111
|
1 /* multimedia.c -- Builtins for HSAIL multimedia instructions.
|
|
2
|
145
|
3 Copyright (C) 2015-2020 Free Software Foundation, Inc.
|
111
|
4 Contributed by Pekka Jaaskelainen <pekka.jaaskelainen@parmance.com>
|
|
5 for General Processor Tech.
|
|
6
|
|
7 Permission is hereby granted, free of charge, to any person obtaining a
|
|
8 copy of this software and associated documentation files
|
|
9 (the "Software"), to deal in the Software without restriction, including
|
|
10 without limitation the rights to use, copy, modify, merge, publish,
|
|
11 distribute, sublicense, and/or sell copies of the Software, and to
|
|
12 permit persons to whom the Software is furnished to do so, subject to
|
|
13 the following conditions:
|
|
14
|
|
15 The above copyright notice and this permission notice shall be included
|
|
16 in all copies or substantial portions of the Software.
|
|
17
|
|
18 THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
|
|
19 OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
|
|
20 MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
|
|
21 IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,
|
|
22 DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
|
|
23 OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
|
|
24 USE OR OTHER DEALINGS IN THE SOFTWARE.
|
|
25 */
|
|
26
|
|
27 #include <math.h>
|
|
28 #include <stdint.h>
|
|
29
|
|
30 uint32_t
|
|
31 __hsail_bitalign (uint64_t lower, uint64_t upper, uint32_t shift_amount)
|
|
32 {
|
|
33 shift_amount = shift_amount & 31;
|
|
34 uint64_t packed_value = (upper << 32) | lower;
|
|
35 return (packed_value >> shift_amount) & 0xFFFFFFFF;
|
|
36 }
|
|
37
|
|
38 uint32_t
|
|
39 __hsail_bytealign (uint64_t lower, uint64_t upper, uint32_t shift_amount)
|
|
40 {
|
|
41 shift_amount = (shift_amount & 3) * 8;
|
|
42 uint64_t packed_value = (upper << 32) | lower;
|
|
43 return (packed_value >> shift_amount) & 0xFFFFFFFF;
|
|
44 }
|
|
45
|
|
46 uint32_t
|
|
47 __hsail_lerp (uint32_t a, uint32_t b, uint32_t c)
|
|
48 {
|
|
49 uint32_t e3
|
|
50 = (((((a >> 24) & 0xff) + ((b >> 24) & 0xff) + ((c >> 24) & 0x1)) / 2)
|
|
51 & 0xff)
|
|
52 << 24;
|
|
53 uint32_t e2
|
|
54 = (((((a >> 16) & 0xff) + ((b >> 16) & 0xff) + ((c >> 16) & 0x1)) / 2)
|
|
55 & 0xff)
|
|
56 << 16;
|
|
57 uint32_t e1
|
|
58 = (((((a >> 8) & 0xff) + ((b >> 8) & 0xff) + ((c >> 8) & 0x1)) / 2) & 0xff)
|
|
59 << 8;
|
|
60 uint32_t e0 = (((a & 0xff) + (b & 0xff) + (c & 0x1)) / 2) & 0xff;
|
|
61
|
|
62 return e3 | e2 | e1 | e0;
|
|
63 }
|
|
64
|
|
65 static uint8_t
|
|
66 cvt_neari_sat_u8_f32 (float a)
|
|
67 {
|
|
68 if (isinf (a))
|
|
69 {
|
|
70 if (signbit (a)) return 0;
|
|
71 else return 255;
|
|
72 }
|
|
73 else if (isnan (a)) return 0;
|
|
74 else if (a < 0.0)
|
|
75 return 0;
|
|
76 else if (a > 255.0)
|
|
77 return 255;
|
|
78 else
|
|
79 return (uint8_t) a;
|
|
80 }
|
|
81
|
|
82 uint32_t
|
|
83 __hsail_packcvt (float a, float b, float c, float d)
|
|
84 {
|
|
85 return (uint32_t) cvt_neari_sat_u8_f32 (a)
|
|
86 | (uint32_t) cvt_neari_sat_u8_f32 (b) << 8
|
|
87 | (uint32_t) cvt_neari_sat_u8_f32 (c) << 16
|
|
88 | (uint32_t) cvt_neari_sat_u8_f32 (d) << 24;
|
|
89 }
|
|
90
|
|
91 float
|
|
92 __hsail_unpackcvt (uint32_t val, uint32_t index)
|
|
93 {
|
|
94 return (float) ((val >> (index * 8)) & 0xff);
|
|
95 }
|
|
96
|
|
97 static uint32_t
|
|
98 abs_diff (uint32_t a, uint32_t b)
|
|
99 {
|
|
100 if (a < b)
|
|
101 return b - a;
|
|
102 else
|
|
103 return a - b;
|
|
104 }
|
|
105
|
|
106 uint32_t
|
|
107 __hsail_sad_u8x4 (uint32_t a, uint32_t b, uint32_t add)
|
|
108 {
|
|
109 return abs_diff ((a >> 24) & 0xff, (b >> 24) & 0xff)
|
|
110 + abs_diff ((a >> 16) & 0xff, (b >> 16) & 0xff)
|
|
111 + abs_diff ((a >> 8) & 0xff, (b >> 8) & 0xff)
|
|
112 + abs_diff ((a >> 0) & 0xff, (b >> 0) & 0xff) + add;
|
|
113 }
|
|
114
|
|
115 uint32_t
|
|
116 __hsail_sad_u16x2 (uint32_t a, uint32_t b, uint32_t add)
|
|
117 {
|
|
118 return abs_diff ((a >> 16) & 0xffff, (b >> 16) & 0xffff)
|
|
119 + abs_diff ((a >> 0) & 0xffff, (b >> 0) & 0xffff) + add;
|
|
120 }
|
|
121
|
|
122 uint32_t
|
|
123 __hsail_sad_u32 (uint32_t a, uint32_t b, uint32_t add)
|
|
124 {
|
|
125 return abs_diff (a, b) + add;
|
|
126 }
|
|
127
|
|
128 uint32_t
|
|
129 __hsail_sadhi_u16x2_u8x4 (uint32_t a, uint32_t b, uint32_t add)
|
|
130 {
|
|
131 return (abs_diff ((a >> 24) & 0xff, (b >> 24) & 0xff) << 16)
|
|
132 + (abs_diff ((a >> 16) & 0xff, (b >> 16) & 0xff) << 16)
|
|
133 + (abs_diff ((a >> 8) & 0xff, (b >> 8) & 0xff) << 16)
|
|
134 + (abs_diff ((a >> 0) & 0xff, (b >> 0) & 0xff) << 16) + add;
|
|
135 }
|