1 /*
2 * Copyright (c) 2017 The WebM project authors. All Rights Reserved.
3 *
4 * Use of this source code is governed by a BSD-style license
5 * that can be found in the LICENSE file in the root of the source
6 * tree. An additional intellectual property rights grant can be found
7 * in the file PATENTS. All contributing project authors may
8 * be found in the AUTHORS file in the root of the source tree.
9 */
10 #ifndef VPX_DSP_X86_BITDEPTH_CONVERSION_SSE2_H_
11 #define VPX_DSP_X86_BITDEPTH_CONVERSION_SSE2_H_
12
13 #include <xmmintrin.h>
14
15 #include "./vpx_config.h"
16 #include "vpx/vpx_integer.h"
17 #include "vpx_dsp/vpx_dsp_common.h"
18
19 // Load 8 16 bit values. If the source is 32 bits then pack down with
20 // saturation.
load_tran_low(const tran_low_t * a)21 static INLINE __m128i load_tran_low(const tran_low_t *a) {
22 #if CONFIG_VP9_HIGHBITDEPTH
23 const __m128i a_low = _mm_load_si128((const __m128i *)a);
24 return _mm_packs_epi32(a_low, *(const __m128i *)(a + 4));
25 #else
26 return _mm_load_si128((const __m128i *)a);
27 #endif
28 }
29
30 // Store 8 16 bit values. If the destination is 32 bits then sign extend the
31 // values by multiplying by 1.
store_tran_low(__m128i a,tran_low_t * b)32 static INLINE void store_tran_low(__m128i a, tran_low_t *b) {
33 #if CONFIG_VP9_HIGHBITDEPTH
34 const __m128i one = _mm_set1_epi16(1);
35 const __m128i a_hi = _mm_mulhi_epi16(a, one);
36 const __m128i a_lo = _mm_mullo_epi16(a, one);
37 const __m128i a_1 = _mm_unpacklo_epi16(a_lo, a_hi);
38 const __m128i a_2 = _mm_unpackhi_epi16(a_lo, a_hi);
39 _mm_store_si128((__m128i *)(b), a_1);
40 _mm_store_si128((__m128i *)(b + 4), a_2);
41 #else
42 _mm_store_si128((__m128i *)(b), a);
43 #endif
44 }
45
46 // Zero fill 8 positions in the output buffer.
store_zero_tran_low(tran_low_t * a)47 static INLINE void store_zero_tran_low(tran_low_t *a) {
48 const __m128i zero = _mm_setzero_si128();
49 #if CONFIG_VP9_HIGHBITDEPTH
50 _mm_store_si128((__m128i *)(a), zero);
51 _mm_store_si128((__m128i *)(a + 4), zero);
52 #else
53 _mm_store_si128((__m128i *)(a), zero);
54 #endif
55 }
56 #endif // VPX_DSP_X86_BITDEPTH_CONVERSION_SSE2_H_
57