This is a preparation for complete ARMv6 optimisations.
Unified · Split
| @@ -13,7 +13,9 @@ ARMV6-OBJS-$(CONFIG_MPEGAUDIODSP) += arm/mpegaudiodsp_fixed_armv6.o · 12 unmodified lines | ||
| 13 | 13 | OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_init_arm.o |
| 14 | 14 | OBJS-$(CONFIG_VP6_DECODER) += arm/vp56dsp_init_arm.o |
| 15 | 15 | OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_init_arm.o |
| 16 | −ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o | |
| 16 | +ARMV6-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8_armv6.o \ | |
| 17 | + arm/vp8dsp_init_armv6.o \ | |
| 18 | + arm/vp8dsp_armv6.o | |
| 17 | 19 | |
| 18 | 20 | OBJS-$(CONFIG_H264DSP) += arm/h264dsp_init_arm.o |
| 19 | 21 | OBJS-$(CONFIG_H264PRED) += arm/h264pred_init_arm.o |
| @@ -79,7 +81,8 @@ NEON-OBJS-$(CONFIG_VP5_DECODER) += arm/vp56dsp_neon.o \ · 59 unmodified lines | ||
| 79 | 81 | NEON-OBJS-$(CONFIG_VP6_DECODER) += arm/vp56dsp_neon.o \ |
| 80 | 82 | arm/vp3dsp_neon.o \ |
| 81 | 83 | |
| 82 | −NEON-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_neon.o | |
| 84 | +NEON-OBJS-$(CONFIG_VP8_DECODER) += arm/vp8dsp_init_neon.o \ | |
| 85 | + arm/vp8dsp_neon.o | |
| 83 | 86 | |
| 84 | 87 | NEON-OBJS += arm/dsputil_init_neon.o \ |
| 85 | 88 | arm/dsputil_neon.o \ |
| @@ -0,0 +1,78 @@ | ||
| 1 | +/* | |
| 2 | + * This file is part of Libav. | |
| 3 | + * | |
| 4 | + * Libav is free software; you can redistribute it and/or | |
| 5 | + * modify it under the terms of the GNU Lesser General Public | |
| 6 | + * License as published by the Free Software Foundation; either | |
| 7 | + * version 2.1 of the License, or (at your option) any later version. | |
| 8 | + * | |
| 9 | + * Libav is distributed in the hope that it will be useful, | |
| 10 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 11 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 12 | + * Lesser General Public License for more details. | |
| 13 | + * | |
| 14 | + * You should have received a copy of the GNU Lesser General Public | |
| 15 | + * License along with Libav; if not, write to the Free Software | |
| 16 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 17 | + */ | |
| 18 | + | |
| 19 | +#ifndef AVCODEC_ARM_VP8DSP_H | |
| 20 | +#define AVCODEC_ARM_VP8DSP_H | |
| 21 | + | |
| 22 | +#include "libavcodec/vp8dsp.h" | |
| 23 | + | |
| 24 | +void ff_vp8dsp_init_armv6(VP8DSPContext *dsp); | |
| 25 | +void ff_vp8dsp_init_neon(VP8DSPContext *dsp); | |
| 26 | + | |
| 27 | +#define VP8_LF_Y(hv, inner, opt) \ | |
| 28 | + void ff_vp8_##hv##_loop_filter16##inner##_##opt(uint8_t *dst, \ | |
| 29 | + ptrdiff_t stride, \ | |
| 30 | + int flim_E, int flim_I, \ | |
| 31 | + int hev_thresh) | |
| 32 | + | |
| 33 | +#define VP8_LF_UV(hv, inner, opt) \ | |
| 34 | + void ff_vp8_##hv##_loop_filter8uv##inner##_##opt(uint8_t *dstU, \ | |
| 35 | + uint8_t *dstV, \ | |
| 36 | + ptrdiff_t stride, \ | |
| 37 | + int flim_E, int flim_I, \ | |
| 38 | + int hev_thresh) | |
| 39 | + | |
| 40 | +#define VP8_LF_SIMPLE(hv, opt) \ | |
| 41 | + void ff_vp8_##hv##_loop_filter16_simple_##opt(uint8_t *dst, \ | |
| 42 | + ptrdiff_t stride, \ | |
| 43 | + int flim) | |
| 44 | + | |
| 45 | +#define VP8_LF_HV(inner, opt) \ | |
| 46 | + VP8_LF_Y(h, inner, opt); \ | |
| 47 | + VP8_LF_Y(v, inner, opt); \ | |
| 48 | + VP8_LF_UV(h, inner, opt); \ | |
| 49 | + VP8_LF_UV(v, inner, opt) | |
| 50 | + | |
| 51 | +#define VP8_LF(opt) \ | |
| 52 | + VP8_LF_HV(, opt); \ | |
| 53 | + VP8_LF_HV(_inner, opt); \ | |
| 54 | + VP8_LF_SIMPLE(h, opt); \ | |
| 55 | + VP8_LF_SIMPLE(v, opt) | |
| 56 | + | |
| 57 | +#define VP8_MC(n, opt) \ | |
| 58 | + void ff_put_vp8_##n##_##opt(uint8_t *dst, ptrdiff_t dststride, \ | |
| 59 | + uint8_t *src, ptrdiff_t srcstride, \ | |
| 60 | + int h, int x, int y) | |
| 61 | + | |
| 62 | +#define VP8_EPEL(w, opt) \ | |
| 63 | + VP8_MC(pixels ## w, opt); \ | |
| 64 | + VP8_MC(epel ## w ## _h4, opt); \ | |
| 65 | + VP8_MC(epel ## w ## _h6, opt); \ | |
| 66 | + VP8_MC(epel ## w ## _v4, opt); \ | |
| 67 | + VP8_MC(epel ## w ## _h4v4, opt); \ | |
| 68 | + VP8_MC(epel ## w ## _h6v4, opt); \ | |
| 69 | + VP8_MC(epel ## w ## _v6, opt); \ | |
| 70 | + VP8_MC(epel ## w ## _h4v6, opt); \ | |
| 71 | + VP8_MC(epel ## w ## _h6v6, opt) | |
| 72 | + | |
| 73 | +#define VP8_BILIN(w, opt) \ | |
| 74 | + VP8_MC(bilin ## w ## _h, opt); \ | |
| 75 | + VP8_MC(bilin ## w ## _v, opt); \ | |
| 76 | + VP8_MC(bilin ## w ## _hv, opt) | |
| 77 | + | |
| 78 | +#endif /* AVCODEC_ARM_VP8DSP_H */ | |
| @@ -0,0 +1,53 @@ | ||
| 1 | +/* | |
| 2 | + * VP8 ARMv6 optimisations | |
| 3 | + * | |
| 4 | + * Copyright (c) 2010 Rob Clark <rob@ti.com> | |
| 5 | + * Copyright (c) 2011 Mans Rullgard <mans@mansr.com> | |
| 6 | + * | |
| 7 | + * This file is part of Libav. | |
| 8 | + * | |
| 9 | + * Libav is free software; you can redistribute it and/or | |
| 10 | + * modify it under the terms of the GNU Lesser General Public | |
| 11 | + * License as published by the Free Software Foundation; either | |
| 12 | + * version 2.1 of the License, or (at your option) any later version. | |
| 13 | + * | |
| 14 | + * Libav is distributed in the hope that it will be useful, | |
| 15 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 16 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 17 | + * Lesser General Public License for more details. | |
| 18 | + * | |
| 19 | + * You should have received a copy of the GNU Lesser General Public | |
| 20 | + * License along with Libav; if not, write to the Free Software | |
| 21 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 22 | + */ | |
| 23 | + | |
| 24 | +#include "asm.S" | |
| 25 | + | |
| 26 | +function ff_vp8_luma_dc_wht_dc_armv6, export=1 | |
| 27 | + ldrsh r2, [r1] | |
| 28 | + mov r3, #0 | |
| 29 | + add r2, r2, #3 | |
| 30 | + strh r3, [r1] | |
| 31 | + asr r2, r2, #3 | |
| 32 | + .rept 16 | |
| 33 | + strh r2, [r0], #32 | |
| 34 | + .endr | |
| 35 | + bx lr | |
| 36 | +endfunc | |
| 37 | + | |
| 38 | +function ff_put_vp8_pixels4_armv6, export=1 | |
| 39 | + ldr r12, [sp, #0] @ h | |
| 40 | + push {r4-r6,lr} | |
| 41 | +1: | |
| 42 | + subs r12, r12, #4 | |
| 43 | + ldr_post r4, r2, r3 | |
| 44 | + ldr_post r5, r2, r3 | |
| 45 | + ldr_post r6, r2, r3 | |
| 46 | + ldr_post lr, r2, r3 | |
| 47 | + str_post r4, r0, r1 | |
| 48 | + str_post r5, r0, r1 | |
| 49 | + str_post r6, r0, r1 | |
| 50 | + str_post lr, r0, r1 | |
| 51 | + bgt 1b | |
| 52 | + pop {r4-r6,pc} | |
| 53 | +endfunc | |
| @@ -20,148 +20,14 @@ · 19 unmodified lines | ||
| 20 | 20 | |
| 21 | 21 | #include "libavutil/arm/cpu.h" |
| 22 | 22 | #include "libavcodec/vp8dsp.h" |
| 23 | − | |
| 24 | −void ff_vp8_luma_dc_wht_neon(DCTELEM block[4][4][16], DCTELEM dc[16]); | |
| 25 | −void ff_vp8_luma_dc_wht_dc_neon(DCTELEM block[4][4][16], DCTELEM dc[16]); | |
| 26 | − | |
| 27 | −void ff_vp8_idct_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); | |
| 28 | −void ff_vp8_idct_dc_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); | |
| 29 | −void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride); | |
| 30 | −void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride); | |
| 31 | − | |
| 32 | −void ff_vp8_v_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride, | |
| 33 | − int flim_E, int flim_I, int hev_thresh); | |
| 34 | −void ff_vp8_h_loop_filter16_neon(uint8_t *dst, ptrdiff_t stride, | |
| 35 | − int flim_E, int flim_I, int hev_thresh); | |
| 36 | −void ff_vp8_v_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, | |
| 37 | − int flim_E, int flim_I, int hev_thresh); | |
| 38 | −void ff_vp8_h_loop_filter8uv_neon(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride, | |
| 39 | − int flim_E, int flim_I, int hev_thresh); | |
| 40 | − | |
| 41 | −void ff_vp8_v_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride, | |
| 42 | − int flim_E, int flim_I, int hev_thresh); | |
| 43 | −void ff_vp8_h_loop_filter16_inner_neon(uint8_t *dst, ptrdiff_t stride, | |
| 44 | − int flim_E, int flim_I, int hev_thresh); | |
| 45 | −void ff_vp8_v_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV, | |
| 46 | − ptrdiff_t stride, int flim_E, int flim_I, | |
| 47 | − int hev_thresh); | |
| 48 | −void ff_vp8_h_loop_filter8uv_inner_neon(uint8_t *dstU, uint8_t *dstV, | |
| 49 | − ptrdiff_t stride, int flim_E, int flim_I, | |
| 50 | − int hev_thresh); | |
| 51 | − | |
| 52 | −void ff_vp8_v_loop_filter16_simple_neon(uint8_t *dst, ptrdiff_t stride, int flim); | |
| 53 | −void ff_vp8_h_loop_filter16_simple_neon(uint8_t *dst, ptrdiff_t stride, int flim); | |
| 54 | − | |
| 55 | − | |
| 56 | −#define VP8_MC(n) \ | |
| 57 | − void ff_put_vp8_##n##_neon(uint8_t *dst, ptrdiff_t dststride, \ | |
| 58 | − uint8_t *src, ptrdiff_t srcstride, \ | |
| 59 | − int h, int x, int y) | |
| 60 | − | |
| 61 | −#define VP8_EPEL(w) \ | |
| 62 | − VP8_MC(pixels ## w); \ | |
| 63 | − VP8_MC(epel ## w ## _h4); \ | |
| 64 | − VP8_MC(epel ## w ## _h6); \ | |
| 65 | − VP8_MC(epel ## w ## _v4); \ | |
| 66 | − VP8_MC(epel ## w ## _h4v4); \ | |
| 67 | − VP8_MC(epel ## w ## _h6v4); \ | |
| 68 | − VP8_MC(epel ## w ## _v6); \ | |
| 69 | − VP8_MC(epel ## w ## _h4v6); \ | |
| 70 | − VP8_MC(epel ## w ## _h6v6) | |
| 71 | − | |
| 72 | −VP8_EPEL(16); | |
| 73 | −VP8_EPEL(8); | |
| 74 | −VP8_EPEL(4); | |
| 75 | − | |
| 76 | −VP8_MC(bilin16_h); | |
| 77 | −VP8_MC(bilin16_v); | |
| 78 | −VP8_MC(bilin16_hv); | |
| 79 | −VP8_MC(bilin8_h); | |
| 80 | −VP8_MC(bilin8_v); | |
| 81 | −VP8_MC(bilin8_hv); | |
| 82 | −VP8_MC(bilin4_h); | |
| 83 | −VP8_MC(bilin4_v); | |
| 84 | −VP8_MC(bilin4_hv); | |
| 23 | +#include "vp8dsp.h" | |
| 85 | 24 | |
| 86 | 25 | av_cold void ff_vp8dsp_init_arm(VP8DSPContext *dsp) |
| 87 | 26 | { |
| 88 | 27 | int cpu_flags = av_get_cpu_flags(); |
| 89 | 28 | |
| 90 | − if (have_neon(cpu_flags)) { | |
| 91 | − dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; | |
| 92 | − dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_neon; | |
| 93 | − | |
| 94 | − dsp->vp8_idct_add = ff_vp8_idct_add_neon; | |
| 95 | − dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; | |
| 96 | − dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; | |
| 97 | − dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; | |
| 98 | − | |
| 99 | − dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; | |
| 100 | − dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; | |
| 101 | − dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon; | |
| 102 | − dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon; | |
| 103 | − | |
| 104 | − dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon; | |
| 105 | − dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon; | |
| 106 | − dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon; | |
| 107 | − dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon; | |
| 108 | − | |
| 109 | − dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon; | |
| 110 | − dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon; | |
| 111 | − | |
| 112 | − dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; | |
| 113 | − dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon; | |
| 114 | − dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon; | |
| 115 | − dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon; | |
| 116 | − | |
| 117 | − dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; | |
| 118 | − dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon; | |
| 119 | − dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon; | |
| 120 | − dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon; | |
| 121 | − dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon; | |
| 122 | − dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon; | |
| 123 | − dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon; | |
| 124 | − dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon; | |
| 125 | − dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon; | |
| 126 | − | |
| 127 | − dsp->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_neon; | |
| 128 | − dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon; | |
| 129 | − dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon; | |
| 130 | − dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon; | |
| 131 | − dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon; | |
| 132 | − dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon; | |
| 133 | − dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon; | |
| 134 | − dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon; | |
| 135 | − dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon; | |
| 136 | − | |
| 137 | − dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; | |
| 138 | − dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon; | |
| 139 | − dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon; | |
| 140 | − dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon; | |
| 141 | − dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon; | |
| 142 | − dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon; | |
| 143 | − dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon; | |
| 144 | − dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon; | |
| 145 | − dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon; | |
| 146 | − | |
| 147 | − dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; | |
| 148 | − dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon; | |
| 149 | − dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon; | |
| 150 | − dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon; | |
| 151 | − dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon; | |
| 152 | − dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon; | |
| 153 | − dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon; | |
| 154 | − dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon; | |
| 155 | − dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon; | |
| 156 | − | |
| 157 | − dsp->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_neon; | |
| 158 | − dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon; | |
| 159 | − dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon; | |
| 160 | − dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon; | |
| 161 | − dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon; | |
| 162 | − dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon; | |
| 163 | − dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon; | |
| 164 | − dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon; | |
| 165 | − dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon; | |
| 166 | − } | |
| 29 | + if (have_armv6(cpu_flags)) | |
| 30 | + ff_vp8dsp_init_armv6(dsp); | |
| 31 | + if (have_neon(cpu_flags)) | |
| 32 | + ff_vp8dsp_init_neon(dsp); | |
| 167 | 33 | } |
| @@ -0,0 +1,34 @@ | ||
| 1 | +/* | |
| 2 | + * This file is part of Libav. | |
| 3 | + * | |
| 4 | + * Libav is free software; you can redistribute it and/or | |
| 5 | + * modify it under the terms of the GNU Lesser General Public | |
| 6 | + * License as published by the Free Software Foundation; either | |
| 7 | + * version 2.1 of the License, or (at your option) any later version. | |
| 8 | + * | |
| 9 | + * Libav is distributed in the hope that it will be useful, | |
| 10 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 11 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 12 | + * Lesser General Public License for more details. | |
| 13 | + * | |
| 14 | + * You should have received a copy of the GNU Lesser General Public | |
| 15 | + * License along with Libav; if not, write to the Free Software | |
| 16 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 17 | + */ | |
| 18 | + | |
| 19 | +#include <stdint.h> | |
| 20 | +#include "libavcodec/vp8dsp.h" | |
| 21 | +#include "vp8dsp.h" | |
| 22 | + | |
| 23 | +void ff_vp8_luma_dc_wht_dc_armv6(DCTELEM block[4][4][16], DCTELEM dc[16]); | |
| 24 | + | |
| 25 | +VP8_MC(pixels4, armv6); | |
| 26 | + | |
| 27 | +av_cold void ff_vp8dsp_init_armv6(VP8DSPContext *dsp) | |
| 28 | +{ | |
| 29 | + dsp->vp8_luma_dc_wht_dc = ff_vp8_luma_dc_wht_dc_armv6; | |
| 30 | + | |
| 31 | + dsp->put_vp8_epel_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6; | |
| 32 | + | |
| 33 | + dsp->put_vp8_bilinear_pixels_tab[2][0][0] = ff_put_vp8_pixels4_armv6; | |
| 34 | +} | |
| @@ -0,0 +1,114 @@ | ||
| 1 | +/* | |
| 2 | + * This file is part of Libav. | |
| 3 | + * | |
| 4 | + * Libav is free software; you can redistribute it and/or | |
| 5 | + * modify it under the terms of the GNU Lesser General Public | |
| 6 | + * License as published by the Free Software Foundation; either | |
| 7 | + * version 2.1 of the License, or (at your option) any later version. | |
| 8 | + * | |
| 9 | + * Libav is distributed in the hope that it will be useful, | |
| 10 | + * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
| 11 | + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
| 12 | + * Lesser General Public License for more details. | |
| 13 | + * | |
| 14 | + * You should have received a copy of the GNU Lesser General Public | |
| 15 | + * License along with Libav; if not, write to the Free Software | |
| 16 | + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | |
| 17 | + */ | |
| 18 | + | |
| 19 | +#include <stdint.h> | |
| 20 | +#include "libavcodec/vp8dsp.h" | |
| 21 | +#include "vp8dsp.h" | |
| 22 | + | |
| 23 | +void ff_vp8_luma_dc_wht_neon(DCTELEM block[4][4][16], DCTELEM dc[16]); | |
| 24 | + | |
| 25 | +void ff_vp8_idct_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); | |
| 26 | +void ff_vp8_idct_dc_add_neon(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride); | |
| 27 | +void ff_vp8_idct_dc_add4y_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride); | |
| 28 | +void ff_vp8_idct_dc_add4uv_neon(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride); | |
| 29 | + | |
| 30 | +VP8_LF(neon); | |
| 31 | + | |
| 32 | +VP8_EPEL(16, neon); | |
| 33 | +VP8_EPEL(8, neon); | |
| 34 | +VP8_EPEL(4, neon); | |
| 35 | + | |
| 36 | +VP8_BILIN(16, neon); | |
| 37 | +VP8_BILIN(8, neon); | |
| 38 | +VP8_BILIN(4, neon); | |
| 39 | + | |
| 40 | +av_cold void ff_vp8dsp_init_neon(VP8DSPContext *dsp) | |
| 41 | +{ | |
| 42 | + dsp->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_neon; | |
| 43 | + | |
| 44 | + dsp->vp8_idct_add = ff_vp8_idct_add_neon; | |
| 45 | + dsp->vp8_idct_dc_add = ff_vp8_idct_dc_add_neon; | |
| 46 | + dsp->vp8_idct_dc_add4y = ff_vp8_idct_dc_add4y_neon; | |
| 47 | + dsp->vp8_idct_dc_add4uv = ff_vp8_idct_dc_add4uv_neon; | |
| 48 | + | |
| 49 | + dsp->vp8_v_loop_filter16y = ff_vp8_v_loop_filter16_neon; | |
| 50 | + dsp->vp8_h_loop_filter16y = ff_vp8_h_loop_filter16_neon; | |
| 51 | + dsp->vp8_v_loop_filter8uv = ff_vp8_v_loop_filter8uv_neon; | |
| 52 | + dsp->vp8_h_loop_filter8uv = ff_vp8_h_loop_filter8uv_neon; | |
| 53 | + | |
| 54 | + dsp->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16_inner_neon; | |
| 55 | + dsp->vp8_h_loop_filter16y_inner = ff_vp8_h_loop_filter16_inner_neon; | |
| 56 | + dsp->vp8_v_loop_filter8uv_inner = ff_vp8_v_loop_filter8uv_inner_neon; | |
| 57 | + dsp->vp8_h_loop_filter8uv_inner = ff_vp8_h_loop_filter8uv_inner_neon; | |
| 58 | + | |
| 59 | + dsp->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter16_simple_neon; | |
| 60 | + dsp->vp8_h_loop_filter_simple = ff_vp8_h_loop_filter16_simple_neon; | |
| 61 | + | |
| 62 | + dsp->put_vp8_epel_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; | |
| 63 | + dsp->put_vp8_epel_pixels_tab[0][0][2] = ff_put_vp8_epel16_h6_neon; | |
| 64 | + dsp->put_vp8_epel_pixels_tab[0][2][0] = ff_put_vp8_epel16_v6_neon; | |
| 65 | + dsp->put_vp8_epel_pixels_tab[0][2][2] = ff_put_vp8_epel16_h6v6_neon; | |
| 66 | + | |
| 67 | + dsp->put_vp8_epel_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; | |
| 68 | + dsp->put_vp8_epel_pixels_tab[1][0][1] = ff_put_vp8_epel8_h4_neon; | |
| 69 | + dsp->put_vp8_epel_pixels_tab[1][0][2] = ff_put_vp8_epel8_h6_neon; | |
| 70 | + dsp->put_vp8_epel_pixels_tab[1][1][0] = ff_put_vp8_epel8_v4_neon; | |
| 71 | + dsp->put_vp8_epel_pixels_tab[1][1][1] = ff_put_vp8_epel8_h4v4_neon; | |
| 72 | + dsp->put_vp8_epel_pixels_tab[1][1][2] = ff_put_vp8_epel8_h6v4_neon; | |
| 73 | + dsp->put_vp8_epel_pixels_tab[1][2][0] = ff_put_vp8_epel8_v6_neon; | |
| 74 | + dsp->put_vp8_epel_pixels_tab[1][2][1] = ff_put_vp8_epel8_h4v6_neon; | |
| 75 | + dsp->put_vp8_epel_pixels_tab[1][2][2] = ff_put_vp8_epel8_h6v6_neon; | |
| 76 | + | |
| 77 | + dsp->put_vp8_epel_pixels_tab[2][0][1] = ff_put_vp8_epel4_h4_neon; | |
| 78 | + dsp->put_vp8_epel_pixels_tab[2][0][2] = ff_put_vp8_epel4_h6_neon; | |
| 79 | + dsp->put_vp8_epel_pixels_tab[2][1][0] = ff_put_vp8_epel4_v4_neon; | |
| 80 | + dsp->put_vp8_epel_pixels_tab[2][1][1] = ff_put_vp8_epel4_h4v4_neon; | |
| 81 | + dsp->put_vp8_epel_pixels_tab[2][1][2] = ff_put_vp8_epel4_h6v4_neon; | |
| 82 | + dsp->put_vp8_epel_pixels_tab[2][2][0] = ff_put_vp8_epel4_v6_neon; | |
| 83 | + dsp->put_vp8_epel_pixels_tab[2][2][1] = ff_put_vp8_epel4_h4v6_neon; | |
| 84 | + dsp->put_vp8_epel_pixels_tab[2][2][2] = ff_put_vp8_epel4_h6v6_neon; | |
| 85 | + | |
| 86 | + dsp->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_neon; | |
| 87 | + dsp->put_vp8_bilinear_pixels_tab[0][0][1] = ff_put_vp8_bilin16_h_neon; | |
| 88 | + dsp->put_vp8_bilinear_pixels_tab[0][0][2] = ff_put_vp8_bilin16_h_neon; | |
| 89 | + dsp->put_vp8_bilinear_pixels_tab[0][1][0] = ff_put_vp8_bilin16_v_neon; | |
| 90 | + dsp->put_vp8_bilinear_pixels_tab[0][1][1] = ff_put_vp8_bilin16_hv_neon; | |
| 91 | + dsp->put_vp8_bilinear_pixels_tab[0][1][2] = ff_put_vp8_bilin16_hv_neon; | |
| 92 | + dsp->put_vp8_bilinear_pixels_tab[0][2][0] = ff_put_vp8_bilin16_v_neon; | |
| 93 | + dsp->put_vp8_bilinear_pixels_tab[0][2][1] = ff_put_vp8_bilin16_hv_neon; | |
| 94 | + dsp->put_vp8_bilinear_pixels_tab[0][2][2] = ff_put_vp8_bilin16_hv_neon; | |
| 95 | + | |
| 96 | + dsp->put_vp8_bilinear_pixels_tab[1][0][0] = ff_put_vp8_pixels8_neon; | |
| 97 | + dsp->put_vp8_bilinear_pixels_tab[1][0][1] = ff_put_vp8_bilin8_h_neon; | |
| 98 | + dsp->put_vp8_bilinear_pixels_tab[1][0][2] = ff_put_vp8_bilin8_h_neon; | |
| 99 | + dsp->put_vp8_bilinear_pixels_tab[1][1][0] = ff_put_vp8_bilin8_v_neon; | |
| 100 | + dsp->put_vp8_bilinear_pixels_tab[1][1][1] = ff_put_vp8_bilin8_hv_neon; | |
| 101 | + dsp->put_vp8_bilinear_pixels_tab[1][1][2] = ff_put_vp8_bilin8_hv_neon; | |
| 102 | + dsp->put_vp8_bilinear_pixels_tab[1][2][0] = ff_put_vp8_bilin8_v_neon; | |
| 103 | + dsp->put_vp8_bilinear_pixels_tab[1][2][1] = ff_put_vp8_bilin8_hv_neon; | |
| 104 | + dsp->put_vp8_bilinear_pixels_tab[1][2][2] = ff_put_vp8_bilin8_hv_neon; | |
| 105 | + | |
| 106 | + dsp->put_vp8_bilinear_pixels_tab[2][0][1] = ff_put_vp8_bilin4_h_neon; | |
| 107 | + dsp->put_vp8_bilinear_pixels_tab[2][0][2] = ff_put_vp8_bilin4_h_neon; | |
| 108 | + dsp->put_vp8_bilinear_pixels_tab[2][1][0] = ff_put_vp8_bilin4_v_neon; | |
| 109 | + dsp->put_vp8_bilinear_pixels_tab[2][1][1] = ff_put_vp8_bilin4_hv_neon; | |
| 110 | + dsp->put_vp8_bilinear_pixels_tab[2][1][2] = ff_put_vp8_bilin4_hv_neon; | |
| 111 | + dsp->put_vp8_bilinear_pixels_tab[2][2][0] = ff_put_vp8_bilin4_v_neon; | |
| 112 | + dsp->put_vp8_bilinear_pixels_tab[2][2][1] = ff_put_vp8_bilin4_hv_neon; | |
| 113 | + dsp->put_vp8_bilinear_pixels_tab[2][2][2] = ff_put_vp8_bilin4_hv_neon; | |
| 114 | +} | |
| @@ -77,18 +77,6 @@ function ff_vp8_luma_dc_wht_neon, export=1 · 76 unmodified lines | ||
| 77 | 77 | bx lr |
| 78 | 78 | endfunc |
| 79 | 79 | |
| 80 | −function ff_vp8_luma_dc_wht_dc_neon, export=1 | |
| 81 | − ldrsh r2, [r1] | |
| 82 | − mov r3, #0 | |
| 83 | − add r2, r2, #3 | |
| 84 | − strh r3, [r1] | |
| 85 | − asr r2, r2, #3 | |
| 86 | − .rept 16 | |
| 87 | − strh r2, [r0], #32 | |
| 88 | − .endr | |
| 89 | − bx lr | |
| 90 | −endfunc | |
| 91 | − | |
| 92 | 80 | function ff_vp8_idct_add_neon, export=1 |
| 93 | 81 | vld1.16 {q0-q1}, [r1,:128] |
| 94 | 82 | movw r3, #20091 |
| @@ -725,23 +713,6 @@ function ff_put_vp8_pixels8_neon, export=1 · 630 unmodified lines | ||
| 725 | 713 | bx lr |
| 726 | 714 | endfunc |
| 727 | 715 | |
| 728 | −function ff_put_vp8_pixels4_neon, export=1 | |
| 729 | − ldr r12, [sp, #0] @ h | |
| 730 | − push {r4-r6,lr} | |
| 731 | −1: | |
| 732 | − subs r12, r12, #4 | |
| 733 | − ldr_post r4, r2, r3 | |
| 734 | − ldr_post r5, r2, r3 | |
| 735 | − ldr_post r6, r2, r3 | |
| 736 | − ldr_post lr, r2, r3 | |
| 737 | − str_post r4, r0, r1 | |
| 738 | − str_post r5, r0, r1 | |
| 739 | − str_post r6, r0, r1 | |
| 740 | − str_post lr, r0, r1 | |
| 741 | − bgt 1b | |
| 742 | − pop {r4-r6,pc} | |
| 743 | −endfunc | |
| 744 | − | |
| 745 | 716 | /* 4/6-tap 8th-pel MC */ |
| 746 | 717 | |
| 747 | 718 | .macro vp8_epel8_h6 d, a, b |