7 Commits

Author SHA1 Message Date
Florian Hahn
3bdf9a0880
[EquivalenceClasses] Use SmallVector for deterministic iteration order. (#134075)
Currently iterators over EquivalenceClasses will iterate over std::set,
which guarantees the order specified by the comperator. Unfortunately in
many cases, EquivalenceClasses are used with pointers, so iterating over
std::set of pointers will not be deterministic across runs.

There are multiple places that explicitly try to sort the equivalence
classes before using them to try to get a deterministic order
(LowerTypeTests, SplitModule), but there are others that do not at the
moment and this can result at least in non-determinstic value naming in
Float2Int.

This patch updates EquivalenceClasses to keep track of all members via a
extra SmallVector and removes code from LowerTypeTests and SplitModule
to sort the classes before processing.

Overall it looks like compile-time slightly decreases in most cases, but
close to noise:

https://llvm-compile-time-tracker.com/compare.php?from=7d441d9892295a6eb8aaf481e1715f039f6f224f&to=b0c2ac67a88d3ef86987e2f82115ea0170675a17&stat=instructions

PR: https://github.com/llvm/llvm-project/pull/134075
2025-04-02 20:27:43 +01:00
Florian Hahn
9e5bfbf77d
[EquivalenceClasses] Update member_begin to take ECValue (NFC).
Remove a level of indirection and update code to use range-based for
loops.
2025-04-01 09:28:46 +01:00
Florian Hahn
64d493f987
[EquivalenceClasses] Return ECValue directly from insert (NFC).
Removes a redundant lookup in the mapping.:
2025-04-01 08:45:46 +01:00
Florian Hahn
32f24029c7
Reapply "[EquivalenceClasses] Replace findValue with contains (NFC)."
This reverts the revert commit 616f447fc84bdc7655117f1b303d895dc3b93e4d.

It includes updates to remaining users in Polly and Clang, to avoid
failures when building those projects.
2025-03-31 22:27:59 +01:00
martinboehme
1b1b525147
[clang][dataflow] Use Formula::isLiteral() in a couple more places. (#78404) 2024-01-18 08:58:47 +01:00
Aaron Ballman
24060db3a9 Silence diagnostics about not all control paths returning a value; NFC 2023-11-07 10:42:50 -05:00
martinboehme
7c636728c0
[clang][dataflow] Simplify flow conditions displayed in HTMLLogger. (#70848)
This can make the flow condition significantly easier to interpret; see
below
for an example.

I had hoped that adding the simplification as a preprocessing step
before the
SAT solver (in `DataflowAnalysisContext::querySolver()`) would also
speed up SAT
solving and maybe even eliminate SAT solver timeouts, but in my testing,
this
actually turns out to be a pessimization. It appears that these
simplifications
are easy enough for the SAT solver to perform itself.

Nevertheless, the improvement in debugging alone makes this a worthwhile
change.

Example of flow condition output with these changes:

```
Flow condition token: V37
Constraints:
(V16 = (((V15 & (V19 = V12)) & V22) & V25))
(V15 = ((V12 & ((V14 = V9) | (V14 = V4))) & (V13 = V14)))
True atoms: (V0, V1, V2, V5, V6, V7, V29, V30, V32, V34, V35, V37)
False atoms: (V3, V8, V17)
Equivalent atoms:
(V11, V15)

Flow condition constraints before simplification:
V37
((!V3 & !V8) & !V17)
(V37 = V34)
(V34 = (V29 & (V35 = V30)))
(V29 = (((V16 | V2) & V32) & (V30 = V32)))
(V16 = (((V15 & (V19 = V12)) & V22) & V25))
(V15 = V11)
(V11 = ((((V7 | V2) & V12) & ((V7 & (V14 = V9)) | (V2 & (V14 = V4)))) & (V13 = V14)))
(V2 = V1)
(V1 = V0)
V0
(V7 = V6)
(V6 = V5)
(V5 = V2)
```
2023-11-07 15:18:34 +01:00