Revert "Revert "VP8 for ARMv8 by using NEON intrinsics 10""

This reverts commit 677fb5123e0ece1d6c30be9d0282e1e1f594a56d

Compiles with 4.6.

Change-Id: I7f87048911b6bc28a61741d95501fa45ee97b819

4 files changed, 284 insertions, 163 deletions

diff --git a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
deleted file mode 100644
index 78d13c895..000000000
--- a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.asm
+++ /dev/null
@@ -1,156 +0,0 @@
-;
-;  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.
-;
-
-
-    EXPORT |vp8_loop_filter_bvs_neon|
-    EXPORT |vp8_loop_filter_mbvs_neon|
-    ARM
-    PRESERVE8
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-
-; r0    unsigned char *s, PRESERVE
-; r1    int p, PRESERVE
-; q1    limit, PRESERVE
-
-|vp8_loop_filter_simple_vertical_edge_neon| PROC
-    vpush       {d8-d15}
-
-    sub         r0, r0, #2                  ; move src pointer down by 2 columns
-    add         r12, r1, r1
-    add         r3, r0, r1
-
-    vld4.8      {d6[0], d7[0], d8[0], d9[0]}, [r0], r12
-    vld4.8      {d6[1], d7[1], d8[1], d9[1]}, [r3], r12
-    vld4.8      {d6[2], d7[2], d8[2], d9[2]}, [r0], r12
-    vld4.8      {d6[3], d7[3], d8[3], d9[3]}, [r3], r12
-    vld4.8      {d6[4], d7[4], d8[4], d9[4]}, [r0], r12
-    vld4.8      {d6[5], d7[5], d8[5], d9[5]}, [r3], r12
-    vld4.8      {d6[6], d7[6], d8[6], d9[6]}, [r0], r12
-    vld4.8      {d6[7], d7[7], d8[7], d9[7]}, [r3], r12
-
-    vld4.8      {d10[0], d11[0], d12[0], d13[0]}, [r0], r12
-    vld4.8      {d10[1], d11[1], d12[1], d13[1]}, [r3], r12
-    vld4.8      {d10[2], d11[2], d12[2], d13[2]}, [r0], r12
-    vld4.8      {d10[3], d11[3], d12[3], d13[3]}, [r3], r12
-    vld4.8      {d10[4], d11[4], d12[4], d13[4]}, [r0], r12
-    vld4.8      {d10[5], d11[5], d12[5], d13[5]}, [r3], r12
-    vld4.8      {d10[6], d11[6], d12[6], d13[6]}, [r0], r12
-    vld4.8      {d10[7], d11[7], d12[7], d13[7]}, [r3]
-
-    vswp        d7, d10
-    vswp        d12, d9
-
-    ;vp8_filter_mask() function
-    ;vp8_hevmask() function
-    sub         r0, r0, r1, lsl #4
-    vabd.u8     q15, q5, q4                 ; abs(p0 - q0)
-    vabd.u8     q14, q3, q6                 ; abs(p1 - q1)
-
-    vqadd.u8    q15, q15, q15               ; abs(p0 - q0) * 2
-    vshr.u8     q14, q14, #1                ; abs(p1 - q1) / 2
-    vmov.u8     q0, #0x80                   ; 0x80
-    vmov.s16    q11, #3
-    vqadd.u8    q15, q15, q14               ; abs(p0 - q0) * 2 + abs(p1 - q1) / 2
-
-    veor        q4, q4, q0                  ; qs0: q0 offset to convert to a signed value
-    veor        q5, q5, q0                  ; ps0: p0 offset to convert to a signed value
-    veor        q3, q3, q0                  ; ps1: p1 offset to convert to a signed value
-    veor        q6, q6, q0                  ; qs1: q1 offset to convert to a signed value
-
-    vcge.u8     q15, q1, q15                ; abs(p0 - q0)*2 + abs(p1-q1)/2 > flimit*2 + limit)*-1
-
-    vsubl.s8    q2, d8, d10                 ; ( qs0 - ps0)
-    vsubl.s8    q13, d9, d11
-
-    vqsub.s8    q14, q3, q6                  ; vp8_filter = vp8_signed_char_clamp(ps1-qs1)
-
-    vmul.s16    q2, q2, q11                 ;  3 * ( qs0 - ps0)
-    vmul.s16    q13, q13, q11
-
-    vmov.u8     q11, #0x03                  ; 0x03
-    vmov.u8     q12, #0x04                  ; 0x04
-
-    vaddw.s8    q2, q2, d28                  ; vp8_filter + 3 * ( qs0 - ps0)
-    vaddw.s8    q13, q13, d29
-
-    vqmovn.s16  d28, q2                      ; vp8_filter = vp8_signed_char_clamp(vp8_filter + 3 * ( qs0 - ps0))
-    vqmovn.s16  d29, q13
-
-    add         r0, r0, #1
-    add         r3, r0, r1
-
-    vand        q14, q14, q15                 ; vp8_filter &= mask
-
-    vqadd.s8    q2, q14, q11                 ; Filter2 = vp8_signed_char_clamp(vp8_filter+3)
-    vqadd.s8    q3, q14, q12                 ; Filter1 = vp8_signed_char_clamp(vp8_filter+4)
-    vshr.s8     q2, q2, #3                  ; Filter2 >>= 3
-    vshr.s8     q14, q3, #3                  ; Filter1 >>= 3
-
-    ;calculate output
-    vqadd.s8    q11, q5, q2                 ; u = vp8_signed_char_clamp(ps0 + Filter2)
-    vqsub.s8    q10, q4, q14                 ; u = vp8_signed_char_clamp(qs0 - Filter1)
-
-    veor        q6, q11, q0                 ; *op0 = u^0x80
-    veor        q7, q10, q0                 ; *oq0 = u^0x80
-    add         r12, r1, r1
-    vswp        d13, d14
-
-    ;store op1, op0, oq0, oq1
-    vst2.8      {d12[0], d13[0]}, [r0], r12
-    vst2.8      {d12[1], d13[1]}, [r3], r12
-    vst2.8      {d12[2], d13[2]}, [r0], r12
-    vst2.8      {d12[3], d13[3]}, [r3], r12
-    vst2.8      {d12[4], d13[4]}, [r0], r12
-    vst2.8      {d12[5], d13[5]}, [r3], r12
-    vst2.8      {d12[6], d13[6]}, [r0], r12
-    vst2.8      {d12[7], d13[7]}, [r3], r12
-    vst2.8      {d14[0], d15[0]}, [r0], r12
-    vst2.8      {d14[1], d15[1]}, [r3], r12
-    vst2.8      {d14[2], d15[2]}, [r0], r12
-    vst2.8      {d14[3], d15[3]}, [r3], r12
-    vst2.8      {d14[4], d15[4]}, [r0], r12
-    vst2.8      {d14[5], d15[5]}, [r3], r12
-    vst2.8      {d14[6], d15[6]}, [r0], r12
-    vst2.8      {d14[7], d15[7]}, [r3]
-
-    vpop        {d8-d15}
-    bx          lr
-    ENDP        ; |vp8_loop_filter_simple_vertical_edge_neon|
-
-; r0    unsigned char *y
-; r1    int ystride
-; r2    const unsigned char *blimit
-
-|vp8_loop_filter_bvs_neon| PROC
-    push        {r4, lr}
-    ldrb        r3, [r2]                   ; load blim from mem
-    mov         r4, r0
-    add         r0, r0, #4
-    vdup.s8     q1, r3                     ; duplicate blim
-    bl          vp8_loop_filter_simple_vertical_edge_neon
-    ; vp8_loop_filter_simple_vertical_edge_neon preserves  r1 and q1
-    add         r0, r4, #8
-    bl          vp8_loop_filter_simple_vertical_edge_neon
-    add         r0, r4, #12
-    pop         {r4, lr}
-    b           vp8_loop_filter_simple_vertical_edge_neon
-    ENDP        ;|vp8_loop_filter_bvs_neon|
-
-; r0    unsigned char *y
-; r1    int ystride
-; r2    const unsigned char *blimit
-
-|vp8_loop_filter_mbvs_neon| PROC
-    ldrb        r3, [r2]                   ; load mblim from mem
-    vdup.s8     q1, r3                     ; duplicate mblim
-    b           vp8_loop_filter_simple_vertical_edge_neon
-    ENDP        ;|vp8_loop_filter_bvs_neon|
-    END
diff --git a/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
new file mode 100644
index 000000000..d5178bbae
--- /dev/null
+++ b/vp8/common/arm/neon/loopfiltersimpleverticaledge_neon.c
@@ -0,0 +1,279 @@
+/*
+ *  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"
+
+#if (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
+static INLINE void write_2x8(unsigned char *dst, int pitch,
+                             const uint8x8x2_t result,
+                             const uint8x8x2_t result2) {
+  vst2_lane_u8(dst, result, 0);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 1);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 2);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 3);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 4);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 5);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 6);
+  dst += pitch;
+  vst2_lane_u8(dst, result, 7);
+  dst += pitch;
+
+  vst2_lane_u8(dst, result2, 0);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 1);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 2);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 3);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 4);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 5);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 6);
+  dst += pitch;
+  vst2_lane_u8(dst, result2, 7);
+}
+#else
+static INLINE void write_2x4(unsigned char *dst, int pitch,
+                             const uint8x8x2_t result) {
+    /*
+     * uint8x8x2_t result
+    00 01 02 03 | 04 05 06 07
+    10 11 12 13 | 14 15 16 17
+    ---
+    * after vtrn_u8
+    00 10 02 12 | 04 14 06 16
+    01 11 03 13 | 05 15 07 17
+    */
+    const uint8x8x2_t r01_u8 = vtrn_u8(result.val[0],
+                                       result.val[1]);
+    const uint16x4_t x_0_4 = vreinterpret_u16_u8(r01_u8.val[0]);
+    const uint16x4_t x_1_5 = vreinterpret_u16_u8(r01_u8.val[1]);
+    vst1_lane_u16((uint16_t *)dst, x_0_4, 0);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_1_5, 0);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_0_4, 1);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_1_5, 1);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_0_4, 2);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_1_5, 2);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_0_4, 3);
+    dst += pitch;
+    vst1_lane_u16((uint16_t *)dst, x_1_5, 3);
+}
+
+static INLINE void write_2x8(unsigned char *dst, int pitch,
+                             const uint8x8x2_t result,
+                             const uint8x8x2_t result2) {
+  write_2x4(dst, pitch, result);
+  dst += pitch * 8;
+  write_2x4(dst, pitch, result2);
+}
+#endif
+
+
+#if (__GNUC__ == 4 && (__GNUC_MINOR__ >= 7))
+static INLINE
+uint8x8x4_t read_4x8(unsigned char *src, int pitch, uint8x8x4_t x) {
+    x = vld4_lane_u8(src, x, 0);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 1);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 2);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 3);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 4);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 5);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 6);
+    src += pitch;
+    x = vld4_lane_u8(src, x, 7);
+    return x;
+}
+#else
+static INLINE
+uint8x8x4_t read_4x8(unsigned char *src, int pitch, uint8x8x4_t x) {
+    const uint8x8_t a = vld1_u8(src);
+    const uint8x8_t b = vld1_u8(src + pitch * 1);
+    const uint8x8_t c = vld1_u8(src + pitch * 2);
+    const uint8x8_t d = vld1_u8(src + pitch * 3);
+    const uint8x8_t e = vld1_u8(src + pitch * 4);
+    const uint8x8_t f = vld1_u8(src + pitch * 5);
+    const uint8x8_t g = vld1_u8(src + pitch * 6);
+    const uint8x8_t h = vld1_u8(src + pitch * 7);
+    const uint32x2x2_t r04_u32 = vtrn_u32(vreinterpret_u32_u8(a),
+                                          vreinterpret_u32_u8(e));
+    const uint32x2x2_t r15_u32 = vtrn_u32(vreinterpret_u32_u8(b),
+                                          vreinterpret_u32_u8(f));
+    const uint32x2x2_t r26_u32 = vtrn_u32(vreinterpret_u32_u8(c),
+                                          vreinterpret_u32_u8(g));
+    const uint32x2x2_t r37_u32 = vtrn_u32(vreinterpret_u32_u8(d),
+                                          vreinterpret_u32_u8(h));
+    const uint16x4x2_t r02_u16 = vtrn_u16(vreinterpret_u16_u32(r04_u32.val[0]),
+                                          vreinterpret_u16_u32(r26_u32.val[0]));
+    const uint16x4x2_t r13_u16 = vtrn_u16(vreinterpret_u16_u32(r15_u32.val[0]),
+                                          vreinterpret_u16_u32(r37_u32.val[0]));
+    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]));
+    /*
+     * after vtrn_u32
+    00 01 02 03 | 40 41 42 43
+    10 11 12 13 | 50 51 52 53
+    20 21 22 23 | 60 61 62 63
+    30 31 32 33 | 70 71 72 73
+    ---
+    * after vtrn_u16
+    00 01 20 21 | 40 41 60 61
+    02 03 22 23 | 42 43 62 63
+    10 11 30 31 | 50 51 70 71
+    12 13 32 33 | 52 52 72 73
+
+    00 01 20 21 | 40 41 60 61
+    10 11 30 31 | 50 51 70 71
+    02 03 22 23 | 42 43 62 63
+    12 13 32 33 | 52 52 72 73
+    ---
+    * after vtrn_u8
+    00 10 20 30 | 40 50 60 70
+    01 11 21 31 | 41 51 61 71
+    02 12 22 32 | 42 52 62 72
+    03 13 23 33 | 43 53 63 73
+    */
+    x.val[0] = r01_u8.val[0];
+    x.val[1] = r01_u8.val[1];
+    x.val[2] = r23_u8.val[0];
+    x.val[3] = r23_u8.val[1];
+
+    return x;
+}
+#endif
+
+static INLINE void vp8_loop_filter_simple_vertical_edge_neon(
+        unsigned char *s,
+        int p,
+        const unsigned char *blimit) {
+    unsigned char *src1;
+    uint8x16_t qblimit, q0u8;
+    uint8x16_t q3u8, q4u8, q5u8, q6u8, q7u8, q11u8, q12u8, q14u8, q15u8;
+    int16x8_t q2s16, q13s16, q11s16;
+    int8x8_t d28s8, d29s8;
+    int8x16_t q2s8, q3s8, q10s8, q11s8, q14s8;
+    uint8x8x4_t d0u8x4;  // d6, d7, d8, d9
+    uint8x8x4_t d1u8x4;  // d10, d11, d12, d13
+    uint8x8x2_t d2u8x2;  // d12, d13
+    uint8x8x2_t d3u8x2;  // d14, d15
+
+    qblimit = vdupq_n_u8(*blimit);
+
+    src1 = s - 2;
+    d0u8x4 = read_4x8(src1, p, d0u8x4);
+    src1 += p * 8;
+    d1u8x4 = read_4x8(src1, p, d1u8x4);
+
+    q3u8 = vcombine_u8(d0u8x4.val[0], d1u8x4.val[0]);  // d6 d10
+    q4u8 = vcombine_u8(d0u8x4.val[2], d1u8x4.val[2]);  // d8 d12
+    q5u8 = vcombine_u8(d0u8x4.val[1], d1u8x4.val[1]);  // d7 d11
+    q6u8 = vcombine_u8(d0u8x4.val[3], d1u8x4.val[3]);  // d9 d13
+
+    q15u8 = vabdq_u8(q5u8, q4u8);
+    q14u8 = vabdq_u8(q3u8, q6u8);
+
+    q15u8 = vqaddq_u8(q15u8, q15u8);
+    q14u8 = vshrq_n_u8(q14u8, 1);
+    q0u8 = vdupq_n_u8(0x80);
+    q11s16 = vdupq_n_s16(3);
+    q15u8 = vqaddq_u8(q15u8, q14u8);
+
+    q3u8 = veorq_u8(q3u8, q0u8);
+    q4u8 = veorq_u8(q4u8, q0u8);
+    q5u8 = veorq_u8(q5u8, q0u8);
+    q6u8 = veorq_u8(q6u8, q0u8);
+
+    q15u8 = vcgeq_u8(qblimit, q15u8);
+
+    q2s16 = vsubl_s8(vget_low_s8(vreinterpretq_s8_u8(q4u8)),
+                     vget_low_s8(vreinterpretq_s8_u8(q5u8)));
+    q13s16 = vsubl_s8(vget_high_s8(vreinterpretq_s8_u8(q4u8)),
+                      vget_high_s8(vreinterpretq_s8_u8(q5u8)));
+
+    q14s8 = vqsubq_s8(vreinterpretq_s8_u8(q3u8),
+                      vreinterpretq_s8_u8(q6u8));
+
+    q2s16 = vmulq_s16(q2s16, q11s16);
+    q13s16 = vmulq_s16(q13s16, q11s16);
+
+    q11u8 = vdupq_n_u8(3);
+    q12u8 = vdupq_n_u8(4);
+
+    q2s16 = vaddw_s8(q2s16, vget_low_s8(q14s8));
+    q13s16 = vaddw_s8(q13s16, vget_high_s8(q14s8));
+
+    d28s8 = vqmovn_s16(q2s16);
+    d29s8 = vqmovn_s16(q13s16);
+    q14s8 = vcombine_s8(d28s8, d29s8);
+
+    q14s8 = vandq_s8(q14s8, vreinterpretq_s8_u8(q15u8));
+
+    q2s8 = vqaddq_s8(q14s8, vreinterpretq_s8_u8(q11u8));
+    q3s8 = vqaddq_s8(q14s8, vreinterpretq_s8_u8(q12u8));
+    q2s8 = vshrq_n_s8(q2s8, 3);
+    q14s8 = vshrq_n_s8(q3s8, 3);
+
+    q11s8 = vqaddq_s8(vreinterpretq_s8_u8(q5u8), q2s8);
+    q10s8 = vqsubq_s8(vreinterpretq_s8_u8(q4u8), q14s8);
+
+    q6u8 = veorq_u8(vreinterpretq_u8_s8(q11s8), q0u8);
+    q7u8 = veorq_u8(vreinterpretq_u8_s8(q10s8), q0u8);
+
+    d2u8x2.val[0] = vget_low_u8(q6u8);   // d12
+    d2u8x2.val[1] = vget_low_u8(q7u8);   // d14
+    d3u8x2.val[0] = vget_high_u8(q6u8);  // d13
+    d3u8x2.val[1] = vget_high_u8(q7u8);  // d15
+
+    src1 = s - 1;
+    write_2x8(src1, p, d2u8x2, d3u8x2);
+}
+
+void vp8_loop_filter_bvs_neon(
+        unsigned char *y_ptr,
+        int y_stride,
+        const unsigned char *blimit) {
+    y_ptr += 4;
+    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, blimit);
+    y_ptr += 4;
+    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, blimit);
+    y_ptr += 4;
+    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, blimit);
+    return;
+}
+
+void vp8_loop_filter_mbvs_neon(
+        unsigned char *y_ptr,
+        int y_stride,
+        const unsigned char *blimit) {
+    vp8_loop_filter_simple_vertical_edge_neon(y_ptr, y_stride, blimit);
+    return;
+}
diff --git a/vp8/common/rtcd_defs.pl b/vp8/common/rtcd_defs.pl
index ea260b961..651809dc5 100644
--- a/vp8/common/rtcd_defs.pl
+++ b/vp8/common/rtcd_defs.pl
@@ -72,12 +72,12 @@ $vp8_loop_filter_bh_dspr2=vp8_loop_filter_bh_dspr2;
 
 
 add_proto qw/void vp8_loop_filter_simple_mbv/, "unsigned char *y, int ystride, const unsigned char *blimit";
-specialize qw/vp8_loop_filter_simple_mbv mmx sse2 media neon_asm/;
+specialize qw/vp8_loop_filter_simple_mbv mmx sse2 media neon/;
 $vp8_loop_filter_simple_mbv_c=vp8_loop_filter_simple_vertical_edge_c;
 $vp8_loop_filter_simple_mbv_mmx=vp8_loop_filter_simple_vertical_edge_mmx;
 $vp8_loop_filter_simple_mbv_sse2=vp8_loop_filter_simple_vertical_edge_sse2;
 $vp8_loop_filter_simple_mbv_media=vp8_loop_filter_simple_vertical_edge_armv6;
-$vp8_loop_filter_simple_mbv_neon_asm=vp8_loop_filter_mbvs_neon;
+$vp8_loop_filter_simple_mbv_neon=vp8_loop_filter_mbvs_neon;
 
 add_proto qw/void vp8_loop_filter_simple_mbh/, "unsigned char *y, int ystride, const unsigned char *blimit";
 specialize qw/vp8_loop_filter_simple_mbh mmx sse2 media neon/;
@@ -88,12 +88,12 @@ $vp8_loop_filter_simple_mbh_media=vp8_loop_filter_simple_horizontal_edge_armv6;
 $vp8_loop_filter_simple_mbh_neon=vp8_loop_filter_mbhs_neon;
 
 add_proto qw/void vp8_loop_filter_simple_bv/, "unsigned char *y, int ystride, const unsigned char *blimit";
-specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon_asm/;
+specialize qw/vp8_loop_filter_simple_bv mmx sse2 media neon/;
 $vp8_loop_filter_simple_bv_c=vp8_loop_filter_bvs_c;
 $vp8_loop_filter_simple_bv_mmx=vp8_loop_filter_bvs_mmx;
 $vp8_loop_filter_simple_bv_sse2=vp8_loop_filter_bvs_sse2;
 $vp8_loop_filter_simple_bv_media=vp8_loop_filter_bvs_armv6;
-$vp8_loop_filter_simple_bv_neon_asm=vp8_loop_filter_bvs_neon;
+$vp8_loop_filter_simple_bv_neon=vp8_loop_filter_bvs_neon;
 
 add_proto qw/void vp8_loop_filter_simple_bh/, "unsigned char *y, int ystride, const unsigned char *blimit";
 specialize qw/vp8_loop_filter_simple_bh mmx sse2 media neon/;
diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index 5bb1f2713..9b11c0da3 100644
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -155,9 +155,6 @@ VP8_COMMON_SRCS-$(HAVE_MEDIA)  += common/arm/armv6/vp8_variance_halfpixvar16x16_
 VP8_COMMON_SRCS-$(HAVE_MEDIA)  += common/arm/armv6/vp8_variance_halfpixvar16x16_v_armv6$(ASM)
 VP8_COMMON_SRCS-$(HAVE_MEDIA)  += common/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
 
-# common (neon)
-VP8_COMMON_SRCS-$(HAVE_NEON_ASM)  += common/arm/neon/loopfiltersimpleverticaledge_neon$(ASM)
-
 # common (neon intrinsics)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/bilinearpredict_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem_neon.c
@@ -170,6 +167,7 @@ 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/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
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/reconintra_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/sad_neon.c