X-Git-Url: https://git.nihav.org/?a=blobdiff_plain;f=nihav-itu%2Fsrc%2Fcodecs%2Fh264%2Fdsp%2Fmc%2Fmod.rs;h=27bffe5e96041dd27bbaa840bd7fab3c234787fc;hb=932ae27bc58abef098a4be6f05fdd731d47a7653;hp=8763d6cd806ac947d588b6466ec181f2e7076e55;hpb=65997a44789b7d91e14d7197bececab534e76499;p=nihav.git

diff --git a/nihav-itu/src/codecs/h264/dsp/mc/mod.rs b/nihav-itu/src/codecs/h264/dsp/mc/mod.rs
index 8763d6c..27bffe5 100644
--- a/nihav-itu/src/codecs/h264/dsp/mc/mod.rs
+++ b/nihav-itu/src/codecs/h264/dsp/mc/mod.rs
@@ -2,14 +2,31 @@ use nihav_core::frame::*;
 use nihav_codec_support::codecs::MV;
 use nihav_codec_support::codecs::blockdsp::*;
 
-#[cfg(not(debug_assertions))]
-mod release;
-#[cfg(not(debug_assertions))]
-use release::*;
-#[cfg(debug_assertions)]
-mod debug;
-#[cfg(debug_assertions)]
-use debug::*;
+macro_rules! module_selector {
+    ($( ($cond:meta, $module:ident) ),*) => {
+        module_selector!(list; r#false; $(($cond, $module)),*);
+    };
+    (list; $nocond:meta; ($ccar:meta, $carmod:ident), $(($condcdr:meta, $cdrmod:ident)),*) => {
+        module_selector!(single; $nocond; $ccar; $carmod);
+        module_selector!(list; any($nocond, $ccar); $(($condcdr, $cdrmod)),*);
+    };
+    (list; $nocond:meta; ($yescond:meta, $module:ident)) => {
+        module_selector!(single; $nocond; $yescond; $module);
+    };
+    (list; $_:meta; ) => {};
+    (single; $nocond:meta; $yescond:meta; $module:ident) => {
+        #[cfg(all(not($nocond), $yescond))]
+        mod $module;
+        #[cfg(all(not($nocond), $yescond))]
+        use $module::*;
+    };
+}
+
+module_selector! (
+    (all(feature = "simd", target_arch = "x86_64"), x86),
+    (debug_assertions, debug),
+    (not(debug_assertions), release)
+);
 
 type MCFunc = fn (dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, h: usize);
 
@@ -19,6 +36,7 @@ trait RegisterSIMD {
     fn register_simd(&mut self);
 }
 
+#[allow(clippy::type_complexity)]
 pub struct H264MC {
     avg_buf:    NAVideoBufferRef<u8>,
     pub put_block_weighted:     [fn (dst: &mut [u8], stride: usize, src: &[u8], h: usize, wparams: [i8; 3]); 4],
@@ -62,7 +80,7 @@ impl H264MC {
         let (ysrc, ystride) = if (src_x - pre < 0) || (src_x + (w as isize) + post > (yw as isize)) || (src_y - pre < 0) || (src_y + (h as isize) + post > (yh as isize)) {
                 let add = (pre + post) as usize;
                 edge_emu(&refpic, src_x - pre, src_y - pre, w + add, h + add, &mut ebuf, 22, 0, 0);
-                (ebuf.as_slice(), 22)
+                (&ebuf[..], 22)
             } else {
                 (&src[refpic.get_offset(0) + ((src_x - pre) as usize) + ((src_y - pre) as usize) * systride..], systride)
             };
@@ -208,7 +226,20 @@ fn avg(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bw: usize, bh
 }
 
 fn avg_2(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
-    avg(dst, dstride, src, sstride, 2, bh);
+    let _ = src[sstride + 1];
+    let _ = dst[dstride + 1];
+    dst[0]           = ((u16::from(dst[0])           + u16::from(src[0])           + 1) >> 1) as u8;
+    dst[1]           = ((u16::from(dst[1])           + u16::from(src[1])           + 1) >> 1) as u8;
+    dst[dstride]     = ((u16::from(dst[dstride])     + u16::from(src[sstride])     + 1) >> 1) as u8;
+    dst[dstride + 1] = ((u16::from(dst[dstride + 1]) + u16::from(src[sstride + 1]) + 1) >> 1) as u8;
+    if bh == 4 {
+        let _ = src[sstride * 3 + 1];
+        let _ = dst[dstride * 3 + 1];
+        dst[dstride * 2]     = ((u16::from(dst[dstride * 2])     + u16::from(src[sstride * 2])     + 1) >> 1) as u8;
+        dst[dstride * 2 + 1] = ((u16::from(dst[dstride * 2 + 1]) + u16::from(src[sstride * 2 + 1]) + 1) >> 1) as u8;
+        dst[dstride * 3]     = ((u16::from(dst[dstride * 3])     + u16::from(src[sstride * 3])     + 1) >> 1) as u8;
+        dst[dstride * 3 + 1] = ((u16::from(dst[dstride * 3 + 1]) + u16::from(src[sstride * 3 + 1]) + 1) >> 1) as u8;
+    }
 }
 fn avg_4(dst: &mut [u8], dstride: usize, src: &[u8], sstride: usize, bh: usize) {
     avg(dst, dstride, src, sstride, 4, bh);
@@ -263,7 +294,30 @@ fn put_block_weighted2(dst: &mut [u8], stride: usize, src0: &[u8], src1: &[u8],
 }
 
 fn put_blk_w2_2(dst: &mut [u8], stride: usize, src0: &[u8], src1: &[u8], h: usize, wparams: [i8; 5]) {
-    put_block_weighted2(dst, stride, src0, src1, 2, h, wparams);
+    let weight0 = i16::from(wparams[0]);
+    let offset0 = i16::from(wparams[1]);
+    let weight1 = i16::from(wparams[2]);
+    let offset1 = i16::from(wparams[3]);
+    let wshift = (wparams[4] as u8) + 1;
+    let offset = (offset0 + offset1 + 1) >> 1;
+    let bias = (1 << wshift) >> 1;
+
+    let _ = src0[16 + 1];
+    let _ = src1[16 + 1];
+    let _ = dst[stride + 1];
+    dst[0]          = clip_u8(((i16::from(src0[ 0]) * weight0 + i16::from(src1[ 0]) * weight1 + bias) >> wshift) + offset);
+    dst[1]          = clip_u8(((i16::from(src0[ 1]) * weight0 + i16::from(src1[ 1]) * weight1 + bias) >> wshift) + offset);
+    dst[stride]     = clip_u8(((i16::from(src0[16]) * weight0 + i16::from(src1[16]) * weight1 + bias) >> wshift) + offset);
+    dst[stride + 1] = clip_u8(((i16::from(src0[17]) * weight0 + i16::from(src1[17]) * weight1 + bias) >> wshift) + offset);
+    if h == 4 {
+        let _ = src0[16 * 3 + 1];
+        let _ = src1[16 * 3 + 1];
+        let _ = dst[stride * 3 + 1];
+        dst[stride * 2]     = clip_u8(((i16::from(src0[32]) * weight0 + i16::from(src1[32]) * weight1 + bias) >> wshift) + offset);
+        dst[stride * 2 + 1] = clip_u8(((i16::from(src0[33]) * weight0 + i16::from(src1[33]) * weight1 + bias) >> wshift) + offset);
+        dst[stride * 3]     = clip_u8(((i16::from(src0[48]) * weight0 + i16::from(src1[48]) * weight1 + bias) >> wshift) + offset);
+        dst[stride * 3 + 1] = clip_u8(((i16::from(src0[49]) * weight0 + i16::from(src1[49]) * weight1 + bias) >> wshift) + offset);
+    }
 }
 fn put_blk_w2_4(dst: &mut [u8], stride: usize, src0: &[u8], src1: &[u8], h: usize, wparams: [i8; 5]) {
     put_block_weighted2(dst, stride, src0, src1, 4, h, wparams);