Skip to content

Grow native buffers with realloc - #1983

Merged
sbillig merged 3 commits into
argotorg:masterfrom
sbillig:perf/realloc-growth
Nov 11, 2026
Merged

sbillig merged 3 commits into
argotorg:masterfrom
sbillig:perf/realloc-growth

Conversation

@sbillig

@sbillig sbillig commented Nov 11, 2026

Copy link
Copy Markdown
Collaborator

MemVec and MemBuffer grew by allocating the new capacity, copying the elements and freeing the old block. Growth now goes through new core::ptr::realloc_raw and try_realloc_raw. Native code resizes with C's realloc, which extends a block where it lies or, on glibc, moves a mapped one with mremap. EVM memory and WebAssembly modules still allocate and copy.

A native resize to at most 1032 bytes still goes through malloc and free, because realloc bypasses glibc's per-thread cache.

fevm's LOG0 benchmark with 1 KB logs, per transaction: 1.17M to 1.08M instructions, about 390K to 260K user cycles, and with glibc's default malloc settings 956 to 15 page faults.

A program that replaces malloc and free must now replace realloc too.

`MemVec` and `MemBuffer` grew by allocating the new capacity, copying
the elements and freeing the old block. Add `core::ptr::realloc_raw`
and `try_realloc_raw`, which growth now uses. Native code resizes with
C's `realloc`, which extends a block where it lies or, on glibc, moves a
mapped one with `mremap`; EVM memory and WebAssembly modules still
allocate and copy.

A native resize to at most 1032 bytes still moves through `malloc` and
`free`: `realloc` bypasses glibc's per-thread cache, and with it there
a fevm transfer took about 250 more instructions.

fevm's LOG0 benchmark with 1 KB logs, per transaction: 1.17M to 1.08M
instructions and about 390K to 260K user cycles; with glibc's default
malloc settings, 956 to 15 page faults and 1.6M to about 30K kernel
cycles.

A program that replaces `malloc` and `free` must now replace `realloc`
too, as the test allocators do.
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 11, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-11T21:05:06.698227Z 1899ed7 New commits
ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 11d2cde440

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

  • Open a pull request for review
  • Mark a draft as ready
  • Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread ingots/core/src/ptr.fe Resolved
`grown_capacity` returned at least `initial` even when `max`, the
elements that fit in `MAX_ALLOCATION` bytes, was smaller. For an empty
`MemVec` of elements over a quarter of `MAX_ALLOCATION`, such as a
1 GiB array on 32-bit WebAssembly, `try_reserve(1)` asked for four, and
the byte size overflowed with `Panic(0x11)` instead of returning
`false`. `try_alloc_elems` had rejected that capacity before growth
went through `try_realloc_raw`. Clamp the initial capacity to `max`.
@sbillig
sbillig merged commit 08359e3 into argotorg:master Nov 11, 2026
10 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant