Fixes ln, atan2

Cargo.toml

@@ -2,7 +2,7 @@ workspace = { members = ["src/app"] }
 
 [package]
 name = "colorutils-rs"
-version = "0.4.8"
+version = "0.4.9"
 edition = "2021"
 description = "High performance utilities for color format handling and conversion."
 readme = "README.md"

src/app/src/main.rs

@@ -1,4 +1,3 @@
-use std::arch::aarch64::{vdupq_n_f32, vgetq_lane_f32};
 use std::time::Instant;
 
 use image::io::Reader as ImageReader;
@@ -15,13 +14,6 @@ pub const fn shuffle(z: u32, y: u32, x: u32, w: u32) -> i32 {
 }
 
 fn main() {
-    unsafe {
-        let y = vdupq_n_f32(0f32);
-        let x = vdupq_n_f32(0f32);
-        let v = vatan2q_f32(y, x);
-        let val = vgetq_lane_f32::<0>(v);
-        print!("{}", val);
-    }
     let r = 140;
     let g = 164;
     let b = 177;
@@ -31,7 +23,7 @@ fn main() {
     println!("HSL {:?}", hsl);
     println!("Back RGB {:?}", hsl.to_rgb8());
 
-    let img = ImageReader::open("./assets/horse.png")
+    let img = ImageReader::open("./assets/asset_2.jpg")
         .unwrap()
         .decode()
         .unwrap();
@@ -42,7 +34,7 @@ fn main() {
     let mut src_bytes = img.as_bytes();
     let width = dimensions.0;
     let height = dimensions.1;
-    let components = 4;
+    let components = 3;
     //
     // let mut dst_rgba = vec![];
     // dst_rgba.resize(4usize * width as usize * height as usize, 0u8);
@@ -66,13 +58,14 @@ fn main() {
         lab_store.resize(width as usize * components * height as usize, 0f32);
         let src_stride = width * components as u32;
         let start_time = Instant::now();
-        rgba_to_lch_with_alpha(
+        rgb_to_linear(
             src_bytes,
             src_stride,
             &mut lab_store,
             store_stride as u32,
             width,
             height,
+            TransferFunction::Gamma2p2,
         );
         let elapsed_time = start_time.elapsed();
         // Print the elapsed time in milliseconds
@@ -100,13 +93,14 @@ fn main() {
         // }
 
         let start_time = Instant::now();
-        lch_with_alpha_to_rgba(
+        linear_to_rgb(
             &lab_store,
             store_stride as u32,
             &mut dst_slice,
             src_stride,
             width,
             height,
+            TransferFunction::Gamma2p2,
         );
 
         let elapsed_time = start_time.elapsed();

src/avx/math.rs

@@ -1,4 +1,3 @@
-use crate::sse::{_mm_mulsign_ps, _mm_select_ps};
 #[cfg(target_arch = "x86")]
 use std::arch::x86::*;
 #[cfg(target_arch = "x86_64")]
@@ -210,13 +209,13 @@ unsafe fn _mm256_exp_ps_impl<const PROCESS_NAN: bool>(x: __m256) -> __m256 {
 #[inline(always)]
 #[allow(dead_code)]
 pub unsafe fn _mm256_pow_ps(x: __m256, n: __m256) -> __m256 {
-    _mm256_exp_ps(_mm256_mul_ps(n, _mm256_log_ps(x)))
+    _mm256_exp_ps(_mm256_mul_ps(n, _mm256_log_ps::<false>(x)))
 }
 
 #[inline(always)]
 #[allow(dead_code)]
 pub unsafe fn _mm256_pow_n_ps(x: __m256, n: f32) -> __m256 {
-    _mm256_exp_ps(_mm256_mul_ps(_mm256_set1_ps(n), _mm256_log_ps(x)))
+    _mm256_exp_ps(_mm256_mul_ps(_mm256_set1_ps(n), _mm256_log_ps::<false>(x)))
 }
 
 #[inline(always)]

src/avx/sigmoidal.rs

@@ -21,7 +21,7 @@ pub(crate) unsafe fn avx_sigmoidal_to_color(x: __m256) -> __m256 {
     let k = _mm256_mul_ps(x, _mm256_rcp_ps(den));
     let zeros = _mm256_setzero_ps();
     let zero_mask_2 = _mm256_cmp_ps::<_CMP_LT_OS>(k, zeros);
-    let ln = _mm256_log_ps(k);
+    let ln = _mm256_log_ps::<false>(k);
     let rs = _mm256_select_ps(_mm256_and_ps(zero_mask_1, zero_mask_2), zeros, ln);
     return rs;
 }

src/lib.rs

@@ -109,7 +109,6 @@ pub use xyza_laba_to_image::xyz_with_alpha_to_rgba;
 pub use image_to_sigmoidal::bgra_to_sigmoidal;
 pub use image_to_sigmoidal::rgb_to_sigmoidal;
 pub use image_to_sigmoidal::rgba_to_sigmoidal;
-pub use neon::*;
 pub use rgb_expand::*;
 pub use sigmoidal::Sigmoidal;
 pub use sigmoidal_to_image::sigmoidal_to_bgra;

src/neon/mod.rs

@@ -21,7 +21,6 @@ pub use gamma_curves::*;
 pub use hsv_to_image::*;
 pub use image_to_hsv::*;
 pub use linear_to_image::*;
-pub use math::vatan2q_f32;
 pub use to_linear::*;
 pub use to_linear_u8::*;
 pub use to_sigmoidal::neon_image_to_sigmoidal;

src/sse/math.rs

@@ -193,12 +193,12 @@ unsafe fn _mm_exp_ps_ulp_5_impl<const PROCESS_NAN: bool>(x: __m128) -> __m128 {
 
 #[inline(always)]
 pub unsafe fn _mm_pow_ps(x: __m128, n: __m128) -> __m128 {
-    _mm_exp_ps(_mm_mul_ps(n, _mm_log_ps(x)))
+    _mm_exp_ps(_mm_mul_ps(n, _mm_log_ps::<false>(x)))
 }
 
 #[inline(always)]
 pub unsafe fn _mm_pow_n_ps(x: __m128, n: f32) -> __m128 {
-    _mm_exp_ps(_mm_mul_ps(_mm_set1_ps(n), _mm_log_ps(x)))
+    _mm_exp_ps(_mm_mul_ps(_mm_set1_ps(n), _mm_log_ps::<false>(x)))
 }
 
 #[inline(always)]

src/sse/sigmoidal.rs

@@ -21,7 +21,7 @@ pub(crate) unsafe fn sse_sigmoidal_to_color(x: __m128) -> __m128 {
     let k = _mm_mul_ps(x, _mm_rcp_ps(den));
     let zeros = _mm_setzero_ps();
     let zero_mask_2 = _mm_cmple_ps(k, zeros);
-    let ln = _mm_log_ps(k);
+    let ln = _mm_log_ps::<false>(k);
     let rs = _mm_select_ps(_mm_and_ps(zero_mask_1, zero_mask_2), zeros, ln);
     return rs;
 }