Skip to content

Commit

Permalink
Added stackblur, alpha handling improvements
Browse files Browse the repository at this point in the history
  • Loading branch information
@awxkee
awxkee committed Jun 11, 2024
1 parent 9d186b1 commit 01ec35e
Show file tree
Hide file tree
Showing 9 changed files with 737 additions and 54 deletions.
Binary file added assets/abstract_alpha.png
View file
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
79 changes: 67 additions & 12 deletions src/lib/fast_gaussian.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -63,7 +63,10 @@ fn fast_gaussian_vertical_pass<
);
return;
}
#[cfg(all(any(target_arch = "x86_64", target_arch = "x86"), target_feature = "sse4.1"))]
#[cfg(all(
any(target_arch = "x86_64", target_arch = "x86"),
target_feature = "sse4.1"
))]
{
let slice: &UnsafeSlice<'_, u8> = unsafe { std::mem::transmute(bytes) };
sse_support::fast_gaussian_vertical_pass_sse_u8::<CHANNELS_CONFIGURATION>(
Expand All @@ -75,6 +78,7 @@ fn fast_gaussian_vertical_pass<
let mut buffer_r: [i32; 1024] = [0; 1024];
let mut buffer_g: [i32; 1024] = [0; 1024];
let mut buffer_b: [i32; 1024] = [0; 1024];
let mut buffer_a: [i32; 1024] = [0; 1024];
let radius_64 = radius as i64;
let height_wide = height as i64;
let mul_value = MUL_TABLE_DOUBLE[radius as usize];
Expand All @@ -87,6 +91,8 @@ fn fast_gaussian_vertical_pass<
let mut sum_g: i32 = initial;
let mut dif_b: i32 = 0;
let mut sum_b: i32 = initial;
let mut dif_a: i32 = 0;
let mut sum_a: i32 = initial;

let current_px = (x * CHANNELS_CONFIGURATION as u32) as usize;

Expand All @@ -105,6 +111,11 @@ fn fast_gaussian_vertical_pass<
bytes.write(current_y + current_px, new_r);
bytes.write(current_y + current_px + 1, new_g);
bytes.write(current_y + current_px + 2, new_b);
if CHANNELS_CONFIGURATION == 4 {
let new_a = T::from_u32((sum_a * mul_value) as u32 >> shr_value)
.unwrap_or_default();
bytes.write(current_y + current_px + 3, new_a);
}
}

let arr_index = ((y - radius_64) & 1023) as usize;
Expand All @@ -115,11 +126,18 @@ fn fast_gaussian_vertical_pass<
- 2 * unsafe { *buffer_g.get_unchecked(d_arr_index) };
dif_b += unsafe { *buffer_b.get_unchecked(arr_index) }
- 2 * unsafe { *buffer_b.get_unchecked(d_arr_index) };
if CHANNELS_CONFIGURATION == 4 {
dif_a += unsafe { *buffer_a.get_unchecked(arr_index) }
- 2 * unsafe { *buffer_a.get_unchecked(d_arr_index) };
}
} else if y + radius_64 >= 0 {
let arr_index = (y & 1023) as usize;
dif_r -= 2 * unsafe { *buffer_r.get_unchecked(arr_index) };
dif_g -= 2 * unsafe { *buffer_g.get_unchecked(arr_index) };
dif_b -= 2 * unsafe { *buffer_b.get_unchecked(arr_index) };
if CHANNELS_CONFIGURATION == 4 {
dif_a -= 2 * unsafe { *buffer_a.get_unchecked(arr_index) };
}
}

let next_row_y = (std::cmp::min(std::cmp::max(y + radius_64, 0), height_wide - 1)
Expand Down Expand Up @@ -152,6 +170,15 @@ fn fast_gaussian_vertical_pass<
unsafe {
*buffer_b.get_unchecked_mut(arr_index) = ub8.into();
}

if CHANNELS_CONFIGURATION == 4 {
let ua8 = bytes[px_idx + 3];
dif_a += ua8.into();
sum_a += dif_a;
unsafe {
*buffer_a.get_unchecked_mut(arr_index) = ua8.into();
}
}
}
}
}
Expand Down Expand Up @@ -195,6 +222,7 @@ fn fast_gaussian_horizontal_pass<
let mut buffer_r: [i32; 1024] = [0; 1024];
let mut buffer_g: [i32; 1024] = [0; 1024];
let mut buffer_b: [i32; 1024] = [0; 1024];
let mut buffer_a: [i32; 1024] = [0; 1024];
let radius_64 = radius as i64;
let width_wide = width as i64;
let mul_value = MUL_TABLE_DOUBLE[radius as usize];
Expand All @@ -211,21 +239,32 @@ fn fast_gaussian_horizontal_pass<
let mut sum_g: i32 = initial_sum;
let mut dif_b: i32 = 0;
let mut sum_b: i32 = initial_sum;
let mut dif_a: i32 = 0;
let mut sum_a: i32 = initial_sum;

let current_y = ((y as i64) * (stride as i64)) as usize;

let start_x = 0 - 2 * radius_64;
for x in start_x..(width as i64) {
if x >= 0 {
let current_px = ((std::cmp::max(x, 0) as u32) * channels_count) as usize;
let new_r = T::from_u32((sum_r * mul_value) as u32 >> shr_value).unwrap_or_default();
let new_g = T::from_u32((sum_g * mul_value) as u32 >> shr_value).unwrap_or_default();
let new_b = T::from_u32((sum_b * mul_value) as u32 >> shr_value).unwrap_or_default();
let new_r =
T::from_u32((sum_r * mul_value) as u32 >> shr_value).unwrap_or_default();
let new_g =
T::from_u32((sum_g * mul_value) as u32 >> shr_value).unwrap_or_default();
let new_b =
T::from_u32((sum_b * mul_value) as u32 >> shr_value).unwrap_or_default();

unsafe {
bytes.write(current_y + current_px, new_r);
bytes.write(current_y + current_px + 1, new_g);
bytes.write(current_y + current_px + 2, new_b);
let offset = current_y + current_px;
bytes.write(offset, new_r);
bytes.write(offset + 1, new_g);
bytes.write(offset + 2, new_b);
if CHANNEL_CONFIGURATION == 4 {
let new_a = T::from_u32((sum_a * mul_value) as u32 >> shr_value)
.unwrap_or_default();
bytes.write(offset + 3, new_a);
}
}

let arr_index = ((x - radius_64) & 1023) as usize;
Expand All @@ -236,21 +275,30 @@ fn fast_gaussian_horizontal_pass<
- 2 * unsafe { *buffer_g.get_unchecked(d_arr_index) };
dif_b += unsafe { *buffer_b.get_unchecked(arr_index) }
- 2 * unsafe { *buffer_b.get_unchecked(d_arr_index) };
if CHANNEL_CONFIGURATION == 4 {
dif_a += unsafe { *buffer_a.get_unchecked(arr_index) }
- 2 * unsafe { *buffer_a.get_unchecked(d_arr_index) };
}
} else if x + radius_64 >= 0 {
let arr_index = (x & 1023) as usize;
dif_r -= 2 * unsafe { *buffer_r.get_unchecked(arr_index) };
dif_g -= 2 * unsafe { *buffer_g.get_unchecked(arr_index) };
dif_b -= 2 * unsafe { *buffer_b.get_unchecked(arr_index) };
if CHANNEL_CONFIGURATION == 4 {
dif_a -= 2 * unsafe { *buffer_a.get_unchecked(arr_index) };
}
}

let next_row_y = (y as usize) * (stride as usize);
let next_row_x = ((std::cmp::min(std::cmp::max(x + radius_64, 0), width_wide - 1)
as u32)
* channels_count) as usize;

let ur8 = bytes[next_row_y + next_row_x];
let ug8 = bytes[next_row_y + next_row_x + 1];
let ub8 = bytes[next_row_y + next_row_x + 2];
let bytes_offset = next_row_y + next_row_x;

let ur8 = bytes[bytes_offset];
let ug8 = bytes[bytes_offset + 1];
let ub8 = bytes[bytes_offset + 2];

let arr_index = ((x + radius_64) & 1023) as usize;

Expand All @@ -271,6 +319,15 @@ fn fast_gaussian_horizontal_pass<
unsafe {
*buffer_b.get_unchecked_mut(arr_index) = ub8.into();
}

if CHANNEL_CONFIGURATION == 4 {
let ua8 = bytes[bytes_offset + 3];
dif_a += ua8.into();
sum_a += dif_a;
unsafe {
*buffer_a.get_unchecked_mut(arr_index) = ua8.into();
}
}
}
}
}
Expand Down Expand Up @@ -415,7 +472,6 @@ pub fn fast_gaussian_u16(
/// * `stride` - Lane length, default is width * channels_count if not aligned
/// * `radius` - almost any radius is supported
#[no_mangle]
#[allow(dead_code)]
pub fn fast_gaussian_f32(
bytes: &mut [f32],
stride: u32,
Expand All @@ -434,7 +490,6 @@ pub fn fast_gaussian_f32(
/// * `stride` - Lane length, default is width * channels_count if not aligned
/// * `radius` - almost any radius is supported
#[no_mangle]
#[allow(dead_code)]
pub fn fast_gaussian_f16(
bytes: &mut [u16],
stride: u32,
Expand Down
41 changes: 30 additions & 11 deletions src/lib/fast_gaussian_neon.rs
View file
Original file line number Diff line number Diff line change
Expand Up @@ -75,12 +75,20 @@ pub mod neon_support {

let casted_u32 = unsafe { vreinterpret_u32_u8(prepared_u8) };
let pixel = unsafe { vget_lane_u32::<0>(casted_u32) };
let bits = pixel.to_le_bytes();

unsafe {
bytes.write(current_y + current_px, bits[0]);
bytes.write(current_y + current_px + 1, bits[1]);
bytes.write(current_y + current_px + 2, bits[2]);
let offset = current_y + current_px;
if CHANNELS_COUNT == 4 {
unsafe {
let dst_ptr = (bytes.slice.as_ptr() as *mut u8).add(offset) as *mut u32;
*dst_ptr = pixel;
}
} else {
let bits = pixel.to_le_bytes();

unsafe {
bytes.write(offset, bits[0]);
bytes.write(offset + 1, bits[1]);
bytes.write(offset + 2, bits[2]);
}
}

let arr_index = ((x - radius_64) & 1023) as usize;
Expand Down Expand Up @@ -165,12 +173,23 @@ pub mod neon_support {

let casted_u32 = unsafe { vreinterpret_u32_u8(prepared_u8) };
let pixel = unsafe { vget_lane_u32::<0>(casted_u32) };
let bits = pixel.to_le_bytes();

unsafe {
bytes.write(current_y + current_px, bits[0]);
bytes.write(current_y + current_px + 1, bits[1]);
bytes.write(current_y + current_px + 2, bits[2]);
let bytes_offset = current_y + current_px;

if CHANNELS_COUNT == 4 {
unsafe {
let dst_ptr =
(bytes.slice.as_ptr() as *mut u8).add(bytes_offset) as *mut u32;
*dst_ptr = pixel;
}
} else {
let bits = pixel.to_le_bytes();

unsafe {
bytes.write(bytes_offset, bits[0]);
bytes.write(bytes_offset + 1, bits[1]);
bytes.write(bytes_offset + 2, bits[2]);
}
}

let arr_index = ((y - radius_64) & 1023) as usize;
Expand Down
Loading

0 comments on commit 01ec35e

Please sign in to comment.