[llvm] [TailDuplicator] Add maximum predecessors and successors to consider tail duplicating blocks (PR #78582)

via llvm-commits llvm-commits at lists.llvm.org
Sun May 12 03:09:41 PDT 2024


DianQK wrote:

Hmm, I tried LLVM 18 and the main (c8864bceeb20582b4e7a739d8ba3e11052f0e49f) branch on Raspberry Pi 4 (arm64), but I didn't find any performance issues:

```

 Performance counter stats for './llvm18':

          1,805.64 msec task-clock:u                     #    0.994 CPUs utilized
                 0      context-switches:u               #    0.000 /sec
                 0      cpu-migrations:u                 #    0.000 /sec
                46      page-faults:u                    #   25.476 /sec
     2,653,746,964      cycles:u                         #    1.470 GHz
     2,656,672,132      instructions:u                   #    1.00  insn per cycle
   <not supported>      branches:u
           247,193      branch-misses:u

       1.815950443 seconds time elapsed

       1.800769000 seconds user
       0.004009000 seconds sys


Sum: 3273600000

 Performance counter stats for './main':

          1,784.68 msec task-clock:u                     #    0.998 CPUs utilized
                 0      context-switches:u               #    0.000 /sec
                 0      cpu-migrations:u                 #    0.000 /sec
                46      page-faults:u                    #   25.775 /sec
     2,653,681,171      cycles:u                         #    1.487 GHz
     2,656,672,132      instructions:u                   #    1.00  insn per cycle
   <not supported>      branches:u
           241,177      branch-misses:u

       1.788022549 seconds time elapsed

       1.780709000 seconds user
       0.003991000 seconds sys
```

https://github.com/llvm/llvm-project/pull/78582


More information about the llvm-commits mailing list