lavc/vp9: split into vp9{block,data,mvs}

This is following Libav layout to ease merges.

libavcodec/Makefile

@@ -610,8 +610,9 @@ OBJS-$(CONFIG_VP8_DECODER)             += vp8.o vp56rac.o
 OBJS-$(CONFIG_VP8_CUVID_DECODER)       += cuvid.o
 OBJS-$(CONFIG_VP8_MEDIACODEC_DECODER)  += mediacodecdec.o
 OBJS-$(CONFIG_VP8_VAAPI_ENCODER)       += vaapi_encode_vp8.o
-OBJS-$(CONFIG_VP9_DECODER)             += vp9.o vp9dsp.o vp56rac.o vp9dsp_8bpp.o \
-                                          vp9dsp_10bpp.o vp9dsp_12bpp.o
+OBJS-$(CONFIG_VP9_DECODER)             += vp9.o vp9data.o vp9dsp.o \
+                                          vp9block.o vp9prob.o vp9mvs.o vp56rac.o \
+                                          vp9dsp_8bpp.o vp9dsp_10bpp.o vp9dsp_12bpp.o
 OBJS-$(CONFIG_VP9_CUVID_DECODER)       += cuvid.o
 OBJS-$(CONFIG_VP9_MEDIACODEC_DECODER)  += mediacodecdec.o
 OBJS-$(CONFIG_VPLAYER_DECODER)         += textdec.o ass.o

libavcodec/aarch64/vp9dsp_init.h

@@ -21,7 +21,7 @@
 #ifndef AVCODEC_AARCH64_VP9DSP_INIT_H
 #define AVCODEC_AARCH64_VP9DSP_INIT_H
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 
 void ff_vp9dsp_init_10bpp_aarch64(VP9DSPContext *dsp);
 void ff_vp9dsp_init_12bpp_aarch64(VP9DSPContext *dsp);

libavcodec/aarch64/vp9dsp_init_aarch64.c

@@ -22,7 +22,7 @@
 
 #include "libavutil/attributes.h"
 #include "libavutil/aarch64/cpu.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "vp9dsp_init.h"
 
 #define declare_fpel(type, sz)                                          \

libavcodec/arm/vp9dsp_init.h

@@ -21,7 +21,7 @@
 #ifndef AVCODEC_ARM_VP9DSP_INIT_H
 #define AVCODEC_ARM_VP9DSP_INIT_H
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 
 void ff_vp9dsp_init_10bpp_arm(VP9DSPContext *dsp);
 void ff_vp9dsp_init_12bpp_arm(VP9DSPContext *dsp);

libavcodec/arm/vp9dsp_init_arm.c

@@ -22,7 +22,7 @@
 
 #include "libavutil/attributes.h"
 #include "libavutil/arm/cpu.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "vp9dsp_init.h"
 
 #define declare_fpel(type, sz)                                          \

libavcodec/mips/vp9_idct_msa.c

@@ -19,7 +19,7 @@
  */
 
 #include <string.h>
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavutil/mips/generic_macros_msa.h"
 #include "vp9dsp_mips.h"
 

libavcodec/mips/vp9_intra_msa.c

@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavutil/mips/generic_macros_msa.h"
 #include "vp9dsp_mips.h"
 

libavcodec/mips/vp9_lpf_msa.c

@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavutil/mips/generic_macros_msa.h"
 #include "vp9dsp_mips.h"
 

libavcodec/mips/vp9_mc_msa.c

@@ -18,7 +18,7 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavutil/mips/generic_macros_msa.h"
 #include "vp9dsp_mips.h"
 

libavcodec/mips/vp9dsp_init_mips.c

@@ -20,7 +20,7 @@
 
 #include "config.h"
 #include "libavutil/common.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "vp9dsp_mips.h"
 
 #if HAVE_MSA

libavcodec/vp9dsp.c

@@ -23,7 +23,7 @@
 
 #include "libavutil/avassert.h"
 #include "libavutil/common.h"
-#include "vp9dsp.h"
+#include "vp9.h"
 
 const DECLARE_ALIGNED(16, int16_t, ff_vp9_subpel_filters)[3][16][8] = {
     [FILTER_8TAP_REGULAR] = {

libavcodec/vp9dsp.h

-/*
- * VP9 compatible video decoder
- *
- * Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
- * Copyright (C) 2013 Clément Bœsch <u pkh me>
- *
- * 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_VP9DSP_H
-#define AVCODEC_VP9DSP_H
-
-#include <stddef.h>
-#include <stdint.h>
-
-#include "vp9.h"
-
-typedef void (*vp9_mc_func)(uint8_t *dst, ptrdiff_t dst_stride,
-                            const uint8_t *ref, ptrdiff_t ref_stride,
-                            int h, int mx, int my);
-typedef void (*vp9_scaled_mc_func)(uint8_t *dst, ptrdiff_t dst_stride,
-                                   const uint8_t *ref, ptrdiff_t ref_stride,
-                                   int h, int mx, int my, int dx, int dy);
-
-typedef struct VP9DSPContext {
-    /*
-     * dimension 1: 0=4x4, 1=8x8, 2=16x16, 3=32x32
-     * dimension 2: intra prediction modes
-     *
-     * dst/left/top is aligned by transform-size (i.e. 4, 8, 16 or 32 pixels)
-     * stride is aligned by 16 pixels
-     * top[-1] is top/left; top[4,7] is top-right for 4x4
-     */
-    // FIXME(rbultje) maybe replace left/top pointers with HAVE_TOP/
-    // HAVE_LEFT/HAVE_TOPRIGHT flags instead, and then handle it in-place?
-    // also needs to fit in with what H.264/VP8/etc do
-    void (*intra_pred[N_TXFM_SIZES][N_INTRA_PRED_MODES])(uint8_t *dst,
-                                                         ptrdiff_t stride,
-                                                         const uint8_t *left,
-                                                         const uint8_t *top);
-
-    /*
-     * dimension 1: 0=4x4, 1=8x8, 2=16x16, 3=32x32, 4=lossless (3-4=dct only)
-     * dimension 2: 0=dct/dct, 1=dct/adst, 2=adst/dct, 3=adst/adst
-     *
-     * dst is aligned by transform-size (i.e. 4, 8, 16 or 32 pixels)
-     * stride is aligned by 16 pixels
-     * block is 16-byte aligned
-     * eob indicates the position (+1) of the last non-zero coefficient,
-     * in scan-order. This can be used to write faster versions, e.g. a
-     * dc-only 4x4/8x8/16x16/32x32, or a 4x4-only (eob<10) 8x8/16x16/32x32,
-     * etc.
-     */
-    // FIXME also write idct_add_block() versions for whole (inter) pred
-    // blocks, so we can do 2 4x4s at once
-    void (*itxfm_add[N_TXFM_SIZES + 1][N_TXFM_TYPES])(uint8_t *dst,
-                                                      ptrdiff_t stride,
-                                                      int16_t *block, int eob);
-
-    /*
-     * dimension 1: width of filter (0=4, 1=8, 2=16)
-     * dimension 2: 0=col-edge filter (h), 1=row-edge filter (v)
-     *
-     * dst/stride are aligned by 8
-     */
-    void (*loop_filter_8[3][2])(uint8_t *dst, ptrdiff_t stride,
-                                int mb_lim, int lim, int hev_thr);
-
-    /*
-     * dimension 1: 0=col-edge filter (h), 1=row-edge filter (v)
-     *
-     * The width of filter is assumed to be 16; dst/stride are aligned by 16
-     */
-    void (*loop_filter_16[2])(uint8_t *dst, ptrdiff_t stride,
-                              int mb_lim, int lim, int hev_thr);
-
-    /*
-     * dimension 1/2: width of filter (0=4, 1=8) for each filter half
-     * dimension 3: 0=col-edge filter (h), 1=row-edge filter (v)
-     *
-     * dst/stride are aligned by operation size
-     * this basically calls loop_filter[d1][d3][0](), followed by
-     * loop_filter[d2][d3][0]() on the next 8 pixels
-     * mb_lim/lim/hev_thr contain two values in the lowest two bytes of the
-     * integer.
-     */
-    // FIXME perhaps a mix4 that operates on 32px (for AVX2)
-    void (*loop_filter_mix2[2][2][2])(uint8_t *dst, ptrdiff_t stride,
-                                      int mb_lim, int lim, int hev_thr);
-
-    /*
-     * dimension 1: hsize (0: 64, 1: 32, 2: 16, 3: 8, 4: 4)
-     * dimension 2: filter type (0: smooth, 1: regular, 2: sharp, 3: bilin)
-     * dimension 3: averaging type (0: put, 1: avg)
-     * dimension 4: x subpel interpolation (0: none, 1: 8tap/bilin)
-     * dimension 5: y subpel interpolation (0: none, 1: 8tap/bilin)
-     *
-     * dst/stride are aligned by hsize
-     */
-    vp9_mc_func mc[5][4][2][2][2];
-
-    /*
-     * for scalable MC, first 3 dimensions identical to above, the other two
-     * don't exist since it changes per stepsize.
-     */
-    vp9_scaled_mc_func smc[5][4][2];
-} VP9DSPContext;
-
-
-extern const int16_t ff_vp9_subpel_filters[3][16][8];
-
-void ff_vp9dsp_init(VP9DSPContext *dsp, int bpp, int bitexact);
-
-void ff_vp9dsp_init_8(VP9DSPContext *dsp);
-void ff_vp9dsp_init_10(VP9DSPContext *dsp);
-void ff_vp9dsp_init_12(VP9DSPContext *dsp);
-
-void ff_vp9dsp_init_aarch64(VP9DSPContext *dsp, int bpp);
-void ff_vp9dsp_init_arm(VP9DSPContext *dsp, int bpp);
-void ff_vp9dsp_init_x86(VP9DSPContext *dsp, int bpp, int bitexact);
-void ff_vp9dsp_init_mips(VP9DSPContext *dsp, int bpp);
-
-#endif /* AVCODEC_VP9DSP_H */

libavcodec/vp9dsp_template.c

@@ -23,7 +23,7 @@
 
 #include "libavutil/common.h"
 #include "bit_depth_template.c"
-#include "vp9dsp.h"
+#include "vp9.h"
 
 #if BIT_DEPTH != 12
 

libavcodec/vp9prob.c

+/*
+ * VP9 compatible video decoder
+ *
+ * Copyright (C) 2013 Ronald S. Bultje <rsbultje gmail com>
+ * Copyright (C) 2013 Clément Bœsch <u pkh me>
+ *
+ * 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 "vp56.h"
+#include "vp9.h"
+#include "vp9data.h"
+
+static av_always_inline void adapt_prob(uint8_t *p, unsigned ct0, unsigned ct1,
+                                        int max_count, int update_factor)
+{
+    unsigned ct = ct0 + ct1, p2, p1;
+
+    if (!ct)
+        return;
+
+    update_factor = FASTDIV(update_factor * FFMIN(ct, max_count), max_count);
+    p1 = *p;
+    p2 = ((((int64_t) ct0) << 8) + (ct >> 1)) / ct;
+    p2 = av_clip(p2, 1, 255);
+
+    // (p1 * (256 - update_factor) + p2 * update_factor + 128) >> 8
+    *p = p1 + (((p2 - p1) * update_factor + 128) >> 8);
+}
+
+void ff_vp9_adapt_probs(VP9Context *s)
+{
+    int i, j, k, l, m;
+    ProbContext *p = &s->prob_ctx[s->s.h.framectxid].p;
+    int uf = (s->s.h.keyframe || s->s.h.intraonly || !s->last_keyframe) ? 112 : 128;
+
+    // coefficients
+    for (i = 0; i < 4; i++)
+        for (j = 0; j < 2; j++)
+            for (k = 0; k < 2; k++)
+                for (l = 0; l < 6; l++)
+                    for (m = 0; m < 6; m++) {
+                        uint8_t *pp = s->prob_ctx[s->s.h.framectxid].coef[i][j][k][l][m];
+                        unsigned *e = s->counts.eob[i][j][k][l][m];
+                        unsigned *c = s->counts.coef[i][j][k][l][m];
+
+                        if (l == 0 && m >= 3) // dc only has 3 pt
+                            break;
+
+                        adapt_prob(&pp[0], e[0], e[1], 24, uf);
+                        adapt_prob(&pp[1], c[0], c[1] + c[2], 24, uf);
+                        adapt_prob(&pp[2], c[1], c[2], 24, uf);
+                    }
+
+    if (s->s.h.keyframe || s->s.h.intraonly) {
+        memcpy(p->skip,  s->prob.p.skip,  sizeof(p->skip));
+        memcpy(p->tx32p, s->prob.p.tx32p, sizeof(p->tx32p));
+        memcpy(p->tx16p, s->prob.p.tx16p, sizeof(p->tx16p));
+        memcpy(p->tx8p,  s->prob.p.tx8p,  sizeof(p->tx8p));
+        return;
+    }
+
+    // skip flag
+    for (i = 0; i < 3; i++)
+        adapt_prob(&p->skip[i], s->counts.skip[i][0], s->counts.skip[i][1], 20, 128);
+
+    // intra/inter flag
+    for (i = 0; i < 4; i++)
+        adapt_prob(&p->intra[i], s->counts.intra[i][0], s->counts.intra[i][1], 20, 128);
+
+    // comppred flag
+    if (s->s.h.comppredmode == PRED_SWITCHABLE) {
+      for (i = 0; i < 5; i++)
+          adapt_prob(&p->comp[i], s->counts.comp[i][0], s->counts.comp[i][1], 20, 128);
+    }
+
+    // reference frames
+    if (s->s.h.comppredmode != PRED_SINGLEREF) {
+      for (i = 0; i < 5; i++)
+          adapt_prob(&p->comp_ref[i], s->counts.comp_ref[i][0],
+                     s->counts.comp_ref[i][1], 20, 128);
+    }
+
+    if (s->s.h.comppredmode != PRED_COMPREF) {
+      for (i = 0; i < 5; i++) {
+          uint8_t *pp = p->single_ref[i];
+          unsigned (*c)[2] = s->counts.single_ref[i];
+
+          adapt_prob(&pp[0], c[0][0], c[0][1], 20, 128);
+          adapt_prob(&pp[1], c[1][0], c[1][1], 20, 128);
+      }
+    }
+
+    // block partitioning
+    for (i = 0; i < 4; i++)
+        for (j = 0; j < 4; j++) {
+            uint8_t *pp = p->partition[i][j];
+            unsigned *c = s->counts.partition[i][j];
+
+            adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
+            adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
+            adapt_prob(&pp[2], c[2], c[3], 20, 128);
+        }
+
+    // tx size
+    if (s->s.h.txfmmode == TX_SWITCHABLE) {
+      for (i = 0; i < 2; i++) {
+          unsigned *c16 = s->counts.tx16p[i], *c32 = s->counts.tx32p[i];
+
+          adapt_prob(&p->tx8p[i], s->counts.tx8p[i][0], s->counts.tx8p[i][1], 20, 128);
+          adapt_prob(&p->tx16p[i][0], c16[0], c16[1] + c16[2], 20, 128);
+          adapt_prob(&p->tx16p[i][1], c16[1], c16[2], 20, 128);
+          adapt_prob(&p->tx32p[i][0], c32[0], c32[1] + c32[2] + c32[3], 20, 128);
+          adapt_prob(&p->tx32p[i][1], c32[1], c32[2] + c32[3], 20, 128);
+          adapt_prob(&p->tx32p[i][2], c32[2], c32[3], 20, 128);
+      }
+    }
+
+    // interpolation filter
+    if (s->s.h.filtermode == FILTER_SWITCHABLE) {
+        for (i = 0; i < 4; i++) {
+            uint8_t *pp = p->filter[i];
+            unsigned *c = s->counts.filter[i];
+
+            adapt_prob(&pp[0], c[0], c[1] + c[2], 20, 128);
+            adapt_prob(&pp[1], c[1], c[2], 20, 128);
+        }
+    }
+
+    // inter modes
+    for (i = 0; i < 7; i++) {
+        uint8_t *pp = p->mv_mode[i];
+        unsigned *c = s->counts.mv_mode[i];
+
+        adapt_prob(&pp[0], c[2], c[1] + c[0] + c[3], 20, 128);
+        adapt_prob(&pp[1], c[0], c[1] + c[3], 20, 128);
+        adapt_prob(&pp[2], c[1], c[3], 20, 128);
+    }
+
+    // mv joints
+    {
+        uint8_t *pp = p->mv_joint;
+        unsigned *c = s->counts.mv_joint;
+
+        adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
+        adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
+        adapt_prob(&pp[2], c[2], c[3], 20, 128);
+    }
+
+    // mv components
+    for (i = 0; i < 2; i++) {
+        uint8_t *pp;
+        unsigned *c, (*c2)[2], sum;
+
+        adapt_prob(&p->mv_comp[i].sign, s->counts.mv_comp[i].sign[0],
+                   s->counts.mv_comp[i].sign[1], 20, 128);
+
+        pp = p->mv_comp[i].classes;
+        c = s->counts.mv_comp[i].classes;
+        sum = c[1] + c[2] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9] + c[10];
+        adapt_prob(&pp[0], c[0], sum, 20, 128);
+        sum -= c[1];
+        adapt_prob(&pp[1], c[1], sum, 20, 128);
+        sum -= c[2] + c[3];
+        adapt_prob(&pp[2], c[2] + c[3], sum, 20, 128);
+        adapt_prob(&pp[3], c[2], c[3], 20, 128);
+        sum -= c[4] + c[5];
+        adapt_prob(&pp[4], c[4] + c[5], sum, 20, 128);
+        adapt_prob(&pp[5], c[4], c[5], 20, 128);
+        sum -= c[6];
+        adapt_prob(&pp[6], c[6], sum, 20, 128);
+        adapt_prob(&pp[7], c[7] + c[8], c[9] + c[10], 20, 128);
+        adapt_prob(&pp[8], c[7], c[8], 20, 128);
+        adapt_prob(&pp[9], c[9], c[10], 20, 128);
+
+        adapt_prob(&p->mv_comp[i].class0, s->counts.mv_comp[i].class0[0],
+                   s->counts.mv_comp[i].class0[1], 20, 128);
+        pp = p->mv_comp[i].bits;
+        c2 = s->counts.mv_comp[i].bits;
+        for (j = 0; j < 10; j++)
+            adapt_prob(&pp[j], c2[j][0], c2[j][1], 20, 128);
+
+        for (j = 0; j < 2; j++) {
+            pp = p->mv_comp[i].class0_fp[j];
+            c = s->counts.mv_comp[i].class0_fp[j];
+            adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
+            adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
+            adapt_prob(&pp[2], c[2], c[3], 20, 128);
+        }
+        pp = p->mv_comp[i].fp;
+        c = s->counts.mv_comp[i].fp;
+        adapt_prob(&pp[0], c[0], c[1] + c[2] + c[3], 20, 128);
+        adapt_prob(&pp[1], c[1], c[2] + c[3], 20, 128);
+        adapt_prob(&pp[2], c[2], c[3], 20, 128);
+
+        if (s->s.h.highprecisionmvs) {
+            adapt_prob(&p->mv_comp[i].class0_hp, s->counts.mv_comp[i].class0_hp[0],
+                       s->counts.mv_comp[i].class0_hp[1], 20, 128);
+            adapt_prob(&p->mv_comp[i].hp, s->counts.mv_comp[i].hp[0],
+                       s->counts.mv_comp[i].hp[1], 20, 128);
+        }
+    }
+
+    // y intra modes
+    for (i = 0; i < 4; i++) {
+        uint8_t *pp = p->y_mode[i];
+        unsigned *c = s->counts.y_mode[i], sum, s2;
+
+        sum = c[0] + c[1] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9];
+        adapt_prob(&pp[0], c[DC_PRED], sum, 20, 128);
+        sum -= c[TM_VP8_PRED];
+        adapt_prob(&pp[1], c[TM_VP8_PRED], sum, 20, 128);
+        sum -= c[VERT_PRED];
+        adapt_prob(&pp[2], c[VERT_PRED], sum, 20, 128);
+        s2 = c[HOR_PRED] + c[DIAG_DOWN_RIGHT_PRED] + c[VERT_RIGHT_PRED];
+        sum -= s2;
+        adapt_prob(&pp[3], s2, sum, 20, 128);
+        s2 -= c[HOR_PRED];
+        adapt_prob(&pp[4], c[HOR_PRED], s2, 20, 128);
+        adapt_prob(&pp[5], c[DIAG_DOWN_RIGHT_PRED], c[VERT_RIGHT_PRED], 20, 128);
+        sum -= c[DIAG_DOWN_LEFT_PRED];
+        adapt_prob(&pp[6], c[DIAG_DOWN_LEFT_PRED], sum, 20, 128);
+        sum -= c[VERT_LEFT_PRED];
+        adapt_prob(&pp[7], c[VERT_LEFT_PRED], sum, 20, 128);
+        adapt_prob(&pp[8], c[HOR_DOWN_PRED], c[HOR_UP_PRED], 20, 128);
+    }
+
+    // uv intra modes
+    for (i = 0; i < 10; i++) {
+        uint8_t *pp = p->uv_mode[i];
+        unsigned *c = s->counts.uv_mode[i], sum, s2;
+
+        sum = c[0] + c[1] + c[3] + c[4] + c[5] + c[6] + c[7] + c[8] + c[9];
+        adapt_prob(&pp[0], c[DC_PRED], sum, 20, 128);
+        sum -= c[TM_VP8_PRED];
+        adapt_prob(&pp[1], c[TM_VP8_PRED], sum, 20, 128);
+        sum -= c[VERT_PRED];
+        adapt_prob(&pp[2], c[VERT_PRED], sum, 20, 128);
+        s2 = c[HOR_PRED] + c[DIAG_DOWN_RIGHT_PRED] + c[VERT_RIGHT_PRED];
+        sum -= s2;
+        adapt_prob(&pp[3], s2, sum, 20, 128);
+        s2 -= c[HOR_PRED];
+        adapt_prob(&pp[4], c[HOR_PRED], s2, 20, 128);
+        adapt_prob(&pp[5], c[DIAG_DOWN_RIGHT_PRED], c[VERT_RIGHT_PRED], 20, 128);
+        sum -= c[DIAG_DOWN_LEFT_PRED];
+        adapt_prob(&pp[6], c[DIAG_DOWN_LEFT_PRED], sum, 20, 128);
+        sum -= c[VERT_LEFT_PRED];
+        adapt_prob(&pp[7], c[VERT_LEFT_PRED], sum, 20, 128);
+        adapt_prob(&pp[8], c[HOR_DOWN_PRED], c[HOR_UP_PRED], 20, 128);
+    }
+}

libavcodec/x86/vp9dsp_init.c

@@ -24,7 +24,7 @@
 #include "libavutil/cpu.h"
 #include "libavutil/mem.h"
 #include "libavutil/x86/cpu.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavcodec/x86/vp9dsp_init.h"
 
 #if HAVE_YASM

libavcodec/x86/vp9dsp_init.h

@@ -23,7 +23,7 @@
 #ifndef AVCODEC_X86_VP9DSP_INIT_H
 #define AVCODEC_X86_VP9DSP_INIT_H
 
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 
 // hack to force-expand BPC
 #define cat(a, bpp, b) a##bpp##b

libavcodec/x86/vp9dsp_init_16bpp.c

@@ -24,7 +24,7 @@
 #include "libavutil/cpu.h"
 #include "libavutil/mem.h"
 #include "libavutil/x86/cpu.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavcodec/x86/vp9dsp_init.h"
 
 #if HAVE_YASM

libavcodec/x86/vp9dsp_init_16bpp_template.c

@@ -24,7 +24,7 @@
 #include "libavutil/cpu.h"
 #include "libavutil/mem.h"
 #include "libavutil/x86/cpu.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavcodec/x86/vp9dsp_init.h"
 
 #if HAVE_YASM

tests/checkasm/vp9dsp.c

@@ -22,7 +22,7 @@
 #include <string.h>
 #include "checkasm.h"
 #include "libavcodec/vp9data.h"
-#include "libavcodec/vp9dsp.h"
+#include "libavcodec/vp9.h"
 #include "libavutil/common.h"
 #include "libavutil/internal.h"
 #include "libavutil/intreadwrite.h"
@@ -259,7 +259,7 @@ static int copy_subcoefs(int16_t *out, const int16_t *in, enum TxfmMode tx,
     // test
 
     int n;
-    const int16_t *scan = vp9_scans[tx][txtp];
+    const int16_t *scan = ff_vp9_scans[tx][txtp];
     int eob;
 
     for (n = 0; n < sz * sz; n++) {