Added avx difference u8

app/Cargo.toml

@@ -7,7 +7,7 @@ edition = "2021"
 image = "0.25.2"
 fast_morphology = {path = "../", features = ["image"]}
 imageproc = "0.25.0"
-opencv = {version = "0.93.0", features = ["imgproc", "clang-runtime"]}
+opencv = {version = "0.93.0", features = ["imgproc"]}
 
 [dev-dependencies]
 criterion = {version = "0.5.1", features = ["html_reports"]}

src/avx/gradient_unsigned_8.rs

@@ -0,0 +1,131 @@
+/*
+ * Copyright (c) Radzivon Bartoshyk. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without modification,
+ * are permitted provided that the following conditions are met:
+ *
+ * 1.  Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2.  Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3.  Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+#[cfg(target_arch = "x86")]
+use std::arch::x86::*;
+#[cfg(target_arch = "x86_64")]
+use std::arch::x86_64::*;
+pub fn morph_gradient_avx(dilation: &[u8], erosion: &[u8], dst: &mut [u8]) {
+    unsafe {
+        morph_gradient_avx_impl(dilation, erosion, dst);
+    }
+}
+#[inline]
+#[target_feature(enable = "avx2")]
+unsafe fn morph_gradient_avx_impl(dilation: &[u8], erosion: &[u8], dst: &mut [u8]) {
+    if dilation.len() != erosion.len() || erosion.len() != dst.len() {
+        panic!(
+            "All array must match in size for gradient but received v0: {}, v1: {}, v2: {}",
+            dilation.len(),
+            erosion.len(),
+            dst.len()
+        );
+    }
+    let length = dilation.len();
+    let mut _cx = 0usize;
+    unsafe {
+        while _cx + 128 < length {
+            let v0_ptr = dilation.get_unchecked(_cx..).as_ptr();
+            let v0_set = (
+                _mm256_loadu_si256(v0_ptr as *const __m256i),
+                _mm256_loadu_si256(v0_ptr.add(32) as *const __m256i),
+                _mm256_loadu_si256(v0_ptr.add(64) as *const __m256i),
+                _mm256_loadu_si256(v0_ptr.add(96) as *const __m256i),
+            );
+            let v1_ptr = erosion.get_unchecked(_cx..).as_ptr();
+            let v1_set = (
+                _mm256_loadu_si256(v1_ptr as *const __m256i),
+                _mm256_loadu_si256(v1_ptr.add(32) as *const __m256i),
+                _mm256_loadu_si256(v1_ptr.add(64) as *const __m256i),
+                _mm256_loadu_si256(v1_ptr.add(96) as *const __m256i),
+            );
+            let result_set = (
+                _mm256_subs_epu8(v0_set.0, v1_set.0),
+                _mm256_subs_epu8(v0_set.1, v1_set.1),
+                _mm256_subs_epu8(v0_set.2, v1_set.2),
+                _mm256_subs_epu8(v0_set.3, v1_set.3),
+            );
+            let v_dst_ptr = dst.get_unchecked_mut(_cx..).as_mut_ptr();
+            _mm256_storeu_si256(v_dst_ptr as *mut __m256i, result_set.0);
+            _mm256_storeu_si256(v_dst_ptr.add(32) as *mut __m256i, result_set.1);
+            _mm256_storeu_si256(v_dst_ptr.add(64) as *mut __m256i, result_set.2);
+            _mm256_storeu_si256(v_dst_ptr.add(96) as *mut __m256i, result_set.3);
+            _cx += 128;
+        }
+
+        while _cx + 64 < length {
+            let v0_ptr = dilation.get_unchecked(_cx..).as_ptr();
+            let v0_set = (
+                _mm256_loadu_si256(v0_ptr as *const __m256i),
+                _mm256_loadu_si256(v0_ptr.add(32) as *const __m256i),
+            );
+            let v1_ptr = erosion.get_unchecked(_cx..).as_ptr();
+            let v1_set = (
+                _mm256_loadu_si256(v1_ptr as *const __m256i),
+                _mm256_loadu_si256(v1_ptr.add(32) as *const __m256i),
+            );
+            let result_set = (
+                _mm256_subs_epu8(v0_set.0, v1_set.0),
+                _mm256_subs_epu8(v0_set.1, v1_set.1),
+            );
+            let v_dst_ptr = dst.get_unchecked_mut(_cx..).as_mut_ptr();
+            _mm256_storeu_si256(v_dst_ptr as *mut __m256i, result_set.0);
+            _mm256_storeu_si256(v_dst_ptr.add(32) as *mut __m256i, result_set.1);
+            _cx += 64;
+        }
+
+        while _cx + 32 < length {
+            let v0_ptr = dilation.get_unchecked(_cx..).as_ptr();
+            let v0_set = _mm256_loadu_si256(v0_ptr as *const __m256i);
+            let v1_ptr = erosion.get_unchecked(_cx..).as_ptr();
+            let v1_set = _mm256_loadu_si256(v1_ptr as *const __m256i);
+            let result_set = _mm256_subs_epu8(v0_set, v1_set);
+            let v_dst_ptr = dst.get_unchecked_mut(_cx..).as_mut_ptr();
+            _mm256_storeu_si256(v_dst_ptr as *mut __m256i, result_set);
+            _cx += 32;
+        }
+
+        while _cx + 16 < length {
+            let v0_ptr = dilation.get_unchecked(_cx..).as_ptr();
+            let v0_set = _mm_loadu_si128(v0_ptr as *const __m128i);
+            let v1_ptr = erosion.get_unchecked(_cx..).as_ptr();
+            let v1_set = _mm_loadu_si128(v1_ptr as *const __m128i);
+            let result_set = _mm_subs_epu8(v0_set, v1_set);
+            let v_dst_ptr = dst.get_unchecked_mut(_cx..).as_mut_ptr();
+            _mm_storeu_si128(v_dst_ptr as *mut __m128i, result_set);
+            _cx += 16;
+        }
+
+        while _cx < length {
+            *dst.get_unchecked_mut(_cx) = dilation
+                .get_unchecked(_cx)
+                .saturating_sub(*erosion.get_unchecked(_cx));
+            _cx += 1;
+        }
+    }
+}

src/avx/mod.rs

@@ -0,0 +1,32 @@
+/*
+ * Copyright (c) Radzivon Bartoshyk. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without modification,
+ * are permitted provided that the following conditions are met:
+ *
+ * 1.  Redistributions of source code must retain the above copyright notice, this
+ * list of conditions and the following disclaimer.
+ *
+ * 2.  Redistributions in binary form must reproduce the above copyright notice,
+ * this list of conditions and the following disclaimer in the documentation
+ * and/or other materials provided with the distribution.
+ *
+ * 3.  Neither the name of the copyright holder nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+ * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+ * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+ * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+mod gradient_unsigned_8;
+
+pub use gradient_unsigned_8::morph_gradient_avx;

src/difference.rs

@@ -26,6 +26,8 @@
  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
+#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+use crate::avx::morph_gradient_avx;
 #[cfg(all(target_arch = "aarch64", target_feature = "neon"))]
 use crate::neon::morph_gradient_neon;
 #[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
@@ -67,6 +69,9 @@ impl MorphGradient<u8> for u8 {
             if std::arch::is_x86_feature_detected!("sse4.1") {
                 _dispatcher = morph_gradient_sse;
             }
+            if std::arch::is_x86_feature_detected!("avx2") {
+                _dispatcher = morph_gradient_avx;
+            }
         }
         _dispatcher(dilation, erosion, dst)
     }

src/lib.rs

@@ -31,6 +31,8 @@ extern crate core;
 
 mod arena;
 mod arena_roi;
+#[cfg(any(target_arch = "x86", target_arch = "x86_64"))]
+mod avx;
 mod border_mode;
 mod difference;
 #[cfg(feature = "image")]
@@ -43,6 +45,7 @@ mod morph_base;
 mod morph_gray_alpha;
 mod morph_rgb;
 mod morph_rgba;
+#[cfg(all(target_arch = "aarch64", target_feature = "neon"))]
 mod neon;
 mod op;
 mod op_f32;

src/op.rs

@@ -26,6 +26,7 @@
  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  */
+use std::time::Instant;
 use crate::border_mode::BorderMode;
 use crate::difference::MorphGradient;
 use crate::morph_gray_alpha::make_morphology_gray_alpha;
@@ -554,7 +555,9 @@ pub fn morphology_rgb(
                 border_mode,
                 threading_policy,
             )?;
+            let start_time = Instant::now();
             u8::morph_gradient(&dilation, &erosion, dst);
+            println!("end time {:?}", start_time.elapsed());
             Ok(())
         }
         MorphExOp::TopHat => {