diff options
author | Matt A. Tobin <email@mattatobin.com> | 2020-04-07 23:30:51 -0400 |
---|---|---|
committer | Matt A. Tobin <email@mattatobin.com> | 2020-04-07 23:30:51 -0400 |
commit | 5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb (patch) | |
tree | 45d55e3e5e73c4255c4d71258d9be5b2d004d28f /media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h | |
parent | 50f1986697a7412e4160976fa5e11217b4ef1f44 (diff) | |
download | UXP-5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb.tar UXP-5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb.tar.gz UXP-5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb.tar.lz UXP-5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb.tar.xz UXP-5545a8983ff0ef1fb52e64aef8e66fa9b13c1cbb.zip |
Move aom source to a sub-directory under media/libaom
There is no damned reason to treat this differently than any other media lib given its license and there never was.
Diffstat (limited to 'media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h')
-rw-r--r-- | media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h | 54 |
1 files changed, 54 insertions, 0 deletions
diff --git a/media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h b/media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h new file mode 100644 index 000000000..48486c6c4 --- /dev/null +++ b/media/libaom/src/aom_dsp/x86/obmc_intrinsic_ssse3.h @@ -0,0 +1,54 @@ +/* + * Copyright (c) 2017, Alliance for Open Media. All rights reserved + * + * This source code is subject to the terms of the BSD 2 Clause License and + * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License + * was not distributed with this source code in the LICENSE file, you can + * obtain it at www.aomedia.org/license/software. If the Alliance for Open + * Media Patent License 1.0 was not distributed with this source code in the + * PATENTS file, you can obtain it at www.aomedia.org/license/patent. + */ + +#ifndef AOM_AOM_DSP_X86_OBMC_INTRINSIC_SSSE3_H_ +#define AOM_AOM_DSP_X86_OBMC_INTRINSIC_SSSE3_H_ + +#include <immintrin.h> + +#include "config/aom_config.h" + +static INLINE int32_t xx_hsum_epi32_si32(__m128i v_d) { + v_d = _mm_hadd_epi32(v_d, v_d); + v_d = _mm_hadd_epi32(v_d, v_d); + return _mm_cvtsi128_si32(v_d); +} + +static INLINE int64_t xx_hsum_epi64_si64(__m128i v_q) { + v_q = _mm_add_epi64(v_q, _mm_srli_si128(v_q, 8)); +#if ARCH_X86_64 + return _mm_cvtsi128_si64(v_q); +#else + { + int64_t tmp; + _mm_storel_epi64((__m128i *)&tmp, v_q); + return tmp; + } +#endif +} + +static INLINE int64_t xx_hsum_epi32_si64(__m128i v_d) { + const __m128i v_sign_d = _mm_cmplt_epi32(v_d, _mm_setzero_si128()); + const __m128i v_0_q = _mm_unpacklo_epi32(v_d, v_sign_d); + const __m128i v_1_q = _mm_unpackhi_epi32(v_d, v_sign_d); + return xx_hsum_epi64_si64(_mm_add_epi64(v_0_q, v_1_q)); +} + +// This is equivalent to ROUND_POWER_OF_TWO_SIGNED(v_val_d, bits) +static INLINE __m128i xx_roundn_epi32(__m128i v_val_d, int bits) { + const __m128i v_bias_d = _mm_set1_epi32((1 << bits) >> 1); + const __m128i v_sign_d = _mm_srai_epi32(v_val_d, 31); + const __m128i v_tmp_d = + _mm_add_epi32(_mm_add_epi32(v_val_d, v_bias_d), v_sign_d); + return _mm_srai_epi32(v_tmp_d, bits); +} + +#endif // AOM_AOM_DSP_X86_OBMC_INTRINSIC_SSSE3_H_ |