`exp` does not perform any form of unbiasing, so there isn't any reason
it should be signed. Change this.
Additionally, add `EPSILON` to the `Float` trait.
Musl commit 97e9b73d59 ("math: new software sqrt") adds a new algorithm
using Goldschmidt division. Port this algorithm to Rust and make it
generic, which shows a notable performance improvement over the existing
algorithm.
This also allows adding square root routines for `f16` and `f128`.
Any architecture-specific float operations are likely to consist of only
a few instructions, but the softfloat implementations are much more
complex. Ensure this is what gets tested.
`cc` automatically reads this from Cargo's `OPT_LEVEL` variable so we
don't need to set it explicitly. Remove this so running in a debugger
makes more sense.
Introduce a way to ignore the results of icount regression tests, by
specifying `allow-regressions` in the pull request body. This should
apply to both pull requests and the merges based on them, since `gh pr
view` automatically handles both.
These benchmarks are fast to run, so the time cost here is pretty
minimal. Running softfloat benchmarks just ensures that we don't e.g.
test the performance of `_mm_sqrt_ss` rather than our implementation,
and running without softfloat gives us a way to see the effect of arch
intrinsics.
`--limit=1` seems to apply before `jq` filtering, meaning our
`WORKFLOW_NAME` ("CI") workflow may not appear in the input to the jq
query. Removing `--limit` provides a default amount of inputs that jq
can then filter from, so this works better.
This failed a couple of times recently in CI, once on i686 and once on
aarch64-apple:
thread 'main' panicked at crates/libm-test/benches/random.rs:76:65:
called `Result::unwrap()` on an `Err` value: ynf
Caused by:
0:
input: (681, 509.90924) (0x000002a9, 0x43fef462)
expected: -3.2161271e38 0xff71f45b
actual: -inf 0xff800000
1: mismatched infinities
thread 'main' panicked at crates/libm-test/benches/random.rs:76:65:
called `Result::unwrap()` on an `Err` value: ynf
Caused by:
0:
input: (132, 50.46604) (0x00000084, 0x4249dd3a)
expected: -3.3364996e38 0xff7b02a5
actual: -inf 0xff800000
1: mismatched infinities
Add a new override to account for this.
The icount benchmarks are what we will be relying on in CI more than the
existing benchmarks. There isn't much reason to keep these around, but
there isn't much point in dropping them either. So, just reduce the
runtime.
Add support in `ci-util.py` for finding the most recent baseline and
downloading it, which new tests can then be compared against.
Arbitrarily select nightly-2025-01-16 as the rustc version to pin to in
benchmarks.
Running walltime benchmarks in CI is notoriously unstable, Introduce
benchmarks that instead use instruction count and other more
reproducible metrics, using `iai-callgrind` [1], which we are able to
run in CI with a high degree of reproducibility.
Inputs to this benchmark are a logspace sweep, which gives an
approximation for real-world use, but may fail to indicate outlier
cases.
[1]: https://github.com/iai-callgrind/iai-callgrind
This also allows reusing the same generator logic between logspace tests
and extensive tests, so comes with a nice bit of cleanup.
Changes:
* Make the generator part of `CheckCtx` since a `Generator` and
`CheckCtx` are almost always passed together.
* Rename `domain_logspace` to `spaced` since this no longer only
operates within a domain and we may want to handle integer spacing.
* Domain is now calculated at runtime rather than using traits, which is
much easier to work with.
* With the above, domains for multidimensional functions are added.
* The extensive test generator code tests has been combined with the
domain_logspace generator code. With this, the domain tests have just
become a subset of extensive tests. These were renamed to "quickspace"
since, technically, the extensive tests are also "domain" or "domain
logspace" tests.
* Edge case generators now handle functions with multiple inputs.
* The test runners can be significantly cleaned up and deduplicated.
The test suite for this repo has quite a lot of tests, and it is
difficult to tell which contribute the most to the long CI runtime.
libtest does have an unstable flag to report test times, but that is
inconvenient to use because it needs to be passed only to libtest
binaries.
Switch to cargo-nextest [1] which provides time reporting and, overall,
a better test UI. It may also improve test runtime, though this seems
unlikely since we have larger test binaries with many small tests
(nextest benefits the most when there are larger binaries that can be
run in parallel).
For anyone running locally without, `run.sh` should still fall back to
`cargo test` if `cargo-nextest` is not available.
This diff includes some cleanup and consistency changes to other
CI-related files.
[1]: https://nexte.st
These crates take time building in CI, especially with the release
profile having LTO enabled, but there isn't really any reason to test
them with different features or in release mode. Disable this to save
some CI runtime.
Introduce a simple binary that can run arbitrary input against any of
the available implementations (musl, MPFR, our libm). This provides an
easy way to check results, or run specific cases against a debugger.
Examples:
$ cargo run -p util -- eval libm pow 1.6 2.4
3.089498284311124
$ cargo run -p util -- eval mpfr pow 1.6 2.4
3.089498284311124
$ cargo run -p util -- eval musl tgamma 1.2344597839132
0.9097442657960874
$ cargo run -p util -- eval mpfr tgamma 1.2344597839132
0.9097442657960874
$ cargo run -p util -- eval libm tgamma 1.2344597839132
0.9097442657960871
$ cargo run -p util -- eval musl sincos 3.1415926535
(8.979318433952318e-11, -1.0)
Most users who are developing this crate are likely running on a Unix
system, since there isn't much to test against otherwise. For
convenience, enable the features required to run these tests by default.
Currently the features that control what we test against are
`build-musl` and `test-multiprecision`. I didn't name them very
consistently and there isn't really any reason for that.
Rename `test-multiprecision` to `build-mpfr` to better reflect what it
actually does and to be more consistent with `build-musl`.
There was a recent failure from the random tests:
---- mp_random_exp2f stdout ----
Random Mpfr exp2f arg 1/1: 10000 iterations (10000 total) using `LIBM_SEED=fqgMuzs6eqH1VZSEmQpLnThnaIyRUOWe`
thread 'mp_random_exp2f' panicked at crates/libm-test/tests/multiprecision.rs:41:49:
called `Result::unwrap()` on an `Err` value:
input: (127.97238,) (0x42fff1dc,)
expected: 3.3383009e38 0x7f7b2556
actual: inf 0x7f800000
Caused by:
mismatched infinities
Add an xfail for mismatched infinities on i586.
`rint` had a couple recent failures from the random tests:
---- mp_random_rint stdout ----
Random Mpfr rint arg 1/1: 10000 iterations (10000 total) using `LIBM_SEED=Fl1f69DaJnwkHN2FeuCXaBFRvJYsPvEY`
thread 'mp_random_rint' panicked at crates/libm-test/tests/multiprecision.rs:41:49:
called `Result::unwrap()` on an `Err` value:
input: (-849751480.5001163,) (0xc1c95316dc4003d0,)
expected: -849751481.0 0xc1c95316dc800000
actual: -849751480.0 0xc1c95316dc000000
Caused by:
ulp 8388608 > 100000
And:
---- mp_random_rint stdout ----
Random Mpfr rint arg 1/1: 10000 iterations (10000 total) using `LIBM_SEED=XN7VCGhX3Wu6Mzn8COvJPITyZlGP7gN7`
thread 'mp_random_rint' panicked at crates/libm-test/tests/multiprecision.rs:41:49:
called `Result::unwrap()` on an `Err` value:
input: (-12493089.499809155,) (0xc167d4242ffe6fc5,)
expected: -12493089.0 0xc167d42420000000
actual: -12493090.0 0xc167d42440000000
Caused by:
ulp 536870912 > 100000
It seems we just implement an incorrect rounding mode. Replace the
existing `rint` override with an xfail if the difference is 0.0 <= ε <=
1.0.
`compiler_builtins` exposes an `extern "C"` version of `libm` routines,
so add the same here. There really isn't much to test here (unless we
later add tests against C `libm` suites), but one nice benefit is this
gives us a library with unmangled names that is easy to `objdump`. In
accordance with that, also update `cb` to be a `staticlib`.
Unfortunately this also means we have to remove it from the workspace,
since Cargo doesn't allow setting `panic = "abort"` for a single crate.
It would be preferable to switch to a different generator, or at least
set the seed within the benchmark, but this is the most straightforward
way to make things simple.
A failing debug assertion or overflow without correctly wrapping or
saturating is a bug, but the `debug` profile that has these enabled does
not run enough test cases to hit edge cases that may trigger these. Add
a new `release-checked` profile that enables debug assertions and
overflow checks. This seems to only extend per-function test time by a
few seconds (or around a minute on longer extensive tests), so enable
this as the default on CI.
In order to ensure `no_panic` still gets checked, add a build-only step
to CI.
`ExpInt` is likely to only have performance benefits on 16-bit
platforms, but makes working with the exponent more difficult. It seems
like a worthwhile tradeoff to instead just use `i32`, so do that here.
I do not believe Cargo separately caches crates with different sets of
features enabled. So, ensuring that tests run with `unstable-intrinsics`
are always grouped should slightly reduce runtime.
As an added benefit, all the debug mode tests run first so initial
feedback is available faster.
