filters/intel/resample_avx512.cpp
| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | // Avisynth v2.5. Copyright 2002 Ben Rudiak-Gould et al. | ||
| 2 | // http://avisynth.nl | ||
| 3 | |||
| 4 | // This program is free software; you can redistribute it and/or modify | ||
| 5 | // it under the terms of the GNU General Public License as published by | ||
| 6 | // the Free Software Foundation; either version 2 of the License, or | ||
| 7 | // (at your option) any later version. | ||
| 8 | // | ||
| 9 | // This program 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 | ||
| 12 | // GNU General Public License for more details. | ||
| 13 | // | ||
| 14 | // You should have received a copy of the GNU General Public License | ||
| 15 | // along with this program; if not, write to the Free Software | ||
| 16 | // Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA, or visit | ||
| 17 | // http://www.gnu.org/copyleft/gpl.html . | ||
| 18 | // | ||
| 19 | // Linking Avisynth statically or dynamically with other modules is making a | ||
| 20 | // combined work based on Avisynth. Thus, the terms and conditions of the GNU | ||
| 21 | // General Public License cover the whole combination. | ||
| 22 | // | ||
| 23 | // As a special exception, the copyright holders of Avisynth give you | ||
| 24 | // permission to link Avisynth with independent modules that communicate with | ||
| 25 | // Avisynth solely through the interfaces defined in avisynth.h, regardless of the license | ||
| 26 | // terms of these independent modules, and to copy and distribute the | ||
| 27 | // resulting combined work under terms of your choice, provided that | ||
| 28 | // every copy of the combined work is accompanied by a complete copy of | ||
| 29 | // the source code of Avisynth (the version of Avisynth used to produce the | ||
| 30 | // combined work), being distributed under the terms of the GNU General | ||
| 31 | // Public License plus this exception. An independent module is a module | ||
| 32 | // which is not derived from or based on Avisynth, such as 3rd-party filters, | ||
| 33 | // import and export plugins, or graphical user interfaces. | ||
| 34 | |||
| 35 | #include <avs/config.h> | ||
| 36 | #include "../core/internal.h" | ||
| 37 | |||
| 38 | #include <avs/alignment.h> | ||
| 39 | #include <avs/minmax.h> | ||
| 40 | |||
| 41 | #include "check_avx512.h" // compiler avx512 directives check | ||
| 42 | #include "resample_avx512.h" | ||
| 43 | #include <type_traits> | ||
| 44 | |||
| 45 | #include <immintrin.h> // Includes AVX-512 intrinsics | ||
| 46 | |||
| 47 | #include "resample_avx512.hpp" | ||
| 48 | |||
| 49 | // uint8_t h pretransposed_coeffs _vnni wrappers | ||
| 50 | |||
| 51 | 75 | void resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks4_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 52 | // template parameter true: UseVNNI | ||
| 53 | 75 | resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks4_pretransposed_coeffs_internal<true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 54 | 75 | } | |
| 55 | 4 | void resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks8_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 56 | 4 | resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks8_pretransposed_coeffs_internal<true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 57 | 4 | } | |
| 58 | 1 | void resize_h_planar_uint8_avx512_permutex_vstripe_2s32_ks8_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 59 | 1 | resize_h_planar_uint8_avx512_permutex_vstripe_2s32_ks8_pretransposed_coeffs_internal<true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 60 | 1 | } | |
| 61 | 10 | void resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks16_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 62 | 10 | resize_h_planar_uint8_avx512_permutex_vstripe_mpz_ks16_pretransposed_coeffs_internal<true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 63 | 10 | } | |
| 64 | 1 | void resize_h_planar_uint8_avx512_permutex_vstripe_mpz_2s32_ks64_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 65 | 1 | resize_h_planar_uint8_avx512_permutex_vstripe_mpz_2s32_ks64_pretransposed_coeffs_internal<true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 66 | 1 | } | |
| 67 | |||
| 68 | // uint16_t h pretransposed_coeffs _vnni wrappers | ||
| 69 | |||
| 70 | template<bool lessthan16bit> | ||
| 71 | 7 | void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks4_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 72 | 7 | resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks4_pretransposed_coeffs_internal<lessthan16bit, true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 73 | 7 | } | |
| 74 | template<bool lessthan16bit> | ||
| 75 | 5 | void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks8_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 76 | 5 | resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks8_pretransposed_coeffs_internal<lessthan16bit, true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 77 | 5 | } | |
| 78 | template<bool lessthan16bit> | ||
| 79 | 2 | void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks8_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 80 | 2 | resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks8_pretransposed_coeffs_internal<lessthan16bit, true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 81 | 2 | } | |
| 82 | template<bool lessthan16bit> | ||
| 83 | 2 | void resize_h_planar_uint16_avx512_permutex_vstripe_mp_ks16_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 84 | 2 | resize_h_planar_uint16_avx512_permutex_vstripe_mp_ks16_pretransposed_coeffs_internal<lessthan16bit, true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 85 | 2 | } | |
| 86 | template<bool lessthan16bit> | ||
| 87 | 2 | void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks48_pretransposed_coeffs_vnni(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel) { | |
| 88 | 2 | resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks48_pretransposed_coeffs_internal<lessthan16bit, true>(dst8, src8, dst_pitch, src_pitch, program, width, height, bits_per_pixel); | |
| 89 | 2 | } | |
| 90 | |||
| 91 | // Explicit template instantiations for pretransposed uint16 _vnni variants | ||
| 92 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks4_pretransposed_coeffs_vnni<false>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 93 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks4_pretransposed_coeffs_vnni<true>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 94 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks8_pretransposed_coeffs_vnni<false>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 95 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_2s32_ks8_pretransposed_coeffs_vnni<true>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 96 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks8_pretransposed_coeffs_vnni<false>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 97 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks8_pretransposed_coeffs_vnni<true>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 98 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_ks16_pretransposed_coeffs_vnni<false>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 99 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_ks16_pretransposed_coeffs_vnni<true>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 100 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks48_pretransposed_coeffs_vnni<false>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 101 | template void resize_h_planar_uint16_avx512_permutex_vstripe_mp_4s16_ks48_pretransposed_coeffs_vnni<true>(BYTE* dst8, const BYTE* src8, int dst_pitch, int src_pitch, ResamplingProgram* program, int width, int height, int bits_per_pixel); | ||
| 102 |