1/*
2 * Copyright (c) 2011 Mans Rullgard <mans@mansr.com>
3 *
4 * This file is part of Libav.
5 *
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21#ifndef AVCODEC_ARM_DCA_H
22#define AVCODEC_ARM_DCA_H
23
24#include <stdint.h>
25#include "config.h"
26#include "libavutil/intmath.h"
27
28#if HAVE_ARMV6 && HAVE_INLINE_ASM && AV_GCC_VERSION_AT_LEAST(4,4)
29
30#define decode_blockcodes decode_blockcodes
31static inline int decode_blockcodes(int code1, int code2, int levels,
32                                    int *values)
33{
34    int v0, v1, v2, v3, v4, v5;
35
36    __asm__ ("smmul   %8,  %14, %18           \n"
37             "smmul   %11, %15, %18           \n"
38             "smlabb  %14, %8,  %17, %14      \n"
39             "smlabb  %15, %11, %17, %15      \n"
40             "smmul   %9,  %8,  %18           \n"
41             "smmul   %12, %11, %18           \n"
42             "sub     %14, %14, %16, lsr #1   \n"
43             "sub     %15, %15, %16, lsr #1   \n"
44             "smlabb  %8,  %9,  %17, %8       \n"
45             "smlabb  %11, %12, %17, %11      \n"
46             "smmul   %10, %9,  %18           \n"
47             "smmul   %13, %12, %18           \n"
48             "str     %14, %0                 \n"
49             "str     %15, %4                 \n"
50             "sub     %8,  %8,  %16, lsr #1   \n"
51             "sub     %11, %11, %16, lsr #1   \n"
52             "smlabb  %9,  %10, %17, %9       \n"
53             "smlabb  %12, %13, %17, %12      \n"
54             "smmul   %14, %10, %18           \n"
55             "smmul   %15, %13, %18           \n"
56             "str     %8,  %1                 \n"
57             "str     %11, %5                 \n"
58             "sub     %9,  %9,  %16, lsr #1   \n"
59             "sub     %12, %12, %16, lsr #1   \n"
60             "smlabb  %10, %14, %17, %10      \n"
61             "smlabb  %13, %15, %17, %13      \n"
62             "str     %9,  %2                 \n"
63             "str     %12, %6                 \n"
64             "sub     %10, %10, %16, lsr #1   \n"
65             "sub     %13, %13, %16, lsr #1   \n"
66             "str     %10, %3                 \n"
67             "str     %13, %7                 \n"
68             : "=m"(values[0]), "=m"(values[1]),
69               "=m"(values[2]), "=m"(values[3]),
70               "=m"(values[4]), "=m"(values[5]),
71               "=m"(values[6]), "=m"(values[7]),
72               "=&r"(v0), "=&r"(v1), "=&r"(v2),
73               "=&r"(v3), "=&r"(v4), "=&r"(v5),
74               "+&r"(code1), "+&r"(code2)
75             : "r"(levels - 1), "r"(-levels), "r"(ff_inverse[levels]));
76
77    return code1 | code2;
78}
79
80#endif
81
82#if HAVE_NEON && HAVE_INLINE_ASM && HAVE_ASM_MOD_Y
83
84#define int8x8_fmul_int32 int8x8_fmul_int32
85static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
86{
87    __asm__ ("vcvt.f32.s32 %2,  %2,  #4         \n"
88             "vld1.8       {d0},     [%1,:64]   \n"
89             "vmovl.s8     q0,  d0              \n"
90             "vmovl.s16    q1,  d1              \n"
91             "vmovl.s16    q0,  d0              \n"
92             "vcvt.f32.s32 q0,  q0              \n"
93             "vcvt.f32.s32 q1,  q1              \n"
94             "vmul.f32     q0,  q0,  %y2        \n"
95             "vmul.f32     q1,  q1,  %y2        \n"
96             "vst1.32      {q0-q1},  [%m0,:128] \n"
97             : "=Um"(*(float (*)[8])dst)
98             : "r"(src), "x"(scale)
99             : "d0", "d1", "d2", "d3");
100}
101
102#endif
103
104#endif /* AVCODEC_ARM_DCA_H */
105