libavcodec: vp8 neon optimizations for aarch64

Partial port of the ARM Neon for aarch64.

Benchmarks from fate:

benchmarking with Linux Perf Monitoring API
nop: 58.6
checkasm: using random seed 1760970128
NEON:
 - vp8dsp.idct       [OK]
 - vp8dsp.mc         [OK]
 - vp8dsp.loopfilter [OK]
checkasm: all 21 tests passed
vp8_idct_add_c: 201.6
vp8_idct_add_neon: 83.1
vp8_idct_dc_add_c: 107.6
vp8_idct_dc_add_neon: 33.8
vp8_idct_dc_add4y_c: 426.4
vp8_idct_dc_add4y_neon: 59.4
vp8_loop_filter8uv_h_c: 688.1
vp8_loop_filter8uv_h_neon: 216.3
vp8_loop_filter8uv_inner_h_c: 649.3
vp8_loop_filter8uv_inner_h_neon: 195.3
vp8_loop_filter8uv_inner_v_c: 544.8
vp8_loop_filter8uv_inner_v_neon: 131.3
vp8_loop_filter8uv_v_c: 706.1
vp8_loop_filter8uv_v_neon: 141.1
vp8_loop_filter16y_h_c: 668.8
vp8_loop_filter16y_h_neon: 242.8
vp8_loop_filter16y_inner_h_c: 647.3
vp8_loop_filter16y_inner_h_neon: 224.6
vp8_loop_filter16y_inner_v_c: 647.8
vp8_loop_filter16y_inner_v_neon: 128.8
vp8_loop_filter16y_v_c: 721.8
vp8_loop_filter16y_v_neon: 154.3
vp8_loop_filter_simple_h_c: 387.8
vp8_loop_filter_simple_h_neon: 187.6
vp8_loop_filter_simple_v_c: 384.1
vp8_loop_filter_simple_v_neon: 78.6
vp8_put_epel8_h4v4_c: 3971.1
vp8_put_epel8_h4v4_neon: 855.1
vp8_put_epel8_h4v6_c: 5060.1
vp8_put_epel8_h4v6_neon: 989.6
vp8_put_epel8_h6v4_c: 4320.8
vp8_put_epel8_h6v4_neon: 1007.3
vp8_put_epel8_h6v6_c: 5449.3
vp8_put_epel8_h6v6_neon: 1158.1
vp8_put_epel16_h6_c: 6683.8
vp8_put_epel16_h6_neon: 831.8
vp8_put_epel16_h6v6_c: 11110.8
vp8_put_epel16_h6v6_neon: 2214.8
vp8_put_epel16_v6_c: 7024.8
vp8_put_epel16_v6_neon: 799.6
vp8_put_pixels8_c: 112.8
vp8_put_pixels8_neon: 78.1
vp8_put_pixels16_c: 131.3
vp8_put_pixels16_neon: 129.8

This contains a fix to include guards by Carl Eugen Hoyos.
Signed-off-by: Martin Storsjö <martin@martin.st>

libavcodec/aarch64/Makefile

@@ -44,6 +44,8 @@ NEON-OBJS-$(CONFIG_MPEGAUDIODSP)        += aarch64/mpegaudiodsp_neon.o
 NEON-OBJS-$(CONFIG_DCA_DECODER)         += aarch64/dcadsp_neon.o               \
                                            aarch64/synth_filter_neon.o
 NEON-OBJS-$(CONFIG_VORBIS_DECODER)      += aarch64/vorbisdsp_neon.o
+NEON-OBJS-$(CONFIG_VP8DSP)              += aarch64/vp8dsp_init_aarch64.o       \
+                                           aarch64/vp8dsp_neon.o
 NEON-OBJS-$(CONFIG_VP9_DECODER)         += aarch64/vp9itxfm_neon.o             \
                                            aarch64/vp9lpf_neon.o               \
                                            aarch64/vp9mc_neon.o

libavcodec/aarch64/vp8dsp.h

+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_AARCH64_VP8DSP_H
+#define AVCODEC_AARCH64_VP8DSP_H
+
+#include "libavcodec/vp8dsp.h"
+
+#define VP8_LF_Y(hv, inner, opt)                                             \
+    void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst,            \
+                                                    ptrdiff_t stride,        \
+                                                    int flim_E, int flim_I,  \
+                                                    int hev_thresh)
+
+#define VP8_LF_UV(hv, inner, opt)                                            \
+    void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU,          \
+                                                     uint8_t *dstV,          \
+                                                     ptrdiff_t stride,       \
+                                                     int flim_E, int flim_I, \
+                                                     int hev_thresh)
+
+#define VP8_LF_SIMPLE(hv, opt)                                          \
+    void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst,         \
+                                                  ptrdiff_t stride,     \
+                                                  int flim)
+
+#define VP8_LF_HV(inner, opt)                   \
+    VP8_LF_Y(h,  inner, opt);                   \
+    VP8_LF_Y(v,  inner, opt);                   \
+    VP8_LF_UV(h, inner, opt);                   \
+    VP8_LF_UV(v, inner, opt)
+
+#define VP8_LF(opt)                             \
+    VP8_LF_HV(,       opt);                     \
+    VP8_LF_HV(_inner, opt);                     \
+    VP8_LF_SIMPLE(h, opt);                      \
+    VP8_LF_SIMPLE(v, opt)
+
+#define VP8_MC(n, opt)                                                  \
+    void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride,      \
+                                uint8_t *src, ptrdiff_t srcstride,      \
+                                int h, int x, int y)
+
+#define VP8_EPEL(w, opt)                        \
+    VP8_MC(pixels ## w, opt);                   \
+    VP8_MC(epel ## w ## _h4, opt);              \
+    VP8_MC(epel ## w ## _h6, opt);              \
+    VP8_MC(epel ## w ## _v4, opt);              \
+    VP8_MC(epel ## w ## _h4v4, opt);            \
+    VP8_MC(epel ## w ## _h6v4, opt);            \
+    VP8_MC(epel ## w ## _v6, opt);              \
+    VP8_MC(epel ## w ## _h4v6, opt);            \
+    VP8_MC(epel ## w ## _h6v6, opt)
+
+#endif /* AVCODEC_AARCH64_VP8DSP_H */

libavcodec/aarch64/vp8dsp_init_aarch64.c

+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stdint.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+
+#include "libavutil/attributes.h"
+#include "libavutil/aarch64/cpu.h"
+#include "libavcodec/vp8dsp.h"
+#include "vp8dsp.h"
+
+void ff_vp8_luma_dc_wht_neon(int16_t block[4][4][16], int16_t dc[16]);
+
+void ff_vp8_idct_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add_neon(uint8_t *dst, int16_t block[16], ptrdiff_t stride);
+void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride);
+
+VP8_LF(neon);
+
+VP8_EPEL(16, neon);
+VP8_EPEL(8,  neon);
+
+
+av_cold void ff_vp78dsp_init_aarch64(VP8DSPContext *dsp)
+{
+    if (!have_neon(av_get_cpu_flags()))
+        return;
+    dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon;
+    dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon;
+    dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon;
+
+    dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon;
+    dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon;
+}
+
+av_cold void ff_vp8dsp_init_aarch64(VP8DSPContext *dsp)
+{
+    if (!have_neon(av_get_cpu_flags()))
+        return;
+
+    dsp->vp8_idct_add       = ff_vp8_idct_add_neon;
+    dsp->vp8_idct_dc_add    = ff_vp8_idct_dc_add_neon;
+    dsp->vp8_idct_dc_add4y  = ff_vp8_idct_dc_add4y_neon;
+
+    dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon;
+    dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon;
+    dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon;
+    dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon;
+
+    dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon;
+    dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon;
+    dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon;
+    dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon;
+
+    dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon;
+    dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon;
+}

libavcodec/vp8dsp.c

@@ -679,6 +679,8 @@ av_cold void ff_vp78dsp_init(VP8DSPContext *dsp)
     VP78_BILINEAR_MC_FUNC(1, 8);
     VP78_BILINEAR_MC_FUNC(2, 4);
 
+    if (ARCH_AARCH64)
+        ff_vp78dsp_init_aarch64(dsp);
     if (ARCH_ARM)
         ff_vp78dsp_init_arm(dsp);
     if (ARCH_PPC)
@@ -739,6 +741,8 @@ av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
     dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
     dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
 
+    if (ARCH_AARCH64)
+        ff_vp8dsp_init_aarch64(dsp);
     if (ARCH_ARM)
         ff_vp8dsp_init_arm(dsp);
     if (ARCH_X86)

libavcodec/vp8dsp.h

@@ -91,11 +91,13 @@ void ff_put_vp8_pixels4_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride,
 void ff_vp7dsp_init(VP8DSPContext *c);
 
 void ff_vp78dsp_init(VP8DSPContext *c);
+void ff_vp78dsp_init_aarch64(VP8DSPContext *c);
 void ff_vp78dsp_init_arm(VP8DSPContext *c);
 void ff_vp78dsp_init_ppc(VP8DSPContext *c);
 void ff_vp78dsp_init_x86(VP8DSPContext *c);
 
 void ff_vp8dsp_init(VP8DSPContext *c);
+void ff_vp8dsp_init_aarch64(VP8DSPContext *c);
 void ff_vp8dsp_init_arm(VP8DSPContext *c);
 void ff_vp8dsp_init_x86(VP8DSPContext *c);