vp8: arm: separate ARMv6 functions from NEON

This is a preparation for complete ARMv6 optimisations.

Signed-off-by: Mans Rullgard <mans@mansr.com>

7 files changed, 289 insertions, 170 deletions

diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index e6d92189de..3d002ccdcf 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -13,7 +13,9 @@ ARMV6-OBJS-$(CONFIG_MPEGAUDIODSP)      += arm/mpegaudiodsp_fixed_armv6.o
 OBJS-$(CONFIG_VP5_DECODER)             += arm/vp56dsp_init_arm.o
 OBJS-$(CONFIG_VP6_DECODER)             += arm/vp56dsp_init_arm.o
 OBJS-$(CONFIG_VP8_DECODER)             += arm/vp8dsp_init_arm.o
-ARMV6-OBJS-$(CONFIG_VP8_DECODER)       += arm/vp8_armv6.o
+ARMV6-OBJS-$(CONFIG_VP8_DECODER)       += arm/vp8_armv6.o               \
+                                          arm/vp8dsp_init_armv6.o       \
+                                          arm/vp8dsp_armv6.o
 
 OBJS-$(CONFIG_H264DSP)                 += arm/h264dsp_init_arm.o
 OBJS-$(CONFIG_H264PRED)                += arm/h264pred_init_arm.o
@@ -79,7 +81,8 @@ NEON-OBJS-$(CONFIG_VP5_DECODER)        += arm/vp56dsp_neon.o            \
 NEON-OBJS-$(CONFIG_VP6_DECODER)        += arm/vp56dsp_neon.o            \
                                           arm/vp3dsp_neon.o             \
 
-NEON-OBJS-$(CONFIG_VP8_DECODER)        += arm/vp8dsp_neon.o
+NEON-OBJS-$(CONFIG_VP8_DECODER)        += arm/vp8dsp_init_neon.o        \
+                                          arm/vp8dsp_neon.o
 
 NEON-OBJS                              += arm/dsputil_init_neon.o       \
                                           arm/dsputil_neon.o            \
diff --git a/libavcodec/arm/vp8dsp.h b/libavcodec/arm/vp8dsp.h
new file mode 100644
index 0000000000..ce00e4abf5
--- /dev/null
+++ b/libavcodec/arm/vp8dsp.h
@@ -0,0 +1,78 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_ARM_VP8DSP_H
+#define AVCODEC_ARM_VP8DSP_H
+
+#include "libavcodec/vp8dsp.h"
+
+void ff_vp8dsp_init_armv6(VP8DSPContext *dsp);
+void ff_vp8dsp_init_neon(VP8DSPContext *dsp);
+
+#define VP8_LF_Y(hv, inner, opt)                                             \
+    void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst,            \
+                                                    ptrdiff_t stride,        \
+                                                    int flim_E, int flim_I,  \
+                                                    int hev_thresh)
+
+#define VP8_LF_UV(hv, inner, opt)                                            \
+    void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU,          \
+                                                     uint8_t *dstV,          \
+                                                     ptrdiff_t stride,       \
+                                                     int flim_E, int flim_I, \
+                                                     int hev_thresh)
+
+#define VP8_LF_SIMPLE(hv, opt)                                          \
+    void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst,         \
+                                                  ptrdiff_t stride,     \
+                                                  int flim)
+
+#define VP8_LF_HV(inner, opt)                   \
+    VP8_LF_Y(h,  inner, opt);                   \
+    VP8_LF_Y(v,  inner, opt);                   \
+    VP8_LF_UV(h, inner, opt);                   \
+    VP8_LF_UV(v, inner, opt)
+
+#define VP8_LF(opt)                             \
+    VP8_LF_HV(,       opt);                     \
+    VP8_LF_HV(_inner, opt);                     \
+    VP8_LF_SIMPLE(h, opt);                      \
+    VP8_LF_SIMPLE(v, opt)
+
+#define VP8_MC(n, opt)                                                  \
+    void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride,      \
+                                uint8_t *src, ptrdiff_t srcstride,      \
+                                int h, int x, int y)
+
+#define VP8_EPEL(w, opt)                        \
+    VP8_MC(pixels ## w, opt);                   \
+    VP8_MC(epel ## w ## _h4, opt);              \
+    VP8_MC(epel ## w ## _h6, opt);              \
+    VP8_MC(epel ## w ## _v4, opt);              \
+    VP8_MC(epel ## w ## _h4v4, opt);            \
+    VP8_MC(epel ## w ## _h6v4, opt);            \
+    VP8_MC(epel ## w ## _v6, opt);              \
+    VP8_MC(epel ## w ## _h4v6, opt);            \
+    VP8_MC(epel ## w ## _h6v6, opt)
+
+#define VP8_BILIN(w, opt)                       \
+    VP8_MC(bilin ## w ## _h, opt);              \
+    VP8_MC(bilin ## w ## _v, opt);              \
+    VP8_MC(bilin ## w ## _hv, opt)
+
+#endif /* AVCODEC_ARM_VP8DSP_H */
diff --git a/libavcodec/arm/vp8dsp_armv6.S b/libavcodec/arm/vp8dsp_armv6.S
new file mode 100644
index 0000000000..1ed020c069
--- /dev/null
+++ b/libavcodec/arm/vp8dsp_armv6.S
@@ -0,0 +1,53 @@
+/*
+ * VP8 ARMv6 optimisations
+ *
+ * Copyright (c) 2010 Rob Clark <rob@ti.com>
+ * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
+ *
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "asm.S"
+
+function ff_vp8_luma_dc_wht_dc_armv6, export=1
+        ldrsh           r2,  [r1]
+        mov             r3,  #0
+        add             r2,  r2,  #3
+        strh            r3,  [r1]
+        asr             r2,  r2,  #3
+    .rept 16
+        strh            r2,  [r0], #32
+    .endr
+        bx              lr
+endfunc
+
+function ff_put_vp8_pixels4_armv6, export=1
+        ldr             r12, [sp, #0]           @ h
+        push            {r4-r6,lr}
+1:
+        subs            r12, r12, #4
+        ldr_post        r4,  r2,  r3
+        ldr_post        r5,  r2,  r3
+        ldr_post        r6,  r2,  r3
+        ldr_post        lr,  r2,  r3
+        str_post        r4,  r0,  r1
+        str_post        r5,  r0,  r1
+        str_post        r6,  r0,  r1
+        str_post        lr,  r0,  r1
+        bgt             1b
+        pop             {r4-r6,pc}
+endfunc
diff --git a/libavcodec/arm/vp8dsp_init_arm.c b/libavcodec/arm/vp8dsp_init_arm.c
index 29c64d4079..603f68cd24 100644
--- a/libavcodec/arm/vp8dsp_init_arm.c
+++ b/libavcodec/arm/vp8dsp_init_arm.c
@@ -20,148 +20,14 @@
 
 #include "libavutil/arm/cpu.h"
 #include "libavcodec/vp8dsp.h"
-
-void ff_vp8_luma_dc_wht_neon(DCTELEM block[4][4][16], DCTELEM dc[16]);
-void ff_vp8_luma_dc_wht_dc_neon(DCTELEM block[4][4][16], DCTELEM dc[16]);
-
-void ff_vp8_idct_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride);
-void ff_vp8_idct_dc_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride);
-void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride);
-void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride);
-
-void ff_vp8_v_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride,
-                                 int flim_E, int flim_I, int hev_thresh);
-void ff_vp8_h_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride,
-                                 int flim_E, int flim_I, int hev_thresh);
-void ff_vp8_v_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
-                                  int flim_E, int flim_I, int hev_thresh);
-void ff_vp8_h_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,
-                                  int flim_E, int flim_I, int hev_thresh);
-
-void ff_vp8_v_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride,
-                                       int flim_E, int flim_I, int hev_thresh);
-void ff_vp8_h_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride,
-                                       int flim_E, int flim_I, int hev_thresh);
-void ff_vp8_v_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV,
-                                        ptrdiff_t stride, int flim_E, int flim_I,
-                                        int hev_thresh);
-void ff_vp8_h_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV,
-                                        ptrdiff_t stride, int flim_E, int flim_I,
-                                        int hev_thresh);
-
-void ff_vp8_v_loop_filter16_simple_neon(uint8_t *dst, ptrdiff_t stride, int flim);
-void ff_vp8_h_loop_filter16_simple_neon(uint8_t *dst, ptrdiff_t stride, int flim);
-
-
-#define VP8_MC(n)                                                       \
-    void ff_put_vp8_##n##_neon(uint8_t *dst, ptrdiff_t dststride,       \
-                               uint8_t *src, ptrdiff_t srcstride,       \
-                               int h, int x, int y)
-
-#define VP8_EPEL(w)                             \
-    VP8_MC(pixels ## w);                        \
-    VP8_MC(epel ## w ## _h4);                   \
-    VP8_MC(epel ## w ## _h6);                   \
-    VP8_MC(epel ## w ## _v4);                   \
-    VP8_MC(epel ## w ## _h4v4);                 \
-    VP8_MC(epel ## w ## _h6v4);                 \
-    VP8_MC(epel ## w ## _v6);                   \
-    VP8_MC(epel ## w ## _h4v6);                 \
-    VP8_MC(epel ## w ## _h6v6)
-
-VP8_EPEL(16);
-VP8_EPEL(8);
-VP8_EPEL(4);
-
-VP8_MC(bilin16_h);
-VP8_MC(bilin16_v);
-VP8_MC(bilin16_hv);
-VP8_MC(bilin8_h);
-VP8_MC(bilin8_v);
-VP8_MC(bilin8_hv);
-VP8_MC(bilin4_h);
-VP8_MC(bilin4_v);
-VP8_MC(bilin4_hv);
+#include "vp8dsp.h"
 
 av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp)
 {
     int cpu_flags = av_get_cpu_flags();
 
-    if (have_neon(cpu_flags)) {
-        dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_neon;
-        dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_neon;
-
-        dsp->vp8_idct_add       = ff_vp8_idct_add_neon;
-        dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_neon;
-        dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_neon;
-        dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon;
-
-        dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon;
-        dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon;
-        dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon;
-        dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon;
-
-        dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon;
-        dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon;
-        dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon;
-        dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon;
-
-        dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon;
-        dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon;
-
-        dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
-        dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon;
-        dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon;
-        dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon;
-
-        dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
-        dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon;
-        dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon;
-        dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon;
-        dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon;
-        dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon;
-        dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon;
-        dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon;
-        dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon;
-
-        dsp->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_neon;
-        dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon;
-        dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon;
-        dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon;
-        dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon;
-        dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon;
-        dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon;
-        dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon;
-        dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon;
-
-        dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon;
-
-        dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon;
-
-        dsp->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon;
-        dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon;
-    }
+    if (have_armv6(cpu_flags))
+        ff_vp8dsp_init_armv6(dsp);
+    if (have_neon(cpu_flags))
+        ff_vp8dsp_init_neon(dsp);
 }
diff --git a/libavcodec/arm/vp8dsp_init_armv6.c b/libavcodec/arm/vp8dsp_init_armv6.c
new file mode 100644
index 0000000000..a1dcef57ee
--- /dev/null
+++ b/libavcodec/arm/vp8dsp_init_armv6.c
@@ -0,0 +1,34 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+#include "libavcodec/vp8dsp.h"
+#include "vp8dsp.h"
+
+void ff_vp8_luma_dc_wht_dc_armv6(DCTELEM block[4][4][16], DCTELEM dc[16]);
+
+VP8_MC(pixels4, armv6);
+
+av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp)
+{
+    dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6;
+
+    dsp->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6;
+
+    dsp->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6;
+}
diff --git a/libavcodec/arm/vp8dsp_init_neon.c b/libavcodec/arm/vp8dsp_init_neon.c
new file mode 100644
index 0000000000..dbe5b9f961
--- /dev/null
+++ b/libavcodec/arm/vp8dsp_init_neon.c
@@ -0,0 +1,114 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+#include "libavcodec/vp8dsp.h"
+#include "vp8dsp.h"
+
+void ff_vp8_luma_dc_wht_neon(DCTELEM block[4][4][16], DCTELEM dc[16]);
+
+void ff_vp8_idct_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride);
+
+VP8_LF(neon);
+
+VP8_EPEL(16, neon);
+VP8_EPEL(8,  neon);
+VP8_EPEL(4,  neon);
+
+VP8_BILIN(16, neon);
+VP8_BILIN(8,  neon);
+VP8_BILIN(4,  neon);
+
+av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp)
+{
+    dsp->vp8_luma_dc_wht    = ff_vp8_luma_dc_wht_neon;
+
+    dsp->vp8_idct_add       = ff_vp8_idct_add_neon;
+    dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_neon;
+    dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_neon;
+    dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon;
+
+    dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon;
+    dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon;
+    dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon;
+    dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon;
+
+    dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon;
+    dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon;
+    dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon;
+    dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon;
+
+    dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon;
+    dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon;
+
+    dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
+    dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon;
+
+    dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
+    dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon;
+
+    dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon;
+    dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon;
+
+    dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon;
+    dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon;
+}
diff --git a/libavcodec/arm/vp8dsp_neon.S b/libavcodec/arm/vp8dsp_neon.S
index e9f5b298ce..c1a91e1461 100644
--- a/libavcodec/arm/vp8dsp_neon.S
+++ b/libavcodec/arm/vp8dsp_neon.S
@@ -77,18 +77,6 @@ function ff_vp8_luma_dc_wht_neon, export=1
         bx              lr
 endfunc
 
-function ff_vp8_luma_dc_wht_dc_neon, export=1
-        ldrsh           r2,  [r1]
-        mov             r3,  #0
-        add             r2,  r2,  #3
-        strh            r3,  [r1]
-        asr             r2,  r2,  #3
-    .rept 16
-        strh            r2,  [r0], #32
-    .endr
-        bx              lr
-endfunc
-
 function ff_vp8_idct_add_neon, export=1
         vld1.16         {q0-q1},  [r1,:128]
         movw            r3,  #20091
@@ -725,23 +713,6 @@ function ff_put_vp8_pixels8_neon, export=1
         bx              lr
 endfunc
 
-function ff_put_vp8_pixels4_neon, export=1
-        ldr             r12, [sp, #0]           @ h
-        push            {r4-r6,lr}
-1:
-        subs            r12, r12, #4
-        ldr_post        r4,  r2,  r3
-        ldr_post        r5,  r2,  r3
-        ldr_post        r6,  r2,  r3
-        ldr_post        lr,  r2,  r3
-        str_post        r4,  r0,  r1
-        str_post        r5,  r0,  r1
-        str_post        r6,  r0,  r1
-        str_post        lr,  r0,  r1
-        bgt             1b
-        pop             {r4-r6,pc}
-endfunc
-
 /* 4/6-tap 8th-pel MC */
 
 .macro  vp8_epel8_h6    d,   a,   b