shithub: libvpx

Download patch

ref: 950b68288a4908037bd7b74c105724d813fc4eb1
parent: 798b54a006bf971e206555a563e218f0b57dcb5a
parent: 4e5e5fc52b9b8d4ba1a3a712dfea6418e1bd4fc9
author: Johann Koenig <[email protected]>
date: Tue Aug 18 19:08:51 EDT 2015

Merge "Rename vp8 loopfilter[_neon.c]"

--- a/vp8/common/arm/neon/loopfilter_neon.c
+++ /dev/null
@@ -1,550 +1,0 @@
-/*
- *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-#include <arm_neon.h>
-#include "./vpx_config.h"
-#include "vpx_ports/arm.h"
-
-static INLINE void vp8_loop_filter_neon(
-        uint8x16_t qblimit,  // flimit
-        uint8x16_t qlimit,   // limit
-        uint8x16_t qthresh,  // thresh
-        uint8x16_t q3,       // p3
-        uint8x16_t q4,       // p2
-        uint8x16_t q5,       // p1
-        uint8x16_t q6,       // p0
-        uint8x16_t q7,       // q0
-        uint8x16_t q8,       // q1
-        uint8x16_t q9,       // q2
-        uint8x16_t q10,      // q3
-        uint8x16_t *q5r,     // p1
-        uint8x16_t *q6r,     // p0
-        uint8x16_t *q7r,     // q0
-        uint8x16_t *q8r) {   // q1
-    uint8x16_t q0u8, q1u8, q2u8, q11u8, q12u8, q13u8, q14u8, q15u8;
-    int16x8_t q2s16, q11s16;
-    uint16x8_t q4u16;
-    int8x16_t q1s8, q2s8, q10s8, q11s8, q12s8, q13s8;
-    int8x8_t d2s8, d3s8;
-
-    q11u8 = vabdq_u8(q3, q4);
-    q12u8 = vabdq_u8(q4, q5);
-    q13u8 = vabdq_u8(q5, q6);
-    q14u8 = vabdq_u8(q8, q7);
-    q3    = vabdq_u8(q9, q8);
-    q4    = vabdq_u8(q10, q9);
-
-    q11u8 = vmaxq_u8(q11u8, q12u8);
-    q12u8 = vmaxq_u8(q13u8, q14u8);
-    q3    = vmaxq_u8(q3, q4);
-    q15u8 = vmaxq_u8(q11u8, q12u8);
-
-    q9 = vabdq_u8(q6, q7);
-
-    // vp8_hevmask
-    q13u8 = vcgtq_u8(q13u8, qthresh);
-    q14u8 = vcgtq_u8(q14u8, qthresh);
-    q15u8 = vmaxq_u8(q15u8, q3);
-
-    q2u8 = vabdq_u8(q5, q8);
-    q9 = vqaddq_u8(q9, q9);
-
-    q15u8 = vcgeq_u8(qlimit, q15u8);
-
-    // vp8_filter() function
-    // convert to signed
-    q10 = vdupq_n_u8(0x80);
-    q8 = veorq_u8(q8, q10);
-    q7 = veorq_u8(q7, q10);
-    q6 = veorq_u8(q6, q10);
-    q5 = veorq_u8(q5, q10);
-
-    q2u8 = vshrq_n_u8(q2u8, 1);
-    q9 = vqaddq_u8(q9, q2u8);
-
-    q10 = vdupq_n_u8(3);
-
-    q2s16 = vsubl_s8(vget_low_s8(vreinterpretq_s8_u8(q7)),
-                     vget_low_s8(vreinterpretq_s8_u8(q6)));
-    q11s16 = vsubl_s8(vget_high_s8(vreinterpretq_s8_u8(q7)),
-                      vget_high_s8(vreinterpretq_s8_u8(q6)));
-
-    q9 = vcgeq_u8(qblimit, q9);
-
-    q1s8 = vqsubq_s8(vreinterpretq_s8_u8(q5),
-                    vreinterpretq_s8_u8(q8));
-
-    q14u8 = vorrq_u8(q13u8, q14u8);
-
-    q4u16 = vmovl_u8(vget_low_u8(q10));
-    q2s16 = vmulq_s16(q2s16, vreinterpretq_s16_u16(q4u16));
-    q11s16 = vmulq_s16(q11s16, vreinterpretq_s16_u16(q4u16));
-
-    q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q14u8);
-    q15u8 = vandq_u8(q15u8, q9);
-
-    q1s8 = vreinterpretq_s8_u8(q1u8);
-    q2s16 = vaddw_s8(q2s16, vget_low_s8(q1s8));
-    q11s16 = vaddw_s8(q11s16, vget_high_s8(q1s8));
-
-    q9 = vdupq_n_u8(4);
-    // vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
-    d2s8 = vqmovn_s16(q2s16);
-    d3s8 = vqmovn_s16(q11s16);
-    q1s8 = vcombine_s8(d2s8, d3s8);
-    q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q15u8);
-    q1s8 = vreinterpretq_s8_u8(q1u8);
-
-    q2s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q10));
-    q1s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q9));
-    q2s8 = vshrq_n_s8(q2s8, 3);
-    q1s8 = vshrq_n_s8(q1s8, 3);
-
-    q11s8 = vqaddq_s8(vreinterpretq_s8_u8(q6), q2s8);
-    q10s8 = vqsubq_s8(vreinterpretq_s8_u8(q7), q1s8);
-
-    q1s8 = vrshrq_n_s8(q1s8, 1);
-    q1s8 = vbicq_s8(q1s8, vreinterpretq_s8_u8(q14u8));
-
-    q13s8 = vqaddq_s8(vreinterpretq_s8_u8(q5), q1s8);
-    q12s8 = vqsubq_s8(vreinterpretq_s8_u8(q8), q1s8);
-
-    q0u8 = vdupq_n_u8(0x80);
-    *q8r = veorq_u8(vreinterpretq_u8_s8(q12s8), q0u8);
-    *q7r = veorq_u8(vreinterpretq_u8_s8(q10s8), q0u8);
-    *q6r = veorq_u8(vreinterpretq_u8_s8(q11s8), q0u8);
-    *q5r = veorq_u8(vreinterpretq_u8_s8(q13s8), q0u8);
-    return;
-}
-
-void vp8_loop_filter_horizontal_edge_y_neon(
-        unsigned char *src,
-        int pitch,
-        unsigned char blimit,
-        unsigned char limit,
-        unsigned char thresh) {
-    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
-    uint8x16_t q5, q6, q7, q8, q9, q10;
-
-    qblimit = vdupq_n_u8(blimit);
-    qlimit  = vdupq_n_u8(limit);
-    qthresh = vdupq_n_u8(thresh);
-    src -= (pitch << 2);
-
-    q3 = vld1q_u8(src);
-    src += pitch;
-    q4 = vld1q_u8(src);
-    src += pitch;
-    q5 = vld1q_u8(src);
-    src += pitch;
-    q6 = vld1q_u8(src);
-    src += pitch;
-    q7 = vld1q_u8(src);
-    src += pitch;
-    q8 = vld1q_u8(src);
-    src += pitch;
-    q9 = vld1q_u8(src);
-    src += pitch;
-    q10 = vld1q_u8(src);
-
-    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
-                         q5, q6, q7, q8, q9, q10,
-                         &q5, &q6, &q7, &q8);
-
-    src -= (pitch * 5);
-    vst1q_u8(src, q5);
-    src += pitch;
-    vst1q_u8(src, q6);
-    src += pitch;
-    vst1q_u8(src, q7);
-    src += pitch;
-    vst1q_u8(src, q8);
-    return;
-}
-
-void vp8_loop_filter_horizontal_edge_uv_neon(
-        unsigned char *u,
-        int pitch,
-        unsigned char blimit,
-        unsigned char limit,
-        unsigned char thresh,
-        unsigned char *v) {
-    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
-    uint8x16_t q5, q6, q7, q8, q9, q10;
-    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
-    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
-
-    qblimit = vdupq_n_u8(blimit);
-    qlimit  = vdupq_n_u8(limit);
-    qthresh = vdupq_n_u8(thresh);
-
-    u -= (pitch << 2);
-    v -= (pitch << 2);
-
-    d6  = vld1_u8(u);
-    u += pitch;
-    d7  = vld1_u8(v);
-    v += pitch;
-    d8  = vld1_u8(u);
-    u += pitch;
-    d9  = vld1_u8(v);
-    v += pitch;
-    d10 = vld1_u8(u);
-    u += pitch;
-    d11 = vld1_u8(v);
-    v += pitch;
-    d12 = vld1_u8(u);
-    u += pitch;
-    d13 = vld1_u8(v);
-    v += pitch;
-    d14 = vld1_u8(u);
-    u += pitch;
-    d15 = vld1_u8(v);
-    v += pitch;
-    d16 = vld1_u8(u);
-    u += pitch;
-    d17 = vld1_u8(v);
-    v += pitch;
-    d18 = vld1_u8(u);
-    u += pitch;
-    d19 = vld1_u8(v);
-    v += pitch;
-    d20 = vld1_u8(u);
-    d21 = vld1_u8(v);
-
-    q3 = vcombine_u8(d6, d7);
-    q4 = vcombine_u8(d8, d9);
-    q5 = vcombine_u8(d10, d11);
-    q6 = vcombine_u8(d12, d13);
-    q7 = vcombine_u8(d14, d15);
-    q8 = vcombine_u8(d16, d17);
-    q9 = vcombine_u8(d18, d19);
-    q10 = vcombine_u8(d20, d21);
-
-    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
-                         q5, q6, q7, q8, q9, q10,
-                         &q5, &q6, &q7, &q8);
-
-    u -= (pitch * 5);
-    vst1_u8(u, vget_low_u8(q5));
-    u += pitch;
-    vst1_u8(u, vget_low_u8(q6));
-    u += pitch;
-    vst1_u8(u, vget_low_u8(q7));
-    u += pitch;
-    vst1_u8(u, vget_low_u8(q8));
-
-    v -= (pitch * 5);
-    vst1_u8(v, vget_high_u8(q5));
-    v += pitch;
-    vst1_u8(v, vget_high_u8(q6));
-    v += pitch;
-    vst1_u8(v, vget_high_u8(q7));
-    v += pitch;
-    vst1_u8(v, vget_high_u8(q8));
-    return;
-}
-
-static INLINE void write_4x8(unsigned char *dst, int pitch,
-                             const uint8x8x4_t result) {
-#ifdef VPX_INCOMPATIBLE_GCC
-    /*
-     * uint8x8x4_t result
-    00 01 02 03 | 04 05 06 07
-    10 11 12 13 | 14 15 16 17
-    20 21 22 23 | 24 25 26 27
-    30 31 32 33 | 34 35 36 37
-    ---
-    * after vtrn_u16
-    00 01 20 21 | 04 05 24 25
-    02 03 22 23 | 06 07 26 27
-    10 11 30 31 | 14 15 34 35
-    12 13 32 33 | 16 17 36 37
-    ---
-    * after vtrn_u8
-    00 10 20 30 | 04 14 24 34
-    01 11 21 31 | 05 15 25 35
-    02 12 22 32 | 06 16 26 36
-    03 13 23 33 | 07 17 27 37
-    */
-    const uint16x4x2_t r02_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[0]),
-                                          vreinterpret_u16_u8(result.val[2]));
-    const uint16x4x2_t r13_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[1]),
-                                          vreinterpret_u16_u8(result.val[3]));
-    const uint8x8x2_t r01_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[0]),
-                                       vreinterpret_u8_u16(r13_u16.val[0]));
-    const uint8x8x2_t r23_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[1]),
-                                       vreinterpret_u8_u16(r13_u16.val[1]));
-    const uint32x2_t x_0_4 = vreinterpret_u32_u8(r01_u8.val[0]);
-    const uint32x2_t x_1_5 = vreinterpret_u32_u8(r01_u8.val[1]);
-    const uint32x2_t x_2_6 = vreinterpret_u32_u8(r23_u8.val[0]);
-    const uint32x2_t x_3_7 = vreinterpret_u32_u8(r23_u8.val[1]);
-    vst1_lane_u32((uint32_t *)dst, x_0_4, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_1_5, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_2_6, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_3_7, 0);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_0_4, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_1_5, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_2_6, 1);
-    dst += pitch;
-    vst1_lane_u32((uint32_t *)dst, x_3_7, 1);
-#else
-    vst4_lane_u8(dst, result, 0);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 1);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 2);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 3);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 4);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 5);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 6);
-    dst += pitch;
-    vst4_lane_u8(dst, result, 7);
-#endif  // VPX_INCOMPATIBLE_GCC
-}
-
-void vp8_loop_filter_vertical_edge_y_neon(
-        unsigned char *src,
-        int pitch,
-        unsigned char blimit,
-        unsigned char limit,
-        unsigned char thresh) {
-    unsigned char *s, *d;
-    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
-    uint8x16_t q5, q6, q7, q8, q9, q10;
-    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
-    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
-    uint32x4x2_t q2tmp0, q2tmp1, q2tmp2, q2tmp3;
-    uint16x8x2_t q2tmp4, q2tmp5, q2tmp6, q2tmp7;
-    uint8x16x2_t q2tmp8, q2tmp9, q2tmp10, q2tmp11;
-    uint8x8x4_t q4ResultH, q4ResultL;
-
-    qblimit = vdupq_n_u8(blimit);
-    qlimit  = vdupq_n_u8(limit);
-    qthresh = vdupq_n_u8(thresh);
-
-    s = src - 4;
-    d6  = vld1_u8(s);
-    s += pitch;
-    d8  = vld1_u8(s);
-    s += pitch;
-    d10 = vld1_u8(s);
-    s += pitch;
-    d12 = vld1_u8(s);
-    s += pitch;
-    d14 = vld1_u8(s);
-    s += pitch;
-    d16 = vld1_u8(s);
-    s += pitch;
-    d18 = vld1_u8(s);
-    s += pitch;
-    d20 = vld1_u8(s);
-    s += pitch;
-    d7  = vld1_u8(s);
-    s += pitch;
-    d9  = vld1_u8(s);
-    s += pitch;
-    d11 = vld1_u8(s);
-    s += pitch;
-    d13 = vld1_u8(s);
-    s += pitch;
-    d15 = vld1_u8(s);
-    s += pitch;
-    d17 = vld1_u8(s);
-    s += pitch;
-    d19 = vld1_u8(s);
-    s += pitch;
-    d21 = vld1_u8(s);
-
-    q3 = vcombine_u8(d6, d7);
-    q4 = vcombine_u8(d8, d9);
-    q5 = vcombine_u8(d10, d11);
-    q6 = vcombine_u8(d12, d13);
-    q7 = vcombine_u8(d14, d15);
-    q8 = vcombine_u8(d16, d17);
-    q9 = vcombine_u8(d18, d19);
-    q10 = vcombine_u8(d20, d21);
-
-    q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
-    q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
-    q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
-    q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
-
-    q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
-                       vreinterpretq_u16_u32(q2tmp2.val[0]));
-    q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
-                       vreinterpretq_u16_u32(q2tmp3.val[0]));
-    q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
-                       vreinterpretq_u16_u32(q2tmp2.val[1]));
-    q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
-                       vreinterpretq_u16_u32(q2tmp3.val[1]));
-
-    q2tmp8  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
-                       vreinterpretq_u8_u16(q2tmp5.val[0]));
-    q2tmp9  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
-                       vreinterpretq_u8_u16(q2tmp5.val[1]));
-    q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
-                       vreinterpretq_u8_u16(q2tmp7.val[0]));
-    q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
-                       vreinterpretq_u8_u16(q2tmp7.val[1]));
-
-    q3 = q2tmp8.val[0];
-    q4 = q2tmp8.val[1];
-    q5 = q2tmp9.val[0];
-    q6 = q2tmp9.val[1];
-    q7 = q2tmp10.val[0];
-    q8 = q2tmp10.val[1];
-    q9 = q2tmp11.val[0];
-    q10 = q2tmp11.val[1];
-
-    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
-                         q5, q6, q7, q8, q9, q10,
-                         &q5, &q6, &q7, &q8);
-
-    q4ResultL.val[0] = vget_low_u8(q5);   // d10
-    q4ResultL.val[1] = vget_low_u8(q6);   // d12
-    q4ResultL.val[2] = vget_low_u8(q7);   // d14
-    q4ResultL.val[3] = vget_low_u8(q8);   // d16
-    q4ResultH.val[0] = vget_high_u8(q5);  // d11
-    q4ResultH.val[1] = vget_high_u8(q6);  // d13
-    q4ResultH.val[2] = vget_high_u8(q7);  // d15
-    q4ResultH.val[3] = vget_high_u8(q8);  // d17
-
-    d = src - 2;
-    write_4x8(d, pitch, q4ResultL);
-    d += pitch * 8;
-    write_4x8(d, pitch, q4ResultH);
-}
-
-void vp8_loop_filter_vertical_edge_uv_neon(
-        unsigned char *u,
-        int pitch,
-        unsigned char blimit,
-        unsigned char limit,
-        unsigned char thresh,
-        unsigned char *v) {
-    unsigned char *us, *ud;
-    unsigned char *vs, *vd;
-    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
-    uint8x16_t q5, q6, q7, q8, q9, q10;
-    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
-    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
-    uint32x4x2_t q2tmp0, q2tmp1, q2tmp2, q2tmp3;
-    uint16x8x2_t q2tmp4, q2tmp5, q2tmp6, q2tmp7;
-    uint8x16x2_t q2tmp8, q2tmp9, q2tmp10, q2tmp11;
-    uint8x8x4_t q4ResultH, q4ResultL;
-
-    qblimit = vdupq_n_u8(blimit);
-    qlimit  = vdupq_n_u8(limit);
-    qthresh = vdupq_n_u8(thresh);
-
-    us = u - 4;
-    d6 = vld1_u8(us);
-    us += pitch;
-    d8 = vld1_u8(us);
-    us += pitch;
-    d10 = vld1_u8(us);
-    us += pitch;
-    d12 = vld1_u8(us);
-    us += pitch;
-    d14 = vld1_u8(us);
-    us += pitch;
-    d16 = vld1_u8(us);
-    us += pitch;
-    d18 = vld1_u8(us);
-    us += pitch;
-    d20 = vld1_u8(us);
-
-    vs = v - 4;
-    d7 = vld1_u8(vs);
-    vs += pitch;
-    d9 = vld1_u8(vs);
-    vs += pitch;
-    d11 = vld1_u8(vs);
-    vs += pitch;
-    d13 = vld1_u8(vs);
-    vs += pitch;
-    d15 = vld1_u8(vs);
-    vs += pitch;
-    d17 = vld1_u8(vs);
-    vs += pitch;
-    d19 = vld1_u8(vs);
-    vs += pitch;
-    d21 = vld1_u8(vs);
-
-    q3 = vcombine_u8(d6, d7);
-    q4 = vcombine_u8(d8, d9);
-    q5 = vcombine_u8(d10, d11);
-    q6 = vcombine_u8(d12, d13);
-    q7 = vcombine_u8(d14, d15);
-    q8 = vcombine_u8(d16, d17);
-    q9 = vcombine_u8(d18, d19);
-    q10 = vcombine_u8(d20, d21);
-
-    q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
-    q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
-    q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
-    q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
-
-    q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
-                       vreinterpretq_u16_u32(q2tmp2.val[0]));
-    q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
-                       vreinterpretq_u16_u32(q2tmp3.val[0]));
-    q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
-                       vreinterpretq_u16_u32(q2tmp2.val[1]));
-    q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
-                       vreinterpretq_u16_u32(q2tmp3.val[1]));
-
-    q2tmp8  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
-                       vreinterpretq_u8_u16(q2tmp5.val[0]));
-    q2tmp9  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
-                       vreinterpretq_u8_u16(q2tmp5.val[1]));
-    q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
-                       vreinterpretq_u8_u16(q2tmp7.val[0]));
-    q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
-                       vreinterpretq_u8_u16(q2tmp7.val[1]));
-
-    q3 = q2tmp8.val[0];
-    q4 = q2tmp8.val[1];
-    q5 = q2tmp9.val[0];
-    q6 = q2tmp9.val[1];
-    q7 = q2tmp10.val[0];
-    q8 = q2tmp10.val[1];
-    q9 = q2tmp11.val[0];
-    q10 = q2tmp11.val[1];
-
-    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
-                         q5, q6, q7, q8, q9, q10,
-                         &q5, &q6, &q7, &q8);
-
-    q4ResultL.val[0] = vget_low_u8(q5);   // d10
-    q4ResultL.val[1] = vget_low_u8(q6);   // d12
-    q4ResultL.val[2] = vget_low_u8(q7);   // d14
-    q4ResultL.val[3] = vget_low_u8(q8);   // d16
-    ud = u - 2;
-    write_4x8(ud, pitch, q4ResultL);
-
-    q4ResultH.val[0] = vget_high_u8(q5);  // d11
-    q4ResultH.val[1] = vget_high_u8(q6);  // d13
-    q4ResultH.val[2] = vget_high_u8(q7);  // d15
-    q4ResultH.val[3] = vget_high_u8(q8);  // d17
-    vd = v - 2;
-    write_4x8(vd, pitch, q4ResultH);
-}
--- /dev/null
+++ b/vp8/common/arm/neon/vp8_loopfilter_neon.c
@@ -1,0 +1,550 @@
+/*
+ *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <arm_neon.h>
+#include "./vpx_config.h"
+#include "vpx_ports/arm.h"
+
+static INLINE void vp8_loop_filter_neon(
+        uint8x16_t qblimit,  // flimit
+        uint8x16_t qlimit,   // limit
+        uint8x16_t qthresh,  // thresh
+        uint8x16_t q3,       // p3
+        uint8x16_t q4,       // p2
+        uint8x16_t q5,       // p1
+        uint8x16_t q6,       // p0
+        uint8x16_t q7,       // q0
+        uint8x16_t q8,       // q1
+        uint8x16_t q9,       // q2
+        uint8x16_t q10,      // q3
+        uint8x16_t *q5r,     // p1
+        uint8x16_t *q6r,     // p0
+        uint8x16_t *q7r,     // q0
+        uint8x16_t *q8r) {   // q1
+    uint8x16_t q0u8, q1u8, q2u8, q11u8, q12u8, q13u8, q14u8, q15u8;
+    int16x8_t q2s16, q11s16;
+    uint16x8_t q4u16;
+    int8x16_t q1s8, q2s8, q10s8, q11s8, q12s8, q13s8;
+    int8x8_t d2s8, d3s8;
+
+    q11u8 = vabdq_u8(q3, q4);
+    q12u8 = vabdq_u8(q4, q5);
+    q13u8 = vabdq_u8(q5, q6);
+    q14u8 = vabdq_u8(q8, q7);
+    q3    = vabdq_u8(q9, q8);
+    q4    = vabdq_u8(q10, q9);
+
+    q11u8 = vmaxq_u8(q11u8, q12u8);
+    q12u8 = vmaxq_u8(q13u8, q14u8);
+    q3    = vmaxq_u8(q3, q4);
+    q15u8 = vmaxq_u8(q11u8, q12u8);
+
+    q9 = vabdq_u8(q6, q7);
+
+    // vp8_hevmask
+    q13u8 = vcgtq_u8(q13u8, qthresh);
+    q14u8 = vcgtq_u8(q14u8, qthresh);
+    q15u8 = vmaxq_u8(q15u8, q3);
+
+    q2u8 = vabdq_u8(q5, q8);
+    q9 = vqaddq_u8(q9, q9);
+
+    q15u8 = vcgeq_u8(qlimit, q15u8);
+
+    // vp8_filter() function
+    // convert to signed
+    q10 = vdupq_n_u8(0x80);
+    q8 = veorq_u8(q8, q10);
+    q7 = veorq_u8(q7, q10);
+    q6 = veorq_u8(q6, q10);
+    q5 = veorq_u8(q5, q10);
+
+    q2u8 = vshrq_n_u8(q2u8, 1);
+    q9 = vqaddq_u8(q9, q2u8);
+
+    q10 = vdupq_n_u8(3);
+
+    q2s16 = vsubl_s8(vget_low_s8(vreinterpretq_s8_u8(q7)),
+                     vget_low_s8(vreinterpretq_s8_u8(q6)));
+    q11s16 = vsubl_s8(vget_high_s8(vreinterpretq_s8_u8(q7)),
+                      vget_high_s8(vreinterpretq_s8_u8(q6)));
+
+    q9 = vcgeq_u8(qblimit, q9);
+
+    q1s8 = vqsubq_s8(vreinterpretq_s8_u8(q5),
+                    vreinterpretq_s8_u8(q8));
+
+    q14u8 = vorrq_u8(q13u8, q14u8);
+
+    q4u16 = vmovl_u8(vget_low_u8(q10));
+    q2s16 = vmulq_s16(q2s16, vreinterpretq_s16_u16(q4u16));
+    q11s16 = vmulq_s16(q11s16, vreinterpretq_s16_u16(q4u16));
+
+    q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q14u8);
+    q15u8 = vandq_u8(q15u8, q9);
+
+    q1s8 = vreinterpretq_s8_u8(q1u8);
+    q2s16 = vaddw_s8(q2s16, vget_low_s8(q1s8));
+    q11s16 = vaddw_s8(q11s16, vget_high_s8(q1s8));
+
+    q9 = vdupq_n_u8(4);
+    // vp8_filter = clamp(vp8_filter + 3 * ( qs0 - ps0))
+    d2s8 = vqmovn_s16(q2s16);
+    d3s8 = vqmovn_s16(q11s16);
+    q1s8 = vcombine_s8(d2s8, d3s8);
+    q1u8 = vandq_u8(vreinterpretq_u8_s8(q1s8), q15u8);
+    q1s8 = vreinterpretq_s8_u8(q1u8);
+
+    q2s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q10));
+    q1s8 = vqaddq_s8(q1s8, vreinterpretq_s8_u8(q9));
+    q2s8 = vshrq_n_s8(q2s8, 3);
+    q1s8 = vshrq_n_s8(q1s8, 3);
+
+    q11s8 = vqaddq_s8(vreinterpretq_s8_u8(q6), q2s8);
+    q10s8 = vqsubq_s8(vreinterpretq_s8_u8(q7), q1s8);
+
+    q1s8 = vrshrq_n_s8(q1s8, 1);
+    q1s8 = vbicq_s8(q1s8, vreinterpretq_s8_u8(q14u8));
+
+    q13s8 = vqaddq_s8(vreinterpretq_s8_u8(q5), q1s8);
+    q12s8 = vqsubq_s8(vreinterpretq_s8_u8(q8), q1s8);
+
+    q0u8 = vdupq_n_u8(0x80);
+    *q8r = veorq_u8(vreinterpretq_u8_s8(q12s8), q0u8);
+    *q7r = veorq_u8(vreinterpretq_u8_s8(q10s8), q0u8);
+    *q6r = veorq_u8(vreinterpretq_u8_s8(q11s8), q0u8);
+    *q5r = veorq_u8(vreinterpretq_u8_s8(q13s8), q0u8);
+    return;
+}
+
+void vp8_loop_filter_horizontal_edge_y_neon(
+        unsigned char *src,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh) {
+    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
+    uint8x16_t q5, q6, q7, q8, q9, q10;
+
+    qblimit = vdupq_n_u8(blimit);
+    qlimit  = vdupq_n_u8(limit);
+    qthresh = vdupq_n_u8(thresh);
+    src -= (pitch << 2);
+
+    q3 = vld1q_u8(src);
+    src += pitch;
+    q4 = vld1q_u8(src);
+    src += pitch;
+    q5 = vld1q_u8(src);
+    src += pitch;
+    q6 = vld1q_u8(src);
+    src += pitch;
+    q7 = vld1q_u8(src);
+    src += pitch;
+    q8 = vld1q_u8(src);
+    src += pitch;
+    q9 = vld1q_u8(src);
+    src += pitch;
+    q10 = vld1q_u8(src);
+
+    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
+                         q5, q6, q7, q8, q9, q10,
+                         &q5, &q6, &q7, &q8);
+
+    src -= (pitch * 5);
+    vst1q_u8(src, q5);
+    src += pitch;
+    vst1q_u8(src, q6);
+    src += pitch;
+    vst1q_u8(src, q7);
+    src += pitch;
+    vst1q_u8(src, q8);
+    return;
+}
+
+void vp8_loop_filter_horizontal_edge_uv_neon(
+        unsigned char *u,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh,
+        unsigned char *v) {
+    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
+    uint8x16_t q5, q6, q7, q8, q9, q10;
+    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
+    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
+
+    qblimit = vdupq_n_u8(blimit);
+    qlimit  = vdupq_n_u8(limit);
+    qthresh = vdupq_n_u8(thresh);
+
+    u -= (pitch << 2);
+    v -= (pitch << 2);
+
+    d6  = vld1_u8(u);
+    u += pitch;
+    d7  = vld1_u8(v);
+    v += pitch;
+    d8  = vld1_u8(u);
+    u += pitch;
+    d9  = vld1_u8(v);
+    v += pitch;
+    d10 = vld1_u8(u);
+    u += pitch;
+    d11 = vld1_u8(v);
+    v += pitch;
+    d12 = vld1_u8(u);
+    u += pitch;
+    d13 = vld1_u8(v);
+    v += pitch;
+    d14 = vld1_u8(u);
+    u += pitch;
+    d15 = vld1_u8(v);
+    v += pitch;
+    d16 = vld1_u8(u);
+    u += pitch;
+    d17 = vld1_u8(v);
+    v += pitch;
+    d18 = vld1_u8(u);
+    u += pitch;
+    d19 = vld1_u8(v);
+    v += pitch;
+    d20 = vld1_u8(u);
+    d21 = vld1_u8(v);
+
+    q3 = vcombine_u8(d6, d7);
+    q4 = vcombine_u8(d8, d9);
+    q5 = vcombine_u8(d10, d11);
+    q6 = vcombine_u8(d12, d13);
+    q7 = vcombine_u8(d14, d15);
+    q8 = vcombine_u8(d16, d17);
+    q9 = vcombine_u8(d18, d19);
+    q10 = vcombine_u8(d20, d21);
+
+    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
+                         q5, q6, q7, q8, q9, q10,
+                         &q5, &q6, &q7, &q8);
+
+    u -= (pitch * 5);
+    vst1_u8(u, vget_low_u8(q5));
+    u += pitch;
+    vst1_u8(u, vget_low_u8(q6));
+    u += pitch;
+    vst1_u8(u, vget_low_u8(q7));
+    u += pitch;
+    vst1_u8(u, vget_low_u8(q8));
+
+    v -= (pitch * 5);
+    vst1_u8(v, vget_high_u8(q5));
+    v += pitch;
+    vst1_u8(v, vget_high_u8(q6));
+    v += pitch;
+    vst1_u8(v, vget_high_u8(q7));
+    v += pitch;
+    vst1_u8(v, vget_high_u8(q8));
+    return;
+}
+
+static INLINE void write_4x8(unsigned char *dst, int pitch,
+                             const uint8x8x4_t result) {
+#ifdef VPX_INCOMPATIBLE_GCC
+    /*
+     * uint8x8x4_t result
+    00 01 02 03 | 04 05 06 07
+    10 11 12 13 | 14 15 16 17
+    20 21 22 23 | 24 25 26 27
+    30 31 32 33 | 34 35 36 37
+    ---
+    * after vtrn_u16
+    00 01 20 21 | 04 05 24 25
+    02 03 22 23 | 06 07 26 27
+    10 11 30 31 | 14 15 34 35
+    12 13 32 33 | 16 17 36 37
+    ---
+    * after vtrn_u8
+    00 10 20 30 | 04 14 24 34
+    01 11 21 31 | 05 15 25 35
+    02 12 22 32 | 06 16 26 36
+    03 13 23 33 | 07 17 27 37
+    */
+    const uint16x4x2_t r02_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[0]),
+                                          vreinterpret_u16_u8(result.val[2]));
+    const uint16x4x2_t r13_u16 = vtrn_u16(vreinterpret_u16_u8(result.val[1]),
+                                          vreinterpret_u16_u8(result.val[3]));
+    const uint8x8x2_t r01_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[0]),
+                                       vreinterpret_u8_u16(r13_u16.val[0]));
+    const uint8x8x2_t r23_u8 = vtrn_u8(vreinterpret_u8_u16(r02_u16.val[1]),
+                                       vreinterpret_u8_u16(r13_u16.val[1]));
+    const uint32x2_t x_0_4 = vreinterpret_u32_u8(r01_u8.val[0]);
+    const uint32x2_t x_1_5 = vreinterpret_u32_u8(r01_u8.val[1]);
+    const uint32x2_t x_2_6 = vreinterpret_u32_u8(r23_u8.val[0]);
+    const uint32x2_t x_3_7 = vreinterpret_u32_u8(r23_u8.val[1]);
+    vst1_lane_u32((uint32_t *)dst, x_0_4, 0);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_1_5, 0);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_2_6, 0);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_3_7, 0);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_0_4, 1);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_1_5, 1);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_2_6, 1);
+    dst += pitch;
+    vst1_lane_u32((uint32_t *)dst, x_3_7, 1);
+#else
+    vst4_lane_u8(dst, result, 0);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 1);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 2);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 3);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 4);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 5);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 6);
+    dst += pitch;
+    vst4_lane_u8(dst, result, 7);
+#endif  // VPX_INCOMPATIBLE_GCC
+}
+
+void vp8_loop_filter_vertical_edge_y_neon(
+        unsigned char *src,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh) {
+    unsigned char *s, *d;
+    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
+    uint8x16_t q5, q6, q7, q8, q9, q10;
+    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
+    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
+    uint32x4x2_t q2tmp0, q2tmp1, q2tmp2, q2tmp3;
+    uint16x8x2_t q2tmp4, q2tmp5, q2tmp6, q2tmp7;
+    uint8x16x2_t q2tmp8, q2tmp9, q2tmp10, q2tmp11;
+    uint8x8x4_t q4ResultH, q4ResultL;
+
+    qblimit = vdupq_n_u8(blimit);
+    qlimit  = vdupq_n_u8(limit);
+    qthresh = vdupq_n_u8(thresh);
+
+    s = src - 4;
+    d6  = vld1_u8(s);
+    s += pitch;
+    d8  = vld1_u8(s);
+    s += pitch;
+    d10 = vld1_u8(s);
+    s += pitch;
+    d12 = vld1_u8(s);
+    s += pitch;
+    d14 = vld1_u8(s);
+    s += pitch;
+    d16 = vld1_u8(s);
+    s += pitch;
+    d18 = vld1_u8(s);
+    s += pitch;
+    d20 = vld1_u8(s);
+    s += pitch;
+    d7  = vld1_u8(s);
+    s += pitch;
+    d9  = vld1_u8(s);
+    s += pitch;
+    d11 = vld1_u8(s);
+    s += pitch;
+    d13 = vld1_u8(s);
+    s += pitch;
+    d15 = vld1_u8(s);
+    s += pitch;
+    d17 = vld1_u8(s);
+    s += pitch;
+    d19 = vld1_u8(s);
+    s += pitch;
+    d21 = vld1_u8(s);
+
+    q3 = vcombine_u8(d6, d7);
+    q4 = vcombine_u8(d8, d9);
+    q5 = vcombine_u8(d10, d11);
+    q6 = vcombine_u8(d12, d13);
+    q7 = vcombine_u8(d14, d15);
+    q8 = vcombine_u8(d16, d17);
+    q9 = vcombine_u8(d18, d19);
+    q10 = vcombine_u8(d20, d21);
+
+    q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+    q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+    q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+    q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
+
+    q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+                       vreinterpretq_u16_u32(q2tmp2.val[0]));
+    q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+                       vreinterpretq_u16_u32(q2tmp3.val[0]));
+    q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+                       vreinterpretq_u16_u32(q2tmp2.val[1]));
+    q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+                       vreinterpretq_u16_u32(q2tmp3.val[1]));
+
+    q2tmp8  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+                       vreinterpretq_u8_u16(q2tmp5.val[0]));
+    q2tmp9  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+                       vreinterpretq_u8_u16(q2tmp5.val[1]));
+    q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+                       vreinterpretq_u8_u16(q2tmp7.val[0]));
+    q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+                       vreinterpretq_u8_u16(q2tmp7.val[1]));
+
+    q3 = q2tmp8.val[0];
+    q4 = q2tmp8.val[1];
+    q5 = q2tmp9.val[0];
+    q6 = q2tmp9.val[1];
+    q7 = q2tmp10.val[0];
+    q8 = q2tmp10.val[1];
+    q9 = q2tmp11.val[0];
+    q10 = q2tmp11.val[1];
+
+    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
+                         q5, q6, q7, q8, q9, q10,
+                         &q5, &q6, &q7, &q8);
+
+    q4ResultL.val[0] = vget_low_u8(q5);   // d10
+    q4ResultL.val[1] = vget_low_u8(q6);   // d12
+    q4ResultL.val[2] = vget_low_u8(q7);   // d14
+    q4ResultL.val[3] = vget_low_u8(q8);   // d16
+    q4ResultH.val[0] = vget_high_u8(q5);  // d11
+    q4ResultH.val[1] = vget_high_u8(q6);  // d13
+    q4ResultH.val[2] = vget_high_u8(q7);  // d15
+    q4ResultH.val[3] = vget_high_u8(q8);  // d17
+
+    d = src - 2;
+    write_4x8(d, pitch, q4ResultL);
+    d += pitch * 8;
+    write_4x8(d, pitch, q4ResultH);
+}
+
+void vp8_loop_filter_vertical_edge_uv_neon(
+        unsigned char *u,
+        int pitch,
+        unsigned char blimit,
+        unsigned char limit,
+        unsigned char thresh,
+        unsigned char *v) {
+    unsigned char *us, *ud;
+    unsigned char *vs, *vd;
+    uint8x16_t qblimit, qlimit, qthresh, q3, q4;
+    uint8x16_t q5, q6, q7, q8, q9, q10;
+    uint8x8_t d6, d7, d8, d9, d10, d11, d12, d13, d14;
+    uint8x8_t d15, d16, d17, d18, d19, d20, d21;
+    uint32x4x2_t q2tmp0, q2tmp1, q2tmp2, q2tmp3;
+    uint16x8x2_t q2tmp4, q2tmp5, q2tmp6, q2tmp7;
+    uint8x16x2_t q2tmp8, q2tmp9, q2tmp10, q2tmp11;
+    uint8x8x4_t q4ResultH, q4ResultL;
+
+    qblimit = vdupq_n_u8(blimit);
+    qlimit  = vdupq_n_u8(limit);
+    qthresh = vdupq_n_u8(thresh);
+
+    us = u - 4;
+    d6 = vld1_u8(us);
+    us += pitch;
+    d8 = vld1_u8(us);
+    us += pitch;
+    d10 = vld1_u8(us);
+    us += pitch;
+    d12 = vld1_u8(us);
+    us += pitch;
+    d14 = vld1_u8(us);
+    us += pitch;
+    d16 = vld1_u8(us);
+    us += pitch;
+    d18 = vld1_u8(us);
+    us += pitch;
+    d20 = vld1_u8(us);
+
+    vs = v - 4;
+    d7 = vld1_u8(vs);
+    vs += pitch;
+    d9 = vld1_u8(vs);
+    vs += pitch;
+    d11 = vld1_u8(vs);
+    vs += pitch;
+    d13 = vld1_u8(vs);
+    vs += pitch;
+    d15 = vld1_u8(vs);
+    vs += pitch;
+    d17 = vld1_u8(vs);
+    vs += pitch;
+    d19 = vld1_u8(vs);
+    vs += pitch;
+    d21 = vld1_u8(vs);
+
+    q3 = vcombine_u8(d6, d7);
+    q4 = vcombine_u8(d8, d9);
+    q5 = vcombine_u8(d10, d11);
+    q6 = vcombine_u8(d12, d13);
+    q7 = vcombine_u8(d14, d15);
+    q8 = vcombine_u8(d16, d17);
+    q9 = vcombine_u8(d18, d19);
+    q10 = vcombine_u8(d20, d21);
+
+    q2tmp0 = vtrnq_u32(vreinterpretq_u32_u8(q3), vreinterpretq_u32_u8(q7));
+    q2tmp1 = vtrnq_u32(vreinterpretq_u32_u8(q4), vreinterpretq_u32_u8(q8));
+    q2tmp2 = vtrnq_u32(vreinterpretq_u32_u8(q5), vreinterpretq_u32_u8(q9));
+    q2tmp3 = vtrnq_u32(vreinterpretq_u32_u8(q6), vreinterpretq_u32_u8(q10));
+
+    q2tmp4 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[0]),
+                       vreinterpretq_u16_u32(q2tmp2.val[0]));
+    q2tmp5 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[0]),
+                       vreinterpretq_u16_u32(q2tmp3.val[0]));
+    q2tmp6 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp0.val[1]),
+                       vreinterpretq_u16_u32(q2tmp2.val[1]));
+    q2tmp7 = vtrnq_u16(vreinterpretq_u16_u32(q2tmp1.val[1]),
+                       vreinterpretq_u16_u32(q2tmp3.val[1]));
+
+    q2tmp8  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[0]),
+                       vreinterpretq_u8_u16(q2tmp5.val[0]));
+    q2tmp9  = vtrnq_u8(vreinterpretq_u8_u16(q2tmp4.val[1]),
+                       vreinterpretq_u8_u16(q2tmp5.val[1]));
+    q2tmp10 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[0]),
+                       vreinterpretq_u8_u16(q2tmp7.val[0]));
+    q2tmp11 = vtrnq_u8(vreinterpretq_u8_u16(q2tmp6.val[1]),
+                       vreinterpretq_u8_u16(q2tmp7.val[1]));
+
+    q3 = q2tmp8.val[0];
+    q4 = q2tmp8.val[1];
+    q5 = q2tmp9.val[0];
+    q6 = q2tmp9.val[1];
+    q7 = q2tmp10.val[0];
+    q8 = q2tmp10.val[1];
+    q9 = q2tmp11.val[0];
+    q10 = q2tmp11.val[1];
+
+    vp8_loop_filter_neon(qblimit, qlimit, qthresh, q3, q4,
+                         q5, q6, q7, q8, q9, q10,
+                         &q5, &q6, &q7, &q8);
+
+    q4ResultL.val[0] = vget_low_u8(q5);   // d10
+    q4ResultL.val[1] = vget_low_u8(q6);   // d12
+    q4ResultL.val[2] = vget_low_u8(q7);   // d14
+    q4ResultL.val[3] = vget_low_u8(q8);   // d16
+    ud = u - 2;
+    write_4x8(ud, pitch, q4ResultL);
+
+    q4ResultH.val[0] = vget_high_u8(q5);  // d11
+    q4ResultH.val[1] = vget_high_u8(q6);  // d13
+    q4ResultH.val[2] = vget_high_u8(q7);  // d15
+    q4ResultH.val[3] = vget_high_u8(q8);  // d17
+    vd = v - 2;
+    write_4x8(vd, pitch, q4ResultH);
+}
--- a/vp8/common/loopfilter.c
+++ /dev/null
@@ -1,661 +1,0 @@
-/*
- *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
- *
- *  Use of this source code is governed by a BSD-style license
- *  that can be found in the LICENSE file in the root of the source
- *  tree. An additional intellectual property rights grant can be found
- *  in the file PATENTS.  All contributing project authors may
- *  be found in the AUTHORS file in the root of the source tree.
- */
-
-
-#include "vpx_config.h"
-#include "vp8_rtcd.h"
-#include "loopfilter.h"
-#include "onyxc_int.h"
-#include "vpx_mem/vpx_mem.h"
-
-
-static void lf_init_lut(loop_filter_info_n *lfi)
-{
-    int filt_lvl;
-
-    for (filt_lvl = 0; filt_lvl <= MAX_LOOP_FILTER; filt_lvl++)
-    {
-        if (filt_lvl >= 40)
-        {
-            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 2;
-            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 3;
-        }
-        else if (filt_lvl >= 20)
-        {
-            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
-            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 2;
-        }
-        else if (filt_lvl >= 15)
-        {
-            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
-            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 1;
-        }
-        else
-        {
-            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 0;
-            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 0;
-        }
-    }
-
-    lfi->mode_lf_lut[DC_PRED] = 1;
-    lfi->mode_lf_lut[V_PRED] = 1;
-    lfi->mode_lf_lut[H_PRED] = 1;
-    lfi->mode_lf_lut[TM_PRED] = 1;
-    lfi->mode_lf_lut[B_PRED]  = 0;
-
-    lfi->mode_lf_lut[ZEROMV]  = 1;
-    lfi->mode_lf_lut[NEARESTMV] = 2;
-    lfi->mode_lf_lut[NEARMV] = 2;
-    lfi->mode_lf_lut[NEWMV] = 2;
-    lfi->mode_lf_lut[SPLITMV] = 3;
-
-}
-
-void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
-                                      int sharpness_lvl)
-{
-    int i;
-
-    /* For each possible value for the loop filter fill out limits */
-    for (i = 0; i <= MAX_LOOP_FILTER; i++)
-    {
-        int filt_lvl = i;
-        int block_inside_limit = 0;
-
-        /* Set loop filter paramaeters that control sharpness. */
-        block_inside_limit = filt_lvl >> (sharpness_lvl > 0);
-        block_inside_limit = block_inside_limit >> (sharpness_lvl > 4);
-
-        if (sharpness_lvl > 0)
-        {
-            if (block_inside_limit > (9 - sharpness_lvl))
-                block_inside_limit = (9 - sharpness_lvl);
-        }
-
-        if (block_inside_limit < 1)
-            block_inside_limit = 1;
-
-        memset(lfi->lim[i], block_inside_limit, SIMD_WIDTH);
-        memset(lfi->blim[i], (2 * filt_lvl + block_inside_limit), SIMD_WIDTH);
-        memset(lfi->mblim[i], (2 * (filt_lvl + 2) + block_inside_limit),
-               SIMD_WIDTH);
-    }
-}
-
-void vp8_loop_filter_init(VP8_COMMON *cm)
-{
-    loop_filter_info_n *lfi = &cm->lf_info;
-    int i;
-
-    /* init limits for given sharpness*/
-    vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
-    cm->last_sharpness_level = cm->sharpness_level;
-
-    /* init LUT for lvl  and hev thr picking */
-    lf_init_lut(lfi);
-
-    /* init hev threshold const vectors */
-    for(i = 0; i < 4 ; i++)
-    {
-        memset(lfi->hev_thr[i], i, SIMD_WIDTH);
-    }
-}
-
-void vp8_loop_filter_frame_init(VP8_COMMON *cm,
-                                MACROBLOCKD *mbd,
-                                int default_filt_lvl)
-{
-    int seg,  /* segment number */
-        ref,  /* index in ref_lf_deltas */
-        mode; /* index in mode_lf_deltas */
-
-    loop_filter_info_n *lfi = &cm->lf_info;
-
-    /* update limits if sharpness has changed */
-    if(cm->last_sharpness_level != cm->sharpness_level)
-    {
-        vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
-        cm->last_sharpness_level = cm->sharpness_level;
-    }
-
-    for(seg = 0; seg < MAX_MB_SEGMENTS; seg++)
-    {
-        int lvl_seg = default_filt_lvl;
-        int lvl_ref, lvl_mode;
-
-        /* Note the baseline filter values for each segment */
-        if (mbd->segmentation_enabled)
-        {
-            /* Abs value */
-            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
-            {
-                lvl_seg = mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
-            }
-            else  /* Delta Value */
-            {
-                lvl_seg += mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
-                lvl_seg = (lvl_seg > 0) ? ((lvl_seg > 63) ? 63: lvl_seg) : 0;
-            }
-        }
-
-        if (!mbd->mode_ref_lf_delta_enabled)
-        {
-            /* we could get rid of this if we assume that deltas are set to
-             * zero when not in use; encoder always uses deltas
-             */
-            memset(lfi->lvl[seg][0], lvl_seg, 4 * 4 );
-            continue;
-        }
-
-        /* INTRA_FRAME */
-        ref = INTRA_FRAME;
-
-        /* Apply delta for reference frame */
-        lvl_ref = lvl_seg + mbd->ref_lf_deltas[ref];
-
-        /* Apply delta for Intra modes */
-        mode = 0; /* B_PRED */
-        /* Only the split mode BPRED has a further special case */
-        lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
-        /* clamp */
-        lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0;
-
-        lfi->lvl[seg][ref][mode] = lvl_mode;
-
-        mode = 1; /* all the rest of Intra modes */
-        /* clamp */
-        lvl_mode = (lvl_ref > 0) ? (lvl_ref > 63 ? 63 : lvl_ref) : 0;
-        lfi->lvl[seg][ref][mode] = lvl_mode;
-
-        /* LAST, GOLDEN, ALT */
-        for(ref = 1; ref < MAX_REF_FRAMES; ref++)
-        {
-            /* Apply delta for reference frame */
-            lvl_ref = lvl_seg + mbd->ref_lf_deltas[ref];
-
-            /* Apply delta for Inter modes */
-            for (mode = 1; mode < 4; mode++)
-            {
-                lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
-                /* clamp */
-                lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0;
-
-                lfi->lvl[seg][ref][mode] = lvl_mode;
-            }
-        }
-    }
-}
-
-
-void vp8_loop_filter_row_normal(VP8_COMMON *cm, MODE_INFO *mode_info_context,
-                         int mb_row, int post_ystride, int post_uvstride,
-                         unsigned char *y_ptr, unsigned char *u_ptr,
-                         unsigned char *v_ptr)
-{
-    int mb_col;
-    int filter_level;
-    loop_filter_info_n *lfi_n = &cm->lf_info;
-    loop_filter_info lfi;
-    FRAME_TYPE frame_type = cm->frame_type;
-
-    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
-    {
-        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                        mode_info_context->mbmi.mode != SPLITMV &&
-                        mode_info_context->mbmi.mb_skip_coeff);
-
-        const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-        const int seg = mode_info_context->mbmi.segment_id;
-        const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-        filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-
-        if (filter_level)
-        {
-            const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
-            lfi.mblim = lfi_n->mblim[filter_level];
-            lfi.blim = lfi_n->blim[filter_level];
-            lfi.lim = lfi_n->lim[filter_level];
-            lfi.hev_thr = lfi_n->hev_thr[hev_index];
-
-            if (mb_col > 0)
-                vp8_loop_filter_mbv
-                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
-
-            if (!skip_lf)
-                vp8_loop_filter_bv
-                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
-
-            /* don't apply across umv border */
-            if (mb_row > 0)
-                vp8_loop_filter_mbh
-                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
-
-            if (!skip_lf)
-                vp8_loop_filter_bh
-                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
-        }
-
-        y_ptr += 16;
-        u_ptr += 8;
-        v_ptr += 8;
-
-        mode_info_context++;     /* step to next MB */
-    }
-
-}
-
-void vp8_loop_filter_row_simple(VP8_COMMON *cm, MODE_INFO *mode_info_context,
-                         int mb_row, int post_ystride, int post_uvstride,
-                         unsigned char *y_ptr, unsigned char *u_ptr,
-                         unsigned char *v_ptr)
-{
-    int mb_col;
-    int filter_level;
-    loop_filter_info_n *lfi_n = &cm->lf_info;
-    (void)post_uvstride;
-
-    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
-    {
-        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                        mode_info_context->mbmi.mode != SPLITMV &&
-                        mode_info_context->mbmi.mb_skip_coeff);
-
-        const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-        const int seg = mode_info_context->mbmi.segment_id;
-        const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-        filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-
-        if (filter_level)
-        {
-            if (mb_col > 0)
-                vp8_loop_filter_simple_mbv
-                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
-
-            if (!skip_lf)
-                vp8_loop_filter_simple_bv
-                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
-
-            /* don't apply across umv border */
-            if (mb_row > 0)
-                vp8_loop_filter_simple_mbh
-                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
-
-            if (!skip_lf)
-                vp8_loop_filter_simple_bh
-                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
-        }
-
-        y_ptr += 16;
-        u_ptr += 8;
-        v_ptr += 8;
-
-        mode_info_context++;     /* step to next MB */
-    }
-
-}
-void vp8_loop_filter_frame(VP8_COMMON *cm,
-                           MACROBLOCKD *mbd,
-                           int frame_type)
-{
-    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
-    loop_filter_info_n *lfi_n = &cm->lf_info;
-    loop_filter_info lfi;
-
-    int mb_row;
-    int mb_col;
-    int mb_rows = cm->mb_rows;
-    int mb_cols = cm->mb_cols;
-
-    int filter_level;
-
-    unsigned char *y_ptr, *u_ptr, *v_ptr;
-
-    /* Point at base of Mb MODE_INFO list */
-    const MODE_INFO *mode_info_context = cm->mi;
-    int post_y_stride = post->y_stride;
-    int post_uv_stride = post->uv_stride;
-
-    /* Initialize the loop filter for this frame. */
-    vp8_loop_filter_frame_init(cm, mbd, cm->filter_level);
-
-    /* Set up the buffer pointers */
-    y_ptr = post->y_buffer;
-    u_ptr = post->u_buffer;
-    v_ptr = post->v_buffer;
-
-    /* vp8_filter each macro block */
-    if (cm->filter_type == NORMAL_LOOPFILTER)
-    {
-        for (mb_row = 0; mb_row < mb_rows; mb_row++)
-        {
-            for (mb_col = 0; mb_col < mb_cols; mb_col++)
-            {
-                int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                                mode_info_context->mbmi.mode != SPLITMV &&
-                                mode_info_context->mbmi.mb_skip_coeff);
-
-                const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-                const int seg = mode_info_context->mbmi.segment_id;
-                const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-                filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-
-                if (filter_level)
-                {
-                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
-                    lfi.mblim = lfi_n->mblim[filter_level];
-                    lfi.blim = lfi_n->blim[filter_level];
-                    lfi.lim = lfi_n->lim[filter_level];
-                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
-
-                    if (mb_col > 0)
-                        vp8_loop_filter_mbv
-                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bv
-                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
-
-                    /* don't apply across umv border */
-                    if (mb_row > 0)
-                        vp8_loop_filter_mbh
-                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bh
-                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
-                }
-
-                y_ptr += 16;
-                u_ptr += 8;
-                v_ptr += 8;
-
-                mode_info_context++;     /* step to next MB */
-            }
-            y_ptr += post_y_stride  * 16 - post->y_width;
-            u_ptr += post_uv_stride *  8 - post->uv_width;
-            v_ptr += post_uv_stride *  8 - post->uv_width;
-
-            mode_info_context++;         /* Skip border mb */
-
-        }
-    }
-    else /* SIMPLE_LOOPFILTER */
-    {
-        for (mb_row = 0; mb_row < mb_rows; mb_row++)
-        {
-            for (mb_col = 0; mb_col < mb_cols; mb_col++)
-            {
-                int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                                mode_info_context->mbmi.mode != SPLITMV &&
-                                mode_info_context->mbmi.mb_skip_coeff);
-
-                const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-                const int seg = mode_info_context->mbmi.segment_id;
-                const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-                filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-                if (filter_level)
-                {
-                    const unsigned char * mblim = lfi_n->mblim[filter_level];
-                    const unsigned char * blim = lfi_n->blim[filter_level];
-
-                    if (mb_col > 0)
-                        vp8_loop_filter_simple_mbv
-                        (y_ptr, post_y_stride, mblim);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bv
-                        (y_ptr, post_y_stride, blim);
-
-                    /* don't apply across umv border */
-                    if (mb_row > 0)
-                        vp8_loop_filter_simple_mbh
-                        (y_ptr, post_y_stride, mblim);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bh
-                        (y_ptr, post_y_stride, blim);
-                }
-
-                y_ptr += 16;
-                u_ptr += 8;
-                v_ptr += 8;
-
-                mode_info_context++;     /* step to next MB */
-            }
-            y_ptr += post_y_stride  * 16 - post->y_width;
-            u_ptr += post_uv_stride *  8 - post->uv_width;
-            v_ptr += post_uv_stride *  8 - post->uv_width;
-
-            mode_info_context++;         /* Skip border mb */
-
-        }
-    }
-}
-
-void vp8_loop_filter_frame_yonly
-(
-    VP8_COMMON *cm,
-    MACROBLOCKD *mbd,
-    int default_filt_lvl
-)
-{
-    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
-
-    unsigned char *y_ptr;
-    int mb_row;
-    int mb_col;
-
-    loop_filter_info_n *lfi_n = &cm->lf_info;
-    loop_filter_info lfi;
-
-    int filter_level;
-    FRAME_TYPE frame_type = cm->frame_type;
-
-    /* Point at base of Mb MODE_INFO list */
-    const MODE_INFO *mode_info_context = cm->mi;
-
-#if 0
-    if(default_filt_lvl == 0) /* no filter applied */
-        return;
-#endif
-
-    /* Initialize the loop filter for this frame. */
-    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
-
-    /* Set up the buffer pointers */
-    y_ptr = post->y_buffer;
-
-    /* vp8_filter each macro block */
-    for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
-    {
-        for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
-        {
-            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                            mode_info_context->mbmi.mode != SPLITMV &&
-                            mode_info_context->mbmi.mb_skip_coeff);
-
-            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-            const int seg = mode_info_context->mbmi.segment_id;
-            const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-
-            if (filter_level)
-            {
-                if (cm->filter_type == NORMAL_LOOPFILTER)
-                {
-                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
-                    lfi.mblim = lfi_n->mblim[filter_level];
-                    lfi.blim = lfi_n->blim[filter_level];
-                    lfi.lim = lfi_n->lim[filter_level];
-                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
-
-                    if (mb_col > 0)
-                        vp8_loop_filter_mbv
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bv
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    /* don't apply across umv border */
-                    if (mb_row > 0)
-                        vp8_loop_filter_mbh
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bh
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-                }
-                else
-                {
-                    if (mb_col > 0)
-                        vp8_loop_filter_simple_mbv
-                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bv
-                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
-
-                    /* don't apply across umv border */
-                    if (mb_row > 0)
-                        vp8_loop_filter_simple_mbh
-                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bh
-                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
-                }
-            }
-
-            y_ptr += 16;
-            mode_info_context ++;        /* step to next MB */
-
-        }
-
-        y_ptr += post->y_stride  * 16 - post->y_width;
-        mode_info_context ++;            /* Skip border mb */
-    }
-
-}
-
-void vp8_loop_filter_partial_frame
-(
-    VP8_COMMON *cm,
-    MACROBLOCKD *mbd,
-    int default_filt_lvl
-)
-{
-    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
-
-    unsigned char *y_ptr;
-    int mb_row;
-    int mb_col;
-    int mb_cols = post->y_width >> 4;
-    int mb_rows = post->y_height >> 4;
-
-    int linestocopy;
-
-    loop_filter_info_n *lfi_n = &cm->lf_info;
-    loop_filter_info lfi;
-
-    int filter_level;
-    FRAME_TYPE frame_type = cm->frame_type;
-
-    const MODE_INFO *mode_info_context;
-
-#if 0
-    if(default_filt_lvl == 0) /* no filter applied */
-        return;
-#endif
-
-    /* Initialize the loop filter for this frame. */
-    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
-
-    /* number of MB rows to use in partial filtering */
-    linestocopy = mb_rows / PARTIAL_FRAME_FRACTION;
-    linestocopy = linestocopy ? linestocopy << 4 : 16;     /* 16 lines per MB */
-
-    /* Set up the buffer pointers; partial image starts at ~middle of frame */
-    y_ptr = post->y_buffer + ((post->y_height >> 5) * 16) * post->y_stride;
-    mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);
-
-    /* vp8_filter each macro block */
-    for (mb_row = 0; mb_row<(linestocopy >> 4); mb_row++)
-    {
-        for (mb_col = 0; mb_col < mb_cols; mb_col++)
-        {
-            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
-                           mode_info_context->mbmi.mode != SPLITMV &&
-                           mode_info_context->mbmi.mb_skip_coeff);
-
-            const int mode_index =
-                lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
-            const int seg = mode_info_context->mbmi.segment_id;
-            const int ref_frame = mode_info_context->mbmi.ref_frame;
-
-            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
-
-            if (filter_level)
-            {
-                if (cm->filter_type == NORMAL_LOOPFILTER)
-                {
-                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
-                    lfi.mblim = lfi_n->mblim[filter_level];
-                    lfi.blim = lfi_n->blim[filter_level];
-                    lfi.lim = lfi_n->lim[filter_level];
-                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
-
-                    if (mb_col > 0)
-                        vp8_loop_filter_mbv
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bv
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    vp8_loop_filter_mbh
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_bh
-                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
-                }
-                else
-                {
-                    if (mb_col > 0)
-                        vp8_loop_filter_simple_mbv
-                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bv
-                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
-
-                    vp8_loop_filter_simple_mbh
-                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
-
-                    if (!skip_lf)
-                        vp8_loop_filter_simple_bh
-                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
-                }
-            }
-
-            y_ptr += 16;
-            mode_info_context += 1;      /* step to next MB */
-        }
-
-        y_ptr += post->y_stride  * 16 - post->y_width;
-        mode_info_context += 1;          /* Skip border mb */
-    }
-}
--- /dev/null
+++ b/vp8/common/vp8_loopfilter.c
@@ -1,0 +1,661 @@
+/*
+ *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+
+#include "vpx_config.h"
+#include "vp8_rtcd.h"
+#include "loopfilter.h"
+#include "onyxc_int.h"
+#include "vpx_mem/vpx_mem.h"
+
+
+static void lf_init_lut(loop_filter_info_n *lfi)
+{
+    int filt_lvl;
+
+    for (filt_lvl = 0; filt_lvl <= MAX_LOOP_FILTER; filt_lvl++)
+    {
+        if (filt_lvl >= 40)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 2;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 3;
+        }
+        else if (filt_lvl >= 20)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 2;
+        }
+        else if (filt_lvl >= 15)
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 1;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 1;
+        }
+        else
+        {
+            lfi->hev_thr_lut[KEY_FRAME][filt_lvl] = 0;
+            lfi->hev_thr_lut[INTER_FRAME][filt_lvl] = 0;
+        }
+    }
+
+    lfi->mode_lf_lut[DC_PRED] = 1;
+    lfi->mode_lf_lut[V_PRED] = 1;
+    lfi->mode_lf_lut[H_PRED] = 1;
+    lfi->mode_lf_lut[TM_PRED] = 1;
+    lfi->mode_lf_lut[B_PRED]  = 0;
+
+    lfi->mode_lf_lut[ZEROMV]  = 1;
+    lfi->mode_lf_lut[NEARESTMV] = 2;
+    lfi->mode_lf_lut[NEARMV] = 2;
+    lfi->mode_lf_lut[NEWMV] = 2;
+    lfi->mode_lf_lut[SPLITMV] = 3;
+
+}
+
+void vp8_loop_filter_update_sharpness(loop_filter_info_n *lfi,
+                                      int sharpness_lvl)
+{
+    int i;
+
+    /* For each possible value for the loop filter fill out limits */
+    for (i = 0; i <= MAX_LOOP_FILTER; i++)
+    {
+        int filt_lvl = i;
+        int block_inside_limit = 0;
+
+        /* Set loop filter paramaeters that control sharpness. */
+        block_inside_limit = filt_lvl >> (sharpness_lvl > 0);
+        block_inside_limit = block_inside_limit >> (sharpness_lvl > 4);
+
+        if (sharpness_lvl > 0)
+        {
+            if (block_inside_limit > (9 - sharpness_lvl))
+                block_inside_limit = (9 - sharpness_lvl);
+        }
+
+        if (block_inside_limit < 1)
+            block_inside_limit = 1;
+
+        memset(lfi->lim[i], block_inside_limit, SIMD_WIDTH);
+        memset(lfi->blim[i], (2 * filt_lvl + block_inside_limit), SIMD_WIDTH);
+        memset(lfi->mblim[i], (2 * (filt_lvl + 2) + block_inside_limit),
+               SIMD_WIDTH);
+    }
+}
+
+void vp8_loop_filter_init(VP8_COMMON *cm)
+{
+    loop_filter_info_n *lfi = &cm->lf_info;
+    int i;
+
+    /* init limits for given sharpness*/
+    vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+    cm->last_sharpness_level = cm->sharpness_level;
+
+    /* init LUT for lvl  and hev thr picking */
+    lf_init_lut(lfi);
+
+    /* init hev threshold const vectors */
+    for(i = 0; i < 4 ; i++)
+    {
+        memset(lfi->hev_thr[i], i, SIMD_WIDTH);
+    }
+}
+
+void vp8_loop_filter_frame_init(VP8_COMMON *cm,
+                                MACROBLOCKD *mbd,
+                                int default_filt_lvl)
+{
+    int seg,  /* segment number */
+        ref,  /* index in ref_lf_deltas */
+        mode; /* index in mode_lf_deltas */
+
+    loop_filter_info_n *lfi = &cm->lf_info;
+
+    /* update limits if sharpness has changed */
+    if(cm->last_sharpness_level != cm->sharpness_level)
+    {
+        vp8_loop_filter_update_sharpness(lfi, cm->sharpness_level);
+        cm->last_sharpness_level = cm->sharpness_level;
+    }
+
+    for(seg = 0; seg < MAX_MB_SEGMENTS; seg++)
+    {
+        int lvl_seg = default_filt_lvl;
+        int lvl_ref, lvl_mode;
+
+        /* Note the baseline filter values for each segment */
+        if (mbd->segmentation_enabled)
+        {
+            /* Abs value */
+            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
+            {
+                lvl_seg = mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+            }
+            else  /* Delta Value */
+            {
+                lvl_seg += mbd->segment_feature_data[MB_LVL_ALT_LF][seg];
+                lvl_seg = (lvl_seg > 0) ? ((lvl_seg > 63) ? 63: lvl_seg) : 0;
+            }
+        }
+
+        if (!mbd->mode_ref_lf_delta_enabled)
+        {
+            /* we could get rid of this if we assume that deltas are set to
+             * zero when not in use; encoder always uses deltas
+             */
+            memset(lfi->lvl[seg][0], lvl_seg, 4 * 4 );
+            continue;
+        }
+
+        /* INTRA_FRAME */
+        ref = INTRA_FRAME;
+
+        /* Apply delta for reference frame */
+        lvl_ref = lvl_seg + mbd->ref_lf_deltas[ref];
+
+        /* Apply delta for Intra modes */
+        mode = 0; /* B_PRED */
+        /* Only the split mode BPRED has a further special case */
+        lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
+        /* clamp */
+        lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0;
+
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        mode = 1; /* all the rest of Intra modes */
+        /* clamp */
+        lvl_mode = (lvl_ref > 0) ? (lvl_ref > 63 ? 63 : lvl_ref) : 0;
+        lfi->lvl[seg][ref][mode] = lvl_mode;
+
+        /* LAST, GOLDEN, ALT */
+        for(ref = 1; ref < MAX_REF_FRAMES; ref++)
+        {
+            /* Apply delta for reference frame */
+            lvl_ref = lvl_seg + mbd->ref_lf_deltas[ref];
+
+            /* Apply delta for Inter modes */
+            for (mode = 1; mode < 4; mode++)
+            {
+                lvl_mode = lvl_ref + mbd->mode_lf_deltas[mode];
+                /* clamp */
+                lvl_mode = (lvl_mode > 0) ? (lvl_mode > 63 ? 63 : lvl_mode) : 0;
+
+                lfi->lvl[seg][ref][mode] = lvl_mode;
+            }
+        }
+    }
+}
+
+
+void vp8_loop_filter_row_normal(VP8_COMMON *cm, MODE_INFO *mode_info_context,
+                         int mb_row, int post_ystride, int post_uvstride,
+                         unsigned char *y_ptr, unsigned char *u_ptr,
+                         unsigned char *v_ptr)
+{
+    int mb_col;
+    int filter_level;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+    FRAME_TYPE frame_type = cm->frame_type;
+
+    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+    {
+        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                        mode_info_context->mbmi.mode != SPLITMV &&
+                        mode_info_context->mbmi.mb_skip_coeff);
+
+        const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+        const int seg = mode_info_context->mbmi.segment_id;
+        const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+        filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+
+        if (filter_level)
+        {
+            const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+            lfi.mblim = lfi_n->mblim[filter_level];
+            lfi.blim = lfi_n->blim[filter_level];
+            lfi.lim = lfi_n->lim[filter_level];
+            lfi.hev_thr = lfi_n->hev_thr[hev_index];
+
+            if (mb_col > 0)
+                vp8_loop_filter_mbv
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            if (!skip_lf)
+                vp8_loop_filter_bv
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            /* don't apply across umv border */
+            if (mb_row > 0)
+                vp8_loop_filter_mbh
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+
+            if (!skip_lf)
+                vp8_loop_filter_bh
+                (y_ptr, u_ptr, v_ptr, post_ystride, post_uvstride, &lfi);
+        }
+
+        y_ptr += 16;
+        u_ptr += 8;
+        v_ptr += 8;
+
+        mode_info_context++;     /* step to next MB */
+    }
+
+}
+
+void vp8_loop_filter_row_simple(VP8_COMMON *cm, MODE_INFO *mode_info_context,
+                         int mb_row, int post_ystride, int post_uvstride,
+                         unsigned char *y_ptr, unsigned char *u_ptr,
+                         unsigned char *v_ptr)
+{
+    int mb_col;
+    int filter_level;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    (void)post_uvstride;
+
+    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+    {
+        int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                        mode_info_context->mbmi.mode != SPLITMV &&
+                        mode_info_context->mbmi.mb_skip_coeff);
+
+        const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+        const int seg = mode_info_context->mbmi.segment_id;
+        const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+        filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+
+        if (filter_level)
+        {
+            if (mb_col > 0)
+                vp8_loop_filter_simple_mbv
+                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
+
+            if (!skip_lf)
+                vp8_loop_filter_simple_bv
+                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
+
+            /* don't apply across umv border */
+            if (mb_row > 0)
+                vp8_loop_filter_simple_mbh
+                (y_ptr, post_ystride, lfi_n->mblim[filter_level]);
+
+            if (!skip_lf)
+                vp8_loop_filter_simple_bh
+                (y_ptr, post_ystride, lfi_n->blim[filter_level]);
+        }
+
+        y_ptr += 16;
+        u_ptr += 8;
+        v_ptr += 8;
+
+        mode_info_context++;     /* step to next MB */
+    }
+
+}
+void vp8_loop_filter_frame(VP8_COMMON *cm,
+                           MACROBLOCKD *mbd,
+                           int frame_type)
+{
+    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
+    int mb_row;
+    int mb_col;
+    int mb_rows = cm->mb_rows;
+    int mb_cols = cm->mb_cols;
+
+    int filter_level;
+
+    unsigned char *y_ptr, *u_ptr, *v_ptr;
+
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
+    int post_y_stride = post->y_stride;
+    int post_uv_stride = post->uv_stride;
+
+    /* Initialize the loop filter for this frame. */
+    vp8_loop_filter_frame_init(cm, mbd, cm->filter_level);
+
+    /* Set up the buffer pointers */
+    y_ptr = post->y_buffer;
+    u_ptr = post->u_buffer;
+    v_ptr = post->v_buffer;
+
+    /* vp8_filter each macro block */
+    if (cm->filter_type == NORMAL_LOOPFILTER)
+    {
+        for (mb_row = 0; mb_row < mb_rows; mb_row++)
+        {
+            for (mb_col = 0; mb_col < mb_cols; mb_col++)
+            {
+                int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                                mode_info_context->mbmi.mode != SPLITMV &&
+                                mode_info_context->mbmi.mb_skip_coeff);
+
+                const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+                const int seg = mode_info_context->mbmi.segment_id;
+                const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+                filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+
+                if (filter_level)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
+
+                    if (mb_col > 0)
+                        vp8_loop_filter_mbv
+                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bv
+                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        vp8_loop_filter_mbh
+                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bh
+                        (y_ptr, u_ptr, v_ptr, post_y_stride, post_uv_stride, &lfi);
+                }
+
+                y_ptr += 16;
+                u_ptr += 8;
+                v_ptr += 8;
+
+                mode_info_context++;     /* step to next MB */
+            }
+            y_ptr += post_y_stride  * 16 - post->y_width;
+            u_ptr += post_uv_stride *  8 - post->uv_width;
+            v_ptr += post_uv_stride *  8 - post->uv_width;
+
+            mode_info_context++;         /* Skip border mb */
+
+        }
+    }
+    else /* SIMPLE_LOOPFILTER */
+    {
+        for (mb_row = 0; mb_row < mb_rows; mb_row++)
+        {
+            for (mb_col = 0; mb_col < mb_cols; mb_col++)
+            {
+                int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                                mode_info_context->mbmi.mode != SPLITMV &&
+                                mode_info_context->mbmi.mb_skip_coeff);
+
+                const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+                const int seg = mode_info_context->mbmi.segment_id;
+                const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+                filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+                if (filter_level)
+                {
+                    const unsigned char * mblim = lfi_n->mblim[filter_level];
+                    const unsigned char * blim = lfi_n->blim[filter_level];
+
+                    if (mb_col > 0)
+                        vp8_loop_filter_simple_mbv
+                        (y_ptr, post_y_stride, mblim);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bv
+                        (y_ptr, post_y_stride, blim);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        vp8_loop_filter_simple_mbh
+                        (y_ptr, post_y_stride, mblim);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bh
+                        (y_ptr, post_y_stride, blim);
+                }
+
+                y_ptr += 16;
+                u_ptr += 8;
+                v_ptr += 8;
+
+                mode_info_context++;     /* step to next MB */
+            }
+            y_ptr += post_y_stride  * 16 - post->y_width;
+            u_ptr += post_uv_stride *  8 - post->uv_width;
+            v_ptr += post_uv_stride *  8 - post->uv_width;
+
+            mode_info_context++;         /* Skip border mb */
+
+        }
+    }
+}
+
+void vp8_loop_filter_frame_yonly
+(
+    VP8_COMMON *cm,
+    MACROBLOCKD *mbd,
+    int default_filt_lvl
+)
+{
+    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
+
+    unsigned char *y_ptr;
+    int mb_row;
+    int mb_col;
+
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
+    int filter_level;
+    FRAME_TYPE frame_type = cm->frame_type;
+
+    /* Point at base of Mb MODE_INFO list */
+    const MODE_INFO *mode_info_context = cm->mi;
+
+#if 0
+    if(default_filt_lvl == 0) /* no filter applied */
+        return;
+#endif
+
+    /* Initialize the loop filter for this frame. */
+    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
+
+    /* Set up the buffer pointers */
+    y_ptr = post->y_buffer;
+
+    /* vp8_filter each macro block */
+    for (mb_row = 0; mb_row < cm->mb_rows; mb_row++)
+    {
+        for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
+        {
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                            mode_info_context->mbmi.mode != SPLITMV &&
+                            mode_info_context->mbmi.mb_skip_coeff);
+
+            const int mode_index = lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+
+            if (filter_level)
+            {
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
+
+                    if (mb_col > 0)
+                        vp8_loop_filter_mbv
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bv
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        vp8_loop_filter_mbh
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bh
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        vp8_loop_filter_simple_mbv
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bv
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    /* don't apply across umv border */
+                    if (mb_row > 0)
+                        vp8_loop_filter_simple_mbh
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bh
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
+            }
+
+            y_ptr += 16;
+            mode_info_context ++;        /* step to next MB */
+
+        }
+
+        y_ptr += post->y_stride  * 16 - post->y_width;
+        mode_info_context ++;            /* Skip border mb */
+    }
+
+}
+
+void vp8_loop_filter_partial_frame
+(
+    VP8_COMMON *cm,
+    MACROBLOCKD *mbd,
+    int default_filt_lvl
+)
+{
+    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
+
+    unsigned char *y_ptr;
+    int mb_row;
+    int mb_col;
+    int mb_cols = post->y_width >> 4;
+    int mb_rows = post->y_height >> 4;
+
+    int linestocopy;
+
+    loop_filter_info_n *lfi_n = &cm->lf_info;
+    loop_filter_info lfi;
+
+    int filter_level;
+    FRAME_TYPE frame_type = cm->frame_type;
+
+    const MODE_INFO *mode_info_context;
+
+#if 0
+    if(default_filt_lvl == 0) /* no filter applied */
+        return;
+#endif
+
+    /* Initialize the loop filter for this frame. */
+    vp8_loop_filter_frame_init( cm, mbd, default_filt_lvl);
+
+    /* number of MB rows to use in partial filtering */
+    linestocopy = mb_rows / PARTIAL_FRAME_FRACTION;
+    linestocopy = linestocopy ? linestocopy << 4 : 16;     /* 16 lines per MB */
+
+    /* Set up the buffer pointers; partial image starts at ~middle of frame */
+    y_ptr = post->y_buffer + ((post->y_height >> 5) * 16) * post->y_stride;
+    mode_info_context = cm->mi + (post->y_height >> 5) * (mb_cols + 1);
+
+    /* vp8_filter each macro block */
+    for (mb_row = 0; mb_row<(linestocopy >> 4); mb_row++)
+    {
+        for (mb_col = 0; mb_col < mb_cols; mb_col++)
+        {
+            int skip_lf = (mode_info_context->mbmi.mode != B_PRED &&
+                           mode_info_context->mbmi.mode != SPLITMV &&
+                           mode_info_context->mbmi.mb_skip_coeff);
+
+            const int mode_index =
+                lfi_n->mode_lf_lut[mode_info_context->mbmi.mode];
+            const int seg = mode_info_context->mbmi.segment_id;
+            const int ref_frame = mode_info_context->mbmi.ref_frame;
+
+            filter_level = lfi_n->lvl[seg][ref_frame][mode_index];
+
+            if (filter_level)
+            {
+                if (cm->filter_type == NORMAL_LOOPFILTER)
+                {
+                    const int hev_index = lfi_n->hev_thr_lut[frame_type][filter_level];
+                    lfi.mblim = lfi_n->mblim[filter_level];
+                    lfi.blim = lfi_n->blim[filter_level];
+                    lfi.lim = lfi_n->lim[filter_level];
+                    lfi.hev_thr = lfi_n->hev_thr[hev_index];
+
+                    if (mb_col > 0)
+                        vp8_loop_filter_mbv
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bv
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    vp8_loop_filter_mbh
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_bh
+                        (y_ptr, 0, 0, post->y_stride, 0, &lfi);
+                }
+                else
+                {
+                    if (mb_col > 0)
+                        vp8_loop_filter_simple_mbv
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bv
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+
+                    vp8_loop_filter_simple_mbh
+                        (y_ptr, post->y_stride, lfi_n->mblim[filter_level]);
+
+                    if (!skip_lf)
+                        vp8_loop_filter_simple_bh
+                        (y_ptr, post->y_stride, lfi_n->blim[filter_level]);
+                }
+            }
+
+            y_ptr += 16;
+            mode_info_context += 1;      /* step to next MB */
+        }
+
+        y_ptr += post->y_stride  * 16 - post->y_width;
+        mode_info_context += 1;          /* Skip border mb */
+    }
+}
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -53,7 +53,7 @@
 VP8_COMMON_SRCS-yes += common/systemdependent.h
 VP8_COMMON_SRCS-yes += common/threading.h
 VP8_COMMON_SRCS-yes += common/treecoder.h
-VP8_COMMON_SRCS-yes += common/loopfilter.c
+VP8_COMMON_SRCS-yes += common/vp8_loopfilter.c
 VP8_COMMON_SRCS-yes += common/loopfilter_filters.c
 VP8_COMMON_SRCS-yes += common/mbpitch.c
 VP8_COMMON_SRCS-yes += common/modecont.c
@@ -161,7 +161,7 @@
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/idct_dequant_0_2x_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/idct_dequant_full_2x_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/iwalsh_neon.c
-VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/loopfilter_neon.c
+VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/vp8_loopfilter_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/loopfiltersimplehorizontaledge_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/loopfiltersimpleverticaledge_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/mbloopfilter_neon.c