Optimize C++ generated code for structures with many conditional fields#241
Optimize C++ generated code for structures with many conditional fields#241AaronWebster wants to merge 7 commits intomasterfrom
Conversation
- Remove [[nodiscard]] annotations (C++17 feature) and create separate issue #242 to discuss C++14 vs C++17 support - Remove redundant check in ExpressionScope.add() that was already covered by the parent scope lookup - Revert unnecessary style change in _render_existence_test - Add comprehensive docstring to _generate_optimized_ok_method_body explaining the switch optimization with examples - Refactor switch/if generation to use templates for better readability - Optimize trivially-true conditions to skip if-block wrapper when condition is statically known to be true
Regenerate all golden .emb.h files to reflect changes from the optimized conditionals implementation.
AaronWebster
added
the
optimization
Add golden file and test for many_conditionals.emb to ensure the generated C++ header is tracked for review.
EricRahm
left a comment
EricRahm
left a comment
There was a problem hiding this comment.
Thanks for polishing this up! Looks really good, added some high level comments. I think overall it'd be a little easier to review if we split out the always true optimization.
| both guarded by tag == 0), only the first field for that case value is | ||
| checked in the switch. This is because once we've checked var_0 for case 0, | ||
| we break out of the switch. Fields with duplicate case values will fall | ||
| through to the if-statement path. |
There was a problem hiding this comment.
Overall great comment, I really appreciate the clear explanations. Sorry to harp on this, but can you add a test case to make sure the statements here match up with the code behavior?
I suspect this will lead to a bit more churn so I'll hold off on digging to much deeper.
| emb_file, | ||
| golden_file, | ||
| include_dirs=None, | ||
| compiler_flags=None, |
There was a problem hiding this comment.
The one_golden_test and run_one_golden_test changes seem like they might be a separate issue? If so would you mind splitting that out (and we can land it quickly).
|
|
||
| This module contains types used by the LR(1) parser, which are also used in | ||
| other parts of the compiler: | ||
| other parts of the compiler: |
There was a problem hiding this comment.
nit: this seems unrelated
| and group["expr"].type.boolean.value | ||
| ) | ||
|
|
||
| if is_always_true: |
There was a problem hiding this comment.
I love this change! I wonder if we could split it out out as a separate issue, I think it'd make the golden churn a little bit easier to grok and also highlight the impact of it :)
| const auto emboss_reserved_local_ok_subexpr_23 = ::emboss::support::And</**/bool, bool, bool, bool>(emboss_reserved_local_ok_subexpr_19, emboss_reserved_local_ok_subexpr_22); | ||
| const auto emboss_reserved_local_ok_subexpr_24 = ::emboss::support::LessThan</**/::std::int32_t, bool, ::std::int32_t, ::std::int32_t>(emboss_reserved_local_ok_subexpr_10, ::emboss::support::Maybe</**/::std::int32_t>(static_cast</**/::std::int32_t>(2LL))); | ||
| const auto emboss_reserved_local_ok_subexpr_25 = ::emboss::support::LessThan</**/::std::int32_t, bool, ::std::int32_t, ::std::int32_t>(emboss_reserved_local_ok_subexpr_3, ::emboss::support::Maybe</**/::std::int32_t>(static_cast</**/::std::int32_t>(4LL))); | ||
| const auto emboss_reserved_local_ok_subexpr_26 = ::emboss::support::Or</**/bool, bool, bool, bool>(emboss_reserved_local_ok_subexpr_24, emboss_reserved_local_ok_subexpr_25); |
There was a problem hiding this comment.
This feels like somehow this got worse, maybe the tradeoff is worth it but I wonder if we could do better.
| VALUE = static_cast</**/::std::int32_t>(10LL), | ||
|
|
||
| }; | ||
| template <class Enum> |
There was a problem hiding this comment.
I suspect this is a result of the fix for passing in compiler args in the one_golden_test changes?
| @@ -1,13 +0,0 @@ | |||
| # Copyright 2019 Google LLC | |||
There was a problem hiding this comment.
I don't think we want to delete this?
2 participants
This change improves the performance of the generated C++ code for Emboss structures containing many conditional fields.
Specifically, it optimizes the validation logic (e.g., the
Ok()method) and field access patterns where fields depend on a shared discriminant or tag. This results in faster runtime validation and reduced overhead when interacting with complex structures defined with extensive conditional logic.Performance Impact
Benchmarked on
testdata/many_conditionals.embwith 10,000 iterations x 100 tags, compiled with-c opt:Summary
Recreated from #235 after the source branch was deleted.