1 /******************************************************************************
2  *
3  * Copyright (C) 2015 The Android Open Source Project
4  *
5  * Licensed under the Apache License, Version 2.0 (the "License");
6  * you may not use this file except in compliance with the License.
7  * You may obtain a copy of the License at:
8  *
9  * http://www.apache.org/licenses/LICENSE-2.0
10  *
11  * Unless required by applicable law or agreed to in writing, software
12  * distributed under the License is distributed on an "AS IS" BASIS,
13  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14  * See the License for the specific language governing permissions and
15  * limitations under the License.
16  *
17  *****************************************************************************
18  * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
19 */
20 
21 /**
22  *******************************************************************************
23  * @file
24  *  impeg2_mem_func_sse42_intr.c
25  *
26  * @brief
27  *  Contains utility function definitions for MPEG2 codec
28  *
29  * @author
30  *  Mohit [100664]
31  *
32 * @par List of Functions:
33 * - impeg2_memset0_16bit_8x8_linear_block_sse42()
34 * - impeg2_memset_8bit_8x8_block_sse42()
35  *
36  * @remarks
37  *  None
38  *
39  *******************************************************************************
40  */
41 #include <stdio.h>
42 #include <string.h>
43 #include "iv_datatypedef.h"
44 #include "impeg2_defs.h"
45 
46 #include <immintrin.h>
47 #include <emmintrin.h>
48 #include <smmintrin.h>
49 #include <tmmintrin.h>
50 
51 /*******************************************************************************
52 *  Function Name   : impeg2_memset0_16bit_8x8_linear_block
53 *
54 *  Description     : memsets resudial buf to 0
55 *
56 *  Arguments       : destination buffer
57 *
58 *  Values Returned : None
59 *******************************************************************************/
60 
61 
62 void impeg2_memset0_16bit_8x8_linear_block_sse42 (WORD16 *buf)
63  {
64     __m128i zero_8x8_16b = _mm_set1_epi16(0);
65     _mm_storeu_si128((__m128i *) buf, zero_8x8_16b);
66     _mm_storeu_si128((__m128i *) (buf + 8), zero_8x8_16b);
67     _mm_storeu_si128((__m128i *) (buf + 16), zero_8x8_16b);
68     _mm_storeu_si128((__m128i *) (buf + 24), zero_8x8_16b);
69     _mm_storeu_si128((__m128i *) (buf + 32), zero_8x8_16b);
70     _mm_storeu_si128((__m128i *) (buf + 40), zero_8x8_16b);
71     _mm_storeu_si128((__m128i *) (buf + 48), zero_8x8_16b);
72     _mm_storeu_si128((__m128i *) (buf + 56), zero_8x8_16b);
73 }
74 
75 
76 
77 /*******************************************************************************
78 *  Function Name   : impeg2_memset_8bit_8x8_block
79 *
80 *  Description     : memsets residual buf to value
81 *
82 *  Arguments       : destination buffer, value and stride
83 *
84 *  Values Returned : None
85 *******************************************************************************/
86 
87 
88 void impeg2_memset_8bit_8x8_block_sse42(UWORD8 *dst, WORD32 dc_val, WORD32 dst_wd)
89 {
90     __m128i value = _mm_set1_epi8((WORD8)dc_val);
91 
92     _mm_storel_epi64((__m128i *)dst, value);
93     _mm_storel_epi64((__m128i *) (dst + dst_wd), value);
94     _mm_storel_epi64((__m128i *) (dst + 2 * dst_wd), value);
95     _mm_storel_epi64((__m128i *) (dst + 3 * dst_wd), value);
96     _mm_storel_epi64((__m128i *) (dst + 4 * dst_wd), value);
97     _mm_storel_epi64((__m128i *) (dst + 5 * dst_wd), value);
98     _mm_storel_epi64((__m128i *) (dst + 6 * dst_wd), value);
99     _mm_storel_epi64((__m128i *) (dst + 7 * dst_wd), value);
100 }
101