Commit 260bfe7
committed
C++: Manual magic in inStaticInitializer
Since `runtimeExprInStaticInitializer` only looks at expressions at the
top level of an initializer or directly below some number of top-level
aggregate literals, there is no need for `inStaticInitializer` to
include expressions strictly below those in the AST.
I tested this on Wireshark, which has very large static initializers,
but found no measureable difference in run time. There are some
differences in tuple counts and iteration counts, though:
- `inStaticInitializer` changes from 6,241,153 rows (86 iterations) to
5,031,617 rows (7 iterations).
- `runtimeExprInStaticInitializer` changes from 386,350 rows to 4,705
rows.
- `hasDynamicInitialization` has 410 rows both before and after, which
suggests that this change does not affect results.
Even though there is no impact on this snapshot at this time, things
might look different if/when the restriction on aggregate literals to
100 children is removed in the extractor.1 parent 93c6f8f commit 260bfe7
1 file changed
+5
-2
lines changed| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
402 | 402 | | |
403 | 403 | | |
404 | 404 | | |
405 | | - | |
| 405 | + | |
| 406 | + | |
| 407 | + | |
| 408 | + | |
406 | 409 | | |
407 | 410 | | |
408 | 411 | | |
409 | | - | |
| 412 | + | |
410 | 413 | | |
411 | 414 | | |
412 | 415 | | |
| |||
0 commit comments