What the rusty_alloc note measured at 2 mebibytes
A rusty_alloc field note says a 2 mebibyte allocate and free became 57 percent cheaper. The same note says a rusty_zstd speedup measured the allocator change as neutral.
The short version
An allocate and free that it says became 57 percent cheaper. The same note says rusty_zstd 0.3.0 is faster at levels 1 and 3, and that the allocator change measured neutral for that gain.
What the note measured
The October field note on rusty_alloc says a 2 mebibyte allocate and free became 57 percent cheaper. It also says rusty_zstd 0.3.0 is faster at levels 1 and 3, and that the same release measured the allocator change as neutral for that speedup. The dependency is those two results: a cheaper large allocate, and a compressor whose gain the note does not attribute to the allocator. The measurement note is the rusty_alloc field note.
What the standard library actually specifies
Rust's alloc documentation describes how a program allocates and how a global allocator can be installed. It does not publish a 57 percent figure, and it does not name this crate. A reader who wants the interface reads the docs. A reader who wants the measurement reads the note. Memory safety in a release note, a different software fact, is what memory safety in the release notes requires.
What the standing page is for
The 57 percent figure is one size: 2 mebibytes, allocate and free. It is not a ranking of allocators and not a result the Rust alloc module publishes. The standing crate page can describe how a program embeds the allocator without becoming the benchmark. The neutral result is the sentence a reader is most likely to drop: rusty_zstd 0.3.0 is faster at levels 1 and 3, and the note measured the allocator change as not the cause. Keeping the speedup and discarding the cause would misread the note. A later remeasure can move the percentage. A change to the global-allocator trait would be a language change sitting beside the number, not a correction of it. Firefox's ship date is a different software fact from the same day and does not bear on this table. The interface, the description, and the measurement stay in three documents so the percentage cannot migrate into the standard library. Levels 1 and 3 are the compressor settings the note names. Neutral means the allocator was measured and was not the explanation. A reader who wants a single winner between the allocator and the compressor will not get one from these pages, because the note refused to award one. The standard library page cannot break the tie either, because it never entered the race. The standing rusty_alloc page describes the allocator as a library a program can embed. It is not the benchmark table. Treating the standing page as the source of the 57 percent figure would attach a number to the wrong document. The desk is Software.
Why it matters on this desk
The dependency is a cost at one size, and a non-result beside it. The note says the allocator got cheaper at 2 mebibytes and says the compressor did not get faster because of that change. Both sentences are the finding. A reader who keeps only the 57 percent has dropped the result the note was careful about. A version number, which this percentage is not, is what a version number has to mean.
What this page will not generalize
Read together, the filed pages are the record, and a later paraphrase would not replace them. Fifty-seven percent at 2 mebibytes is not a claim about 64-byte allocations, and it is not a claim about every Rust program. The neutral compressor result is the note's way of saying a nearby speedup had a different cause. Firefox 158, a different software document this week, is what Firefox 158's notes actually date.
Where the documents split
The measurement note has the percentages. The standing page has the description of the crate. The alloc module has the language interface. They overlap only in the topic, allocation. They do not overlap in a shared benchmark. This page will not average them into one number the standard library never published. The front of the paper is Futureweb.
What to watch next
A later note that remeasures 2 mebibytes, or that attributes a compressor change to the allocator after all, would replace these two sentences. A change to Rust's global-allocator trait would be a language change, not a correction of the 57 percent. Another summary of the same note would not.
What would change the record
If the field note revises 57 percent, the figure moves. If it revises the neutral result, the compressor sentence moves. The standard-library page can stay as it is through either revision, because it never hosted the number. That separation is why the brief cites three documents instead of one.
What stays a measurement
What stays specific is an October field note: a 2 mebibyte allocate and free in rusty_alloc, 57 percent cheaper, and a rusty_zstd 0.3.0 result at levels 1 and 3 that the note measured as neutral with respect to the allocator change. The standing crate page describes the allocator and is not the table. Rust's alloc module describes the interface a global allocator plugs into and does not mention the percentage. This page does not paste the note's method, does not invent a second size, and does not move the compressor speedup onto the allocator. The two results are the brief. One is a cost. The other is a cause the note declined to claim. A reader who wants a universal allocator ranking will not find it here, because none of the three documents is that ranking.
Sources
The reports this brief is filing. Futureweb did not republish them.
Questions
What does the October 2026 rusty_alloc note say changed for a 2 mebibyte allocation?
The note says an allocate and free of 2 mebibytes became 57 percent cheaper. That figure is the note's measurement, not a number from the Rust standard library.
What does the same note say the allocator change did to rusty_zstd 0.3.0?
It says rusty_zstd 0.3.0 is faster at compression levels 1 and 3, and that the release measured the allocator change as neutral for that result.
What does Rust's alloc module document, as opposed to the field note?
The standard library's allocation interface, including the global allocator a program can install. It does not mention rusty_alloc or the 57 percent figure.
Does the standing rusty_alloc page replace the measurement note?
No. The standing page describes the allocator. The measurement note is the October claim about a 2 mebibyte allocate and free. This brief keeps them apart.
Should a reader treat 57 percent as a language-wide allocator result?
No. It is the field note's result for one size, on the allocator the note is writing about. Another size, or another allocator, is a different measurement.