Commit Graph

22 Commits

Author SHA1 Message Date
Josh Stone 6161f1ade1 impl 128-bit MulAdd and MulAddAssign 2018-05-07 12:28:35 -07:00
Josh Stone fe53805550 impl 128-bit CheckedRem and CheckedNeg 2018-05-07 12:28:15 -07:00
Josh Stone 261efafe0b Merge branch 'master' into regexident-i128 2018-05-04 12:28:48 -07:00
bors[bot] a49013e338 Merge #59
59: Added `MulAdd` and `MulAddAssign` traits r=cuviper a=regexident

Both `f32` and `f64` implement fused multiply-add, which computes `(self * a) + b` with only one rounding error. This produces a more accurate result with better performance than a separate multiplication operation followed by an add:

```rust
fn mul_add(self, a: f32, b: f32) -> f32[src]
```

It is however not possible to make use of this in a generic context by abstracting over a trait.

My concrete use-case is machine learning, [gradient descent](https://en.wikipedia.org/wiki/Gradient_descent) to be specific,  
where the core operation of updating the gradient could make use of `mul_add` for both its `weights: Vector` as well as its `bias: f32`:

```rust
struct Perceptron {
  weights: Vector,
  bias: f32,
}

impl MulAdd<f32, Self> for Vector {
  // ...
}

impl Perceptron {
  fn learn(&mut self, example: Vector, expected: f32, learning_rate: f32) {
    let alpha = self.error(example, expected, learning_rate);
    self.weights = example.mul_add(alpha, self.weights);
    self.bias = self.bias.mul_add(alpha, self.bias)
  }
}
```

(The actual impl of `Vector` would be generic over its value type: `Vector<T>`, thus requiring the trait.)

Co-authored-by: Vincent Esche <regexident@gmail.com>
Co-authored-by: Josh Stone <cuviper@gmail.com>
2018-05-04 19:12:41 +00:00
Josh Stone 0d358034d9 Test MulAdd explicitly, guarded by std for floats 2018-05-04 12:09:02 -07:00
Vincent Esche 28be885481 Moved impl of `MulAdd`/`MulAddAssign` for `f32`/`f64` behind feature guard 2018-04-18 10:31:37 +02:00
Josh Stone 4fb749a401 typo: taht -> that 2018-04-13 14:29:00 -07:00
Josh Stone b1c4074cc4 Document CheckedRem and CheckedNeg 2018-04-13 14:14:22 -07:00
Josh Stone 5fb3724b69 rename checked_impl_one_param to checked_impl_unary 2018-04-13 14:13:42 -07:00
LEXUGE f74de249c8
remove formats 2018-04-13 16:04:56 +08:00
Vincent Esche 830363024b Added `MulAdd` and `MulAddAssign` traits 2018-04-10 10:08:55 +02:00
Vincent Esche 5ee2570618 Added impls of `Wrapping…` for `i128` and `u128` 2018-04-09 11:11:05 +02:00
Vincent Esche 234706fb97 Added impls of `Saturating…` for `i128` and `u128` 2018-04-09 11:10:57 +02:00
Vincent Esche b44666183d Added impls of `Checked…` for `i128` and `u128` 2018-04-09 11:10:51 +02:00
Clar Charr 5d6933f34a Fix doc tests. 2018-02-27 13:25:53 -05:00
Clar Charr ce3badca57 Move Pow to pow module. 2018-02-27 13:25:53 -05:00
Clar Charr c1f4118b4e Fix Inv trait, add Pow trait. 2018-02-27 13:25:53 -05:00
Clar Charr 5bdff3f0ff Add Inv trait. 2018-02-27 13:25:53 -05:00
Vinzent Steinberg a843027b56 Re-introduce the std feature
This is a port of @vks's rust-num/num#296, but without the feature-
toggled changes to `Float`.
2018-01-31 15:42:55 -08:00
Fabian Schuiki 809ccff63f Fix checked shift RHS to u32, drop from PrimInt
Make the checked left and right shifts take a `u32` as right-hand side,
which is more consistent with the other checked operations. Also drop
`CheckedShl` and `CheckedShr` from the `PrimInt` trait, to not break
existing code. Add doctests for the two traits.
2018-01-13 15:02:38 +01:00
Fabian Schuiki 21dfae004c Add checked shifts
Add traits `CheckedShl` and `CheckedShr` that correspond to the standard
library's `checked_shl` and `checked_shr` functions. Implement the trait
on all primitive integer types by default, akin to what the standard
library does.

The stdlib is somewhat inconsistent when it comes to the type of the
shift amount. The `checked_*` functions have a `u32` shift amount, but
the `std::ops::{Shl,Shr}` traits are generic over the shift amount. Also
the stdlib implements these traits for all primitive integer types as
right-hand sides. Our implementation mimics this behaviour.
2018-01-03 16:25:13 +01:00
Josh Stone 42a610d323 Move num-traits to its own repo
All the prior `num` history is kept, so old `num-traits` tags are still
valid, but future development here will be just for `num-traits`.
2017-12-18 17:35:41 -08:00