Naturally, the smaller the run-times, the better.
See also continuous gcc tuning benchmarking at https://lnt.opensuse.org/db_default/v4/SPEC/spec_report/tuning.
Non-debug executable sizes
Benchmark
Generic
Native
500.perlbench_r
100.00%
102.58%
502.gcc_r
100.00%
103.94%
505.mcf_r
100.00%
117.81%
520.omnetpp_r
100.00%
102.43%
523.xalancbmk_r
100.00%
102.32%
525.x264_r
100.00%
100.20%
531.deepsjeng_r
100.00%
100.13%
541.leela_r
100.00%
100.20%
548.exchange2_r
100.00%
115.98%
557.xz_r
100.00%
101.03%
Smaller binary sizes are always better.
| Benchmark | Generic | Native |
|---|---|---|
| 500.perlbench_r | 100.00% | 102.58% |
| 502.gcc_r | 100.00% | 103.94% |
| 505.mcf_r | 100.00% | 117.81% |
| 520.omnetpp_r | 100.00% | 102.43% |
| 523.xalancbmk_r | 100.00% | 102.32% |
| 525.x264_r | 100.00% | 100.20% |
| 531.deepsjeng_r | 100.00% | 100.13% |
| 541.leela_r | 100.00% | 100.20% |
| 548.exchange2_r | 100.00% | 115.98% |
| 557.xz_r | 100.00% | 101.03% |