The tests were using `rug::ln_gamma` as a reference for `libm::lgamma`,
which actually computes the natural logarithm *of the absolute value* of
the Gamma function.
This changes the range of inputs used for the icount-benchmarks of these
functions, which causes false regressions in [1].
[1]: https://github.com/rust-lang/compiler-builtins/actions/runs/21788698368/job/62864230903?pr=1075#step:7:2710.
Fixes: a1a066611dc2 ("Create interfaces for testing against MPFR")
The acosh functions were incorrectly returning finite values for some
negative inputs (should be NaN for any `x < 1.0`)
The bug was inherited when originally ported from musl, and this patch
follows their fix for single-precision acoshf in [1].
A similar fix is applied to acosh, though musl still has an incorrect
implementation requiring tests against that basis to be skipped.
[1]: https://git.musl-libc.org/cgit/musl/commit/?id=c4c38e6364323b6d83ba3428464e19987b981d7a
[ added context to message - Trevor ]
In recent nightlies we are hitting errors like the following:
error: an associated constant with this name may be added to the standard library in the future
--> libm/src/math/support/float_traits.rs:248:48
|
248 | const SIGN_MASK: Self::Int = 1 << (Self::BITS - 1);
| ^^^^^^^^^^
...
324 | / float_impl!(
325 | | f32,
326 | | u32,
327 | | i32,
... |
333 | | fmaf32
334 | | );
| |_- in this macro invocation
|
= warning: once this associated item is added to the standard library, the ambiguity may cause an error or change in behavior!
= note: for more information, see issue #48919 <https://github.com/rust-lang/rust/issues/48919>
= note: `-D unstable-name-collisions` implied by `-D warnings`
= help: to override `-D warnings` add `#[allow(unstable_name_collisions)]`
= note: this error originates in the macro `float_impl` (in Nightly builds, run with -Z macro-backtrace for more info)
help: use the fully qualified path to the associated const
|
248 - const SIGN_MASK: Self::Int = 1 << (Self::BITS - 1);
248 + const SIGN_MASK: Self::Int = 1 << (<f32 as float_traits::Float>::BITS - 1);
|
help: add `#![feature(float_bits_const)]` to the crate attributes to enable `core::f32::<impl f32>::BITS`
--> libm/src/lib.rs:26:1
|
26 + #![feature(float_bits_const)]
|
Using fully qualified syntax is verbose and `BITS` only exists since
recently, so allow this lint instead.
Due to an erroneous overflow threshold, `expm1f` was incorrectly
returning `inf` for inputs in the range `[88.72169, 88.72283]`. This
additionally caused `sinhf` to return `NaN` for inputs in that range.
The bug was ported from the original in musl, which has since been fixed
in [1].
[1]: https://git.musl-libc.org/cgit/musl/commit/?id=964104f9f0e056cf58d9defa0b716d7756f040f6
`iai-callgrind` was renamed to `gungraun` and had a new release. Update
everything to match.
There shouldn't be any changes to observable behavior here.
New utility in `libm::support`:
- `trait NarrowingDiv` for dividing `u2N / uN` when the quotient fits in
`uN`
- a reasonable implementation of that for primitives up to `u256 / u128`
This is the inverse operation of unsigned widening multiplication:
let xy: u256 = u128::widen_mul(x, y);
assert_eq!(xy.checked_narrowing_div_rem(y), Some((x, 0))); // unless y == 0
The trait API is based on x86's `div`-instruction: quotient overflow
happens exactly when the high half of the dividend is greater or equal
to the divisor, which includes division by zero.
`i256` and `u256`
- operators now use the same overflow convention as primitives
- implement `<<` and `-` (previously just `>>` and `+`)
- implement `Ord` correctly (the previous `PartialOrd` was broken)
- correct `i256::SIGNED` to `true`
The `Int`-trait is extended with `trailing_zeros`, `carrying_add`, and
`borrowing_sub`.
The submodule was causing issues in rust-lang/rust, so eliminiate it
here. `build-musl` is also removed from `libm-test`'s default features
so the crate doesn't need to be built by default.
`binop_common` emits a `SKIP` that is intended to apply only to
`copysign`, but is instead applying to all binary operators. Correct the
general case but leave the currently-failing `maximum_num` tests as a
FIXME, to be resolved separately in [1].
Also simplify skip logic and NaN checking, and add a few more `copysign`
checks.
[1]: https://github.com/rust-lang/compiler-builtins/pull/939
Currently we run logspace tests for extensive tests, but there isn't any
reason we couldn't also run more kinds of tests more extensively (e.g.
more edge cases, combine edge cases with logspace for multi-input
functions, etc). As a first step toward making this possible, make
`extensive` a new field in `CheckCtx`, and rename `QuickSpaced` to
`Spaced`.
It would be nice to reuse some of the macro structure for internal
functions, like `rem_pio2`. To facilitate this, add a `public` field and
make it available in the macro's API.
In order to disambiguate things now that libm is part of the
compiler-builtins repository, do the following:
* Mention libm in LICENSE.txt
* Clarify the default license for crates other than libm and
compiler-builtins
* Add an explicit license field to Cargo.toml for all other crates
Update `run.sh` to start testing `libm`. Currently this is somewhat
inefficient because `builtins-test` gets run more than once on some
targets; this can be cleaned up later.
Distribute everything from `libm/` to better locations in the repo.
`libm/libm/*` has not moved yet to avoid Git seeing the move as an edit
to `Cargo.toml`.
Files that remain to be merged somehow are in `etc/libm`.
