Optimize C++ generated code for structures with many conditional fields by AaronWebster · Pull Request #241 · google/emboss

AaronWebster · 2026-01-29T17:47:19Z

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.emb with 10,000 iterations x 100 tags, compiled with -c opt:

Metric	Master	PR	Improvement
Compiled Binary Text Section	64,519 bytes	30,079 bytes	53.4% reduction
Runtime (avg of 3 runs)	~1.31s	~0.11s	11.8x faster
Header File Size	611,751 bytes	612,180 bytes	+429 bytes (+0.07%)

Summary

Runtime Speedup: 11.8x (from ~1.31s to ~0.11s)
Binary Size Reduction: 53.4% (text section reduced from 64.5KB to 30KB, saving ~34KB)
Header source size is essentially unchanged (trivial 0.07% increase)

Recreated from #235 after the source branch was deleted.

- 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.

Add golden file and test for many_conditionals.emb to ensure the generated C++ header is tracked for review.

EricRahm

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.

EricRahm · 2026-01-30T22:05:25Z

+    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.


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.

EricRahm · 2026-01-30T22:08:56Z

        emb_file,
        golden_file,
        include_dirs=None,
+        compiler_flags=None,


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).

EricRahm · 2026-01-30T22:09:21Z


 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:


nit: this seems unrelated

EricRahm · 2026-01-30T22:12:05Z

+                and group["expr"].type.boolean.value
+            )
+
+            if is_always_true:


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 :)

EricRahm · 2026-01-30T22:13:29Z

+      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);


This feels like somehow this got worse, maybe the tradeoff is worth it but I wonder if we could do better.

EricRahm · 2026-01-30T22:15:05Z

  VALUE = static_cast</**/::std::int32_t>(10LL),

 };
-template <class Enum>


I suspect this is a result of the fix for passing in compiler args in the one_golden_test changes?

EricRahm · 2026-01-30T22:15:52Z

@@ -1,13 +0,0 @@
-# Copyright 2019 Google LLC


I don't think we want to delete this?

- Revert compiler_flags additions in one_golden_test.py (to be split out) - Restore testdata/__init__.py - Revert unrelated whitespace change in parser_types.py - Revert 'is_always_true' optimization to reduce golden churn - Add C++ test case `DuplicateCaseValueFallthrough` to verify switch behavior

…als-v3

Merge current master, which landed clang-format for the C++ back end (#262) and the optimized-conditionals work (#241). Both rewrote generated_code_templates and every testdata/golden_cpp/*.h. The template conflict is resolved by re-applying the JSON text-output logic on top of master's version, and all goldens are regenerated with clang-format 22.1.5. Fixes that turn the failing "Run Bazel tests (clang)" job green: - Remove internal-ism include paths from emboss_array_view_test.cc (testing/base/public/gunit.h, third_party/absl/..., third_party/ emboss/...). These do not resolve in the open-source build and broke compilation -- the same class of issue already corrected in the template. - Size JsonTestStruct's struct_array to two elements ([+40] -> [+8]) so the 87-byte struct becomes 55 bytes and fits the 57-byte test buffer. The oversized fixed array made WriteToTextStream read past the buffer (the tests assert exactly two elements), causing check failures. - Update the array JSON test expectation to compact form ("[-3,2,...]" rather than "[-3, 2, ...]") to match the compact JSON the implementation emits, resolving the test/implementation version skew noted in review. Additionally, emit newlines between fields in multiline JSON output so multiline JSON is properly formatted; the change is guarded so non-JSON text output is byte-for-byte unchanged. Adds a JsonMultilineOutput test.

AaronWebster and others added 6 commits December 2, 2025 15:27

Port optimized conditionals v2

1a14004

feat(cpp): Optimize generated C++ accessors and control flow

8ea51fb

Fix Black formatting

0f9f76c

Merge branch 'master' into optimized-conditionals-v3

8ac2091

Update C++ golden test files

2d39b63

Regenerate all golden .emb.h files to reflect changes from the optimized conditionals implementation.

AaronWebster requested a review from EricRahm January 29, 2026 19:08

AaronWebster self-assigned this Jan 29, 2026

AaronWebster added the optimization Performance improvements that do not otherwise change behavior label Jan 29, 2026

Add many_conditionals golden test

c8cfee4

Add golden file and test for many_conditionals.emb to ensure the generated C++ header is tracked for review.

EricRahm reviewed Jan 30, 2026

View reviewed changes

AaronWebster mentioned this pull request May 18, 2026

Optimize C++ generated code to skip condition checks for unconditionally present fields #249

Open

AaronWebster added 3 commits May 18, 2026 15:34

Merge remote-tracking branch 'origin/master' into optimized-condition…

8ad4b68

…als-v3

Fix duplicate case values in optimized Ok() switch generation

f11f098

Format python files with black

d4cd926

AaronWebster force-pushed the optimized-conditionals-v3 branch from 8749630 to d4cd926 Compare May 18, 2026 23:55

AaronWebster merged commit 188c9f2 into master May 19, 2026
11 checks passed

This was referenced May 20, 2026

Extend switch matcher: disjunction, conjunction, single-entry demotion #256

Merged

Statically-known existence conditions: drop or simplify the check #259

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Optimize C++ generated code for structures with many conditional fields#241

Optimize C++ generated code for structures with many conditional fields#241
AaronWebster merged 11 commits into
masterfrom
optimized-conditionals-v3

AaronWebster commented Jan 29, 2026 •

edited

Loading

Uh oh!

EricRahm left a comment

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

EricRahm Jan 30, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

AaronWebster commented Jan 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Performance Impact

Summary

Uh oh!

EricRahm left a comment

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

EricRahm Jan 30, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

AaronWebster commented Jan 29, 2026 •

edited

Loading