arm: vp9: Add NEON optimizations of VP9 MC functions

This work is sponsored by, and copyright, Google.

The filter coefficients are signed values, where the product of the
multiplication with one individual filter coefficient doesn't
overflow a 16 bit signed value (the largest filter coefficient is
127). But when the products are accumulated, the resulting sum can
overflow the 16 bit signed range. Instead of accumulating in 32 bit,
we accumulate the largest product (either index 3 or 4) last with a
saturated addition.

(The VP8 MC asm does something similar, but slightly simpler, by
accumulating each half of the filter separately. In the VP9 MC
filters, each half of the filter can also overflow though, so the
largest component has to be handled individually.)

Examples of relative speedup compared to the C version, from checkasm:
                       Cortex      A7     A8     A9    A53
vp9_avg4_neon:                   1.71   1.15   1.42   1.49
vp9_avg8_neon:                   2.51   3.63   3.14   2.58
vp9_avg16_neon:                  2.95   6.76   3.01   2.84
vp9_avg32_neon:                  3.29   6.64   2.85   3.00
vp9_avg64_neon:                  3.47   6.67   3.14   2.80
vp9_avg_8tap_smooth_4h_neon:     3.22   4.73   2.76   4.67
vp9_avg_8tap_smooth_4hv_neon:    3.67   4.76   3.28   4.71
vp9_avg_8tap_smooth_4v_neon:     5.52   7.60   4.60   6.31
vp9_avg_8tap_smooth_8h_neon:     6.22   9.04   5.12   9.32
vp9_avg_8tap_smooth_8hv_neon:    6.38   8.21   5.72   8.17
vp9_avg_8tap_smooth_8v_neon:     9.22  12.66   8.15  11.10
vp9_avg_8tap_smooth_64h_neon:    7.02  10.23   5.54  11.58
vp9_avg_8tap_smooth_64hv_neon:   6.76   9.46   5.93   9.40
vp9_avg_8tap_smooth_64v_neon:   10.76  14.13   9.46  13.37
vp9_put4_neon:                   1.11   1.47   1.00   1.21
vp9_put8_neon:                   1.23   2.17   1.94   1.48
vp9_put16_neon:                  1.63   4.02   1.73   1.97
vp9_put32_neon:                  1.56   4.92   2.00   1.96
vp9_put64_neon:                  2.10   5.28   2.03   2.35
vp9_put_8tap_smooth_4h_neon:     3.11   4.35   2.63   4.35
vp9_put_8tap_smooth_4hv_neon:    3.67   4.69   3.25   4.71
vp9_put_8tap_smooth_4v_neon:     5.45   7.27   4.49   6.52
vp9_put_8tap_smooth_8h_neon:     5.97   8.18   4.81   8.56
vp9_put_8tap_smooth_8hv_neon:    6.39   7.90   5.64   8.15
vp9_put_8tap_smooth_8v_neon:     9.03  11.84   8.07  11.51
vp9_put_8tap_smooth_64h_neon:    6.78   9.48   4.88  10.89
vp9_put_8tap_smooth_64hv_neon:   6.99   8.87   5.94   9.56
vp9_put_8tap_smooth_64v_neon:   10.69  13.30   9.43  14.34

For the larger 8tap filters, the speedup vs C code is around 5-14x.

This is significantly faster than libvpx's implementation of the same
functions, at least when comparing the put_8tap_smooth_64 functions
(compared to vpx_convolve8_horiz_neon and vpx_convolve8_vert_neon from
libvpx).

Absolute runtimes from checkasm:
                          Cortex      A7        A8        A9       A53
vp9_put_8tap_smooth_64h_neon:    20150.3   14489.4   19733.6   10863.7
libvpx vpx_convolve8_horiz_neon: 52623.3   19736.4   21907.7   25027.7

vp9_put_8tap_smooth_64v_neon:    14455.0   12303.9   13746.4    9628.9
libvpx vpx_convolve8_vert_neon:  42090.0   17706.2   17659.9   16941.2

Thus, on the A9, the horizontal filter is only marginally faster than
libvpx, while our version is significantly faster on the other cores,
and the vertical filter is significantly faster on all cores. The
difference is especially large on the A7.

The libvpx implementation does the accumulation in 32 bit, which
probably explains most of the differences.

This is an adapted cherry-pick from libav commits
ffbd1d2b,
392caa65,
557c1675 and
11623217.
Signed-off-by: NRonald S. Bultje <rsbultje@gmail.com>

libavcodec/arm/Makefile

@@ -44,6 +44,7 @@ OBJS-$(CONFIG_MLP_DECODER)             += arm/mlpdsp_init_arm.o
 OBJS-$(CONFIG_RV40_DECODER)            += arm/rv40dsp_init_arm.o
 OBJS-$(CONFIG_VORBIS_DECODER)          += arm/vorbisdsp_init_arm.o
 OBJS-$(CONFIG_VP6_DECODER)             += arm/vp6dsp_init_arm.o
+OBJS-$(CONFIG_VP9_DECODER)             += arm/vp9dsp_init_arm.o
 
 
 # ARMv5 optimizations
@@ -139,3 +140,4 @@ NEON-OBJS-$(CONFIG_RV40_DECODER)       += arm/rv34dsp_neon.o            \
                                           arm/rv40dsp_neon.o
 NEON-OBJS-$(CONFIG_VORBIS_DECODER)     += arm/vorbisdsp_neon.o
 NEON-OBJS-$(CONFIG_VP6_DECODER)        += arm/vp6dsp_neon.o
+NEON-OBJS-$(CONFIG_VP9_DECODER)        += arm/vp9mc_neon.o

libavcodec/arm/vp9dsp_init_arm.c

+/*
+ * Copyright (c) 2016 Google Inc.
+ *
+ * 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 <stdint.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/arm/cpu.h"
+#include "libavcodec/vp9dsp.h"
+
+#define declare_fpel(type, sz)                                          \
+void ff_vp9_##type##sz##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                              const uint8_t *src, ptrdiff_t src_stride, \
+                              int h, int mx, int my)
+
+#define declare_copy_avg(sz) \
+    declare_fpel(copy, sz);  \
+    declare_fpel(avg , sz)
+
+#define decl_mc_func(op, filter, dir, sz)                                                \
+void ff_vp9_##op##_##filter##sz##_##dir##_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                               const uint8_t *src, ptrdiff_t src_stride, \
+                                               int h, int mx, int my)
+
+#define define_8tap_2d_fn(op, filter, sz)                                         \
+static void op##_##filter##sz##_hv_neon(uint8_t *dst, ptrdiff_t dst_stride,       \
+                                        const uint8_t *src, ptrdiff_t src_stride, \
+                                        int h, int mx, int my)                    \
+{                                                                                 \
+    LOCAL_ALIGNED_16(uint8_t, temp, [((1 + (sz < 64)) * sz + 8) * sz]);           \
+    /* We only need h + 7 lines, but the horizontal filter assumes an             \
+     * even number of rows, so filter h + 8 lines here. */                        \
+    ff_vp9_put_##filter##sz##_h_neon(temp, sz,                                    \
+                                     src - 3 * src_stride, src_stride,            \
+                                     h + 8, mx, 0);                               \
+    ff_vp9_##op##_##filter##sz##_v_neon(dst, dst_stride,                          \
+                                        temp + 3 * sz, sz,                        \
+                                        h, 0, my);                                \
+}
+
+#define decl_filter_funcs(op, dir, sz)  \
+    decl_mc_func(op, regular, dir, sz); \
+    decl_mc_func(op, sharp,   dir, sz); \
+    decl_mc_func(op, smooth,  dir, sz)
+
+#define decl_mc_funcs(sz)           \
+    decl_filter_funcs(put, h,  sz); \
+    decl_filter_funcs(avg, h,  sz); \
+    decl_filter_funcs(put, v,  sz); \
+    decl_filter_funcs(avg, v,  sz); \
+    decl_filter_funcs(put, hv, sz); \
+    decl_filter_funcs(avg, hv, sz)
+
+declare_copy_avg(64);
+declare_copy_avg(32);
+declare_copy_avg(16);
+declare_copy_avg(8);
+declare_copy_avg(4);
+
+decl_mc_funcs(64);
+decl_mc_funcs(32);
+decl_mc_funcs(16);
+decl_mc_funcs(8);
+decl_mc_funcs(4);
+
+#define define_8tap_2d_funcs(sz)        \
+    define_8tap_2d_fn(put, regular, sz) \
+    define_8tap_2d_fn(put, sharp,   sz) \
+    define_8tap_2d_fn(put, smooth,  sz) \
+    define_8tap_2d_fn(avg, regular, sz) \
+    define_8tap_2d_fn(avg, sharp,   sz) \
+    define_8tap_2d_fn(avg, smooth,  sz)
+
+define_8tap_2d_funcs(64)
+define_8tap_2d_funcs(32)
+define_8tap_2d_funcs(16)
+define_8tap_2d_funcs(8)
+define_8tap_2d_funcs(4)
+
+
+av_cold void ff_vp9dsp_init_arm(VP9DSPContext *dsp, int bpp)
+{
+    int cpu_flags = av_get_cpu_flags();
+
+    if (bpp != 8)
+        return;
+
+    if (have_neon(cpu_flags)) {
+#define init_fpel(idx1, idx2, sz, type)              \
+    dsp->mc[idx1][FILTER_8TAP_SMOOTH ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_REGULAR][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_8TAP_SHARP  ][idx2][0][0] = \
+    dsp->mc[idx1][FILTER_BILINEAR    ][idx2][0][0] = ff_vp9_##type##sz##_neon
+
+#define init_copy_avg(idx, sz)   \
+    init_fpel(idx, 0, sz, copy); \
+    init_fpel(idx, 1, sz, avg)
+
+#define init_mc_func(idx1, idx2, op, filter, fname, dir, mx, my, sz, pfx) \
+    dsp->mc[idx1][filter][idx2][mx][my] = pfx##op##_##fname##sz##_##dir##_neon
+
+#define init_mc_funcs(idx, dir, mx, my, sz, pfx)                                   \
+    init_mc_func(idx, 0, put, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 0, put, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_REGULAR, regular, dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SHARP,   sharp,   dir, mx, my, sz, pfx); \
+    init_mc_func(idx, 1, avg, FILTER_8TAP_SMOOTH,  smooth,  dir, mx, my, sz, pfx)
+
+#define init_mc_funcs_dirs(idx, sz)            \
+    init_mc_funcs(idx, h,  1, 0, sz, ff_vp9_); \
+    init_mc_funcs(idx, v,  0, 1, sz, ff_vp9_); \
+    init_mc_funcs(idx, hv, 1, 1, sz,)
+
+        init_copy_avg(0, 64);
+        init_copy_avg(1, 32);
+        init_copy_avg(2, 16);
+        init_copy_avg(3, 8);
+        init_copy_avg(4, 4);
+
+        init_mc_funcs_dirs(0, 64);
+        init_mc_funcs_dirs(1, 32);
+        init_mc_funcs_dirs(2, 16);
+        init_mc_funcs_dirs(3, 8);
+        init_mc_funcs_dirs(4, 4);
+    }
+}

libavcodec/vp9.c

@@ -2749,8 +2749,11 @@ static av_always_inline void mc_luma_unscaled(VP9Context *s, vp9_mc_func (*mc)[2
     // the longest loopfilter of the next sbrow
     th = (y + bh + 4 * !!my + 7) >> 6;
     ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
+    // The arm _hv filters read one more row than what actually is
+    // needed, so switch to emulated edge one pixel sooner vertically
+    // (!!my * 5) than horizontally (!!mx * 4).
     if (x < !!mx * 3 || y < !!my * 3 ||
-        x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) {
+        x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
                                  ref - !!my * 3 * ref_stride - !!mx * 3 * bytesperpixel,
                                  160, ref_stride,
@@ -2784,8 +2787,11 @@ static av_always_inline void mc_chroma_unscaled(VP9Context *s, vp9_mc_func (*mc)
     // the longest loopfilter of the next sbrow
     th = (y + bh + 4 * !!my + 7) >> (6 - s->ss_v);
     ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
+    // The arm _hv filters read one more row than what actually is
+    // needed, so switch to emulated edge one pixel sooner vertically
+    // (!!my * 5) than horizontally (!!mx * 4).
     if (x < !!mx * 3 || y < !!my * 3 ||
-        x + !!mx * 4 > w - bw || y + !!my * 4 > h - bh) {
+        x + !!mx * 4 > w - bw || y + !!my * 5 > h - bh) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
                                  ref_u - !!my * 3 * src_stride_u - !!mx * 3 * bytesperpixel,
                                  160, src_stride_u,
@@ -2871,7 +2877,10 @@ static av_always_inline void mc_luma_scaled(VP9Context *s, vp9_scaled_mc_func sm
     // the longest loopfilter of the next sbrow
     th = (y + refbh_m1 + 4 + 7) >> 6;
     ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
-    if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 4 >= h - refbh_m1) {
+    // The arm _hv filters read one more row than what actually is
+    // needed, so switch to emulated edge one pixel sooner vertically
+    // (y + 5 >= h - refbh_m1) than horizontally (x + 4 >= w - refbw_m1).
+    if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 5 >= h - refbh_m1) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
                                  ref - 3 * ref_stride - 3 * bytesperpixel,
                                  288, ref_stride,
@@ -2937,7 +2946,10 @@ static av_always_inline void mc_chroma_scaled(VP9Context *s, vp9_scaled_mc_func
     // the longest loopfilter of the next sbrow
     th = (y + refbh_m1 + 4 + 7) >> (6 - s->ss_v);
     ff_thread_await_progress(ref_frame, FFMAX(th, 0), 0);
-    if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 4 >= h - refbh_m1) {
+    // The arm _hv filters read one more row than what actually is
+    // needed, so switch to emulated edge one pixel sooner vertically
+    // (y + 5 >= h - refbh_m1) than horizontally (x + 4 >= w - refbw_m1).
+    if (x < 3 || y < 3 || x + 4 >= w - refbw_m1 || y + 5 >= h - refbh_m1) {
         s->vdsp.emulated_edge_mc(s->edge_emu_buffer,
                                  ref_u - 3 * src_stride_u - 3 * bytesperpixel,
                                  288, src_stride_u,

libavcodec/vp9dsp.c

@@ -92,6 +92,7 @@ av_cold void ff_vp9dsp_init(VP9DSPContext *dsp, int bpp, int bitexact)
         ff_vp9dsp_init_12(dsp);
     }
 
+    if (ARCH_ARM) ff_vp9dsp_init_arm(dsp, bpp);
     if (ARCH_X86) ff_vp9dsp_init_x86(dsp, bpp, bitexact);
     if (ARCH_MIPS) ff_vp9dsp_init_mips(dsp, bpp);
 }

libavcodec/vp9dsp.h

@@ -129,6 +129,7 @@ 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_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);