h264: make debug and release versions of MC functions for better performance

diff --git a/nihav-itu/src/codecs/h264/dsp/debug.rs b/nihav-itu/src/codecs/h264/dsp/debug.rs
new file mode 100644 (file)
index 0000000..640e597
--- /dev/null
+++ b/nihav-itu/src/codecs/h264/dsp/debug.rs
@@ -0,0 +1,212 @@
+use nihav_codec_support::codecs::blockdsp::*;
+
+use super::clip_u8;
+
+const TMP_BUF_STRIDE: usize = 32;
+
+fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) {
+
+    let step = if hor { 1 } else { sstride };
+    let mut idx = 0;
+    let avgidx = if avg0 { step * 2 } else { step * 3 };
+
+    for dline in dst.chunks_mut(dstride).take(h) {
+        for (x, pix) in dline.iter_mut().take(w).enumerate() {
+            let t = clip_u8((       i16::from(src[idx + x])
+                             - 5  * i16::from(src[idx + x + step])
+                             + 20 * i16::from(src[idx + x + step * 2])
+                             + 20 * i16::from(src[idx + x + step * 3])
+                             - 5  * i16::from(src[idx + x + step * 4])
+                             +      i16::from(src[idx + x + step * 5])
+                             + 16) >> 5);
+            *pix = ((u16::from(t) + u16::from(src[idx + x + avgidx]) + 1) >> 1) as u8;
+        }
+        idx += sstride;
+    }
+}
+
+fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) {
+    let step = if hor { 1 } else { sstride };
+    let mut idx = 0;
+    for dline in dst.chunks_mut(dstride).take(h) {
+        for (x, pix) in dline.iter_mut().take(w).enumerate() {
+            *pix = clip_u8((       i16::from(src[idx + x])
+                            - 5  * i16::from(src[idx + x + step])
+                            + 20 * i16::from(src[idx + x + step * 2])
+                            + 20 * i16::from(src[idx + x + step * 3])
+                            - 5  * i16::from(src[idx + x + step * 4])
+                            +      i16::from(src[idx + x + step * 5])
+                            + 16) >> 5);
+        }
+        idx += sstride;
+    }
+}
+
+fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) {
+    for (dline, (sline0, sline1)) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE).zip(tmp2.chunks(TMP_BUF_STRIDE))).take(h) {
+        for (pix, (&a, &b)) in dline.iter_mut().zip(sline0.iter().zip(sline1.iter())).take(w) {
+            *pix = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8;
+        }
+    }
+}
+
+fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
+        dline[..w].copy_from_slice(&sline[..w]);
+    }
+}
+
+fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true);
+}
+
+fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true);
+}
+
+fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false);
+}
+
+fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true);
+}
+
+fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block2(dst, dstride, &src[2..], sstride, w, h, false);
+}
+
+fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp = [0i32; TMP_BUF_STRIDE * 16];
+    let mut idx = 0;
+    for dline in tmp.chunks_mut(TMP_BUF_STRIDE).take(h) {
+        for (x, pix) in dline.iter_mut().take(w + 5).enumerate() {
+            *pix =        i32::from(src[idx + x])
+                   - 5  * i32::from(src[idx + x + sstride])
+                   + 20 * i32::from(src[idx + x + sstride * 2])
+                   + 20 * i32::from(src[idx + x + sstride * 3])
+                   - 5  * i32::from(src[idx + x + sstride * 4])
+                   +      i32::from(src[idx + x + sstride * 5]);
+        }
+        idx += sstride;
+    }
+    for (dline, sline) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE)).take(h) {
+        for (x, pix) in dline.iter_mut().take(w).enumerate() {
+            *pix = clip_u8(((sline[x] - 5 * sline[x + 1] + 20 * sline[x + 2] + 20 * sline[x + 3] - 5 * sline[x + 4] + sline[x + 5] + 512) >> 10) as i16);
+        }
+    }
+}
+
+fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false);
+}
+
+fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
+    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
+    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+
+pub fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) {
+    let a0 = 8 - dx;
+    let a1 = dx;
+    let b0 = 8 - dy;
+    let b1 = dy;
+
+    let src1 = &src[sstride..];
+    if a0 == 8 && b0 == 8 {
+        for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
+            drow[..w].copy_from_slice(&line[..w]);
+        }
+    } else if a0 == 8 {
+        for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
+            for (pix, (&a, &b)) in drow.iter_mut().take(w).zip(line0.iter().zip(line1.iter())) {
+                *pix = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8;
+            }
+        }
+    } else if b0 == 8 {
+        for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
+            let mut a = line[0];
+            for (pix, &b) in drow.iter_mut().take(w).zip(line.iter().skip(1)) {
+                *pix = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8;
+                a = b;
+            }
+        }
+    } else {
+        for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
+            let mut a = line0[0];
+            let mut c = line1[0];
+            for (pix, (&b, &d)) in drow.iter_mut().take(w).zip(line0[1..].iter().zip(line1[1..].iter())) {
+                *pix = ((u16::from(a) * a0 * b0 + u16::from(b) * a1 * b0 + u16::from(c) * a0 * b1 + u16::from(d) * a1 * b1 + 0x20) >> 6) as u8;
+                a = b;
+                c = d;
+            }
+        }
+    }
+}
+
+pub const H264_LUMA_INTERP: &[BlkInterpFunc] = &[
+    h264_mc00, h264_mc01, h264_mc02, h264_mc03,
+    h264_mc10, h264_mc11, h264_mc12, h264_mc13,
+    h264_mc20, h264_mc21, h264_mc22, h264_mc23,
+    h264_mc30, h264_mc31, h264_mc32, h264_mc33
+];

diff --git a/nihav-itu/src/codecs/h264/dsp.rs b/nihav-itu/src/codecs/h264/dsp/mod.rs
similarity index 81%
rename from nihav-itu/src/codecs/h264/dsp.rs
rename to nihav-itu/src/codecs/h264/dsp/mod.rs
index 87dc8d9fd97b301003d89fe52488eb71d0192f2c..69df7d6eede809e8733b0d015a76853d72b0acdd 100644 (file)
--- a/nihav-itu/src/codecs/h264/dsp.rs
+++ b/nihav-itu/src/codecs/h264/dsp/mod.rs
@@ -2,6 +2,15 @@ use nihav_core::frame::*;
 use nihav_codec_support::codecs::blockdsp::*;
 use nihav_codec_support::codecs::MV;
 
+#[cfg(not(debug_assertions))]
+mod release;
+#[cfg(not(debug_assertions))]
+use release::*;
+#[cfg(debug_assertions)]
+mod debug;
+#[cfg(debug_assertions)]
+use debug::*;
+
 pub const CHROMA_QUANTS: [u8; 52] = [
      0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
     16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 29, 30,
@@ -928,213 +937,6 @@ pub const IPRED_FUNCS16X16: [IPred8x8Func; 7] = [
 
 fn clip_u8(val: i16) -> u8 { val.max(0).min(255) as u8 }
 
-const TMP_BUF_STRIDE: usize = 32;
-
-fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) {
-    let step = if hor { 1 } else { sstride };
-    let mut idx = 0;
-    let avgidx = if avg0 { step * 2 } else { step * 3 };
-    for dline in dst.chunks_mut(dstride).take(h) {
-        for (x, pix) in dline.iter_mut().take(w).enumerate() {
-            let t = clip_u8((       i16::from(src[idx + x])
-                             - 5  * i16::from(src[idx + x + step])
-                             + 20 * i16::from(src[idx + x + step * 2])
-                             + 20 * i16::from(src[idx + x + step * 3])
-                             - 5  * i16::from(src[idx + x + step * 4])
-                             +      i16::from(src[idx + x + step * 5])
-                             + 16) >> 5);
-            *pix = ((u16::from(t) + u16::from(src[idx + x + avgidx]) + 1) >> 1) as u8;
-        }
-        idx += sstride;
-    }
-}
-
-fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) {
-    let step = if hor { 1 } else { sstride };
-    let mut idx = 0;
-    for dline in dst.chunks_mut(dstride).take(h) {
-        for (x, pix) in dline.iter_mut().take(w).enumerate() {
-            *pix = clip_u8((       i16::from(src[idx + x])
-                            - 5  * i16::from(src[idx + x + step])
-                            + 20 * i16::from(src[idx + x + step * 2])
-                            + 20 * i16::from(src[idx + x + step * 3])
-                            - 5  * i16::from(src[idx + x + step * 4])
-                            +      i16::from(src[idx + x + step * 5])
-                            + 16) >> 5);
-        }
-        idx += sstride;
-    }
-}
-
-fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) {
-    for (dline, (sline0, sline1)) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE).zip(tmp2.chunks(TMP_BUF_STRIDE))).take(h) {
-        for (pix, (&a, &b)) in dline.iter_mut().zip(sline0.iter().zip(sline1.iter())).take(w) {
-            *pix = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8;
-        }
-    }
-}
-
-fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    for (dline, sline) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
-        dline[..w].copy_from_slice(&sline[..w]);
-    }
-}
-
-fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true);
-}
-
-fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true);
-}
-
-fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false);
-}
-
-fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true);
-}
-
-fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block2(dst, dstride, &src[2..], sstride, w, h, false);
-}
-
-fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp = [0i32; TMP_BUF_STRIDE * 16];
-    let mut idx = 0;
-    for dline in tmp.chunks_mut(TMP_BUF_STRIDE).take(h) {
-        for (x, pix) in dline.iter_mut().take(w + 5).enumerate() {
-            *pix =        i32::from(src[idx + x])
-                   - 5  * i32::from(src[idx + x + sstride])
-                   + 20 * i32::from(src[idx + x + sstride * 2])
-                   + 20 * i32::from(src[idx + x + sstride * 3])
-                   - 5  * i32::from(src[idx + x + sstride * 4])
-                   +      i32::from(src[idx + x + sstride * 5]);
-        }
-        idx += sstride;
-    }
-    for (dline, sline) in dst.chunks_mut(dstride).zip(tmp.chunks(TMP_BUF_STRIDE)).take(h) {
-        for (x, pix) in dline.iter_mut().take(w).enumerate() {
-            *pix = clip8(((sline[x] - 5 * sline[x + 1] + 20 * sline[x + 2] + 20 * sline[x + 3] - 5 * sline[x + 4] + sline[x + 5] + 512) >> 10) as i16);
-        }
-    }
-}
-
-fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false);
-}
-
-fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
-    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
-    let mut tmp  = [0u8; TMP_BUF_STRIDE * 16];
-    let mut tmp2 = [0u8; TMP_BUF_STRIDE * 16];
-    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, &src[1..], sstride, w, h);
-    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
-    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
-}
-
-
-fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) {
-    let a0 = 8 - dx;
-    let a1 = dx;
-    let b0 = 8 - dy;
-    let b1 = dy;
-
-    let src1 = &src[sstride..];
-    if a0 == 8 && b0 == 8 {
-        for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
-            drow[..w].copy_from_slice(&line[..w]);
-        }
-    } else if a0 == 8 {
-        for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
-            for (pix, (&a, &b)) in drow.iter_mut().take(w).zip(line0.iter().zip(line1.iter())) {
-                *pix = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8;
-            }
-        }
-    } else if b0 == 8 {
-        for (drow, line) in dst.chunks_mut(dstride).zip(src.chunks(sstride)).take(h) {
-            let mut a = line[0];
-            for (pix, &b) in drow.iter_mut().take(w).zip(line.iter().skip(1)) {
-                *pix = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8;
-                a = b;
-            }
-        }
-    } else {
-        for (drow, (line0, line1)) in dst.chunks_mut(dstride).zip(src.chunks(sstride).zip(src1.chunks(sstride))).take(h) {
-            let mut a = line0[0];
-            let mut c = line1[0];
-            for (pix, (&b, &d)) in drow.iter_mut().take(w).zip(line0[1..].iter().zip(line1[1..].iter())) {
-                *pix = ((u16::from(a) * a0 * b0 + u16::from(b) * a1 * b0 + u16::from(c) * a0 * b1 + u16::from(d) * a1 * b1 + 0x20) >> 6) as u8;
-                a = b;
-                c = d;
-            }
-        }
-    }
-}
-
-const H264_LUMA_INTERP: &[BlkInterpFunc] = &[
-    h264_mc00, h264_mc01, h264_mc02, h264_mc03,
-    h264_mc10, h264_mc11, h264_mc12, h264_mc13,
-    h264_mc20, h264_mc21, h264_mc22, h264_mc23,
-    h264_mc30, h264_mc31, h264_mc32, h264_mc33
-];
-
 pub fn do_mc(frm: &mut NASimpleVideoFrame<u8>, refpic: NAVideoBufferRef<u8>, xpos: usize, ypos: usize, w: usize, h: usize, mv: MV) {
     let mode = ((mv.x & 3) + (mv.y & 3) * 4) as usize;
     copy_block(frm, refpic.clone(), 0, xpos, ypos, mv.x >> 2, mv.y >> 2, w, h, 2, 3, mode, H264_LUMA_INTERP);

diff --git a/nihav-itu/src/codecs/h264/dsp/release.rs b/nihav-itu/src/codecs/h264/dsp/release.rs
new file mode 100644 (file)
index 0000000..87b1bc5
--- /dev/null
+++ b/nihav-itu/src/codecs/h264/dsp/release.rs
@@ -0,0 +1,284 @@
+use nihav_codec_support::codecs::blockdsp::*;
+
+use super::clip_u8;
+
+const TMP_BUF_STRIDE: usize = 32;
+
+fn interp_block1(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool, avg0: bool) {
+    unsafe {
+        let step = if hor { 1 } else { sstride };
+        let avgidx = if avg0 { step * 2 } else { step * 3 };
+        let mut src = src.as_ptr();
+        let mut dst = dst.as_mut_ptr();
+        for _ in 0..h {
+            for _ in 0..w {
+                let t = clip_u8((       i16::from(*src)
+                                 - 5  * i16::from(*src.add(step))
+                                 + 20 * i16::from(*src.add(step * 2))
+                                 + 20 * i16::from(*src.add(step * 3))
+                                 - 5  * i16::from(*src.add(step * 4))
+                                 +      i16::from(*src.add(step * 5))
+                                 + 16) >> 5);
+                *dst = ((u16::from(t) + u16::from(*src.add(avgidx)) + 1) >> 1) as u8;
+                src = src.add(1);
+                dst = dst.add(1);
+            }
+            dst = dst.sub(w).add(dstride);
+            src = src.sub(w).add(sstride);
+        }
+    }
+}
+
+fn interp_block2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize, hor: bool) {
+    unsafe {
+        let step = if hor { 1 } else { sstride };
+        let mut pix = dst.as_mut_ptr();
+        let mut src = src.as_ptr();
+        for _ in 0..h {
+            for x in 0..w {
+                *pix.add(x) = clip_u8((       i16::from(*src)
+                                       - 5  * i16::from(*src.add(step))
+                                       + 20 * i16::from(*src.add(step * 2))
+                                       + 20 * i16::from(*src.add(step * 3))
+                                       - 5  * i16::from(*src.add(step * 4))
+                                       +      i16::from(*src.add(step * 5))
+                                       + 16) >> 5);
+                src = src.add(1);
+            }
+            pix = pix.add(dstride);
+            src = src.sub(w);
+            src = src.add(sstride);
+        }
+    }
+}
+
+fn mc_avg_tmp(dst: &mut [u8], dstride: usize, w: usize, h: usize, tmp: &[u8], tmp2: &[u8]) {
+    unsafe {
+        let mut src1 = tmp.as_ptr();
+        let mut src2 = tmp2.as_ptr();
+        let mut dst = dst.as_mut_ptr();
+        for _ in 0..h {
+            for x in 0..w {
+                let a = *src1.add(x);
+                let b = *src2.add(x);
+                *dst.add(x) = ((u16::from(a) + u16::from(b) + 1) >> 1) as u8;
+            }
+            dst = dst.add(dstride);
+            src1 = src1.add(TMP_BUF_STRIDE);
+            src2 = src2.add(TMP_BUF_STRIDE);
+        }
+    }
+}
+
+fn h264_mc00(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    unsafe {
+        let mut src = src.as_ptr();
+        let mut dst = dst.as_mut_ptr();
+        for _ in 0..h {
+            std::ptr::copy_nonoverlapping(src, dst, w);
+            src = src.add(sstride);
+            dst = dst.add(dstride);
+        }
+    }
+}
+
+fn h264_mc01(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, true);
+}
+
+fn h264_mc02(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block2(dst, dstride, &src[sstride * 2..], sstride, w, h, true);
+}
+
+fn h264_mc03(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[sstride * 2..], sstride, w, h, true, false);
+}
+
+fn h264_mc10(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, true);
+}
+
+fn h264_mc11(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc12(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc22(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc13(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc02(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc20(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block2(dst, dstride, &src[2..], sstride, w, h, false);
+}
+
+fn h264_mc21(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, src, sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc22(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp: [i32; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    unsafe {
+        let mut src = src.as_ptr();
+        let mut dst = tmp.as_mut_ptr();
+        for _ in 0..h {
+            for _ in 0..w+5 {
+                *dst =        i32::from(*src)
+                       - 5  * i32::from(*src.add(sstride))
+                       + 20 * i32::from(*src.add(sstride * 2))
+                       + 20 * i32::from(*src.add(sstride * 3))
+                       - 5  * i32::from(*src.add(sstride * 4))
+                       +      i32::from(*src.add(sstride * 5));
+                dst = dst.add(1);
+                src = src.add(1);
+            }
+            src = src.sub(w+5).add(sstride);
+            dst = dst.sub(w+5).add(TMP_BUF_STRIDE);
+        }
+    }
+    unsafe {
+        let mut dst = dst.as_mut_ptr();
+        let mut src = tmp.as_ptr();
+        for _ in 0..h {
+            for _ in 0..w {
+                *dst = clip_u8(((*src - 5 * *src.add(1) + 20 * *src.add(2) + 20 * *src.add(3) - 5 * *src.add(4) + *src.add(5) + 512) >> 10) as i16);
+                dst = dst.add(1);
+                src = src.add(1);
+            }
+            dst = dst.sub(w).add(dstride);
+            src = src.sub(w).add(TMP_BUF_STRIDE);
+        }
+    }
+}
+
+fn h264_mc23(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc20(&mut tmp2, TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc30(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    interp_block1(dst, dstride, &src[2..], sstride, w, h, false, false);
+}
+
+fn h264_mc31(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc32(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc22(&mut tmp,  TMP_BUF_STRIDE, src, sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+fn h264_mc33(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, w: usize, h: usize) {
+    let mut tmp : [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    let mut tmp2: [u8; TMP_BUF_STRIDE * 16] = unsafe { std::mem::MaybeUninit::uninit().assume_init() };
+    h264_mc20(&mut tmp,  TMP_BUF_STRIDE, &src[1..], sstride, w, h);
+    h264_mc02(&mut tmp2, TMP_BUF_STRIDE, &src[sstride..], sstride, w, h);
+    mc_avg_tmp(dst, dstride, w, h, &tmp, &tmp2);
+}
+
+
+pub fn chroma_interp(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, dx: u16, dy: u16, w: usize, h: usize) {
+    let a0 = 8 - dx;
+    let a1 = dx;
+    let b0 = 8 - dy;
+    let b1 = dy;
+
+    if a0 == 8 && b0 == 8 {
+        unsafe {
+            let mut src = src.as_ptr();
+            let mut dst = dst.as_mut_ptr();
+            for _ in 0..h {
+                std::ptr::copy_nonoverlapping(src, dst, w);
+                src = src.add(sstride);
+                dst = dst.add(dstride);
+            }
+        }
+    } else if a0 == 8 {
+        unsafe {
+            let mut src0 = src.as_ptr();
+            let mut src1 = src0.add(sstride);
+            let mut dst = dst.as_mut_ptr();
+            for _ in 0..h {
+                for x in 0..w {
+                    let a = *src0.add(x);
+                    let b = *src1.add(x);
+                    *dst.add(x) = ((u16::from(a) * b0 + u16::from(b) * b1 + 4) >> 3) as u8;
+                }
+                src0 = src0.add(sstride);
+                src1 = src1.add(sstride);
+                dst = dst.add(dstride);
+            }
+        }
+    } else if b0 == 8 {
+        unsafe {
+            let mut src = src.as_ptr();
+            let mut dst = dst.as_mut_ptr();
+            for _ in 0..h {
+                let mut a = *src;
+                for x in 0..w {
+                    let b = *src.add(x + 1);
+                    *dst.add(x) = ((u16::from(a) * a0 + u16::from(b) * a1 + 4) >> 3) as u8;
+                    a = b;
+                }
+                src = src.add(sstride);
+                dst = dst.add(dstride);
+            }
+        }
+    } else {
+        unsafe {
+            let mut src0 = src.as_ptr();
+            let mut src1 = src0.add(sstride);
+            let mut dst = dst.as_mut_ptr();
+            for _ in 0..h {
+                let mut a = *src0;
+                let mut c = *src1;
+                for x in 0..w {
+                    let b = *src0.add(x + 1);
+                    let d = *src1.add(x + 1);
+                    *dst.add(x) = ((u16::from(a) * a0 * b0 + u16::from(b) * a1 * b0 + u16::from(c) * a0 * b1 + u16::from(d) * a1 * b1 + 0x20) >> 6) as u8;
+                    a = b;
+                    c = d;
+                }
+                src0 = src0.add(sstride);
+                src1 = src1.add(sstride);
+                dst = dst.add(dstride);
+            }
+        }
+    }
+}
+
+pub const H264_LUMA_INTERP: &[BlkInterpFunc] = &[
+    h264_mc00, h264_mc01, h264_mc02, h264_mc03,
+    h264_mc10, h264_mc11, h264_mc12, h264_mc13,
+    h264_mc20, h264_mc21, h264_mc22, h264_mc23,
+    h264_mc30, h264_mc31, h264_mc32, h264_mc33
+];