[WIP] SIMD-accelerated IDCT

.github/workflows/rust.yml

@@ -16,6 +16,13 @@ jobs:
         rust: ["1.34.2", stable, beta, nightly]
         features: ["", "rayon"]
         command: [test, benchmark]
+        include:
+          - rust: nightly
+            features: packed_simd
+            command: test
+          - rust: nightly
+            features: packed_simd
+            command: benchmark
     steps:
     - uses: actions/checkout@v2
     - run: rustup default ${{ matrix.rust }}
@@ -24,10 +31,12 @@ jobs:
         cargo build --verbose --no-default-features --features "$FEATURES" &&
         cargo test --tests --benches --no-default-features --features "$FEATURES"
       if: ${{ matrix.command == 'test' }}
+      continue-on-error: ${{ matrix.rust == 'nightly' }}
       env:
         FEATURES: ${{ matrix.features }}
     - name: benchmark
       run: cargo bench --bench decoding_benchmark --no-default-features --features "$FEATURES" -- --warm-up-time 1 --measurement-time 1 --sample-size 25
       if: ${{ matrix.command == 'benchmark' }}
+      continue-on-error: ${{ matrix.rust == 'nightly' }}
       env:
         FEATURES: ${{ matrix.features }}

Cargo.toml

@@ -13,6 +13,8 @@ exclude = ["tests/*"]
 [dependencies]
 byteorder = "1.0"
 rayon = { version = "1.0", optional = true }
+simulated_packed_simd = "0.0.1"
+packed_simd = { version = "0.3", optional = true }
 
 [dev-dependencies]
 png = "0.16"

README.md

@@ -31,5 +31,14 @@ fn main() {
 }
 ```
 
+## Performance
+ This crate uses [rayon](https://github.com/rayon-rs/rayon) to decode images on all available cores by default.
+ This can be disabled by requiring the dependency with `default-features = false`.
+
+ This crate can optionally use [SIMD](https://en.wikipedia.org/wiki/SIMD) instructions
+ to decode images even faster.
+ This is not enabled by default because it requires a nightly compiler,
+ but can be activated with the `packed_simd` feature.
+
 ## Requirements
 This crate compiles only with rust >= 1.34.