float: Replace some approximate assertions with exact #141669

tgross35 · 2025-05-28T01:36:31Z

As was mentioned at 1, we currently use assert_approx_eq for testing
some math functions that guarantee exact results. Replace approximate
assertions with exact ones for the following:

ceil
floor
fract
from_bits
mul_add
round_ties_even
round
trunc

This likely wasn't done in the past to avoid writing out exact decimals
that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004),
but ensuring our results are accurate seems more important here.

The first commit is a small bit of macro cleanup.

try-job: aarch64-gnu
try-job: x86_64-gnu-aux

cc @beetrees as you suggested this

tgross35 · 2025-05-28T01:36:38Z

@bors try

bors · 2025-05-28T01:37:48Z

⌛ Trying commit f12c89e with merge 6f9d015...

float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: #138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu

rustbot · 2025-05-28T01:43:37Z

r? @Mark-Simulacrum

rustbot has assigned @Mark-Simulacrum.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

bors · 2025-05-28T02:41:57Z

💔 Test failed - checks-actions

tgross35 · 2025-05-28T16:47:10Z

@beetrees any idea what is going on here? It looks like the f128 fma and frac tests aren't giving the expected results with clif.

library/coretests/tests/floats/f128.rs

tgross35 · 2025-05-28T20:31:51Z

@bors try

float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: #138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu

bors · 2025-05-28T20:36:28Z

⌛ Trying commit b94d32e with merge f2c471b...

bors · 2025-05-28T22:36:54Z

☀️ Try build successful - checks-actions
Build commit: f2c471b (f2c471b102998ac17e9717af5c6da4647473717b)

tgross35 · 2025-05-28T22:39:13Z

@rustbot ready

RalfJung · 2025-05-29T14:05:20Z

library/coretests/tests/floats/f128.rs

@@ -260,98 +249,98 @@ fn test_classify() {
 #[cfg(not(miri))]


Miri has f16 and f128 implementations for all the rounding functions by now -- I don't think there's still a good reason for these tests to be disabled in Miri?

Not at all, I just wasn't aware of the change. I added 2e4824d to remove the #[cfg].

(which try job runs miri again?)

(which try job runs miri again?)

The gnu-aux one.

RalfJung · 2025-05-29T14:06:34Z

library/coretests/tests/floats/f16.rs

+    assert_eq!(0.0f16.floor(), 0.0f16);
+    assert_eq!((-0.0f16).floor(), -0.0f16);


This test does not compare the sign since == ignores the sign of 0 -- that's pre-existing, but likely unintentional?

You're very right. I added 0a87911 to just switch everything to assert_biteq which should catch things like this, since I don't think there is any downside.

RalfJung · 2025-05-29T14:56:25Z

library/coretests/tests/floats/f128.rs

@@ -257,114 +247,107 @@ fn test_classify() {
 }

 #[test]
-#[cfg(not(miri))]
 #[cfg(target_has_reliable_f128_math)]


Ideally this would become any(miri, target_has_reliable_f128_math) since f16/f128 math is always reliable with Miri... not sure if that's worth the churn though.

Might as well since I'm updating those configs anyway :) Done in a fixup

RalfJung · 2025-05-29T14:56:45Z

library/coretests/tests/floats/f128.rs

@@ -71,23 +60,23 @@ fn test_num_f128() {
 fn test_num_f128_rem() {
    let ten = 10f128;
    let two = 2f128;
-    assert_eq!(ten.rem(two), ten % two);
+    assert_biteq!(ten.rem(two), ten % two);
 }

 #[test]
 #[cfg(not(miri))]
 #[cfg(target_has_reliable_f128_math)]
 fn test_min_nan() {


min and max (and minimum and maximum) should also work in miri for f16 and f128

Updated. powi, fma, and sqrt all still do not work, correct?

Correct (though they'd be fairly easy to add :)

RalfJung · 2025-05-29T14:58:21Z

library/coretests/tests/floats/f128.rs

@@ -71,23 +60,23 @@ fn test_num_f128() {
 fn test_num_f128_rem() {
    let ten = 10f128;
    let two = 2f128;
-    assert_eq!(ten.rem(two), ten % two);
+    assert_biteq!(ten.rem(two), ten % two);


Isn't rem the same as %? I'm a bit confused by the test.^^

Anyway I think this always worked in Miri ever nice basic arithmetic worked, there's no intrinsics involved here.

This test came from

rust/library/coretests/tests/floats/mod.rs

Lines 19 to 35 in 8afd710

pub fn test_num<T>(ten: T, two: T)

where

T: PartialEq

+ Add<Output = T>

+ Sub<Output = T>

+ Mul<Output = T>

+ Div<Output = T>

+ Rem<Output = T>

+ fmt::Debug

+ Copy,

{

assert_eq!(ten.add(two), ten + two);

assert_eq!(ten.sub(two), ten - two);

assert_eq!(ten.mul(two), ten * two);

assert_eq!(ten.div(two), ten / two);

assert_eq!(ten.rem(two), ten % two);

}

, indeed it is a bit dubious. I don't see anything else that exercises Rem but that would probably be something good to have.

float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: #138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu try-job: gnu-aux

tgross35 · 2025-05-29T15:36:07Z

🤦
@bors try

bors · 2025-05-29T15:37:16Z

⌛ Trying commit 9c047e7 with merge a609092...

float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: #138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu try-job: x86_64-gnu-aux

bors · 2025-05-29T17:38:53Z

☀️ Try build successful - checks-actions
Build commit: a609092 (a6090927d4f33b36e2998fdaa6d76908ba43c50e)

tgross35 · 2025-05-29T17:45:18Z

Cleaned up history a bit.

Considering you've already looked at this and other float tests, r? @RalfJung

library/coretests/tests/floats/mod.rs

RalfJung · 2025-05-29T21:12:03Z

Thanks! r=me after squashing

Clean up the separate `assert_f{16,32,64,128}` macros with a single `assert_biteq!` macro that works for all float widths.

As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: rust-lang#138087 (comment)

The rounding tests are now supported, so there is no longer any reason to skip these.

`assert_eq!` ignores the sign of zero, but for any tests involving zeros we do care about this sign. Replace `assert_eq!` with `assert_biteq!` everywhere possible for float tests to ensure we don't miss this. `assert_biteq!` is also updated to check equality on non-NaNs, to catch the unlikely case that bitwise equality works but our `==` implementation is broken. There is one notable output change: we were asserting that `(-0.0).fract()` and `(-1.0).fract()` both return -0.0, but both actually return +0.0.

tgross35 · 2025-05-29T21:14:32Z

Thank you!
@bors r=RalfJung

bors · 2025-05-29T21:14:34Z

📌 Commit 70cce1c has been approved by RalfJung

It is now in the queue for this repository.

Rollup of 8 pull requests Successful merges: - #133823 (Use `cfg_attr_trace` in AST with a placeholder attribute for accurate suggestion) - #141004 (Report text_direction_codepoint_in_literal when parsing) - #141407 (Refactor the two-phase check for impls and impl items) - #141430 (remove `visit_clobber` and move `DummyAstNode` to `rustc_expand`) - #141507 (atomic_load intrinsic: use const generic parameter for ordering) - #141538 (implement `va_arg` for x86_64 systemv) - #141669 (float: Replace some approximate assertions with exact) - #141747 (rustdoc: display doc(cfg(false)) properly) r? `@ghost` `@rustbot` modify labels: rollup

Rollup merge of #141669 - tgross35:float-test-cleanup, r=RalfJung float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: #138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu try-job: x86_64-gnu-aux

Rollup of 8 pull requests Successful merges: - rust-lang/rust#133823 (Use `cfg_attr_trace` in AST with a placeholder attribute for accurate suggestion) - rust-lang/rust#141004 (Report text_direction_codepoint_in_literal when parsing) - rust-lang/rust#141407 (Refactor the two-phase check for impls and impl items) - rust-lang/rust#141430 (remove `visit_clobber` and move `DummyAstNode` to `rustc_expand`) - rust-lang/rust#141507 (atomic_load intrinsic: use const generic parameter for ordering) - rust-lang/rust#141538 (implement `va_arg` for x86_64 systemv) - rust-lang/rust#141669 (float: Replace some approximate assertions with exact) - rust-lang/rust#141747 (rustdoc: display doc(cfg(false)) properly) r? `@ghost` `@rustbot` modify labels: rollup

…fJung float: Replace some approximate assertions with exact As was mentioned at [1], we currently use `assert_approx_eq` for testing some math functions that guarantee exact results. Replace approximate assertions with exact ones for the following: * `ceil` * `floor` * `fract` * `from_bits` * `mul_add` * `round_ties_even` * `round` * `trunc` This likely wasn't done in the past to avoid writing out exact decimals that don't match the intuitive answer (e.g. 1.3 - 1.0 = 0.300...004), but ensuring our results are accurate seems more important here. [1]: rust-lang#138087 (comment) The first commit is a small bit of macro cleanup. try-job: aarch64-gnu try-job: x86_64-gnu-aux

rustbot added the T-libs Relevant to the library team, which will review and decide on the PR/issue. label May 28, 2025

tgross35 marked this pull request as ready for review May 28, 2025 01:43

rustbot assigned Mark-Simulacrum May 28, 2025

rustbot added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label May 28, 2025

This comment has been minimized.

Sign in to view

bors added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 28, 2025

beetrees suggested changes May 28, 2025

View reviewed changes

library/coretests/tests/floats/f128.rs Outdated Show resolved Hide resolved

library/coretests/tests/floats/f128.rs Outdated Show resolved Hide resolved

tgross35 force-pushed the float-test-cleanup branch from b94d32e to 04570ba Compare May 28, 2025 22:39

rustbot added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels May 28, 2025

tgross35 mentioned this pull request May 29, 2025

Unify and deduplicate float tests #141726

Open

RalfJung reviewed May 29, 2025

View reviewed changes

This comment has been minimized.

Sign in to view

This comment was marked as outdated.

Sign in to view

tgross35 force-pushed the float-test-cleanup branch from 9c047e7 to 9f9ac74 Compare May 29, 2025 17:43

rustbot assigned RalfJung and unassigned Mark-Simulacrum May 29, 2025

RalfJung reviewed May 29, 2025

View reviewed changes

library/coretests/tests/floats/mod.rs Outdated Show resolved Hide resolved

tgross35 added 4 commits May 29, 2025 21:13

float: Use a shared assert_biteq! macro for tests

6a79b27

Clean up the separate `assert_f{16,32,64,128}` macros with a single `assert_biteq!` macro that works for all float widths.

float: Enable some f16 and f128 rounding tests on miri

5446ba3

The rounding tests are now supported, so there is no longer any reason to skip these.

tgross35 force-pushed the float-test-cleanup branch from 7de59e6 to 70cce1c Compare May 29, 2025 21:13

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels May 29, 2025

matthiaskrgr mentioned this pull request May 30, 2025

Rollup of 8 pull requests #141753

Merged

bors merged commit 3ebdd59 into rust-lang:master May 30, 2025
9 checks passed

rustbot added this to the 1.89.0 milestone May 30, 2025

tgross35 deleted the float-test-cleanup branch May 30, 2025 14:36

		assert_eq!(0.0f16.floor(), 0.0f16);
		assert_eq!((-0.0f16).floor(), -0.0f16);

	pub fn test_num<T>(ten: T, two: T)
	where
	T: PartialEq
	+ Add<Output = T>
	+ Sub<Output = T>
	+ Mul<Output = T>
	+ Div<Output = T>
	+ Rem<Output = T>
	+ fmt::Debug
	+ Copy,
	{
	assert_eq!(ten.add(two), ten + two);
	assert_eq!(ten.sub(two), ten - two);
	assert_eq!(ten.mul(two), ten * two);
	assert_eq!(ten.div(two), ten / two);
	assert_eq!(ten.rem(two), ten % two);
	}

float: Replace some approximate assertions with exact #141669

float: Replace some approximate assertions with exact #141669

Uh oh!

Conversation

tgross35 commented May 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tgross35 commented May 28, 2025

Uh oh!

bors commented May 28, 2025

Uh oh!

rustbot commented May 28, 2025

Uh oh!

This comment has been minimized.

This comment has been minimized.

bors commented May 28, 2025

Uh oh!

tgross35 commented May 28, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tgross35 commented May 28, 2025

Uh oh!

bors commented May 28, 2025

Uh oh!

bors commented May 28, 2025

Uh oh!

tgross35 commented May 28, 2025

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

This comment has been minimized.

This comment was marked as outdated.

This comment was marked as outdated.

tgross35 commented May 29, 2025

Uh oh!

bors commented May 29, 2025

Uh oh!

bors commented May 29, 2025

Uh oh!

tgross35 commented May 29, 2025

Uh oh!

Uh oh!

RalfJung commented May 29, 2025

Uh oh!

tgross35 commented May 29, 2025

Uh oh!

bors commented May 29, 2025

Uh oh!

Uh oh!

Uh oh!

tgross35 commented May 28, 2025 •

edited

Loading

tgross35 commented May 28, 2025 •

edited

Loading