[wip] experiment with reversing alignment and size validation order by jswrenn · Pull Request #3133 · google/zerocopy

jswrenn · 2026-03-17T16:57:26Z

Latest Update: v6 — Compare vs v5

📚 Full Patch History

Links show the diff between the row version and the column version.

Version	v5	v4	v3	v2	v1	Base
v6	vs v5	vs v4	vs v3	vs v2	vs v1	vs Base
v5		vs v4	vs v3	vs v2	vs v1	vs Base
v4			vs v3	vs v2	vs v1	vs Base
v3				vs v2	vs v1	vs Base
v2					vs v1	vs Base
v1						vs Base

⬇️ Download this PR

Branch

git fetch origin refs/heads/G74bf0db1b7c30669171b5985ce24849a264746e7 && git checkout -b pr-G74bf0db1b7c30669171b5985ce24849a264746e7 FETCH_HEAD

Checkout

git fetch origin refs/heads/G74bf0db1b7c30669171b5985ce24849a264746e7 && git checkout FETCH_HEAD

Cherry Pick

git fetch origin refs/heads/G74bf0db1b7c30669171b5985ce24849a264746e7 && git cherry-pick FETCH_HEAD

Pull

git pull origin refs/heads/G74bf0db1b7c30669171b5985ce24849a264746e7

Stacked PRs enabled by GHerrit.

jswrenn · 2026-03-17T16:58:59Z

benches/ref_from_bytes_dynamic_padding.x86-64.mca

+Instructions:      2200
+Total Cycles:      783
+Total uOps:        2500


jswrenn · 2026-03-17T16:59:33Z

benches/ref_from_prefix_dynamic_padding.x86-64.mca

+Instructions:      2100
+Total Cycles:      673
+Total uOps:        2200


gemini-code-assist · 2026-03-17T16:59:42Z

Summary of Changes

Hello, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request focuses on optimizing the ref_from_bytes functions by reordering validation steps and tweaking assembly implementations. The goal is to enhance performance, particularly in scenarios involving dynamic padding and size calculations. The changes are primarily experimental and include updates to benchmark files to reflect performance analysis.

Highlights

Reordering Validation: This pull request experiments with reversing the order of alignment and size validation in the ref_from_bytes functions to potentially improve performance.
Assembly Optimization: Assembly code in several benchmark files was modified to optimize the ref_from_bytes functions, focusing on dynamic padding and size calculations.
Performance Benchmarking: The changes include updates to .mca files, reflecting performance measurements and resource pressure analysis after the assembly code modifications.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Activity

The pull request is marked as a work in progress (wip).
It is linked to several related issues: [wip] experiment with reversing alignment and size validation order #3133, Optimize bounds checking by validating metadata against maximum #3123, Optimize new_box #3120, [codegen] Break build dependency between tests and lib #3116, and [codegen] Test and document codegen for FromZeros #3114.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	`/gemini review`	Performs a code review for the current pull request in its current state.
Pull Request Summary	`/gemini summary`	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	`/gemini help`	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

jswrenn · 2026-03-17T17:00:17Z

benches/ref_from_suffix_dynamic_padding.x86-64.mca

 Iterations:        100
 Instructions:      2000
-Total Cycles:      682
+Total Cycles:      683


extremely marginal regression

jswrenn · 2026-03-17T17:00:46Z

benches/ref_from_suffix_dynamic_size.x86-64.mca

+Instructions:      1300
+Total Cycles:      405
+Total uOps:        1300


regression?

jswrenn · 2026-03-17T17:04:08Z

benches/ref_from_suffix_static_size.x86-64.mca

-Total Cycles:      338
-Total uOps:        1100
+Instructions:      1000
+Total Cycles:      404


regression?

jswrenn · 2026-03-17T17:04:19Z

benches/ref_from_suffix_with_elems_dynamic_size.x86-64.mca

+Instructions:      2000
+Total Cycles:      604
+Total uOps:        2000


jswrenn · 2026-03-17T17:04:28Z

benches/try_ref_from_bytes_dynamic_padding.x86-64.mca

+Instructions:      2500
+Total Cycles:      1008
+Total uOps:        2800


jswrenn · 2026-03-17T17:04:40Z

benches/try_ref_from_bytes_dynamic_size.x86-64.mca

 Instructions:      2000
-Total Cycles:      639
-Total uOps:        2100
+Total Cycles:      641


marginal regression?

jswrenn · 2026-03-17T17:04:55Z

benches/try_ref_from_prefix_dynamic_padding.x86-64.mca

+Instructions:      2800
+Total Cycles:      910
+Total uOps:        3100


jswrenn · 2026-03-17T17:05:21Z

benches/try_ref_from_prefix_static_size.x86-64.mca

+Instructions:      1300
+Total Cycles:      407
+Total uOps:        1400


jswrenn · 2026-03-17T17:06:44Z

benches/try_ref_from_suffix_dynamic_padding.x86-64.mca

 Iterations:        100
 Instructions:      2300
-Total Cycles:      791
+Total Cycles:      797


marginal regression?

jswrenn · 2026-03-17T17:07:46Z

benches/try_ref_from_suffix_with_elems_dynamic_size.x86-64.mca

+Instructions:      2500
+Total Cycles:      1105
+Total uOps:        2700


gemini-code-assist

Code Review

This pull request experiments with reordering the alignment and size validation checks in validate_cast_and_convert_metadata. While the logical correctness is maintained, the performance impact is mixed. Some benchmarks show significant improvements (e.g., ref_from_bytes_dynamic_size), while others show significant regressions (e.g., ref_from_suffix_with_elems_dynamic_padding). The original order performs the cheaper alignment check before the more expensive size validation, which is a good heuristic for failing fast. The new order might allow for better code generation in some cases by grouping checks differently, but it seems to hurt performance in others. Given the mixed results, it would be beneficial to understand the trade-offs better. Perhaps there's a way to get the benefits without the regressions.

gemini-code-assist · 2026-03-17T17:08:41Z

src/layout.rs

+        // Alignment checks go in their own block to avoid introducing variables
+        // into the top-level scope.
+        {
+            // We check alignment for `addr` (for prefix casts) or `addr +
+            // bytes_len` (for suffix casts). For a prefix cast, the correctness
+            // of this check is trivial - `addr` is the address the object will
+            // live at.
+            //
+            // For a suffix cast, we know that all valid sizes for the type are
+            // a multiple of the alignment (and by safety precondition, we know
+            // `DstLayout` may only describe valid Rust types). Thus, a
+            // validly-sized instance which lives at a validly-aligned address
+            // must also end at a validly-aligned address. Thus, if the end
+            // address for a suffix cast (`addr + bytes_len`) is not aligned,
+            // then no valid start address will be aligned either.
+            let offset = match cast_type {
+                CastType::Prefix => 0,
+                CastType::Suffix => bytes_len,
+            };
+
+            // Addition is guaranteed not to overflow because `offset <=
+            // bytes_len`, and `addr + bytes_len <= usize::MAX` is a
+            // precondition of this method. Modulus is guaranteed not to divide
+            // by 0 because `align` is non-zero.
+            #[allow(clippy::arithmetic_side_effects)]
+            if (addr + offset) % self.align.get() != 0 {
+                return Err(MetadataCastError::Alignment);
+            }
+        }


Moving the alignment check after the size validation has mixed performance results across the benchmarks. While some cases like ref_from_bytes_dynamic_size improve, others like ref_from_suffix_with_elems_dynamic_padding regress significantly.

The original order follows the fail-fast principle by performing the cheaper alignment check first. This reordering seems to be beneficial only in certain scenarios. It might be worth investigating what distinguishes the improved cases from the regressed ones to see if a more nuanced approach is possible, perhaps conditionally ordering the checks based on properties of the DstLayout.

gherrit-pr-id: G74bf0db1b7c30669171b5985ce24849a264746e7

This was referenced Mar 17, 2026

[codegen] Test and document codegen for FromZeros #3114

Merged

[codegen] Break build dependency between tests and lib #3116

Open

Optimize new_box #3120

Closed

Optimize bounds checking by validating metadata against maximum #3123

Open

jswrenn commented Mar 17, 2026

View reviewed changes

gemini-code-assist bot reviewed Mar 17, 2026

View reviewed changes

jswrenn force-pushed the G74bf0db1b7c30669171b5985ce24849a264746e7 branch from ec58f2e to 799dad9 Compare March 17, 2026 19:37

jswrenn force-pushed the Gcf66958135e905f7cd7d2fac87d9f881e5e5185f branch 2 times, most recently from 67f28a5 to a08a5c9 Compare March 20, 2026 17:47

jswrenn force-pushed the G74bf0db1b7c30669171b5985ce24849a264746e7 branch from 799dad9 to 8bd6bea Compare March 20, 2026 17:47

jswrenn force-pushed the Gcf66958135e905f7cd7d2fac87d9f881e5e5185f branch from a08a5c9 to c30e455 Compare March 20, 2026 18:00

jswrenn force-pushed the G74bf0db1b7c30669171b5985ce24849a264746e7 branch from 8bd6bea to 0de1acc Compare March 20, 2026 18:00

[wip] experiment with reversing alignment and size validation order

f03564b

gherrit-pr-id: G74bf0db1b7c30669171b5985ce24849a264746e7

jswrenn force-pushed the Gcf66958135e905f7cd7d2fac87d9f881e5e5185f branch from c30e455 to ac240ed Compare March 20, 2026 19:41

jswrenn force-pushed the G74bf0db1b7c30669171b5985ce24849a264746e7 branch from 0de1acc to f03564b Compare March 20, 2026 19:41

Conversation

jswrenn commented Mar 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot commented Mar 17, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

gemini-code-assist bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

gemini-code-assist bot Mar 17, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

jswrenn commented Mar 17, 2026 •

edited

Loading