diff mbox series

[FFmpeg-devel,v4] avcodec: add Actimagine VX video decoder

Message ID 20210319084947.18024-1-fnouwt2@gmail.com
State New
Headers show
Series [FFmpeg-devel,v4] avcodec: add Actimagine VX video decoder
Related show

Checks

Context Check Description
andriy/x86_make success Make finished
andriy/x86_make_fate success Make fate finished
andriy/PPC64_make success Make finished
andriy/PPC64_make_fate success Make fate finished

Commit Message

Florian Nouwt March 19, 2021, 8:49 a.m. UTC
Signed-off-by: Florian Nouwt <fnouwt2@gmail.com>
---
 * fixed style issues
 * ff_h264_cavlc_data_init_vlc now only uses ff_thread_once if the vx decoder is enabled
 * made tables static which are not used outside h264_cavlc_data.c
---
 Changelog                       |    1 +
 configure                       |    1 +
 doc/general_contents.texi       |    2 +
 libavcodec/Makefile             |    3 +-
 libavcodec/actimagine_vx_data.c |   42 ++
 libavcodec/actimagine_vx_data.h |   28 +
 libavcodec/actimagine_vxdec.c   | 1149 +++++++++++++++++++++++++++++++
 libavcodec/allcodecs.c          |    1 +
 libavcodec/codec_desc.c         |    7 +
 libavcodec/codec_id.h           |    1 +
 libavcodec/h264_cavlc.c         |  205 +-----
 libavcodec/h264_cavlc_data.c    |  231 +++++++
 libavcodec/h264_cavlc_data.h    |   43 ++
 libavcodec/version.h            |    2 +-
 libavformat/riff.c              |    2 +
 15 files changed, 1524 insertions(+), 194 deletions(-)
 create mode 100644 libavcodec/actimagine_vx_data.c
 create mode 100644 libavcodec/actimagine_vx_data.h
 create mode 100644 libavcodec/actimagine_vxdec.c
 create mode 100644 libavcodec/h264_cavlc_data.c
 create mode 100644 libavcodec/h264_cavlc_data.h
diff mbox series

Patch

diff --git a/Changelog b/Changelog
index a96e350e09..8807f3dcb3 100644
--- a/Changelog
+++ b/Changelog
@@ -83,6 +83,7 @@  version <next>:
 - msad video filter
 - gophers protocol
 - RIST protocol via librist
+- Actimagine VX video decoder
 
 
 version 4.3:
diff --git a/configure b/configure
index f0ac719d2d..c38066bc1c 100755
--- a/configure
+++ b/configure
@@ -2662,6 +2662,7 @@  ac3_fixed_decoder_select="ac3_parser ac3dsp bswapdsp mdct"
 ac3_encoder_select="ac3dsp audiodsp mdct me_cmp"
 ac3_fixed_encoder_select="ac3dsp audiodsp mdct me_cmp"
 acelp_kelvin_decoder_select="audiodsp"
+actimagine_vx_decoder_select="bswapdsp golomb h264dsp h264pred"
 adpcm_g722_decoder_select="g722dsp"
 adpcm_g722_encoder_select="g722dsp"
 aic_decoder_select="golomb idctdsp"
diff --git a/doc/general_contents.texi b/doc/general_contents.texi
index 33ece6e884..d4261386fc 100644
--- a/doc/general_contents.texi
+++ b/doc/general_contents.texi
@@ -807,6 +807,8 @@  following image formats are supported:
 @item 8088flex TMV           @tab     @tab  X
 @item A64 multicolor         @tab  X  @tab
     @tab Creates video suitable to be played on a commodore 64 (multicolor mode).
+@item Actimagine VX Video    @tab     @tab  X
+    @tab fourcc: vxs1, VXS1
 @item Amazing Studio PAF Video @tab     @tab  X
 @item American Laser Games MM  @tab    @tab X
     @tab Used in games like Mad Dog McCree.
diff --git a/libavcodec/Makefile b/libavcodec/Makefile
index 81cc16471b..1adae43b40 100644
--- a/libavcodec/Makefile
+++ b/libavcodec/Makefile
@@ -182,6 +182,7 @@  OBJS-$(CONFIG_AC3_ENCODER)             += ac3enc_float.o ac3enc.o ac3tab.o \
 OBJS-$(CONFIG_AC3_FIXED_ENCODER)       += ac3enc_fixed.o ac3enc.o ac3tab.o ac3.o kbdwin.o
 OBJS-$(CONFIG_AC3_MF_ENCODER)          += mfenc.o mf_utils.o
 OBJS-$(CONFIG_ACELP_KELVIN_DECODER)    += g729dec.o lsp.o celp_math.o celp_filters.o acelp_filters.o acelp_pitch_delay.o acelp_vectors.o g729postfilter.o
+OBJS-$(CONFIG_ACTIMAGINE_VX_DECODER)   += actimagine_vxdec.o actimagine_vx_data.o h264_cavlc_data.o
 OBJS-$(CONFIG_AGM_DECODER)             += agm.o
 OBJS-$(CONFIG_AIC_DECODER)             += aic.o
 OBJS-$(CONFIG_ALAC_DECODER)            += alac.o alac_data.o alacdsp.o
@@ -367,7 +368,7 @@  OBJS-$(CONFIG_H264_DECODER)            += h264dec.o h264_cabac.o h264_cavlc.o \
                                           h264_direct.o h264_loopfilter.o  \
                                           h264_mb.o h264_picture.o \
                                           h264_refs.o h264_sei.o \
-                                          h264_slice.o h264data.o
+                                          h264_slice.o h264data.o h264_cavlc_data.o
 OBJS-$(CONFIG_H264_AMF_ENCODER)        += amfenc_h264.o
 OBJS-$(CONFIG_H264_CUVID_DECODER)      += cuviddec.o
 OBJS-$(CONFIG_H264_MEDIACODEC_DECODER) += mediacodecdec.o
diff --git a/libavcodec/actimagine_vx_data.c b/libavcodec/actimagine_vx_data.c
new file mode 100644
index 0000000000..64144de4a6
--- /dev/null
+++ b/libavcodec/actimagine_vx_data.c
@@ -0,0 +1,42 @@ 
+/*
+ * Actimagine VX Video tables
+ * Copyright (c) 2021 Florian Nouwt
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <inttypes.h>
+#include "actimagine_vx_data.h"
+
+const uint8_t ff_actimagine_vx_old_mb_mode_remap_tab[24] = {
+     1,  2,  0,  4,  7,  3, 11, 15,  9,  5, 14,  6,
+    12, 13,  8, 16, 23, 10, 22, 19, 20, 17, 21, 18
+};
+
+const uint8_t ff_actimagine_vx_residu_mask_old_tab[32] = {
+    0x00, 0x01, 0x02, 0x04, 0x08, 0x10, 0x03, 0x05,
+    0x0A, 0x0C, 0x0F, 0x1F, 0x07, 0x0B, 0x0D, 0x0E,
+    0x06, 0x09, 0x13, 0x15, 0x1A, 0x1C, 0x11, 0x12,
+    0x14, 0x18, 0x17, 0x1B, 0x1D, 0x1E, 0x16, 0x19
+};
+
+const uint8_t ff_actimagine_vx_residu_mask_new_tab[32] = {
+    0x00, 0x08, 0x04, 0x02, 0x01, 0x1F, 0x0F, 0x0A,
+    0x05, 0x0C, 0x03, 0x10, 0x0E, 0x0D, 0x0B, 0x07,
+    0x09, 0x06, 0x1E, 0x1B, 0x1A, 0x1D, 0x17, 0x15,
+    0x18, 0x12, 0x11, 0x1C, 0x14, 0x13, 0x16, 0x19
+};
diff --git a/libavcodec/actimagine_vx_data.h b/libavcodec/actimagine_vx_data.h
new file mode 100644
index 0000000000..46a715ca71
--- /dev/null
+++ b/libavcodec/actimagine_vx_data.h
@@ -0,0 +1,28 @@ 
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_ACTIMAGINE_VX_DATA_H
+#define AVCODEC_ACTIMAGINE_VX_DATA_H
+
+#include <stdint.h>
+
+extern const uint8_t ff_actimagine_vx_old_mb_mode_remap_tab[24];
+extern const uint8_t ff_actimagine_vx_residu_mask_old_tab[32];
+extern const uint8_t ff_actimagine_vx_residu_mask_new_tab[32];
+
+#endif /* AVCODEC_ACTIMAGINE_VX_DATA_H */
diff --git a/libavcodec/actimagine_vxdec.c b/libavcodec/actimagine_vxdec.c
new file mode 100644
index 0000000000..5a527a66dd
--- /dev/null
+++ b/libavcodec/actimagine_vxdec.c
@@ -0,0 +1,1149 @@ 
+/*
+ * Actimagine VX Video decoder
+ * Copyright (c) 2021 Florian Nouwt
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <inttypes.h>
+
+#include "libavutil/avassert.h"
+#include "avcodec.h"
+#include "bytestream.h"
+#include "bswapdsp.h"
+#include "get_bits.h"
+#include "golomb.h"
+#include "internal.h"
+#include "libavutil/mem_internal.h"
+#include "h264_cavlc_data.h"
+#include "h264dsp.h"
+#include "h264pred.h"
+#include "actimagine_vx_data.h"
+
+static const uint8_t quant4x4_tab[][8] = {
+    { 0x0A, 0x0D, 0x0A, 0x0D, 0x0D, 0x10, 0x0D, 0x10 },
+    { 0x0B, 0x0E, 0x0B, 0x0E, 0x0E, 0x12, 0x0E, 0x12 },
+    { 0x0D, 0x10, 0x0D, 0x10, 0x10, 0x14, 0x10, 0x14 },
+    { 0x0E, 0x12, 0x0E, 0x12, 0x12, 0x17, 0x12, 0x17 },
+    { 0x10, 0x14, 0x10, 0x14, 0x14, 0x19, 0x14, 0x19 },
+    { 0x12, 0x17, 0x12, 0x17, 0x17, 0x1D, 0x17, 0x1D }
+};
+
+typedef struct MVec {
+    int x, y;
+} MVec;
+
+typedef struct VxContext {
+    AVFrame *out_frame;
+    AVFrame *cur_frame;
+    AVFrame *ref_frames[3];
+    int      ref_frame_count;
+
+    int version;
+    int quantizer;
+    int avi;
+
+    GetBitContext gb;
+
+    uint8_t *bitstream;
+    int bitstream_size;
+
+    uint8_t zigzag_scan[16];
+
+    int qtab[2][4];
+
+    uint8_t pred4_cache[5][5];
+
+    MVec *vectors;
+    int   vectors_stride;
+
+    uint8_t *total_coeff_y;
+    int      total_coeff_y_stride;
+
+    uint8_t *total_coeff_uv;
+    int      total_coeff_uv_stride;
+
+    BswapDSPContext bdsp;
+    H264DSPContext  h264dsp;
+    H264PredContext h264pred;
+} VxContext;
+
+#define PIXEL_REF(s, ref, plane, x, y)\
+    ((s)->ref_frames[(ref)]->data[(plane)]\
+        [(y) * (s)->ref_frames[(ref)]->linesize[(plane)] + (x)])
+
+#define PIXEL_CUR(s, plane, x, y)\
+    ((s)->cur_frame->data[(plane)]\
+        [(y) * (s)->cur_frame->linesize[(plane)] + (x)])
+
+#define VX_VERSION_INVALID  -1
+#define VX_VERSION_OLD       0
+#define VX_VERSION_NEW       1
+
+static int setup_qtables(AVCodecContext *avctx, int quantizer)
+{
+    int qx, qy;
+    VxContext *s = avctx->priv_data;
+
+    if (quantizer < 12 || quantizer > 161)
+        return AVERROR_INVALIDDATA;
+
+    s->quantizer = quantizer;
+
+    qx = quantizer % 6;
+    qy = quantizer / 6;
+
+    for (int i = 0; i < 2; i++)
+        for (int j = 0; j < 4; j++)
+            s->qtab[i][j] = quant4x4_tab[qx][4 * i + j] << qy;
+
+    return 0;
+}
+
+static av_cold int vx_init(AVCodecContext *avctx)
+{
+    int vectors_size;
+    int total_coeff_y_size;
+    int total_coeff_uv_size;
+    VxContext *s = avctx->priv_data;
+
+    if (avctx->width & 15 || avctx->height & 15) {
+        av_log(avctx, AV_LOG_ERROR, "width/height not multiple of 16\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    ff_bswapdsp_init(&s->bdsp);
+    ff_h264dsp_init(&s->h264dsp, 8, 1);
+    ff_h264_pred_init(&s->h264pred, AV_CODEC_ID_ACTIMAGINE_VX, 8, 1);
+
+    avctx->pix_fmt = AV_PIX_FMT_YUV420P;
+    avctx->color_range = AVCOL_RANGE_JPEG;
+    avctx->chroma_sample_location = AVCHROMA_LOC_CENTER;
+    avctx->colorspace = AVCOL_SPC_SMPTE170M;
+
+    // zigzag_scan is a transposed version of ff_zigzag_scan, like in h264
+    for (int i = 0; i < 16; i++)
+        s->zigzag_scan[i] = (ff_zigzag_scan[i] >> 2) | ((ff_zigzag_scan[i] << 2) & 0xF);
+
+    // predict4 cache
+    for (int i = 0; i < 5; i++)
+        for (int j = 0; j < 5; j++)
+            s->pred4_cache[i][j] = 9;
+
+    // motion vector cache
+    s->vectors_stride = (avctx->width >> 4) + 2;
+    vectors_size = ((avctx->height >> 4) + 1) * s->vectors_stride;
+    s->vectors = av_calloc(vectors_size, sizeof(MVec));
+    if (!s->vectors)
+        return AVERROR(ENOMEM);
+
+    // total dct coefficient cache for luma
+    s->total_coeff_y_stride = (avctx->width >> 2) + 1;
+    total_coeff_y_size = ((avctx->height >> 2) + 1) * s->total_coeff_y_stride;
+    s->total_coeff_y = av_mallocz(total_coeff_y_size);
+    if (!s->total_coeff_y)
+        return AVERROR(ENOMEM);
+
+    // total dct coefficient cache for chroma
+    s->total_coeff_uv_stride = (avctx->width >> 3) + 1;
+    total_coeff_uv_size = ((avctx->height >> 3) + 1) * s->total_coeff_uv_stride;
+    s->total_coeff_uv = av_mallocz(total_coeff_uv_size);
+    if (!s->total_coeff_uv)
+        return AVERROR(ENOMEM);
+
+    s->ref_frame_count = 0;
+    for (int i = 0; i < 3; i++) {
+        s->ref_frames[i] = av_frame_alloc();
+        if (!s->ref_frames[i])
+            return AVERROR(ENOMEM);
+    }
+    s->cur_frame = av_frame_alloc();
+    if (!s->cur_frame)
+        return AVERROR(ENOMEM);
+    s->out_frame = av_frame_alloc();
+    if (!s->out_frame)
+        return AVERROR(ENOMEM);
+
+    s->version = VX_VERSION_INVALID;
+    s->quantizer = -1;
+    s->avi = 0;
+
+    // when the source is an avi file, the quantizer is stored in the extradata
+    if (avctx->extradata_size == 4)
+        if (!setup_qtables(avctx, AV_RL32(avctx->extradata)))
+            s->avi = 1;
+
+    ff_h264_cavlc_data_init_vlc();
+
+    return 0;
+}
+
+static void clear_total_coeff(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+
+    // luma
+    uint8_t *total_coeff = &s->total_coeff_y[
+        ((y >> 2) + 1) * s->total_coeff_y_stride + (x >> 2) + 1];
+
+    for (int y2 = 0; y2 < (h >> 2); y2++)
+        for (int x2 = 0; x2 < (w >> 2); x2++)
+            total_coeff[y2 * s->total_coeff_y_stride + x2] = 0;
+
+    // chroma
+    total_coeff = &s->total_coeff_uv[
+        ((y >> 3) + 1) * s->total_coeff_uv_stride + (x >> 3) + 1];
+
+    for (int y2 = 0; y2 < (h >> 3); y2++)
+        for (int x2 = 0; x2 < (w >> 3); x2++)
+            total_coeff[y2 * s->total_coeff_uv_stride + x2] = 0;
+}
+
+static void predict_plane_intern(AVCodecContext *avctx, int x, int y,
+                                 int w, int h, int plane)
+{
+    VxContext *s = avctx->priv_data;
+    if (w == 1 && h == 1)
+        return;
+    if (w == 1 && h != 1) {
+        uint8_t top    = PIXEL_CUR(s, plane, x, y - 1);
+        uint8_t bottom = PIXEL_CUR(s, plane, x, y + h - 1);
+        PIXEL_CUR(s, plane, x, y + (h >> 1) - 1) = (top + bottom) >> 1;
+        predict_plane_intern(avctx, x, y, 1, h >> 1, plane);
+        predict_plane_intern(avctx, x, y + (h >> 1), 1, h >> 1, plane);
+    } else if (w != 1 && h == 1) {
+        uint8_t left  = PIXEL_CUR(s, plane, x - 1, y);
+        uint8_t right = PIXEL_CUR(s, plane, x + w - 1, y);
+        PIXEL_CUR(s, plane, x + (w >> 1) - 1, y) = (left + right) >> 1;
+        predict_plane_intern(avctx, x, y, w >> 1, 1, plane);
+        predict_plane_intern(avctx, x + (w >> 1), y, w >> 1, 1, plane);
+    } else {
+        uint8_t bottom_left  = PIXEL_CUR(s, plane, x - 1, y + h - 1);
+        uint8_t top_right    = PIXEL_CUR(s, plane, x + w - 1, y - 1);
+        uint8_t bottom_right = PIXEL_CUR(s, plane, x + w - 1, y + h - 1);
+        uint8_t bottom_center = (bottom_left + bottom_right) >> 1;
+        uint8_t center_right = (top_right + bottom_right) >> 1;
+        PIXEL_CUR(s, plane, x + (w >> 1) - 1, y + h - 1) = bottom_center;
+        PIXEL_CUR(s, plane, x + w - 1, y + (h >> 1) - 1) = center_right;
+        if ((w == 4 || w == 16) ^ (h == 4 || h == 16)) {
+            uint8_t center_left = PIXEL_CUR(s, plane, x - 1, y + (h >> 1) - 1);
+            PIXEL_CUR(s, plane, x + (w >> 1) - 1, y + (h >> 1) - 1)
+                = (center_left + center_right) >> 1;
+        } else {
+            uint8_t top_center = PIXEL_CUR(s, plane, x + (w >> 1) - 1, y - 1);
+            PIXEL_CUR(s, plane, x + (w >> 1) - 1, y + (h >> 1) - 1)
+                = (top_center + bottom_center) >> 1;
+        }
+        predict_plane_intern(avctx, x, y, w >> 1, h >> 1, plane);
+        predict_plane_intern(avctx, x + (w >> 1), y, w >> 1, h >> 1, plane);
+        predict_plane_intern(avctx, x, y + (h >> 1), w >> 1, h >> 1, plane);
+        predict_plane_intern(avctx, x + (w >> 1), y + (h >> 1), w >> 1, h >> 1,
+                             plane);
+    }
+}
+
+static void predict_plane(AVCodecContext *avctx, int x, int y, int w, int h,
+                          int plane, int param)
+{
+    VxContext *s = avctx->priv_data;
+    uint8_t bottom_left = PIXEL_CUR(s, plane, x - 1, y + h - 1);
+    uint8_t top_right   = PIXEL_CUR(s, plane, x + w - 1, y - 1);
+    PIXEL_CUR(s, plane, x + w - 1, y + h - 1)
+        = ((bottom_left + top_right + 1) >> 1) + param;
+    predict_plane_intern(avctx, x, y, w, h, plane);
+}
+
+static int predict_mb_plane(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    // y
+    int param = get_se_golomb(gb);
+    if (param < -(1 << 16) || param >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid plane param\n");
+        return AVERROR_INVALIDDATA;
+    }
+    predict_plane(avctx, x, y, w, h, 0, param << 1);
+
+    // u
+    param = get_se_golomb(gb);
+    if (param < -(1 << 16) || param >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid plane param\n");
+        return AVERROR_INVALIDDATA;
+    }
+    predict_plane(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 1, param << 1);
+
+    // v
+    param = get_se_golomb(gb);
+    if (param < -(1 << 16) || param >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid plane param\n");
+        return AVERROR_INVALIDDATA;
+    }
+    predict_plane(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 2, param << 1);
+
+    return 0;
+}
+
+static void predict_horizontal(AVCodecContext *avctx, int x, int y,
+                               int w, int h, int plane)
+{
+    VxContext *s = avctx->priv_data;
+    for (int y2 = 0; y2 < h; y2++) {
+        uint8_t pixel = PIXEL_CUR(s, plane, x - 1, y + y2);
+        for (int x2 = 0; x2 < w; x2++)
+            PIXEL_CUR(s, plane, x + x2, y + y2) = pixel;
+    }
+}
+
+static void predict_vertical(AVCodecContext *avctx, int x, int y,
+                             int w, int h, int plane)
+{
+    VxContext *s = avctx->priv_data;
+    for (int y2 = 0; y2 < h; y2++)
+        for (int x2 = 0; x2 < w; x2++)
+            PIXEL_CUR(s, plane, x + x2, y + y2)
+                = PIXEL_CUR(s, plane, x + x2, y - 1);
+}
+
+static void predict_dc(AVCodecContext *avctx, int x, int y, int w, int h,
+                       int plane)
+{
+    static const uint8_t shift_tab[] = { 1, 2, 3, 0, 4 };
+    uint8_t dc;
+    VxContext *s = avctx->priv_data;
+    if (x != 0 && y != 0) {
+        int sum_h, sum_v;
+        sum_h = w >> 1;
+        for (int x2 = 0; x2 < w; x2++)
+            sum_h += PIXEL_CUR(s, plane, x + x2, y - 1);
+        sum_h >>= shift_tab[w >> 2];
+
+        sum_v = h >> 1;
+        for (int y2 = 0; y2 < h; y2++)
+            sum_v += PIXEL_CUR(s, plane, x - 1, y + y2);
+        sum_v >>= shift_tab[h >> 2];
+
+        dc = (sum_h + sum_v + 1) >> 1;
+    } else if (x == 0 && y != 0) {
+        int sum = w >> 1;
+        for (int x2 = 0; x2 < w; x2++)
+            sum += PIXEL_CUR(s, plane, x + x2, y - 1);
+        dc = sum >> shift_tab[w >> 2];
+    } else if (x != 0 && y == 0) {
+        int sum = h >> 1;
+        for (int y2 = 0; y2 < h; y2++)
+            sum += PIXEL_CUR(s, plane, x - 1, y + y2);
+        dc = sum >> shift_tab[h >> 2];
+    } else
+        dc = 128;
+
+    for (int y2 = 0; y2 < h; y2++)
+        for (int x2 = 0; x2 < w; x2++)
+            PIXEL_CUR(s, plane, x + x2, y + y2) = dc;
+}
+
+static int predict_notile_uv(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    switch (get_ue_golomb_31(gb)) {
+    case 0:// dc
+        predict_dc(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 1);
+        predict_dc(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 2);
+        break;
+    case 1:// horizontal
+        predict_horizontal(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 1);
+        predict_horizontal(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 2);
+        break;
+    case 2:// vertical
+        predict_vertical(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 1);
+        predict_vertical(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 2);
+        break;
+    case 3:// plane
+        predict_plane(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 1, 0);
+        predict_plane(avctx, x >> 1, y >> 1, w >> 1, h >> 1, 2, 0);
+        break;
+    default:
+        av_log(avctx, AV_LOG_ERROR, "invalid predict notile uv mode\n");
+        return AVERROR_INVALIDDATA;
+    }
+    return 0;
+}
+
+static int predict_notile(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    switch (get_ue_golomb_31(gb)) {
+    case 0:// vertical
+        predict_vertical(avctx, x, y, w, h, 0);
+        break;
+    case 1:// horizontal
+        predict_horizontal(avctx, x, y, w, h, 0);
+        break;
+    case 2:// dc
+        predict_dc(avctx, x, y, w, h, 0);
+        break;
+    case 3:// plane
+        predict_plane(avctx, x, y, w, h, 0, 0);
+        break;
+    default:
+        av_log(avctx, AV_LOG_ERROR, "invalid predict notile y mode\n");
+        return AVERROR_INVALIDDATA;
+    }
+    return predict_notile_uv(avctx, x, y, w, h);
+}
+
+static int predict4(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    int stride = s->cur_frame->linesize[0];
+    for (int y2 = 0; y2 < h >> 2; y2++) {
+        for (int x2 = 0; x2 < w >> 2; x2++) {
+            uint8_t *dst, *top_right;
+            uint8_t mode = FFMIN(s->pred4_cache[1 + y2 - 1][1 + x2],
+                                 s->pred4_cache[1 + y2][1 + x2 - 1]);
+            if (mode == 9)// if invalid predict dc
+                mode = 2;
+
+            if (!get_bits1(gb)) {
+                uint8_t val = get_bits(gb, 3);
+                mode = val + (val >= mode);
+            }
+
+            s->pred4_cache[1 + y2][1 + x2] = mode;
+
+            dst = &PIXEL_CUR(s, 0, x + x2 * 4, y + y2 * 4);
+            top_right = &PIXEL_CUR(s, 0, x + x2 * 4 + 4, y + y2 * 4 - 1);
+
+            switch (mode) {
+            case 0:// vertical
+                s->h264pred.pred4x4[VERT_PRED](dst, NULL, stride);
+                break;
+            case 1:// horizontal
+                s->h264pred.pred4x4[HOR_PRED](dst, NULL, stride);
+                break;
+            case 2:// dc
+                if (x + x2 * 4 == 0 && y + y2 * 4 == 0)
+                    s->h264pred.pred4x4[DC_128_PRED](dst, NULL, stride);
+                else if (x + x2 * 4 == 0 && y + y2 * 4 != 0)
+                    s->h264pred.pred4x4[TOP_DC_PRED](dst, NULL, stride);
+                else if (x + x2 * 4 != 0 && y + y2 * 4 == 0)
+                    s->h264pred.pred4x4[LEFT_DC_PRED](dst, NULL, stride);
+                else
+                    s->h264pred.pred4x4[DC_PRED](dst, NULL, stride);
+                break;
+            case 3:// diagonal-down-left
+                s->h264pred.pred4x4[DIAG_DOWN_LEFT_PRED](dst, top_right, stride);
+                break;
+            case 4:// diagonal-down-right
+                s->h264pred.pred4x4[DIAG_DOWN_RIGHT_PRED](dst, NULL, stride);
+                break;
+            case 5:// vertical-right
+                s->h264pred.pred4x4[VERT_RIGHT_PRED](dst, NULL, stride);
+                break;
+            case 6:// horizontal-down
+                s->h264pred.pred4x4[HOR_DOWN_PRED](dst, NULL, stride);
+                break;
+            case 7:// vertical-left
+                s->h264pred.pred4x4[VERT_LEFT_PRED](dst, top_right, stride);
+                break;
+            case 8:// horizontal-up
+                s->h264pred.pred4x4[HOR_UP_PRED](dst, NULL, stride);
+                break;
+            default:
+                av_log(avctx, AV_LOG_ERROR, "invalid predict4 mode\n");
+                return AVERROR_INVALIDDATA;
+            }
+        }
+    }
+    return predict_notile_uv(avctx, x, y, w, h);
+}
+
+static void decode_dct(AVCodecContext *avctx, int x, int y, int plane,
+                       const int* level)
+{
+    LOCAL_ALIGNED_16(int16_t, dct, [16]);
+    VxContext *s = avctx->priv_data;
+
+    // dezigzag
+    for (int i = 0; i < 16; i++)
+        dct[s->zigzag_scan[i]] = level[i];
+
+    // dequantize
+    for (int i = 0; i < 2; i++) {
+        for (int j = 0; j < 4; j++) {
+            dct[4 * j + i]     *= s->qtab[i][j];
+            dct[4 * j + i + 2] *= s->qtab[i][j];
+        }
+    }
+
+    s->h264dsp.h264_idct_add(&PIXEL_CUR(s, plane, x, y), dct,
+                             s->cur_frame->linesize[plane]);
+}
+
+static void decode_residu_cavlc(AVCodecContext *avctx, int x, int y, int plane,
+                                int nc, uint8_t *out_total_coeff)
+{
+    int level[16];
+    int coeff_token, total_coeff, trailing_ones, i, zeros_left, suffix_length;
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+
+    coeff_token = get_vlc2(gb, ff_h264_cavlc_coeff_token_vlc[ff_h264_cavlc_coeff_token_table_index[nc]].table,
+                           FF_H264_CAVLC_COEFF_TOKEN_VLC_BITS, 2);
+    trailing_ones = coeff_token & 3;
+    total_coeff   = coeff_token >> 2;
+
+    *out_total_coeff = total_coeff;
+    if (total_coeff == 0)
+        return;
+
+    av_assert2(total_coeff <= 16);
+
+    i = 15;
+    if (total_coeff != 16) {
+        int trailing_zeros;
+        zeros_left = get_vlc2(gb, ff_h264_cavlc_total_zeros_vlc[total_coeff].table,
+                              FF_H264_CAVLC_TOTAL_ZEROS_VLC_BITS, 1);
+        trailing_zeros = 16 - (total_coeff + zeros_left);
+        while(trailing_zeros-- > 0)
+            level[i--] = 0;
+    } else
+        zeros_left = 0;
+
+    suffix_length = 0;
+    while (1) {
+        int level_suffix, level_code, run_before;
+        if (trailing_ones > 0) {
+            trailing_ones--;
+            level[i--] = get_bits1(gb) ? -1 : 1;
+        } else {
+            int level_prefix = 0;
+            while (!get_bits1(gb))
+                level_prefix++;
+
+            if (level_prefix == 15)
+                level_suffix = get_bits(gb, 11);
+            else
+                level_suffix = suffix_length == 0 ? 0 : get_bits(gb, suffix_length);
+
+            level_code = level_suffix + (level_prefix << suffix_length) + 1;
+
+            suffix_length += level_code > ff_h264_cavlc_suffix_limit[suffix_length + 1];
+
+            if (get_bits1(gb))
+                level_code = -level_code;
+            level[i--] = level_code;
+        }
+
+        if (--total_coeff == 0)
+            break;
+
+        if (zeros_left == 0)
+            continue;
+
+        if(zeros_left < 7)
+            run_before = get_vlc2(gb, ff_h264_cavlc_run_vlc[zeros_left].table,
+                                  FF_H264_CAVLC_RUN_VLC_BITS, 1);
+        else
+            run_before = get_vlc2(gb, ff_h264_cavlc_run7_vlc.table,
+                                  FF_H264_CAVLC_RUN7_VLC_BITS, 2);
+        zeros_left -= run_before;
+        while(run_before-- > 0)
+            level[i--] = 0;
+    }
+
+    while(zeros_left-- > 0)
+        level[i--] = 0;
+
+    decode_dct(avctx, x, y, plane, level);
+}
+
+static int decode_residu_blocks(AVCodecContext *avctx, int x, int y,
+                                int w, int h)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    uint8_t *total_coeff_y = &s->total_coeff_y[
+        ((y >> 2) + 1) * s->total_coeff_y_stride + (x >> 2) + 1];
+    uint8_t *total_coeff_uv = &s->total_coeff_uv[
+        ((y >> 3) + 1) * s->total_coeff_uv_stride + (x >> 3) + 1];
+    for (int y2 = 0; y2 < h >> 3; y2++) {
+        for (int x2 = 0; x2 < w >> 3; x2++) {
+            uint8_t residu_mask;
+            unsigned int code = get_ue_golomb_31(gb);
+            if (code > 0x1F) {
+                av_log(avctx, AV_LOG_ERROR, "invalid residu mask code\n");
+                return AVERROR_INVALIDDATA;
+            }
+            if (s->version == VX_VERSION_OLD)
+                residu_mask = ff_actimagine_vx_residu_mask_old_tab[code];
+            else
+                residu_mask = ff_actimagine_vx_residu_mask_new_tab[code];
+
+            if (residu_mask & 1) {
+                int nc = (total_coeff_y[-1] +
+                          total_coeff_y[-s->total_coeff_y_stride] + 1) >> 1;
+                decode_residu_cavlc(avctx, x + x2 * 8, y + y2 * 8, 0, nc,
+                                    &total_coeff_y[0]);
+            } else
+                total_coeff_y[0] = 0;
+
+            if (residu_mask & 2) {
+                int nc = (total_coeff_y[0] +
+                          total_coeff_y[-s->total_coeff_y_stride + 1] + 1) >> 1;
+                decode_residu_cavlc(avctx, x + x2 * 8 + 4, y + y2 * 8, 0, nc,
+                                    &total_coeff_y[1]);
+            } else
+                total_coeff_y[1] = 0;
+
+            if (residu_mask & 4) {
+                int nc = (total_coeff_y[s->total_coeff_y_stride - 1] +
+                          total_coeff_y[0] + 1) >> 1;
+                decode_residu_cavlc(avctx, x + x2 * 8, y + y2 * 8 + 4, 0, nc,
+                                    &total_coeff_y[s->total_coeff_y_stride]);
+            } else
+                total_coeff_y[s->total_coeff_y_stride] = 0;
+
+            if (residu_mask & 8) {
+                int nc = (total_coeff_y[s->total_coeff_y_stride] +
+                          total_coeff_y[1] + 1) >> 1;
+                decode_residu_cavlc(
+                    avctx, x + x2 * 8 + 4, y + y2 * 8 + 4, 0, nc,
+                    &total_coeff_y[s->total_coeff_y_stride + 1]);
+            } else
+                total_coeff_y[s->total_coeff_y_stride + 1] = 0;
+
+            if (residu_mask & 16) {
+                uint8_t total_coeff_u, total_coeff_v;
+                int nc = (total_coeff_uv[-1] +
+                          total_coeff_uv[-s->total_coeff_uv_stride] + 1) >> 1;
+                decode_residu_cavlc(avctx, (x + x2 * 8) >> 1, (y + y2 * 8) >> 1,
+                                    1, nc, &total_coeff_u);
+                decode_residu_cavlc(avctx, (x + x2 * 8) >> 1, (y + y2 * 8) >> 1,
+                                    2, nc, &total_coeff_v);
+                total_coeff_uv[0] = (total_coeff_u + total_coeff_v + 1) >> 1;
+            } else
+                total_coeff_uv[0] = 0;
+
+            total_coeff_y += 2;
+            total_coeff_uv++;
+        }
+        total_coeff_y  += (s->total_coeff_y_stride << 1) - (w >> 2);
+        total_coeff_uv += s->total_coeff_uv_stride - (w >> 3);
+    }
+    return 0;
+}
+
+static int predict_inter(AVCodecContext *avctx, int x, int y, int w, int h,
+                          const MVec *predVec, int has_delta, int ref_frame)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    MVec vec = *predVec;
+
+    if (ref_frame >= s->ref_frame_count) {
+        av_log(avctx, AV_LOG_ERROR, "reference to unavailable frame\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    s->cur_frame->pict_type = AV_PICTURE_TYPE_P;
+    s->cur_frame->key_frame = 0;
+
+    if (has_delta) {
+        vec.x += (unsigned)get_se_golomb(gb);
+        vec.y += (unsigned)get_se_golomb(gb);
+    }
+
+    if (vec.x >= INT_MAX || vec.y >= INT_MAX)
+        return AVERROR_INVALIDDATA;
+
+    s->vectors[(1 + (y >> 4)) * s->vectors_stride + 1 + (x >> 4)] = vec;
+
+    if (x + vec.x < 0 || x + vec.x + w > avctx->width ||
+        y + vec.y < 0 || y + vec.y + h > avctx->height) {
+        av_log(avctx, AV_LOG_ERROR, "motion vector out of bounds\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    // luma
+    for (int y2 = 0; y2 < h; y2++)
+        for (int x2 = 0; x2 < w; x2++)
+            PIXEL_CUR(s, 0, x + x2, y + y2)
+                = PIXEL_REF(s, ref_frame, 0, x + x2 + vec.x, y + y2 + vec.y);
+
+    // chroma
+    for (int y2 = 0; y2 < (h >> 1); y2++) {
+        for (int x2 = 0; x2 < (w >> 1); x2++) {
+            // u
+            PIXEL_CUR(s, 1, (x >> 1) + x2, (y >> 1) + y2)
+                = PIXEL_REF(s, ref_frame, 1, (x >> 1) + x2 + (vec.x >> 1),
+                            (y >> 1) + y2 + (vec.y >> 1));
+            // v
+            PIXEL_CUR(s, 2, (x >> 1) + x2, (y >> 1) + y2)
+                = PIXEL_REF(s, ref_frame, 2, (x >> 1) + x2 + (vec.x >> 1),
+                            (y >> 1) + y2 + (vec.y >> 1));
+        }
+    }
+
+    return 0;
+}
+
+static int predict_inter_dc(AVCodecContext *avctx, int x, int y, int w, int h)
+{
+    int dx, dy, dc_y, dc_u, dc_v;
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+
+    if (s->ref_frame_count == 0) {
+        av_log(avctx, AV_LOG_ERROR, "reference to unavailable frame\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    dx = get_se_golomb(gb);
+    dy = get_se_golomb(gb);
+
+    if (x + dx < 0 || x + dx + w > avctx->width ||
+        y + dy < 0 || y + dy + h > avctx->height) {
+        av_log(avctx, AV_LOG_ERROR, "motion vector out of bounds\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    dc_y = get_se_golomb(gb);
+    if (dc_y < -(1<<16) || dc_y >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid dc offset\n");
+        return AVERROR_INVALIDDATA;
+    }
+    dc_y <<= 1;
+
+    dc_u = get_se_golomb(gb);
+    if (dc_u < -(1<<16) || dc_u >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid dc offset\n");
+        return AVERROR_INVALIDDATA;
+    }
+    dc_u <<= 1;
+
+    dc_v = get_se_golomb(gb);
+    if (dc_v < -(1<<16) || dc_v >= (1 << 16)) {
+        av_log(avctx, AV_LOG_ERROR, "invalid dc offset\n");
+        return AVERROR_INVALIDDATA;
+    }
+    dc_v <<= 1;
+
+    s->cur_frame->pict_type = AV_PICTURE_TYPE_P;
+    s->cur_frame->key_frame = 0;
+
+    // luma
+    for (int y2 = 0; y2 < h; y2++)
+        for (int x2 = 0; x2 < w; x2++)
+            PIXEL_CUR(s, 0, x + x2, y + y2) = av_clip_uint8(
+                PIXEL_REF(s, 0, 0, x + x2 + dx, y + y2 + dy) + dc_y);
+
+    // chroma
+    for (int y2 = 0; y2 < (h >> 1); y2++) {
+        for (int x2 = 0; x2 < (w >> 1); x2++) {
+            // u
+            PIXEL_CUR(s, 1, (x >> 1) + x2, (y >> 1) + y2) = av_clip_uint8(
+                PIXEL_REF(s, 0, 1, (x >> 1) + x2 + (dx >> 1),
+                          (y >> 1) + y2 + (dy >> 1)) + dc_u);
+            // v
+            PIXEL_CUR(s, 2, (x >> 1) + x2, (y >> 1) + y2) = av_clip_uint8(
+                PIXEL_REF(s, 0, 2, (x >> 1) + x2 + (dx >> 1),
+                          (y >> 1) + y2 + (dy >> 1)) + dc_v);
+        }
+    }
+
+    return 0;
+}
+
+static int decode_mb(AVCodecContext *avctx, int x, int y, int w, int h,
+                     const MVec *predVec)
+{
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    int ret = 0;
+
+    int mode = get_ue_golomb_31(gb);
+    if (s->version == VX_VERSION_OLD)
+        mode = ff_actimagine_vx_old_mb_mode_remap_tab[mode];
+
+    switch (mode) {
+    case 0:// v-split, no residu
+        if (w == 2) {
+            av_log(avctx, AV_LOG_ERROR, "invalid macroblock mode\n");
+            return AVERROR_INVALIDDATA;
+        }
+        if ((ret = decode_mb(avctx, x, y, w >> 1, h, predVec)) < 0)
+            return ret;
+        if ((ret = decode_mb(avctx, x + (w >> 1), y, w >> 1, h, predVec)) < 0)
+            return ret;
+        if (w == 8 && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 1:// no delta, no residu, ref 0
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 0, 0)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 2:// h-split, no residu
+        if (h == 2) {
+            av_log(avctx, AV_LOG_ERROR, "invalid macroblock mode\n");
+            return AVERROR_INVALIDDATA;
+        }
+        if ((ret = decode_mb(avctx, x, y, w, h >> 1, predVec)) < 0)
+            return ret;
+        if ((ret = decode_mb(avctx, x, y + (h >> 1), w, h >> 1, predVec)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && h == 8)
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 3:// unpredicted delta ref0 + dc offset, no residu
+        if ((ret = predict_inter_dc(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 4:// delta, no residu, ref 0
+    case 5:// delta, no residu, ref 1
+    case 6:// delta, no residu, ref 2
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 1, mode - 4)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 7:// plane, no residu
+        if ((ret = predict_mb_plane(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 8:// v-split, residu
+        if (w == 2) {
+            av_log(avctx, AV_LOG_ERROR, "invalid macroblock mode\n");
+            return AVERROR_INVALIDDATA;
+        }
+        if ((ret = decode_mb(avctx, x, y, w >> 1, h, predVec)) < 0)
+            return ret;
+        if ((ret = decode_mb(avctx, x + (w >> 1), y, w >> 1, h, predVec)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 9:// no delta, no residu, ref 1
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 0, 1)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 10:// unpredicted delta ref0 + dc offset, residu
+        if ((ret = predict_inter_dc(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 11:// predict notile, no residu
+        if ((ret = predict_notile(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 12:// no delta, residu, ref 0
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 0, 0)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 13:// h-split, residu
+        if (h == 2) {
+            av_log(avctx, AV_LOG_ERROR, "invalid macroblock mode\n");
+            return AVERROR_INVALIDDATA;
+        }
+        if ((ret = decode_mb(avctx, x, y, w, h >> 1, predVec)) < 0)
+            return ret;
+        if ((ret = decode_mb(avctx, x, y + (h >> 1), w, h >> 1, predVec)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 14:// no delta, no residu, ref 2
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 0, 2)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 15:// predict4, no residu
+        if ((ret = predict4(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((w == 8 || w == 16) && (h == 8 || h == 16))
+            clear_total_coeff(avctx, x, y, w, h);
+        break;
+    case 16:// delta, residu, ref 0
+    case 17:// delta, residu, ref 1
+    case 18:// delta, residu, ref 2
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 1, mode - 16)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 19:// predict4, residu
+        if ((ret = predict4(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 20:// no delta, residu, ref 1
+    case 21:// no delta, residu, ref 2
+        if ((ret = predict_inter(avctx, x, y, w, h, predVec, 0, mode - 20 + 1)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 22:// predict notile, residu
+        if ((ret = predict_notile(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    case 23:// plane, residu
+        if ((ret = predict_mb_plane(avctx, x, y, w, h)) < 0)
+            return ret;
+        if ((ret = decode_residu_blocks(avctx, x, y, w, h)) < 0)
+            return ret;
+        break;
+    default:
+        av_log(avctx, AV_LOG_ERROR, "invalid macroblock mode\n");
+        return AVERROR_INVALIDDATA;
+    }
+    return 0;
+}
+
+static int detect_format(AVCodecContext *avctx)
+{
+    // assume the new format, if any incorrect decisions are made for the
+    // first macroblock of a keyframe (ref, non-dc prediction) then it must be
+    // the old format
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    int w = 16;
+    int h = 16;
+    while (1) {
+        int mode = get_ue_golomb_31(gb);
+        if (mode == 0 || mode == 8) { // v-split
+            if (w == 2) // too many splits
+                return VX_VERSION_OLD;
+            w >>= 1;
+            continue;
+        } else if (mode == 2 || mode == 13) { // h-split
+            if (h == 2) // too many splits
+                return VX_VERSION_OLD;
+            h >>= 1;
+            continue;
+        } else if (mode == 11 || mode == 22) { // predict notile            
+            if (get_ue_golomb_31(gb) != 2) // mode_y != dc
+                return VX_VERSION_OLD;
+            if (get_ue_golomb_31(gb) != 0) // mode_uv != dc
+                return VX_VERSION_OLD;
+            break; //we should have enough evidence now
+        } else if (mode == 15 || mode == 19) { // predict4
+            // initial prediction is always dc
+            // we don't expect that prediction to be wrong
+            if (!get_bits1(gb))
+                return VX_VERSION_OLD;
+            break; //we should have enough evidence now
+        } else // inter prediction, plane or any other value
+            return VX_VERSION_OLD;
+    }
+    return VX_VERSION_NEW;
+}
+
+static int convert_colorspace(AVCodecContext *avctx)
+{
+    int ret;
+    VxContext *s = avctx->priv_data;
+    if ((ret = ff_reget_buffer(avctx, s->out_frame, 0)) < 0)
+        return ret;
+    
+    // the color space is not supported by ffmpeg
+    // r = y + 2v
+    // g = y - 0.5u - v
+    // b = y + 2u
+    // convert to the SMPTE170M format
+
+    for (int y = 0; y < avctx->height; y++) {
+        for (int x = 0; x < avctx->width; x++) {
+            s->out_frame->data[0][y * s->out_frame->linesize[0] + x]
+                = av_clip_uint8(PIXEL_CUR(s, 0, x, y) + ((
+                    (PIXEL_CUR(s, 1, x >> 1, y >> 1) - 128) * -68682 +
+                    (PIXEL_CUR(s, 2, x >> 1, y >> 1) - 128) * 11534 +                    
+                    (1 << 19)) >> 20));
+        }
+    }
+
+    for (int y = 0; y < avctx->height >> 1; y++) {
+        for (int x = 0; x < avctx->width >> 1; x++) {
+            s->out_frame->data[1][y * s->out_frame->linesize[1] + x]
+                = av_clip_uint8(128 + ((
+                    (PIXEL_CUR(s, 1, x, y) - 128) * 1222325 +
+                    (PIXEL_CUR(s, 2, x, y) - 128) * -4928) >> 20));
+            s->out_frame->data[2][y * s->out_frame->linesize[2] + x]
+                = av_clip_uint8(128 + ((
+                    (PIXEL_CUR(s, 1, x, y) - 128) * 49073 +
+                    (PIXEL_CUR(s, 2, x, y) - 128) * 1487615) >> 20));
+        }
+    }
+
+    return 0;
+}
+
+static int vx_decode(AVCodecContext *avctx, void *data, int *got_frame,
+                     AVPacket *pkt)
+{
+    MVec *vectors;
+    int ret;
+    VxContext *s = avctx->priv_data;
+    GetBitContext *gb = &s->gb;
+    AVFrame *frame = s->cur_frame;
+
+    // in avi files the frames start with a 32 bit number that seems to
+    // indicate the total number of bits
+    int offset = s->avi ? 4 : 0;
+
+    av_fast_padded_malloc(&s->bitstream, &s->bitstream_size,
+                          pkt->size);
+
+    if ((ret = ff_reget_buffer(avctx, frame, 0)) < 0)
+        return ret;
+
+    s->bdsp.bswap16_buf((uint16_t *)s->bitstream, (uint16_t *)pkt->data,
+                        (pkt->size + 1) >> 1);
+
+    ret = init_get_bits8(gb, s->bitstream + offset,
+                         FFALIGN(pkt->size - offset, 2));
+    if (ret < 0)
+        return ret;
+
+    // determine the bitstream version if this was not done yet
+    if (s->version == VX_VERSION_INVALID) {
+        if (s->ref_frame_count != 0) {
+            av_log(avctx, AV_LOG_ERROR, "can't determine version on p frame\n");
+            return AVERROR_INVALIDDATA;
+        }
+        s->version = detect_format(avctx);
+
+        // reinit bitreader
+        ret = init_get_bits8(gb, s->bitstream + offset,
+                             FFALIGN(pkt->size - offset, 2));
+        if (ret < 0)
+            return ret;
+    }
+
+    vectors = s->vectors + s->vectors_stride + 1;
+
+    // initially we assume this is an i frame
+    // this is corrected when a p block is found
+    frame->pict_type = AV_PICTURE_TYPE_I;
+    frame->key_frame = 1;
+
+    if (s->quantizer == -1) {
+        av_log(avctx, AV_LOG_ERROR, "no quantizer configured\n");
+        return AVERROR_INVALIDDATA;
+    }
+
+    for (int y = 0; y < avctx->height; y += 16) {
+        MVec *vec_cur = vectors;
+        for (int x = 0; x < avctx->width; x += 16) {
+            MVec predVec;
+            vec_cur[0].x = 0;
+            vec_cur[0].y = 0;
+            predVec.x = mid_pred(vec_cur[-1].x, vec_cur[-s->vectors_stride].x,
+                                 vec_cur[-s->vectors_stride + 1].x);
+            predVec.y = mid_pred(vec_cur[-1].y, vec_cur[-s->vectors_stride].y,
+                                 vec_cur[-s->vectors_stride + 1].y);
+            if ((ret = decode_mb(avctx, x, y, 16, 16, &predVec)) < 0)
+                return ret;
+            vec_cur++;
+        }
+        vectors += s->vectors_stride;
+    }
+
+    if ((ret = convert_colorspace(avctx)) < 0)
+        return ret;
+
+    s->cur_frame = s->ref_frames[2];
+    s->ref_frames[2] = s->ref_frames[1];
+    s->ref_frames[1] = s->ref_frames[0];
+    s->ref_frames[0] = frame;
+
+    if (s->ref_frame_count < 3)
+        s->ref_frame_count++;
+
+    if ((ret = av_frame_ref(data, s->out_frame)) < 0)
+        return ret;
+
+    *got_frame = 1;
+
+    return 0;
+}
+
+static void vx_flush(AVCodecContext *avctx)
+{
+    VxContext *s = avctx->priv_data;
+
+    for (int i = 0; i < 3; i++)
+        av_frame_unref(s->ref_frames[i]);
+
+    s->ref_frame_count = 0;
+}
+
+static av_cold int vx_close(AVCodecContext *avctx)
+{
+    VxContext *s = avctx->priv_data;
+
+    av_freep(&s->vectors);
+    s->vectors_stride = 0;
+    av_freep(&s->total_coeff_y);
+    s->total_coeff_y_stride = 0;
+    av_freep(&s->total_coeff_uv);
+    s->total_coeff_uv_stride = 0;
+
+    av_freep(&s->bitstream);
+    s->bitstream_size = 0;
+
+    for (int i = 0; i < 3; i++)
+        av_frame_free(&s->ref_frames[i]);
+    av_frame_free(&s->cur_frame);
+    av_frame_free(&s->out_frame);
+
+    return 0;
+}
+
+AVCodec ff_actimagine_vx_decoder = {
+    .name           = "actimagine_vx",
+    .long_name      = NULL_IF_CONFIG_SMALL("Actimagine VX Video"),
+    .type           = AVMEDIA_TYPE_VIDEO,
+    .id             = AV_CODEC_ID_ACTIMAGINE_VX,
+    .priv_data_size = sizeof(VxContext),
+    .init           = vx_init,
+    .decode         = vx_decode,
+    .flush          = vx_flush,
+    .close          = vx_close,
+    .capabilities   = AV_CODEC_CAP_DR1,
+    .caps_internal  = FF_CODEC_CAP_INIT_THREADSAFE | FF_CODEC_CAP_INIT_CLEANUP,
+};
diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c
index 2e9a3581de..9762521b11 100644
--- a/libavcodec/allcodecs.c
+++ b/libavcodec/allcodecs.c
@@ -32,6 +32,7 @@ 
 extern AVCodec ff_a64multi_encoder;
 extern AVCodec ff_a64multi5_encoder;
 extern AVCodec ff_aasc_decoder;
+extern AVCodec ff_actimagine_vx_decoder;
 extern AVCodec ff_aic_decoder;
 extern AVCodec ff_alias_pix_encoder;
 extern AVCodec ff_alias_pix_decoder;
diff --git a/libavcodec/codec_desc.c b/libavcodec/codec_desc.c
index 17f8a14044..281da17942 100644
--- a/libavcodec/codec_desc.c
+++ b/libavcodec/codec_desc.c
@@ -1856,6 +1856,13 @@  static const AVCodecDescriptor codec_descriptors[] = {
         .long_name = NULL_IF_CONFIG_SMALL("Digital Pictures SGA Video"),
         .props     = AV_CODEC_PROP_LOSSY,
     },
+    {
+        .id        = AV_CODEC_ID_ACTIMAGINE_VX,
+        .type      = AVMEDIA_TYPE_VIDEO,
+        .name      = "actimagine_vx",
+        .long_name = NULL_IF_CONFIG_SMALL("Actimagine VX Video"),
+        .props     = AV_CODEC_PROP_LOSSY,
+    },
 
     /* various PCM "codecs" */
     {
diff --git a/libavcodec/codec_id.h b/libavcodec/codec_id.h
index ab7bc68ee2..172b96ef27 100644
--- a/libavcodec/codec_id.h
+++ b/libavcodec/codec_id.h
@@ -307,6 +307,7 @@  enum AVCodecID {
     AV_CODEC_ID_CRI,
     AV_CODEC_ID_SIMBIOSIS_IMX,
     AV_CODEC_ID_SGA_VIDEO,
+    AV_CODEC_ID_ACTIMAGINE_VX,
 
     /* various PCM "codecs" */
     AV_CODEC_ID_FIRST_AUDIO = 0x10000,     ///< A dummy id pointing at the start of audio codecs
diff --git a/libavcodec/h264_cavlc.c b/libavcodec/h264_cavlc.c
index 9f5f692331..5b74c81379 100644
--- a/libavcodec/h264_cavlc.c
+++ b/libavcodec/h264_cavlc.c
@@ -36,6 +36,7 @@ 
 #include "golomb.h"
 #include "mpegutils.h"
 #include "libavutil/avassert.h"
+#include "h264_cavlc_data.h"
 
 
 static const uint8_t golomb_to_inter_cbp_gray[16]={
@@ -86,104 +87,6 @@  static const uint8_t chroma422_dc_coeff_token_bits[4*9]={
   7,   5,  4, 4,
 };
 
-static const uint8_t coeff_token_len[4][4*17]={
-{
-     1, 0, 0, 0,
-     6, 2, 0, 0,     8, 6, 3, 0,     9, 8, 7, 5,    10, 9, 8, 6,
-    11,10, 9, 7,    13,11,10, 8,    13,13,11, 9,    13,13,13,10,
-    14,14,13,11,    14,14,14,13,    15,15,14,14,    15,15,15,14,
-    16,15,15,15,    16,16,16,15,    16,16,16,16,    16,16,16,16,
-},
-{
-     2, 0, 0, 0,
-     6, 2, 0, 0,     6, 5, 3, 0,     7, 6, 6, 4,     8, 6, 6, 4,
-     8, 7, 7, 5,     9, 8, 8, 6,    11, 9, 9, 6,    11,11,11, 7,
-    12,11,11, 9,    12,12,12,11,    12,12,12,11,    13,13,13,12,
-    13,13,13,13,    13,14,13,13,    14,14,14,13,    14,14,14,14,
-},
-{
-     4, 0, 0, 0,
-     6, 4, 0, 0,     6, 5, 4, 0,     6, 5, 5, 4,     7, 5, 5, 4,
-     7, 5, 5, 4,     7, 6, 6, 4,     7, 6, 6, 4,     8, 7, 7, 5,
-     8, 8, 7, 6,     9, 8, 8, 7,     9, 9, 8, 8,     9, 9, 9, 8,
-    10, 9, 9, 9,    10,10,10,10,    10,10,10,10,    10,10,10,10,
-},
-{
-     6, 0, 0, 0,
-     6, 6, 0, 0,     6, 6, 6, 0,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
-}
-};
-
-static const uint8_t coeff_token_bits[4][4*17]={
-{
-     1, 0, 0, 0,
-     5, 1, 0, 0,     7, 4, 1, 0,     7, 6, 5, 3,     7, 6, 5, 3,
-     7, 6, 5, 4,    15, 6, 5, 4,    11,14, 5, 4,     8,10,13, 4,
-    15,14, 9, 4,    11,10,13,12,    15,14, 9,12,    11,10,13, 8,
-    15, 1, 9,12,    11,14,13, 8,     7,10, 9,12,     4, 6, 5, 8,
-},
-{
-     3, 0, 0, 0,
-    11, 2, 0, 0,     7, 7, 3, 0,     7,10, 9, 5,     7, 6, 5, 4,
-     4, 6, 5, 6,     7, 6, 5, 8,    15, 6, 5, 4,    11,14,13, 4,
-    15,10, 9, 4,    11,14,13,12,     8,10, 9, 8,    15,14,13,12,
-    11,10, 9,12,     7,11, 6, 8,     9, 8,10, 1,     7, 6, 5, 4,
-},
-{
-    15, 0, 0, 0,
-    15,14, 0, 0,    11,15,13, 0,     8,12,14,12,    15,10,11,11,
-    11, 8, 9,10,     9,14,13, 9,     8,10, 9, 8,    15,14,13,13,
-    11,14,10,12,    15,10,13,12,    11,14, 9,12,     8,10,13, 8,
-    13, 7, 9,12,     9,12,11,10,     5, 8, 7, 6,     1, 4, 3, 2,
-},
-{
-     3, 0, 0, 0,
-     0, 1, 0, 0,     4, 5, 6, 0,     8, 9,10,11,    12,13,14,15,
-    16,17,18,19,    20,21,22,23,    24,25,26,27,    28,29,30,31,
-    32,33,34,35,    36,37,38,39,    40,41,42,43,    44,45,46,47,
-    48,49,50,51,    52,53,54,55,    56,57,58,59,    60,61,62,63,
-}
-};
-
-static const uint8_t total_zeros_len[16][16]= {
-    {1,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9},
-    {3,3,3,3,3,4,4,4,4,5,5,6,6,6,6},
-    {4,3,3,3,4,4,3,3,4,5,5,6,5,6},
-    {5,3,4,4,3,3,3,4,3,4,5,5,5},
-    {4,4,4,3,3,3,3,3,4,5,4,5},
-    {6,5,3,3,3,3,3,3,4,3,6},
-    {6,5,3,3,3,2,3,4,3,6},
-    {6,4,5,3,2,2,3,3,6},
-    {6,6,4,2,2,3,2,5},
-    {5,5,3,2,2,2,4},
-    {4,4,3,3,1,3},
-    {4,4,2,1,3},
-    {3,3,1,2},
-    {2,2,1},
-    {1,1},
-};
-
-static const uint8_t total_zeros_bits[16][16]= {
-    {1,3,2,3,2,3,2,3,2,3,2,3,2,3,2,1},
-    {7,6,5,4,3,5,4,3,2,3,2,3,2,1,0},
-    {5,7,6,5,4,3,4,3,2,3,2,1,1,0},
-    {3,7,5,4,6,5,4,3,3,2,2,1,0},
-    {5,4,3,7,6,5,4,3,2,1,1,0},
-    {1,1,7,6,5,4,3,2,1,1,0},
-    {1,1,5,4,3,3,2,1,1,0},
-    {1,1,1,3,3,2,2,1,0},
-    {1,0,1,3,2,1,1,1},
-    {1,0,1,3,2,1,1},
-    {0,1,1,2,1,3},
-    {0,1,1,1,1},
-    {0,1,1,1},
-    {0,1,1},
-    {0,1},
-};
-
 static const uint8_t chroma_dc_total_zeros_len[3][4]= {
     { 1, 2, 3, 3,},
     { 1, 2, 2, 0,},
@@ -216,30 +119,6 @@  static const uint8_t chroma422_dc_total_zeros_bits[7][8]= {
     { 0, 1 },
 };
 
-static const uint8_t run_len[7][16]={
-    {1,1},
-    {1,2,2},
-    {2,2,2,2},
-    {2,2,2,3,3},
-    {2,2,3,3,3,3},
-    {2,3,3,3,3,3,3},
-    {3,3,3,3,3,3,3,4,5,6,7,8,9,10,11},
-};
-
-static const uint8_t run_bits[7][16]={
-    {1,0},
-    {1,1,0},
-    {3,2,1,0},
-    {3,2,1,1,0},
-    {3,2,3,2,1,0},
-    {3,0,1,3,2,5,4},
-    {7,6,5,4,3,2,1,1,1,1,1,1,1,1,1},
-};
-
-static VLC coeff_token_vlc[4];
-static VLC_TYPE coeff_token_vlc_tables[520+332+280+256][2];
-static const int coeff_token_vlc_tables_size[4]={520,332,280,256};
-
 static VLC chroma_dc_coeff_token_vlc;
 static VLC_TYPE chroma_dc_coeff_token_vlc_table[256][2];
 static const int chroma_dc_coeff_token_vlc_table_size = 256;
@@ -248,10 +127,6 @@  static VLC chroma422_dc_coeff_token_vlc;
 static VLC_TYPE chroma422_dc_coeff_token_vlc_table[8192][2];
 static const int chroma422_dc_coeff_token_vlc_table_size = 8192;
 
-static VLC total_zeros_vlc[15+1];
-static VLC_TYPE total_zeros_vlc_tables[15][512][2];
-static const int total_zeros_vlc_tables_size = 512;
-
 static VLC chroma_dc_total_zeros_vlc[3+1];
 static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2];
 static const int chroma_dc_total_zeros_vlc_tables_size = 8;
@@ -260,25 +135,13 @@  static VLC chroma422_dc_total_zeros_vlc[7+1];
 static VLC_TYPE chroma422_dc_total_zeros_vlc_tables[7][32][2];
 static const int chroma422_dc_total_zeros_vlc_tables_size = 32;
 
-static VLC run_vlc[6+1];
-static VLC_TYPE run_vlc_tables[6][8][2];
-static const int run_vlc_tables_size = 8;
-
-static VLC run7_vlc;
-static VLC_TYPE run7_vlc_table[96][2];
-static const int run7_vlc_table_size = 96;
-
 #define LEVEL_TAB_BITS 8
 static int8_t cavlc_level_tab[7][1<<LEVEL_TAB_BITS][2];
 
 #define CHROMA_DC_COEFF_TOKEN_VLC_BITS 8
 #define CHROMA422_DC_COEFF_TOKEN_VLC_BITS 13
-#define COEFF_TOKEN_VLC_BITS           8
-#define TOTAL_ZEROS_VLC_BITS           9
 #define CHROMA_DC_TOTAL_ZEROS_VLC_BITS 3
 #define CHROMA422_DC_TOTAL_ZEROS_VLC_BITS 5
-#define RUN_VLC_BITS                   3
-#define RUN7_VLC_BITS                  6
 
 /**
  * Get the predicted number of non-zero coefficients.
@@ -326,7 +189,7 @@  static av_cold void init_cavlc_level_tab(void){
 
 av_cold void ff_h264_decode_init_vlc(void)
 {
-    int offset;
+    ff_h264_cavlc_data_init_vlc();
 
     chroma_dc_coeff_token_vlc.table = chroma_dc_coeff_token_vlc_table;
     chroma_dc_coeff_token_vlc.table_allocated = chroma_dc_coeff_token_vlc_table_size;
@@ -342,23 +205,6 @@  av_cold void ff_h264_decode_init_vlc(void)
              &chroma422_dc_coeff_token_bits[0], 1, 1,
              INIT_VLC_USE_NEW_STATIC);
 
-    offset = 0;
-    for (int i = 0; i < 4; i++) {
-        coeff_token_vlc[i].table = coeff_token_vlc_tables + offset;
-        coeff_token_vlc[i].table_allocated = coeff_token_vlc_tables_size[i];
-        init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
-                 &coeff_token_len [i][0], 1, 1,
-                 &coeff_token_bits[i][0], 1, 1,
-                 INIT_VLC_USE_NEW_STATIC);
-        offset += coeff_token_vlc_tables_size[i];
-    }
-    /*
-     * This is a one time safety check to make sure that
-     * the packed static coeff_token_vlc table sizes
-     * were initialized correctly.
-     */
-    av_assert0(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
-
     for (int i = 0; i < 3; i++) {
         chroma_dc_total_zeros_vlc[i + 1].table = chroma_dc_total_zeros_vlc_tables[i];
         chroma_dc_total_zeros_vlc[i + 1].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
@@ -379,32 +225,6 @@  av_cold void ff_h264_decode_init_vlc(void)
                  INIT_VLC_USE_NEW_STATIC);
     }
 
-    for (int i = 0; i < 15; i++) {
-        total_zeros_vlc[i + 1].table = total_zeros_vlc_tables[i];
-        total_zeros_vlc[i + 1].table_allocated = total_zeros_vlc_tables_size;
-        init_vlc(&total_zeros_vlc[i + 1],
-                 TOTAL_ZEROS_VLC_BITS, 16,
-                 &total_zeros_len [i][0], 1, 1,
-                 &total_zeros_bits[i][0], 1, 1,
-                 INIT_VLC_USE_NEW_STATIC);
-    }
-
-    for (int i = 0; i < 6; i++) {
-        run_vlc[i + 1].table = run_vlc_tables[i];
-        run_vlc[i + 1].table_allocated = run_vlc_tables_size;
-        init_vlc(&run_vlc[i + 1],
-                 RUN_VLC_BITS, 7,
-                 &run_len [i][0], 1, 1,
-                 &run_bits[i][0], 1, 1,
-                 INIT_VLC_USE_NEW_STATIC);
-    }
-    run7_vlc.table = run7_vlc_table;
-    run7_vlc.table_allocated = run7_vlc_table_size;
-    init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
-             &run_len [6][0], 1, 1,
-             &run_bits[6][0], 1, 1,
-             INIT_VLC_USE_NEW_STATIC);
-
     init_cavlc_level_tab();
 }
 
@@ -436,7 +256,6 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
                            const uint8_t *scantable, const uint32_t *qmul,
                            int max_coeff)
 {
-    static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3};
     int level[16];
     int zeros_left, coeff_token, total_coeff, i, trailing_ones, run_before;
 
@@ -451,11 +270,13 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
     }else{
         if(n >= LUMA_DC_BLOCK_INDEX){
             total_coeff= pred_non_zero_count(h, sl, (n - LUMA_DC_BLOCK_INDEX)*16);
-            coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
+            coeff_token = get_vlc2(gb, ff_h264_cavlc_coeff_token_vlc[ff_h264_cavlc_coeff_token_table_index[total_coeff]].table,
+                                  FF_H264_CAVLC_COEFF_TOKEN_VLC_BITS, 2);
             total_coeff= coeff_token>>2;
         }else{
             total_coeff= pred_non_zero_count(h, sl, n);
-            coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
+            coeff_token = get_vlc2(gb, ff_h264_cavlc_coeff_token_vlc[ff_h264_cavlc_coeff_token_table_index[total_coeff]].table,
+                                  FF_H264_CAVLC_COEFF_TOKEN_VLC_BITS, 2);
             total_coeff= coeff_token>>2;
         }
     }
@@ -529,7 +350,6 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
 
         //remaining coefficients have suffix_length > 0
         for(i=trailing_ones+1;i<total_coeff;i++) {
-            static const unsigned int suffix_limit[7] = {0,3,6,12,24,48,INT_MAX };
             int bitsi= show_bits(gb, LEVEL_TAB_BITS);
             level_code= cavlc_level_tab[suffix_length][bitsi][0];
 
@@ -556,7 +376,7 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
                 level_code= (((2+level_code)>>1) ^ mask) - mask;
             }
             level[i]= level_code;
-            suffix_length+= suffix_limit[suffix_length] + level_code > 2U*suffix_limit[suffix_length];
+            suffix_length += ff_h264_cavlc_suffix_limit[suffix_length] + level_code > 2U * ff_h264_cavlc_suffix_limit[suffix_length];
         }
     }
 
@@ -571,7 +391,8 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
                 zeros_left = get_vlc2(gb, chroma422_dc_total_zeros_vlc[total_coeff].table,
                                       CHROMA422_DC_TOTAL_ZEROS_VLC_BITS, 1);
         } else {
-            zeros_left= get_vlc2(gb, total_zeros_vlc[ total_coeff ].table, TOTAL_ZEROS_VLC_BITS, 1);
+            zeros_left = get_vlc2(gb, ff_h264_cavlc_total_zeros_vlc[total_coeff].table,
+                                  FF_H264_CAVLC_TOTAL_ZEROS_VLC_BITS, 1);
         }
     }
 
@@ -581,9 +402,9 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
         ((type*)block)[*scantable] = level[0]; \
         for(i=1;i<total_coeff && zeros_left > 0;i++) { \
             if(zeros_left < 7) \
-                run_before= get_vlc2(gb, run_vlc[zeros_left].table, RUN_VLC_BITS, 1); \
+                run_before = get_vlc2(gb, ff_h264_cavlc_run_vlc[zeros_left].table, FF_H264_CAVLC_RUN_VLC_BITS, 1); \
             else \
-                run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
+                run_before = get_vlc2(gb, ff_h264_cavlc_run7_vlc.table, FF_H264_CAVLC_RUN7_VLC_BITS, 2); \
             zeros_left -= run_before; \
             scantable -= 1 + run_before; \
             ((type*)block)[*scantable]= level[i]; \
@@ -596,9 +417,9 @@  static int decode_residual(const H264Context *h, H264SliceContext *sl,
         ((type*)block)[*scantable] = ((int)(level[0] * qmul[*scantable] + 32))>>6; \
         for(i=1;i<total_coeff && zeros_left > 0;i++) { \
             if(zeros_left < 7) \
-                run_before= get_vlc2(gb, run_vlc[zeros_left].table, RUN_VLC_BITS, 1); \
+                run_before = get_vlc2(gb, ff_h264_cavlc_run_vlc[zeros_left].table, FF_H264_CAVLC_RUN_VLC_BITS, 1); \
             else \
-                run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2); \
+                run_before = get_vlc2(gb, ff_h264_cavlc_run7_vlc.table, FF_H264_CAVLC_RUN7_VLC_BITS, 2); \
             zeros_left -= run_before; \
             scantable -= 1 + run_before; \
             ((type*)block)[*scantable]= ((int)(level[i] * qmul[*scantable] + 32))>>6; \
diff --git a/libavcodec/h264_cavlc_data.c b/libavcodec/h264_cavlc_data.c
new file mode 100644
index 0000000000..0f2e95ae4c
--- /dev/null
+++ b/libavcodec/h264_cavlc_data.c
@@ -0,0 +1,231 @@ 
+/*
+ * H.264 cavlc tables
+ * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
+ *
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "internal.h"
+#include "libavutil/thread.h"
+#include "libavutil/avassert.h"
+#include "h264_cavlc_data.h"
+
+static const uint8_t coeff_token_len[4][4 * 17] = {
+    {
+         1, 0, 0, 0,
+         6, 2, 0, 0,     8, 6, 3, 0,     9, 8, 7, 5,    10, 9, 8, 6,
+        11,10, 9, 7,    13,11,10, 8,    13,13,11, 9,    13,13,13,10,
+        14,14,13,11,    14,14,14,13,    15,15,14,14,    15,15,15,14,
+        16,15,15,15,    16,16,16,15,    16,16,16,16,    16,16,16,16
+    },
+    {
+         2, 0, 0, 0,
+         6, 2, 0, 0,     6, 5, 3, 0,     7, 6, 6, 4,     8, 6, 6, 4,
+         8, 7, 7, 5,     9, 8, 8, 6,    11, 9, 9, 6,    11,11,11, 7,
+        12,11,11, 9,    12,12,12,11,    12,12,12,11,    13,13,13,12,
+        13,13,13,13,    13,14,13,13,    14,14,14,13,    14,14,14,14,
+    },
+    {
+         4, 0, 0, 0,
+         6, 4, 0, 0,     6, 5, 4, 0,     6, 5, 5, 4,     7, 5, 5, 4,
+         7, 5, 5, 4,     7, 6, 6, 4,     7, 6, 6, 4,     8, 7, 7, 5,
+         8, 8, 7, 6,     9, 8, 8, 7,     9, 9, 8, 8,     9, 9, 9, 8,
+        10, 9, 9, 9,    10,10,10,10,    10,10,10,10,    10,10,10,10,
+    },
+    {
+         6, 0, 0, 0,
+         6, 6, 0, 0,     6, 6, 6, 0,     6, 6, 6, 6,     6, 6, 6, 6,
+         6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
+         6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
+         6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,     6, 6, 6, 6,
+    }
+};
+
+static const uint8_t coeff_token_bits[4][4 * 17] = {
+    {
+         1, 0, 0, 0,
+         5, 1, 0, 0,     7, 4, 1, 0,     7, 6, 5, 3,     7, 6, 5, 3,
+         7, 6, 5, 4,    15, 6, 5, 4,    11,14, 5, 4,     8,10,13, 4,
+        15,14, 9, 4,    11,10,13,12,    15,14, 9,12,    11,10,13, 8,
+        15, 1, 9,12,    11,14,13, 8,     7,10, 9,12,     4, 6, 5, 8,
+    },
+    {
+         3, 0, 0, 0,
+        11, 2, 0, 0,     7, 7, 3, 0,     7,10, 9, 5,     7, 6, 5, 4,
+         4, 6, 5, 6,     7, 6, 5, 8,    15, 6, 5, 4,    11,14,13, 4,
+        15,10, 9, 4,    11,14,13,12,     8,10, 9, 8,    15,14,13,12,
+        11,10, 9,12,     7,11, 6, 8,     9, 8,10, 1,     7, 6, 5, 4,
+    },
+    {
+        15, 0, 0, 0,
+        15,14, 0, 0,    11,15,13, 0,     8,12,14,12,    15,10,11,11,
+        11, 8, 9,10,     9,14,13, 9,     8,10, 9, 8,    15,14,13,13,
+        11,14,10,12,    15,10,13,12,    11,14, 9,12,     8,10,13, 8,
+        13, 7, 9,12,     9,12,11,10,     5, 8, 7, 6,     1, 4, 3, 2,
+    },
+    {
+         3, 0, 0, 0,
+         0, 1, 0, 0,     4, 5, 6, 0,     8, 9,10,11,    12,13,14,15,
+        16,17,18,19,    20,21,22,23,    24,25,26,27,    28,29,30,31,
+        32,33,34,35,    36,37,38,39,    40,41,42,43,    44,45,46,47,
+        48,49,50,51,    52,53,54,55,    56,57,58,59,    60,61,62,63,
+    }
+};
+
+const uint8_t ff_h264_cavlc_coeff_token_table_index[17] = {
+    0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3
+};
+
+static const uint8_t total_zeros_len[16][16] = {
+    {1,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9},
+    {3,3,3,3,3,4,4,4,4,5,5,6,6,6,6},
+    {4,3,3,3,4,4,3,3,4,5,5,6,5,6},
+    {5,3,4,4,3,3,3,4,3,4,5,5,5},
+    {4,4,4,3,3,3,3,3,4,5,4,5},
+    {6,5,3,3,3,3,3,3,4,3,6},
+    {6,5,3,3,3,2,3,4,3,6},
+    {6,4,5,3,2,2,3,3,6},
+    {6,6,4,2,2,3,2,5},
+    {5,5,3,2,2,2,4},
+    {4,4,3,3,1,3},
+    {4,4,2,1,3},
+    {3,3,1,2},
+    {2,2,1},
+    {1,1},
+};
+
+static const uint8_t total_zeros_bits[16][16] = {
+    {1,3,2,3,2,3,2,3,2,3,2,3,2,3,2,1},
+    {7,6,5,4,3,5,4,3,2,3,2,3,2,1,0},
+    {5,7,6,5,4,3,4,3,2,3,2,1,1,0},
+    {3,7,5,4,6,5,4,3,3,2,2,1,0},
+    {5,4,3,7,6,5,4,3,2,1,1,0},
+    {1,1,7,6,5,4,3,2,1,1,0},
+    {1,1,5,4,3,3,2,1,1,0},
+    {1,1,1,3,3,2,2,1,0},
+    {1,0,1,3,2,1,1,1},
+    {1,0,1,3,2,1,1},
+    {0,1,1,2,1,3},
+    {0,1,1,1,1},
+    {0,1,1,1},
+    {0,1,1},
+    {0,1},
+};
+
+static const uint8_t run_len[7][16] = {
+    {1,1},
+    {1,2,2},
+    {2,2,2,2},
+    {2,2,2,3,3},
+    {2,2,3,3,3,3},
+    {2,3,3,3,3,3,3},
+    {3,3,3,3,3,3,3,4,5,6,7,8,9,10,11},
+};
+
+static const uint8_t run_bits[7][16] = {
+    {1,0},
+    {1,1,0},
+    {3,2,1,0},
+    {3,2,1,1,0},
+    {3,2,3,2,1,0},
+    {3,0,1,3,2,5,4},
+    {7,6,5,4,3,2,1,1,1,1,1,1,1,1,1},
+};
+
+const unsigned int ff_h264_cavlc_suffix_limit[7] = {
+    0, 3, 6, 12, 24, 48, INT_MAX
+};
+
+VLC ff_h264_cavlc_coeff_token_vlc[4];
+static VLC_TYPE coeff_token_vlc_tables[520 + 332 + 280 + 256][2];
+static const int coeff_token_vlc_tables_size[4] = { 520, 332, 280, 256 };
+
+VLC ff_h264_cavlc_total_zeros_vlc[15+1];
+static VLC_TYPE total_zeros_vlc_tables[15][512][2];
+static const int total_zeros_vlc_tables_size = 512;
+
+VLC ff_h264_cavlc_run_vlc[6 + 1];
+static VLC_TYPE run_vlc_tables[6][8][2];
+static const int run_vlc_tables_size = 8;
+
+VLC ff_h264_cavlc_run7_vlc;
+static VLC_TYPE run7_vlc_table[96][2];
+static const int run7_vlc_table_size = 96;
+
+static av_cold void cavlc_init_vlc(void)
+{
+    int offset = 0;
+    for (int i = 0; i < 4; i++) {
+        ff_h264_cavlc_coeff_token_vlc[i].table = coeff_token_vlc_tables + offset;
+        ff_h264_cavlc_coeff_token_vlc[i].table_allocated = coeff_token_vlc_tables_size[i];
+        init_vlc(&ff_h264_cavlc_coeff_token_vlc[i], FF_H264_CAVLC_COEFF_TOKEN_VLC_BITS, 4 * 17,
+                 &coeff_token_len [i][0], 1, 1,
+                 &coeff_token_bits[i][0], 1, 1,
+                 INIT_VLC_USE_NEW_STATIC);
+        offset += coeff_token_vlc_tables_size[i];
+    }
+    /*
+     * This is a one time safety check to make sure that
+     * the packed static coeff_token_vlc table sizes
+     * were initialized correctly.
+     */
+    av_assert0(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
+
+    for (int i = 0; i < 15; i++) {
+        ff_h264_cavlc_total_zeros_vlc[i + 1].table = total_zeros_vlc_tables[i];
+        ff_h264_cavlc_total_zeros_vlc[i + 1].table_allocated = total_zeros_vlc_tables_size;
+        init_vlc(&ff_h264_cavlc_total_zeros_vlc[i + 1],
+                 FF_H264_CAVLC_TOTAL_ZEROS_VLC_BITS, 16,
+                 &total_zeros_len [i][0], 1, 1,
+                 &total_zeros_bits[i][0], 1, 1,
+                 INIT_VLC_USE_NEW_STATIC);
+    }
+
+    for (int i = 0; i < 6; i++) {
+        ff_h264_cavlc_run_vlc[i + 1].table = run_vlc_tables[i];
+        ff_h264_cavlc_run_vlc[i + 1].table_allocated = run_vlc_tables_size;
+        init_vlc(&ff_h264_cavlc_run_vlc[i + 1],
+                 FF_H264_CAVLC_RUN_VLC_BITS, 7,
+                 &run_len [i][0], 1, 1,
+                 &run_bits[i][0], 1, 1,
+                 INIT_VLC_USE_NEW_STATIC);
+    }
+
+    ff_h264_cavlc_run7_vlc.table = run7_vlc_table,
+    ff_h264_cavlc_run7_vlc.table_allocated = run7_vlc_table_size;
+    init_vlc(&ff_h264_cavlc_run7_vlc, FF_H264_CAVLC_RUN7_VLC_BITS, 16,
+             &run_len [6][0], 1, 1,
+             &run_bits[6][0], 1, 1,
+             INIT_VLC_USE_NEW_STATIC);
+}
+
+int ff_h264_cavlc_data_init_vlc(void)
+{
+    // at the current time this function is only called by
+    // - vx_init (actimagine_vxdec.c)
+    // - ff_h264_decode_init_vlc (h264_cavlc.c)
+    // since ff_h264_decode_init_vlc is already called using ff_thread_once
+    // we only need to use ff_thread_once here if the actimagine_vx decoder
+    // is enabled as well
+#if CONFIG_ACTIMAGINE_VX_DECODER
+    static AVOnce vlc_init = AV_ONCE_INIT;
+    return ff_thread_once(&vlc_init, cavlc_init_vlc);
+#else
+    cavlc_init_vlc();
+    return 0;
+#endif
+}
\ No newline at end of file
diff --git a/libavcodec/h264_cavlc_data.h b/libavcodec/h264_cavlc_data.h
new file mode 100644
index 0000000000..170da67360
--- /dev/null
+++ b/libavcodec/h264_cavlc_data.h
@@ -0,0 +1,43 @@ 
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg 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.
+ *
+ * FFmpeg 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 FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_H264_CAVLC_DATA_H
+#define AVCODEC_H264_CAVLC_DATA_H
+
+#include <stdint.h>
+#include "vlc.h"
+
+#define FF_H264_CAVLC_COEFF_TOKEN_VLC_BITS    8
+#define FF_H264_CAVLC_TOTAL_ZEROS_VLC_BITS    9
+#define FF_H264_CAVLC_RUN_VLC_BITS            3
+#define FF_H264_CAVLC_RUN7_VLC_BITS           6
+
+extern const uint8_t ff_h264_cavlc_coeff_token_table_index[17];
+
+extern const unsigned int ff_h264_cavlc_suffix_limit[7];
+
+extern VLC ff_h264_cavlc_coeff_token_vlc[4];
+extern VLC ff_h264_cavlc_total_zeros_vlc[15+1];
+extern VLC ff_h264_cavlc_run_vlc[6+1];
+extern VLC ff_h264_cavlc_run7_vlc;
+
+// when using this function, make sure to modify the #if statement in the c
+// file to ensure a thread-safe one-time initialization
+int ff_h264_cavlc_data_init_vlc(void);
+
+#endif /* AVCODEC_H264_CAVLC_DATA_H */
diff --git a/libavcodec/version.h b/libavcodec/version.h
index 662caebc49..cfdde46960 100644
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@ -28,7 +28,7 @@ 
 #include "libavutil/version.h"
 
 #define LIBAVCODEC_VERSION_MAJOR  58
-#define LIBAVCODEC_VERSION_MINOR 133
+#define LIBAVCODEC_VERSION_MINOR 134
 #define LIBAVCODEC_VERSION_MICRO 100
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
diff --git a/libavformat/riff.c b/libavformat/riff.c
index 270ff7c024..848b1d6cfd 100644
--- a/libavformat/riff.c
+++ b/libavformat/riff.c
@@ -496,6 +496,8 @@  const AVCodecTag ff_codec_bmp_tags[] = {
     { AV_CODEC_ID_MVHA,         MKTAG('M', 'V', 'H', 'A') },
     { AV_CODEC_ID_MV30,         MKTAG('M', 'V', '3', '0') },
     { AV_CODEC_ID_NOTCHLC,      MKTAG('n', 'l', 'c', '1') },
+    { AV_CODEC_ID_ACTIMAGINE_VX,MKTAG('V', 'X', 'S', '1') },
+    { AV_CODEC_ID_ACTIMAGINE_VX,MKTAG('v', 'x', 's', '1') },
     { AV_CODEC_ID_NONE,         0 }
 };