arm/aarch64: vp9lpf: Keep the comparison to E within 8 bit

The theoretical maximum value of E is 193, so we can just
saturate the addition to 255.

Before:                     Cortex A7      A8      A9     A53  A53/AArch64
vp9_loop_filter_v_4_8_neon:     143.0   127.7   114.8    88.0         87.7
vp9_loop_filter_v_8_8_neon:     241.0   197.2   173.7   140.0        136.7
vp9_loop_filter_v_16_8_neon:    497.0   419.5   379.7   293.0        275.7
vp9_loop_filter_v_16_16_neon:   965.2   818.7   731.4   579.0        452.0
After:
vp9_loop_filter_v_4_8_neon:     136.0   125.7   112.6    84.0         83.0
vp9_loop_filter_v_8_8_neon:     234.0   195.5   171.5   136.0        133.7
vp9_loop_filter_v_16_8_neon:    490.0   417.5   377.7   289.0        271.0
vp9_loop_filter_v_16_16_neon:   951.2   814.7   732.3   571.0        446.7

This is cherrypicked from libav commit
c582cb8537.

Signed-off-by: Martin Storsjö <martin@martin.st>

libavcodec/aarch64/vp9lpf_neon.S

@@ -51,13 +51,6 @@
 // see the arm version instead.
 
 
-.macro uabdl_sz dst1, dst2, in1, in2, sz
-        uabdl           \dst1,  \in1\().8b,  \in2\().8b
-.ifc \sz, .16b
-        uabdl2          \dst2,  \in1\().16b, \in2\().16b
-.endif
-.endm
-
 .macro add_sz dst1, dst2, in1, in2, in3, in4, sz
         add             \dst1,  \in1,  \in3
 .ifc \sz, .16b
@@ -86,20 +79,6 @@
 .endif
 .endm
 
-.macro cmhs_sz dst1, dst2, in1, in2, in3, in4, sz
-        cmhs            \dst1,  \in1,  \in3
-.ifc \sz, .16b
-        cmhs            \dst2,  \in2,  \in4
-.endif
-.endm
-
-.macro xtn_sz dst, in1, in2, sz
-        xtn             \dst\().8b,  \in1
-.ifc \sz, .16b
-        xtn2            \dst\().16b, \in2
-.endif
-.endm
-
 .macro usubl_sz dst1, dst2, in1, in2, sz
         usubl           \dst1,  \in1\().8b,  \in2\().8b
 .ifc \sz, .16b
@@ -179,20 +158,20 @@
 // tmpq2 == tmp3 + tmp4, etc.
 .macro loop_filter wd, sz, mix, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8
 .if \mix == 0
-        dup             v0.8h,  w2        // E
-        dup             v1.8h,  w2        // E
+        dup             v0\sz,  w2        // E
         dup             v2\sz,  w3        // I
         dup             v3\sz,  w4        // H
 .else
-        dup             v0.8h,  w2        // E
+        dup             v0.8b,  w2        // E
         dup             v2.8b,  w3        // I
         dup             v3.8b,  w4        // H
+        lsr             w5,     w2,  #8
         lsr             w6,     w3,  #8
         lsr             w7,     w4,  #8
-        ushr            v1.8h,  v0.8h, #8 // E
+        dup             v1.8b,  w5        // E
         dup             v4.8b,  w6        // I
-        bic             v0.8h,  #255, lsl 8 // E
         dup             v5.8b,  w7        // H
+        trn1            v0.2d,  v0.2d,  v1.2d
         trn1            v2.2d,  v2.2d,  v4.2d
         trn1            v3.2d,  v3.2d,  v5.2d
 .endif
@@ -206,16 +185,15 @@
         umax            v4\sz,  v4\sz,  v5\sz
         umax            v5\sz,  v6\sz,  v7\sz
         umax            \tmp1\sz, \tmp1\sz, \tmp2\sz
-        uabdl_sz        v6.8h,  v7.8h,  v23, v24, \sz // abs(p0 - q0)
+        uabd            v6\sz,  v23\sz, v24\sz        // abs(p0 - q0)
         umax            v4\sz,  v4\sz,  v5\sz
-        add_sz          v6.8h,  v7.8h,  v6.8h,  v7.8h,  v6.8h,  v7.8h, \sz // abs(p0 - q0) * 2
+        uqadd           v6\sz,  v6\sz,  v6\sz         // abs(p0 - q0) * 2
         uabd            v5\sz,  v22\sz, v25\sz        // abs(p1 - q1)
         umax            v4\sz,  v4\sz,  \tmp1\sz      // max(abs(p3 - p2), ..., abs(q2 - q3))
         ushr            v5\sz,  v5\sz,  #1
         cmhs            v4\sz,  v2\sz,  v4\sz         // max(abs()) <= I
-        uaddw_sz        v6.8h,  v7.8h,  v6.8h,  v7.8h,  v5, \sz // abs(p0 - q0) * 2 + abs(p1 - q1) >> 1
-        cmhs_sz         v6.8h,  v7.8h,  v0.8h,  v1.8h,  v6.8h,  v7.8h, \sz
-        xtn_sz          v5,     v6.8h,  v7.8h,  \sz
+        uqadd           v6\sz,  v6\sz,  v5\sz         // abs(p0 - q0) * 2 + abs(p1 - q1) >> 1
+        cmhs            v5\sz,  v0\sz,  v6\sz
         and             v4\sz,  v4\sz,  v5\sz         // fm
 
         // If no pixels need filtering, just exit as soon as possible

libavcodec/arm/vp9lpf_neon.S

@@ -51,7 +51,7 @@
 @ and d28-d31 as temp registers, or d8-d15.
 @ tmp1,tmp2 = tmpq1, tmp3,tmp4 = tmpq2, tmp5,tmp6 = tmpq3, tmp7,tmp8 = tmpq4
 .macro loop_filter wd, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8, tmpq1, tmpq2, tmpq3, tmpq4
-        vdup.u16        q0,  r2 @ E
+        vdup.u8         d0,  r2 @ E
         vdup.u8         d2,  r3 @ I
         ldr             r3,  [sp]
 
@@ -64,16 +64,15 @@
         vmax.u8         d4,  d4,  d5
         vmax.u8         d5,  d6,  d7
         vmax.u8         \tmp1,  \tmp1,  \tmp2
-        vabdl.u8        q3,  d23, d24    @ abs(p0 - q0)
+        vabd.u8         d6,  d23, d24    @ abs(p0 - q0)
         vmax.u8         d4,  d4,  d5
-        vadd.u16        q3,  q3,  q3     @ abs(p0 - q0) * 2
+        vqadd.u8        d6,  d6,  d6     @ abs(p0 - q0) * 2
         vabd.u8         d5,  d22, d25    @ abs(p1 - q1)
         vmax.u8         d4,  d4,  \tmp1  @ max(abs(p3 - p2), ..., abs(q2 - q3))
         vshr.u8         d5,  d5,  #1
         vcle.u8         d4,  d4,  d2     @ max(abs()) <= I
-        vaddw.u8        q3,  q3,  d5     @ abs(p0 - q0) * 2 + abs(p1 - q1) >> 1
-        vcle.u16        q3,  q3,  q0
-        vmovn.u16       d5,  q3
+        vqadd.u8        d6,  d6,  d5     @ abs(p0 - q0) * 2 + abs(p1 - q1) >> 1
+        vcle.u8         d5,  d6,  d0
         vand            d4,  d4,  d5     @ fm
 
         vdup.u8         d3,  r3          @ H