Make library no_std compatible by gating std use (#59)

* feat: initial `no_std` support attempt

* fix: feature gate some required imports

* fix: unimported `Vec` references in NEON convolution modules

Cargo.lock

@@ -580,6 +580,7 @@ dependencies = [
  "fast_image_resize",
  "image",
  "itertools 0.14.0",
+ "libm",
  "libvips",
  "nix",
  "num-traits",

Cargo.toml

@@ -21,28 +21,32 @@ exclude = ["/data", "/.github"]
 
 [dependencies]
 cfg-if = "1.0"
-num-traits = "0.2.19"
-thiserror = "2.0"
-document-features = "0.2.12"
+num-traits = { version = "0.2.19", default-features = false }
+thiserror = { version = "2.0", default-features = false }
+document-features = { version = "0.2.12", optional = true }
 # Optional dependencies
 image = { version = "0.25.6", optional = true, default-features = false }
 bytemuck = { version = "1.24", optional = true }
 rayon = { version = "1.11", optional = true }
+libm = "0.2.15"
 
 
 [features]
+# default = ["std"]
+## Whether or not to depend on the Rust `std`
+std = ["thiserror/std", "num-traits/std", "dep:document-features"]
 ## Enable this feature to implement traits [IntoImageView](crate::IntoImageView) and
 ## [IntoImageViewMut](crate::IntoImageViewMut) for the
 ## [DynamicImage](https://docs.rs/image/latest/image/enum.DynamicImage.html)
 ## type and some variants of
 ## [ImageBuffer](https://docs.rs/image/latest/image/struct.ImageBuffer.html)
 ## type from the `image` crate.
-image = ["dep:image", "dep:bytemuck"]
+image = ["std", "dep:image", "dep:bytemuck"]
 ## Enable this feature to implement traits [bytemuck::Pod] and
 ## [bytemuck::Zeroable] for the `Pixel` type.
-bytemuck = ["dep:bytemuck"]
+bytemuck = ["std", "dep:bytemuck"]
 ## This feature enables image processing in the ` rayon ` thread pool.
-rayon = ["dep:rayon", "resize/rayon", "image/rayon"]
+rayon = ["std", "dep:rayon", "resize/rayon", "image/rayon"]
 for_testing = ["image", "image/png", "bytemuck"]
 only_u8x4 = []  # This can be used to experiment with the crate's code.
 

src/alpha/f32x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::F32x2;
 use crate::{ImageView, ImageViewMut};

src/alpha/f32x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::F32x2;
 use crate::{ImageView, ImageViewMut};

src/alpha/f32x4/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::F32x4;
 use crate::{ImageView, ImageViewMut};

src/alpha/f32x4/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::F32x4;
 use crate::{ImageView, ImageViewMut};

src/alpha/u16x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use super::sse4;
 use crate::pixels::U16x2;

src/alpha/u16x2/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::neon_utils;
 use crate::pixels::U16x2;

src/alpha/u16x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use super::native;
 use crate::pixels::U16x2;

src/alpha/u16x2/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::pixels::U16x2;
 use crate::utils::foreach_with_pre_reading;

src/alpha/u16x4/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use super::sse4;
 use crate::pixels::U16x4;

src/alpha/u16x4/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::neon_utils;
 use crate::pixels::U16x4;

src/alpha/u16x4/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::U16x4;
 use crate::utils::foreach_with_pre_reading;

src/alpha/u16x4/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::pixels::U16x4;
 use crate::utils::foreach_with_pre_reading;

src/alpha/u8x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::U8x2;
 use crate::utils::foreach_with_pre_reading;

src/alpha/u8x2/native.rs

@@ -56,7 +56,7 @@ pub(crate) fn divide_alpha(
 #[inline]
 pub(crate) fn divide_alpha_inplace(image_view: &mut impl ImageViewMut<Pixel = U8x2>) {
     for dst_row in image_view.iter_rows_mut(0) {
-        let src_row = unsafe { std::slice::from_raw_parts(dst_row.as_ptr(), dst_row.len()) };
+        let src_row = unsafe { core::slice::from_raw_parts(dst_row.as_ptr(), dst_row.len()) };
         divide_alpha_row(src_row, dst_row);
     }
 }

src/alpha/u8x2/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::neon_utils;
 use crate::pixels::U8x2;

src/alpha/u8x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use super::native;
 use crate::pixels::U8x2;

src/alpha/u8x2/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::pixels::U8x2;
 use crate::wasm32_utils::u16x8_mul_add_shr16;

src/alpha/u8x4/avx2.rs

@@ -1,7 +1,7 @@
 use crate::pixels::U8x4;
 use crate::utils::foreach_with_pre_reading;
 use crate::{simd_utils, ImageView, ImageViewMut};
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use super::sse4;
 

src/alpha/u8x4/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::neon_utils;
 use crate::pixels::U8x4;

src/alpha/u8x4/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::U8x4;
 use crate::utils::foreach_with_pre_reading;

src/alpha/u8x4/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::pixels::U8x4;
 use crate::wasm32_utils::u16x8_mul_add_shr16;

src/array_chunks.rs

@@ -1,7 +1,7 @@
 use core::array::IntoIter;
 use core::iter::Take;
 use core::iter::{FusedIterator, Iterator};
-use std::mem::MaybeUninit;
+use core::mem::MaybeUninit;
 
 /// An iterator over `N` elements of the iterator at a time.
 ///

src/color/mappers.rs

@@ -1,3 +1,5 @@
+#[cfg(not(feature = "std"))]
+use crate::float_ext::FloatExt;
 use crate::PixelComponentMapper;
 
 fn gamma_into_linear(input: f32) -> f32 {

src/color/mod.rs

@@ -1,5 +1,10 @@
 //! Functions and structs for working with colorspace and gamma.
+#[cfg(not(feature = "std"))]
+use alloc::boxed::Box;
+
 use num_traits::bounds::UpperBounded;
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
 use num_traits::Zero;
 
 use crate::pixels::{

src/convolution/f32x1/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32;

src/convolution/f32x1/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32;

src/convolution/f32x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32x2;

src/convolution/f32x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32x2;

src/convolution/f32x3/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::{F32x3, InnerPixel};

src/convolution/f32x3/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::{F32x3, InnerPixel};

src/convolution/f32x4/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32x4;

src/convolution/f32x4/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::F32x4;

src/convolution/filters.rs

@@ -1,7 +1,12 @@
-use std::f64::consts::PI;
-use std::fmt::{Debug, Formatter};
+use core::f64::consts::PI;
+use core::fmt::{Debug, Formatter};
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
 use thiserror::Error;
 
+#[cfg(not(feature = "std"))]
+use crate::float_ext::FloatExt;
+
 type FilterFn = fn(f64) -> f64;
 
 /// Description of custom filter for image convolution.
@@ -24,7 +29,7 @@ impl PartialEq for Filter {
 impl Eq for Filter {}
 
 impl Debug for Filter {
-    fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+    fn fmt(&self, f: &mut Formatter<'_>) -> core::fmt::Result {
         f.debug_struct("Filter")
             .field("name", &self.name)
             .field("support", &self.support)

src/convolution/i32x1/native.rs

@@ -2,6 +2,9 @@ use crate::convolution::Coefficients;
 use crate::pixels::I32;
 use crate::{ImageView, ImageViewMut};
 
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
+
 pub(crate) fn horiz_convolution(
     src_view: &impl ImageView<Pixel = I32>,
     dst_view: &mut impl ImageViewMut<Pixel = I32>,

src/convolution/mod.rs

@@ -1,8 +1,14 @@
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+
 pub use filters::*;
 
 use crate::pixels::InnerPixel;
 use crate::{CpuExtensions, ImageView, ImageViewMut};
 
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
+
 #[macro_use]
 mod macros;
 

src/convolution/optimisations.rs

@@ -1,3 +1,11 @@
+#[cfg(not(feature = "std"))]
+use alloc::borrow::ToOwned;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
+
 use crate::convolution::Coefficients;
 // This code is based on C-implementation from Pillow-SIMD package for Python
 // https://github.com/uploadcare/pillow-simd

src/convolution/u16x1/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16;

src/convolution/u16x1/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 

src/convolution/u16x1/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16;

src/convolution/u16x1/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16;

src/convolution/u16x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x2;

src/convolution/u16x2/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::neon_utils;

src/convolution/u16x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x2;

src/convolution/u16x2/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x2;

src/convolution/u16x3/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x3;

src/convolution/u16x3/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::neon_utils;

src/convolution/u16x3/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x3;

src/convolution/u16x3/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x3;

src/convolution/u16x4/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x4;

src/convolution/u16x4/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::neon_utils;

src/convolution/u16x4/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x4;

src/convolution/u16x4/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer32;
 use crate::pixels::U16x4;

src/convolution/u8x1/avx2.rs

@@ -1,4 +1,6 @@
-use std::arch::x86_64::*;
+#[cfg(not(feature = "std"))]
+use alloc::borrow::ToOwned;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8;

src/convolution/u8x1/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::neon_utils;

src/convolution/u8x1/sse4.rs

@@ -1,4 +1,6 @@
-use std::arch::x86_64::*;
+#[cfg(not(feature = "std"))]
+use alloc::borrow::ToOwned;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8;

src/convolution/u8x1/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8;

src/convolution/u8x2/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x2;

src/convolution/u8x2/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::pixels::U8x2;

src/convolution/u8x2/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x2;

src/convolution/u8x2/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x2;

src/convolution/u8x3/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x3;

src/convolution/u8x3/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::neon_utils;

src/convolution/u8x3/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x3;

src/convolution/u8x3/wasm32.rs

@@ -1,5 +1,5 @@
-use std::arch::wasm32::*;
-use std::mem::transmute;
+use core::arch::wasm32::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x3;

src/convolution/u8x4/avx2.rs

@@ -1,5 +1,5 @@
-use std::arch::x86_64::*;
-use std::mem::transmute;
+use core::arch::x86_64::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x4;

src/convolution/u8x4/neon.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::neon_utils;

src/convolution/u8x4/sse4.rs

@@ -1,5 +1,5 @@
-use std::arch::x86_64::*;
-use std::mem::transmute;
+use core::arch::x86_64::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x4;

src/convolution/u8x4/wasm32.rs

@@ -1,5 +1,5 @@
-use std::arch::wasm32::*;
-use std::mem::transmute;
+use core::arch::wasm32::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::Normalizer16;
 use crate::pixels::U8x4;

src/convolution/vertical_f32/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::InnerPixel;

src/convolution/vertical_f32/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::{Coefficients, CoefficientsChunk};
 use crate::pixels::InnerPixel;

src/convolution/vertical_u16/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::pixels::InnerPixel;

src/convolution/vertical_u16/neon.rs

@@ -1,5 +1,8 @@
-use std::arch::aarch64::*;
-use std::mem::transmute;
+#[cfg(not(feature = "std"))]
+use alloc::{vec, vec::Vec};
+
+use core::arch::aarch64::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::neon_utils;

src/convolution/vertical_u16/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::convolution::vertical_u16::native::convolution_by_u16;

src/convolution/vertical_u16/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::{CoefficientsI32Chunk, Normalizer32};
 use crate::convolution::vertical_u16::native::convolution_by_u16;

src/convolution/vertical_u8/avx2.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::convolution::vertical_u8::native;

src/convolution/vertical_u8/neon.rs

@@ -1,5 +1,8 @@
-use std::arch::aarch64::*;
-use std::mem::transmute;
+#[cfg(not(feature = "std"))]
+use alloc::{vec, vec::Vec};
+
+use core::arch::aarch64::*;
+use core::mem::transmute;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::neon_utils;

src/convolution/vertical_u8/sse4.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::convolution::vertical_u8::native;

src/convolution/vertical_u8/wasm32.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::convolution::optimisations::{CoefficientsI16Chunk, Normalizer16};
 use crate::convolution::vertical_u8::native;

src/cpu_extensions.rs

@@ -23,24 +23,63 @@ impl CpuExtensions {
     pub fn is_supported(&self) -> bool {
         match self {
             #[cfg(target_arch = "x86_64")]
-            Self::Avx2 => is_x86_feature_detected!("avx2"),
+            Self::Avx2 => Self::has_avx2(),
             #[cfg(target_arch = "x86_64")]
-            Self::Sse4_1 => is_x86_feature_detected!("sse4.1"),
+            Self::Sse4_1 => Self::has_sse41(),
             #[cfg(target_arch = "aarch64")]
-            Self::Neon => std::arch::is_aarch64_feature_detected!("neon"),
+            // FIXME: Not sure how, I don't know ARM assembly. AFAIK, all ARM compliant
+            // processors must implement NEON
+            Self::Neon => true,
             #[cfg(target_arch = "wasm32")]
             Self::Simd128 => true,
             Self::None => true,
         }
     }
+
+    #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+    #[inline]
+    fn has_avx2() -> bool {
+        use core::arch::x86_64::{__cpuid, __cpuid_count, _xgetbv};
+
+        unsafe {
+            let cpuid1 = __cpuid(1);
+
+            if (cpuid1.ecx & (1 << 27)) == 0 || (cpuid1.ecx & (1 << 28)) == 0 {
+                // OSXSAVE + AVX
+                return false;
+            }
+
+            let xcr0 = _xgetbv(0);
+            if (xcr0 & 0b110) != 0b110 {
+                // XCR0: XMM (bit 1) and YMM (bit 2) enabled
+                return false;
+            }
+
+            // CPUID leaf 7 subleaf 0: AVX2
+            let cpuid7 = __cpuid_count(7, 0);
+            (cpuid7.ebx & (1 << 5)) != 0
+        }
+    }
+
+    #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+    #[inline]
+    fn has_sse41() -> bool {
+        use core::arch::x86_64::__cpuid;
+
+        // CPUID leaf 1: feature bits
+        let res = unsafe { __cpuid(1) };
+
+        // ECX bit 19 = SSE4.1
+        (res.ecx & (1 << 19)) != 0
+    }
 }
 
 impl Default for CpuExtensions {
     #[cfg(target_arch = "x86_64")]
     fn default() -> Self {
-        if is_x86_feature_detected!("avx2") {
+        if Self::has_avx2() {
             Self::Avx2
-        } else if is_x86_feature_detected!("sse4.1") {
+        } else if Self::has_sse41() {
             Self::Sse4_1
         } else {
             Self::None
@@ -49,11 +88,8 @@ impl Default for CpuExtensions {
 
     #[cfg(target_arch = "aarch64")]
     fn default() -> Self {
-        if std::arch::is_aarch64_feature_detected!("neon") {
-            Self::Neon
-        } else {
-            Self::None
-        }
+        // FIXME: Maybe we detect?
+        Self::Neon
     }
     #[cfg(target_arch = "wasm32")]
     fn default() -> Self {

src/float_ext.rs

@@ -0,0 +1,43 @@
+pub trait FloatExt {
+    fn sin(self) -> Self;
+    fn cos(self) -> Self;
+    fn sqrt(self) -> Self;
+    fn exp(self) -> Self;
+    fn powf(self, exp: Self) -> Self;
+}
+
+impl FloatExt for f32 {
+    fn sin(self) -> Self {
+        libm::sinf(self)
+    }
+    fn cos(self) -> Self {
+        libm::cosf(self)
+    }
+    fn sqrt(self) -> Self {
+        libm::sqrtf(self)
+    }
+    fn exp(self) -> Self {
+        libm::expf(self)
+    }
+    fn powf(self, exp: f32) -> Self {
+        libm::powf(self, exp)
+    }
+}
+
+impl FloatExt for f64 {
+    fn sin(self) -> Self {
+        libm::sin(self)
+    }
+    fn cos(self) -> Self {
+        libm::cos(self)
+    }
+    fn sqrt(self) -> Self {
+        libm::sqrt(self)
+    }
+    fn exp(self) -> Self {
+        libm::exp(self)
+    }
+    fn powf(self, exp: f64) -> Self {
+        libm::pow(self, exp)
+    }
+}

src/image_view.rs

@@ -1,9 +1,14 @@
-use std::num::NonZeroU32;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+use core::num::NonZeroU32;
 
 use crate::images::{TypedCroppedImage, TypedCroppedImageMut, UnsafeImageMut};
 use crate::pixels::InnerPixel;
 use crate::{ArrayChunks, ImageError, PixelTrait, PixelType};
 
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
+
 /// A trait for getting access to image data.
 ///
 /// # Safety

src/images/image.rs

@@ -1,3 +1,6 @@
+#[cfg(not(feature = "std"))]
+use alloc::{vec, vec::Vec};
+
 use crate::images::{BufferContainer, TypedImage, TypedImageRef};
 use crate::pixels::InnerPixel;
 use crate::{

src/images/image_crate.rs

@@ -1,4 +1,4 @@
-use std::ops::DerefMut;
+use core::ops::DerefMut;
 
 use crate::image_view::try_pixel_type;
 use crate::images::{TypedImage, TypedImageRef};

src/images/mod.rs

@@ -1,5 +1,7 @@
 //! Contains different types of images and wrappers for them.
-use std::fmt::Debug;
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+use core::fmt::Debug;
 
 pub use cropped_image::*;
 pub use image::*;

src/images/typed_cropped_image.rs

@@ -1,6 +1,9 @@
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+use core::num::NonZeroU32;
+
 use crate::images::{View, ViewMut};
 use crate::{CropBoxError, ImageView, ImageViewMut};
-use std::num::NonZeroU32;
 
 pub(crate) fn check_crop_box(
     img_width: u32,

src/images/typed_image.rs

@@ -1,8 +1,14 @@
+#[cfg(not(feature = "std"))]
+use alloc::{vec, vec::Vec};
+use core::fmt::Debug;
+use core::num::NonZeroU32;
+
 use crate::images::BufferContainer;
 use crate::pixels::InnerPixel;
 use crate::{ImageBufferError, ImageView, ImageViewMut, InvalidPixelsSize};
-use std::fmt::Debug;
-use std::num::NonZeroU32;
+
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
 
 /// Generic reference to image data that provides [ImageView].
 #[derive(Debug)]

src/images/unsafe_image.rs

@@ -1,13 +1,16 @@
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+use core::marker::PhantomData;
+use core::num::NonZeroU32;
+
 use crate::{ArrayChunks, ImageView, ImageViewMut};
-use std::marker::PhantomData;
-use std::num::NonZeroU32;
 
 #[derive(Copy)]
 pub(crate) struct UnsafeImageMut<'a, V>
 where
     V: ImageViewMut,
 {
-    image: std::ptr::NonNull<V>,
+    image: core::ptr::NonNull<V>,
     p: PhantomData<&'a V>,
 }
 
@@ -28,7 +31,7 @@ unsafe impl<'a, V: ImageViewMut> Sync for UnsafeImageMut<'a, V> {}
 
 impl<'a, V: ImageViewMut> UnsafeImageMut<'a, V> {
     pub fn new(image: &'a mut V) -> Self {
-        let ptr = std::ptr::NonNull::new(image as *mut V).unwrap();
+        let ptr = core::ptr::NonNull::new(image as *mut V).unwrap();
         Self {
             image: ptr,
             p: PhantomData,

src/lib.rs

@@ -1,7 +1,12 @@
+#![cfg_attr(not(feature = "std"), no_std)]
 #![doc = include_str!("../README.md")]
+
 //!
 //! ## Feature flags
-#![doc = document_features::document_features!()]
+#![cfg_attr(feature = "std", doc = document_features::document_features!())]
+
+#[cfg(not(feature = "std"))]
+extern crate alloc;
 
 pub use alpha::errors::*;
 pub use array_chunks::*;
@@ -30,6 +35,7 @@ mod convolution;
 mod cpu_extensions;
 mod crop_box;
 mod errors;
+mod float_ext;
 mod image_view;
 pub mod images;
 mod mul_div;

src/neon_utils.rs

@@ -1,4 +1,4 @@
-use std::arch::aarch64::*;
+use core::arch::aarch64::*;
 
 use crate::pixels::InnerPixel;
 
@@ -110,7 +110,7 @@ pub unsafe fn load_deintrel_u16x1x3<T: InnerPixel<Component = u16>>(
     let mut arr = [0u16; 12];
     let src_ptr = buf.get_unchecked(index..).as_ptr() as *const u16;
     let dst_ptr = arr.as_mut_ptr();
-    std::ptr::copy_nonoverlapping(src_ptr, dst_ptr, 3);
+    core::ptr::copy_nonoverlapping(src_ptr, dst_ptr, 3);
     vld3_u16(arr.as_ptr())
 }
 
@@ -122,7 +122,7 @@ pub unsafe fn load_deintrel_u16x2x3<T: InnerPixel<Component = u16>>(
     let mut arr = [0u16; 12];
     let src_ptr = buf.get_unchecked(index..).as_ptr() as *const u16;
     let dst_ptr = arr.as_mut_ptr();
-    std::ptr::copy_nonoverlapping(src_ptr, dst_ptr, 6);
+    core::ptr::copy_nonoverlapping(src_ptr, dst_ptr, 6);
     vld3_u16(arr.as_ptr())
 }
 

src/pixels.rs

@@ -1,8 +1,11 @@
 //! Contains types of pixels.
-use std::fmt::{Debug, Formatter};
-use std::marker::PhantomData;
-use std::mem::size_of;
-use std::slice;
+use core::fmt::{Debug, Formatter};
+use core::marker::PhantomData;
+use core::mem::size_of;
+use core::slice;
+
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
 
 #[derive(Debug, Clone, Copy, PartialEq, Eq)]
 #[non_exhaustive]
@@ -248,7 +251,7 @@ macro_rules! pixel_struct {
         }
 
         impl Debug for $name {
-            fn fmt(&self, f: &mut Formatter<'_>) -> std::fmt::Result {
+            fn fmt(&self, f: &mut Formatter<'_>) -> core::fmt::Result {
                 let components_ptr = self as *const _ as *const $comp_type;
                 let components: &[$comp_type] =
                     unsafe { slice::from_raw_parts(components_ptr, $comp_count) };

src/resizer.rs

@@ -1,3 +1,6 @@
+#[cfg(not(feature = "std"))]
+use alloc::vec::Vec;
+
 use crate::convolution::{self, FilterType};
 use crate::crop_box::CroppedSrcImageView;
 use crate::image_view::{try_pixel_type, ImageView, ImageViewMut, IntoImageView, IntoImageViewMut};
@@ -7,6 +10,9 @@ use crate::{
     CpuExtensions, CropBox, DifferentDimensionsError, MulDiv, PixelTrait, PixelType, ResizeError,
 };
 
+#[cfg(not(feature = "std"))]
+use num_traits::float::FloatCore;
+
 #[derive(Debug, Clone, Copy, Eq, PartialEq)]
 #[non_exhaustive]
 pub enum ResizeAlg {
@@ -320,7 +326,7 @@ impl Resizer {
     ) {
         if use_alpha && self.mul_div.is_supported(P::pixel_type()) {
             let src_view = cropped_src_view.image_view();
-            let mut alpha_buffer = std::mem::take(&mut self.alpha_buffer);
+            let mut alpha_buffer = core::mem::take(&mut self.alpha_buffer);
 
             let mut premultiplied_src =
                 get_temp_image_from_buffer(&mut alpha_buffer, src_view.width(), src_view.height());
@@ -517,7 +523,7 @@ impl Resizer {
             let tmp_width = (crop_box.width / factor).round() as u32;
             let tmp_height = (crop_box.height / factor).round() as u32;
 
-            let mut super_sampling_buffer = std::mem::take(&mut self.super_sampling_buffer);
+            let mut super_sampling_buffer = core::mem::take(&mut self.super_sampling_buffer);
             let mut tmp_img =
                 get_temp_image_from_buffer(&mut super_sampling_buffer, tmp_width, tmp_height);
             resample_nearest(cropped_src_view, &mut tmp_img);

src/simd_utils.rs

@@ -1,4 +1,4 @@
-use std::arch::x86_64::*;
+use core::arch::x86_64::*;
 
 use crate::pixels::{U8x3, U8x4};
 

src/threading.rs

@@ -1,4 +1,4 @@
-use std::num::NonZeroU32;
+use core::num::NonZeroU32;
 
 use rayon::current_num_threads;
 use rayon::prelude::*;

src/wasm32_utils.rs

@@ -1,4 +1,4 @@
-use std::arch::wasm32::*;
+use core::arch::wasm32::*;
 
 use crate::pixels::{U8x3, U8x4};
 

tests/resize_tests.rs

@@ -952,7 +952,7 @@ mod u8x4 {
 
     #[test]
     fn custom_filter_u8x4() {
-        std::env::set_var("DONT_SAVE_RESULT", "1");
+        unsafe { std::env::set_var("DONT_SAVE_RESULT", "1") };
         const LANCZOS3_RESULT: [u64; 4] = [2942547, 2947799, 2885025, 6122718];
         const LANCZOS4_RESULT: [u64; 4] = [2943083, 2948315, 2885436, 6122629];