Change hand-picked RNGs back to SmallRng

While `Xoshiro256+` is faster on 64-bit, it has low linear complexity in the
lower three bits, which *are* used when generating dither.

Also, while `Xoshiro128StarStar` access one less variable from the heap,
multiplication is generally slower than addition in hardware.

Author: roderickvd

Cargo.lock

@@ -47,9 +47,8 @@ lewton = "0.10"
 ogg = "0.8"
 
 # Dithering
-rand = "0.8"
+rand = { version = "0.8", features = ["small_rng"] }
 rand_distr = "0.4"
-rand_xoshiro = "0.6"
 
 [features]
 alsa-backend = ["alsa"]

playback/Cargo.toml

@@ -1,3 +1,4 @@
+use rand::rngs::SmallRng;
 use rand::SeedableRng;
 use rand_distr::{Distribution, Normal, Triangular, Uniform};
 use std::fmt;
@@ -41,29 +42,12 @@ impl fmt::Display for dyn Ditherer {
     }
 }
 
-// `SmallRng` is 33% faster than `ThreadRng`, but we can do even better.
-// `SmallRng` defaults to `Xoshiro256PlusPlus` on 64-bit platforms and
-// `Xoshiro128PlusPlus` on 32-bit platforms. These are excellent for the
-// general case. In our case of just 64-bit floating points, we can make
-// some optimizations. Compared to `SmallRng`, these hand-picked generators
-// improve performance by another 9% on 64-bit platforms and 2% on 32-bit
-// platforms.
-//
-// For reference, see https://prng.di.unimi.it. Note that we do not use
-// `Xoroshiro128Plus` or `Xoshiro128Plus` because they display low linear
-// complexity in the lower four bits, which is not what we want:
-// linearization is the very point of dithering.
-#[cfg(target_pointer_width = "64")]
-type Rng = rand_xoshiro::Xoshiro256Plus;
-#[cfg(not(target_pointer_width = "64"))]
-type Rng = rand_xoshiro::Xoshiro128StarStar;
-
-fn create_rng() -> Rng {
-    Rng::from_entropy()
+fn create_rng() -> SmallRng {
+    SmallRng::from_entropy()
 }
 
 pub struct TriangularDitherer {
-    cached_rng: Rng,
+    cached_rng: SmallRng,
     distribution: Triangular<f64>,
 }
 
@@ -90,7 +74,7 @@ impl TriangularDitherer {
 }
 
 pub struct GaussianDitherer {
-    cached_rng: Rng,
+    cached_rng: SmallRng,
     distribution: Normal<f64>,
 }
 
@@ -119,7 +103,7 @@ impl GaussianDitherer {
 pub struct HighPassDitherer {
     active_channel: usize,
     previous_noises: [f64; NUM_CHANNELS],
-    cached_rng: Rng,
+    cached_rng: SmallRng,
     distribution: Uniform<f64>,
 }