[llvm] [RISCV] Move MachineCombiner to addILPOpts() (PR #158071)

Pengcheng Wang via llvm-commits llvm-commits at lists.llvm.org
Fri Sep 12 00:19:51 PDT 2025


wangpc-pp wrote:

I did a quick evaluation on dynamic instruction count:

| benchmarks                                                                 | before        | after         | after/before |
|----------------------------------------------------------------------------|---------------|---------------|--------------|
| 500.perlbench_r.checkspam.2500.5.25.11.150.1.1.1.1                         | 1511438019326 | 1511490941021 | 1.000035014  |
| 500.perlbench_r.diffmail.4.800.10.17.19.300                                | 925708421637  | 925708376753  | 0.999999952  |
| 500.perlbench_r.splitmail.6400.12.26.16.100.0                              | 880162912655  | 880159320494  | 0.999995919  |
| 502.gcc_r.gcc-pp.opts-O3_-finline-limit_0_-fif-conversion_-fif-conversion2 | 221680021619  | 221627858602  | 0.999764692  |
| 502.gcc_r.gcc-pp.opts-O2_-finline-limit_36000_-fpic                        | 266377405681  | 266277432822  | 0.999624695  |
| 502.gcc_r.gcc-smaller.opts-O3_-fipa-pta                                    | 250452824898  | 250409518608  | 0.999827088  |
| 502.gcc_r.ref32.opts-O3_-fselective-scheduling_-fselective-scheduling2     | 255045713581  | 254964377492  | 0.999681092  |
| 502.gcc_r.ref32.opts-O5                                                    | 209768544305  | 209679045017  | 0.999573343  |
| 505.mcf_r.inp                                                              | 1042193706797 | 1012778135975 | 0.971775333  |
| 520.omnetpp_r.omnetpp.General-0                                            | 1118516266411 | 1118516267920 | 1.000000001  |
| 523.xalancbmk_r.ref-t5                                                     | 1137591968040 | 1137591967921 | 1            |
| 525.x264_r.run_000-1000_x264_r_pass1                                       | 183617996746  | 183791621135  | 1.000945574  |
| 525.x264_r.run_000-1000_x264_r_pass2                                       | 547243857597  | 547956744001  | 1.001302685  |
| 525.x264_r.run_0500-1250_x264_r                                            | 578465188365  | 579235279237  | 1.001331266  |
| 531.deepsjeng_r.ref                                                        | 2001546087182 | 2000678614384 | 0.999566599  |
| 541.leela_r.ref                                                            | 2423960222211 | 2423970707977 | 1.000004326  |
| 557.xz_r.cld.tar-160-6                                                     | 369175467578  | 369034986843  | 0.999619474  |
| 557.xz_r.cpu2006docs.tar-250-6e                                            | 663623243021  | 663506295798  | 0.999823775  |
| 557.xz_r.input.combined-250-7                                              | 483496126716  | 483351690813  | 0.999701268  |
| geomean                                                                    |               |               | 0.998535992  |

The result shows a slight decrease (considered as noise).

https://github.com/llvm/llvm-project/pull/158071


More information about the llvm-commits mailing list