[all-commits] [llvm/llvm-project] 03a0bf: [clang][dataflow] Add an early-out to `flowConditi...

martinboehme via All-commits all-commits at lists.llvm.org
Tue Jan 9 10:19:08 PST 2024


  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 03a0bfa96a6eb09c4bbae344ac3aa062339aa730
      https://github.com/llvm/llvm-project/commit/03a0bfa96a6eb09c4bbae344ac3aa062339aa730
  Author: martinboehme <mboehme at google.com>
  Date:   2024-01-09 (Tue, 09 Jan 2024)

  Changed paths:
    M clang/include/clang/Analysis/FlowSensitive/Formula.h
    M clang/lib/Analysis/FlowSensitive/DataflowAnalysisContext.cpp

  Log Message:
  -----------
  [clang][dataflow] Add an early-out to `flowConditionImplies()` / `flowConditionAllows()`. (#77453)

This saves having to assemble the set of constraints and run the SAT
solver in
the trivial case where `F` is true.

This is a performance win on the benchmarks for the Crubit nullability
checker:

```
name                              old cpu/op   new cpu/op   delta
BM_PointerAnalysisCopyPointer     64.1µs ± 5%  63.1µs ± 0%   -1.56%  (p=0.000 n=20+17)
BM_PointerAnalysisIntLoop          172µs ± 2%   171µs ± 0%     ~     (p=0.752 n=20+17)
BM_PointerAnalysisPointerLoop      408µs ± 3%   355µs ± 0%  -12.99%  (p=0.000 n=20+17)
BM_PointerAnalysisBranch           201µs ± 2%   184µs ± 0%   -8.28%  (p=0.000 n=20+19)
BM_PointerAnalysisLoopAndBranch    684µs ± 2%   613µs ± 2%  -10.38%  (p=0.000 n=20+19)
BM_PointerAnalysisTwoLoops         309µs ± 2%   308µs ± 2%     ~     (p=0.728 n=20+19)
BM_PointerAnalysisJoinFilePath    37.9ms ± 2%  37.9ms ± 2%   +0.06%  (p=0.041 n=20+19)
BM_PointerAnalysisCallInLoop      26.5ms ± 2%  26.4ms ± 4%   -0.59%  (p=0.024 n=20+20)
```

When running clang-tidy on real-world code, the results are less clear.
In
three runs, averaged, on an arbitrarily chosen input file, I get 11.91 s
of user
time without this patch and 11.81 s with it, though with considerable
measurement noise (I'm seeing up to 0.2 s of variation between runs).

Still, this is a very simple change, and it is a clear win in
benchmarks, so I
think it is worth making.




More information about the All-commits mailing list