variable-precision float operations can differ depending on optimization levels #124609
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
rustbot
added
S-waiting-on-review
|
I don't think this is a good idea. Why should libm functions be privileged in this way? They ought to be treated as ordinary library calls. |
|
This describes the current reality. Feel free to make a proposal for changing our semantics, but I don't think that should block documenting the current semantics. In particular this is future compatible with changing them later to be optimization-independent.
Also note that we are not really privileging anything here. What we are effectively saying is that these functions are returning a result with arbitrary nondeterministic precision. It then follows as a consequence that the precision can depend on host, target, compiler version, flags, or literally anything else. This is generally always the case when a function says that there are multiple possible results, so there's nothing special about libm functions here. It's just something people are also often confused by so it may be worth explaining these unexpected consequences of non-determinism.
Maybe we should mention the non-determinism more explicitly.
|
|
How about something like "The precision of this function is non-deterministic, meaning each call to this function may return a value with a different precision even if given the same inputs"? |
RalfJung
force-pushed
the
float-precision
branch
2 times, most recently
from
7de46e5 to
8a012c0
Compare
|
Yeah, something like that -- I updated the PR. The section title "Platform-specific precision" doesn't make a lot of sense any more now, since precision is not "specific" to any platform, so I changed that as well. |
|
Works for me! @bors r+ rollup |
bors
added
S-waiting-on-bors
fmease
added a commit
to fmease/rust
that referenced
this pull request
May 3, 2024
variable-precision float operations can differ depending on optimization levels Follow-up to rust-lang#121793 and rust-lang#118217 that accounts for optimizations changing the precision of these functions. Fixes rust-lang#109118 Fixes rust-lang#71355
|
Ah, I like this a lot better. At least the danger of using them is much clearer now. |
fmease
added a commit
to fmease/rust
that referenced
this pull request
May 3, 2024
variable-precision float operations can differ depending on optimization levels Follow-up to rust-lang#121793 and rust-lang#118217 that accounts for optimizations changing the precision of these functions. Fixes rust-lang#109118 Fixes rust-lang#71355
bors
added a commit
to rust-lang-ci/rust
that referenced
this pull request
May 3, 2024
Rollup of 8 pull requests Successful merges: - rust-lang#124412 (io safety: update Unix explanation to use `Arc`) - rust-lang#124441 (String.truncate comment microfix (greater or equal)) - rust-lang#124594 (run-make-support: preserve tooks.mk behavior for EXTRACXXFLAGS) - rust-lang#124604 (library/std: Remove unused `gimli-symbolize` feature) - rust-lang#124607 (`rustc_expand` cleanups) - rust-lang#124609 (variable-precision float operations can differ depending on optimization levels) - rust-lang#124610 (Tweak `consts_may_unify`) - rust-lang#124612 (Add support for inputing via stdin with run-make-support) r? `@ghost` `@rustbot` modify labels: rollup
fmease
added a commit
to fmease/rust
that referenced
this pull request
May 3, 2024
variable-precision float operations can differ depending on optimization levels Follow-up to rust-lang#121793 and rust-lang#118217 that accounts for optimizations changing the precision of these functions. Fixes rust-lang#109118 Fixes rust-lang#71355
bors
added a commit
to rust-lang-ci/rust
that referenced
this pull request
May 3, 2024
Rollup of 7 pull requests Successful merges: - rust-lang#124441 (String.truncate comment microfix (greater or equal)) - rust-lang#124594 (run-make-support: preserve tooks.mk behavior for EXTRACXXFLAGS) - rust-lang#124604 (library/std: Remove unused `gimli-symbolize` feature) - rust-lang#124607 (`rustc_expand` cleanups) - rust-lang#124609 (variable-precision float operations can differ depending on optimization levels) - rust-lang#124610 (Tweak `consts_may_unify`) - rust-lang#124637 (AST pretty: Use `builtin_syntax` for type ascription) Failed merges: - rust-lang#124638 (Move some tests from `rustc_expand` to `rustc_parse`.) r? `@ghost` `@rustbot` modify labels: rollup
bors
added a commit
to rust-lang-ci/rust
that referenced
this pull request
May 3, 2024
…iaskrgr Rollup of 10 pull requests Successful merges: - rust-lang#123480 (deref patterns: impl `DerefPure` for more std types) - rust-lang#124412 (io safety: update Unix explanation to use `Arc`) - rust-lang#124441 (String.truncate comment microfix (greater or equal)) - rust-lang#124594 (run-make-support: preserve tooks.mk behavior for EXTRACXXFLAGS) - rust-lang#124604 (library/std: Remove unused `gimli-symbolize` feature) - rust-lang#124607 (`rustc_expand` cleanups) - rust-lang#124609 (variable-precision float operations can differ depending on optimization levels) - rust-lang#124610 (Tweak `consts_may_unify`) - rust-lang#124626 (const_eval_select: add tracking issue) - rust-lang#124637 (AST pretty: Use `builtin_syntax` for type ascription) Failed merges: - rust-lang#124638 (Move some tests from `rustc_expand` to `rustc_parse`.) r? `@ghost` `@rustbot` modify labels: rollup
rust-timer
added a commit
to rust-lang-ci/rust
that referenced
this pull request
May 3, 2024
Rollup merge of rust-lang#124609 - RalfJung:float-precision, r=cuviper variable-precision float operations can differ depending on optimization levels Follow-up to rust-lang#121793 and rust-lang#118217 that accounts for optimizations changing the precision of these functions. Fixes rust-lang#109118 Fixes rust-lang#71355
|
@bors r- |
bors
added
S-waiting-on-author
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Follow-up to #121793 and #118217 that accounts for optimizations changing the precision of these functions.
Fixes #109118
Fixes #71355