Added yuv400 16-bit

Cargo.toml

@@ -2,7 +2,7 @@ workspace = { members = ["app"] }
 
 [package]
 name = "yuvutils-rs"
-version = "0.4.9"
+version = "0.4.10"
 edition = "2021"
 description = "High performance utilities for YUV format handling and conversion."
 readme = "README.md"

app/src/main.rs

@@ -5,7 +5,12 @@ use std::io::Read;
 use std::ops::Sub;
 use std::time::Instant;
 
-use yuvutils_rs::{bgra_to_yuv444_p16, rgb_to_sharp_yuv420, rgb_to_yuv420, rgb_to_yuv420_p16, rgba_to_sharp_yuv420, rgba_to_yuv420_p16, rgba_to_yuv444_p16, yuv420_p16_to_rgb16, yuv420_to_rgb, yuv420_to_yuyv422, yuyv422_to_rgb, SharpYuvGammaTransfer, YuvBytesPacking, YuvEndianness, YuvRange, YuvStandardMatrix};
+use yuvutils_rs::{
+    bgra_to_yuv444_p16, rgb_to_sharp_yuv420, rgb_to_yuv420, rgb_to_yuv420_p16,
+    rgba_to_sharp_yuv420, rgba_to_yuv420_p16, rgba_to_yuv444_p16, yuv420_p16_to_rgb16,
+    yuv420_to_rgb, yuv420_to_yuyv422, yuyv422_to_rgb, SharpYuvGammaTransfer, YuvBytesPacking,
+    YuvEndianness, YuvRange, YuvStandardMatrix,
+};
 
 fn read_file_bytes(file_path: &str) -> Result<Vec<u8>, String> {
     // Open the file
@@ -246,7 +251,7 @@ fn main() {
         &y_plane,
         y_stride as u32,
         &u_plane,
-        u_stride as u32 ,
+        u_stride as u32,
         &v_plane,
         v_stride as u32,
         &mut rgba,
@@ -291,8 +296,6 @@ fn main() {
     //
     // rgba = Vec::from(gbr);
 
-    println!("{}", std::mem::align_of::<f32>());
-
     image::save_buffer(
         "converted_sharp.png",
         rgba.as_bytes(),

src/lib.rs

@@ -35,6 +35,7 @@ mod sse;
 mod to_identity;
 #[cfg(all(target_arch = "wasm32", target_feature = "simd128"))]
 mod wasm32;
+mod y_p16_to_rgb16;
 mod y_to_rgb;
 mod ycgco_r_to_rgb;
 mod ycgco_to_rgb;
@@ -392,3 +393,4 @@ pub use yuv_p16_rgba::*;
 pub use yuv_p16_rgba16_alpha::*;
 pub use yuv_p16_rgba_alpha::*;
 pub use yuv_p16_rgba_p16::*;
+pub use y_p16_to_rgb16::*;

src/neon/mod.rs

@@ -17,6 +17,7 @@ mod rgb_to_yuv_p16;
 mod rgba_to_nv;
 mod rgba_to_yuv;
 mod to_identity;
+mod y_p16_to_rgba16;
 mod y_to_rgb;
 mod ycgco_to_rgb;
 mod ycgco_to_rgb_alpha;
@@ -59,3 +60,4 @@ pub use yuv_to_rgba_alpha::neon_yuv_to_rgba_alpha;
 pub use yuv_to_yuy2::yuv_to_yuy2_neon_impl;
 pub use yuy2_to_rgb::yuy2_to_rgb_neon;
 pub use yuy2_to_yuv::yuy2_to_yuv_neon_impl;
+pub use y_p16_to_rgba16::neon_y_p16_to_rgba16_row;

src/neon/y_p16_to_rgba16.rs

@@ -0,0 +1,100 @@
+/*
+ * // Copyright (c) the Radzivon Bartoshyk. All rights reserved.
+ * //
+ * // Use of this source code is governed by a BSD-style
+ * // license that can be found in the LICENSE file.
+ */
+
+use std::arch::aarch64::*;
+
+use crate::internals::ProcessedOffset;
+use crate::yuv_support::{
+    CbCrInverseTransform, YuvBytesPacking, YuvChromaRange, YuvEndianness,
+    YuvSourceChannels,
+};
+
+#[inline(always)]
+pub unsafe fn neon_y_p16_to_rgba16_row<
+    const DESTINATION_CHANNELS: u8,
+    const ENDIANNESS: u8,
+    const BYTES_POSITION: u8,
+>(
+    y_ld_ptr: *const u16,
+    rgba: *mut u16,
+    dst_offset: usize,
+    width: u32,
+    range: &YuvChromaRange,
+    transform: &CbCrInverseTransform<i32>,
+    start_cx: usize,
+    bit_depth: usize,
+) -> ProcessedOffset {
+    let destination_channels: YuvSourceChannels = DESTINATION_CHANNELS.into();
+    let channels = destination_channels.get_channels_count();
+    let endianness: YuvEndianness = ENDIANNESS.into();
+    let bytes_position: YuvBytesPacking = BYTES_POSITION.into();
+    let dst_ptr = rgba;
+
+    let y_corr = vdupq_n_s16(range.bias_y as i16);
+    let v_luma_coeff = vdupq_n_s16(transform.y_coef as i16);
+    let v_min_values = vdupq_n_s16(0i16);
+    let v_alpha = vdupq_n_u16((1 << bit_depth) - 1);
+    let v_msb_shift = vdupq_n_s16(bit_depth as i16 - 16);
+
+    let mut cx = start_cx;
+
+    while cx + 8 < width as usize {
+        let y_values: int16x8_t;
+
+        match endianness {
+            YuvEndianness::BigEndian => {
+                let mut y_u_values = vreinterpretq_u16_u8(vrev16q_u8(vreinterpretq_u8_u16(
+                    vld1q_u16(y_ld_ptr.add(cx)),
+                )));
+                if bytes_position == YuvBytesPacking::MostSignificantBytes {
+                    y_u_values = vshlq_u16(y_u_values, v_msb_shift);
+                }
+                y_values = vsubq_s16(vreinterpretq_s16_u16(y_u_values), y_corr);
+            }
+            YuvEndianness::LittleEndian => {
+                let mut y_vl = vld1q_u16(y_ld_ptr.add(cx));
+                if bytes_position == YuvBytesPacking::MostSignificantBytes {
+                    y_vl = vshlq_u16(y_vl, v_msb_shift);
+                }
+                y_values = vsubq_s16(vreinterpretq_s16_u16(y_vl), y_corr);
+            }
+        }
+
+        let y_high = vmull_high_s16(y_values, v_luma_coeff);
+
+        let r_high = vrshrn_n_s32::<6>(y_high);
+
+        let y_low = vmull_s16(vget_low_s16(y_values), vget_low_s16(v_luma_coeff));
+
+        let r_low = vrshrn_n_s32::<6>(y_low);
+
+        let r_values = vreinterpretq_u16_s16(vmaxq_s16(vcombine_s16(r_low, r_high), v_min_values));
+
+        match destination_channels {
+            YuvSourceChannels::Rgb => {
+                let dst_pack = uint16x8x3_t(r_values, r_values, r_values);
+                vst3q_u16(dst_ptr.add(dst_offset + cx * channels), dst_pack);
+            }
+            YuvSourceChannels::Bgr => {
+                let dst_pack = uint16x8x3_t(r_values, r_values, r_values);
+                vst3q_u16(dst_ptr.add(dst_offset + cx * channels), dst_pack);
+            }
+            YuvSourceChannels::Rgba => {
+                let dst_pack = uint16x8x4_t(r_values, r_values, r_values, v_alpha);
+                vst4q_u16(dst_ptr.add(dst_offset + cx * channels), dst_pack);
+            }
+            YuvSourceChannels::Bgra => {
+                let dst_pack = uint16x8x4_t(r_values, r_values, r_values, v_alpha);
+                vst4q_u16(dst_ptr.add(dst_offset + cx * channels), dst_pack);
+            }
+        }
+
+        cx += 8;
+    }
+
+    ProcessedOffset { cx, ux: 0 }
+}