FFmpeg  4.0
mmiutils.h
Go to the documentation of this file.
1 /*
2  * Loongson SIMD utils
3  *
4  * Copyright (c) 2016 Loongson Technology Corporation Limited
5  * Copyright (c) 2016 Zhou Xiaoyong <zhouxiaoyong@loongson.cn>
6  *
7  * This file is part of FFmpeg.
8  *
9  * FFmpeg is free software; you can redistribute it and/or
10  * modify it under the terms of the GNU Lesser General Public
11  * License as published by the Free Software Foundation; either
12  * version 2.1 of the License, or (at your option) any later version.
13  *
14  * FFmpeg is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17  * Lesser General Public License for more details.
18  *
19  * You should have received a copy of the GNU Lesser General Public
20  * License along with FFmpeg; if not, write to the Free Software
21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22  */
23 
24 #ifndef AVUTIL_MIPS_MMIUTILS_H
25 #define AVUTIL_MIPS_MMIUTILS_H
26 
27 #include "config.h"
28 #include "libavutil/mips/asmdefs.h"
29 
30 #if HAVE_LOONGSON2
31 
32 #define DECLARE_VAR_LOW32 int32_t low32
33 #define RESTRICT_ASM_LOW32 [low32]"=&r"(low32),
34 #define DECLARE_VAR_ALL64 int64_t all64
35 #define RESTRICT_ASM_ALL64 [all64]"=&r"(all64),
36 #define DECLARE_VAR_ADDRT mips_reg addrt
37 #define RESTRICT_ASM_ADDRT [addrt]"=&r"(addrt),
38 
39 #define MMI_LWX(reg, addr, stride, bias) \
40  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
41  "lw "#reg", "#bias"(%[addrt]) \n\t"
42 
43 #define MMI_SWX(reg, addr, stride, bias) \
44  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
45  "sw "#reg", "#bias"(%[addrt]) \n\t"
46 
47 #define MMI_LDX(reg, addr, stride, bias) \
48  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
49  "ld "#reg", "#bias"(%[addrt]) \n\t"
50 
51 #define MMI_SDX(reg, addr, stride, bias) \
52  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
53  "sd "#reg", "#bias"(%[addrt]) \n\t"
54 
55 #define MMI_LWC1(fp, addr, bias) \
56  "lwc1 "#fp", "#bias"("#addr") \n\t"
57 
58 #define MMI_ULWC1(fp, addr, bias) \
59  "ulw %[low32], "#bias"("#addr") \n\t" \
60  "mtc1 %[low32], "#fp" \n\t"
61 
62 #define MMI_LWXC1(fp, addr, stride, bias) \
63  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
64  MMI_LWC1(fp, %[addrt], bias)
65 
66 #define MMI_SWC1(fp, addr, bias) \
67  "swc1 "#fp", "#bias"("#addr") \n\t"
68 
69 #define MMI_USWC1(fp, addr, bias) \
70  "mfc1 %[low32], "#fp" \n\t" \
71  "usw %[low32], "#bias"("#addr") \n\t"
72 
73 #define MMI_SWXC1(fp, addr, stride, bias) \
74  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
75  MMI_SWC1(fp, %[addrt], bias)
76 
77 #define MMI_LDC1(fp, addr, bias) \
78  "ldc1 "#fp", "#bias"("#addr") \n\t"
79 
80 #define MMI_ULDC1(fp, addr, bias) \
81  "uld %[all64], "#bias"("#addr") \n\t" \
82  "dmtc1 %[all64], "#fp" \n\t"
83 
84 #define MMI_LDXC1(fp, addr, stride, bias) \
85  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
86  MMI_LDC1(fp, %[addrt], bias)
87 
88 #define MMI_SDC1(fp, addr, bias) \
89  "sdc1 "#fp", "#bias"("#addr") \n\t"
90 
91 #define MMI_USDC1(fp, addr, bias) \
92  "dmfc1 %[all64], "#fp" \n\t" \
93  "usd %[all64], "#bias"("#addr") \n\t"
94 
95 #define MMI_SDXC1(fp, addr, stride, bias) \
96  PTR_ADDU "%[addrt], "#addr", "#stride" \n\t" \
97  MMI_SDC1(fp, %[addrt], bias)
98 
99 #define MMI_LQ(reg1, reg2, addr, bias) \
100  "ld "#reg1", "#bias"("#addr") \n\t" \
101  "ld "#reg2", 8+"#bias"("#addr") \n\t"
102 
103 #define MMI_SQ(reg1, reg2, addr, bias) \
104  "sd "#reg1", "#bias"("#addr") \n\t" \
105  "sd "#reg2", 8+"#bias"("#addr") \n\t"
106 
107 #define MMI_LQC1(fp1, fp2, addr, bias) \
108  "ldc1 "#fp1", "#bias"("#addr") \n\t" \
109  "ldc1 "#fp2", 8+"#bias"("#addr") \n\t"
110 
111 #define MMI_SQC1(fp1, fp2, addr, bias) \
112  "sdc1 "#fp1", "#bias"("#addr") \n\t" \
113  "sdc1 "#fp2", 8+"#bias"("#addr") \n\t"
114 
115 #elif HAVE_LOONGSON3 /* !HAVE_LOONGSON2 */
116 
117 #define DECLARE_VAR_ALL64
118 #define RESTRICT_ASM_ALL64
119 #define DECLARE_VAR_ADDRT
120 #define RESTRICT_ASM_ADDRT
121 
122 #define MMI_LWX(reg, addr, stride, bias) \
123  "gslwx "#reg", "#bias"("#addr", "#stride") \n\t"
124 
125 #define MMI_SWX(reg, addr, stride, bias) \
126  "gsswx "#reg", "#bias"("#addr", "#stride") \n\t"
127 
128 #define MMI_LDX(reg, addr, stride, bias) \
129  "gsldx "#reg", "#bias"("#addr", "#stride") \n\t"
130 
131 #define MMI_SDX(reg, addr, stride, bias) \
132  "gssdx "#reg", "#bias"("#addr", "#stride") \n\t"
133 
134 #define MMI_LWC1(fp, addr, bias) \
135  "lwc1 "#fp", "#bias"("#addr") \n\t"
136 
137 #if _MIPS_SIM == _ABIO32 /* workaround for 3A2000 gslwlc1 bug */
138 
139 #define DECLARE_VAR_LOW32 int32_t low32
140 #define RESTRICT_ASM_LOW32 [low32]"=&r"(low32),
141 
142 #define MMI_ULWC1(fp, addr, bias) \
143  "ulw %[low32], "#bias"("#addr") \n\t" \
144  "mtc1 %[low32], "#fp" \n\t"
145 
146 #else /* _MIPS_SIM != _ABIO32 */
147 
148 #define DECLARE_VAR_LOW32
149 #define RESTRICT_ASM_LOW32
150 
151 #define MMI_ULWC1(fp, addr, bias) \
152  "gslwlc1 "#fp", 3+"#bias"("#addr") \n\t" \
153  "gslwrc1 "#fp", "#bias"("#addr") \n\t"
154 
155 #endif /* _MIPS_SIM != _ABIO32 */
156 
157 #define MMI_LWXC1(fp, addr, stride, bias) \
158  "gslwxc1 "#fp", "#bias"("#addr", "#stride") \n\t"
159 
160 #define MMI_SWC1(fp, addr, bias) \
161  "swc1 "#fp", "#bias"("#addr") \n\t"
162 
163 #define MMI_USWC1(fp, addr, bias) \
164  "gsswlc1 "#fp", 3+"#bias"("#addr") \n\t" \
165  "gsswrc1 "#fp", "#bias"("#addr") \n\t"
166 
167 #define MMI_SWXC1(fp, addr, stride, bias) \
168  "gsswxc1 "#fp", "#bias"("#addr", "#stride") \n\t"
169 
170 #define MMI_LDC1(fp, addr, bias) \
171  "ldc1 "#fp", "#bias"("#addr") \n\t"
172 
173 #define MMI_ULDC1(fp, addr, bias) \
174  "gsldlc1 "#fp", 7+"#bias"("#addr") \n\t" \
175  "gsldrc1 "#fp", "#bias"("#addr") \n\t"
176 
177 #define MMI_LDXC1(fp, addr, stride, bias) \
178  "gsldxc1 "#fp", "#bias"("#addr", "#stride") \n\t"
179 
180 #define MMI_SDC1(fp, addr, bias) \
181  "sdc1 "#fp", "#bias"("#addr") \n\t"
182 
183 #define MMI_USDC1(fp, addr, bias) \
184  "gssdlc1 "#fp", 7+"#bias"("#addr") \n\t" \
185  "gssdrc1 "#fp", "#bias"("#addr") \n\t"
186 
187 #define MMI_SDXC1(fp, addr, stride, bias) \
188  "gssdxc1 "#fp", "#bias"("#addr", "#stride") \n\t"
189 
190 #define MMI_LQ(reg1, reg2, addr, bias) \
191  "gslq "#reg1", "#reg2", "#bias"("#addr") \n\t"
192 
193 #define MMI_SQ(reg1, reg2, addr, bias) \
194  "gssq "#reg1", "#reg2", "#bias"("#addr") \n\t"
195 
196 #define MMI_LQC1(fp1, fp2, addr, bias) \
197  "gslqc1 "#fp1", "#fp2", "#bias"("#addr") \n\t"
198 
199 #define MMI_SQC1(fp1, fp2, addr, bias) \
200  "gssqc1 "#fp1", "#fp2", "#bias"("#addr") \n\t"
201 
202 #endif /* HAVE_LOONGSON2 */
203 
204 #define TRANSPOSE_4H(m1, m2, m3, m4, t1, t2, t3, t4, t5, r1, zero, shift) \
205  "li "#r1", 0x93 \n\t" \
206  "xor "#zero","#zero","#zero" \n\t" \
207  "mtc1 "#r1", "#shift" \n\t" \
208  "punpcklhw "#t1", "#m1", "#zero" \n\t" \
209  "punpcklhw "#t5", "#m2", "#zero" \n\t" \
210  "pshufh "#t5", "#t5", "#shift" \n\t" \
211  "or "#t1", "#t1", "#t5" \n\t" \
212  "punpckhhw "#t2", "#m1", "#zero" \n\t" \
213  "punpckhhw "#t5", "#m2", "#zero" \n\t" \
214  "pshufh "#t5", "#t5", "#shift" \n\t" \
215  "or "#t2", "#t2", "#t5" \n\t" \
216  "punpcklhw "#t3", "#m3", "#zero" \n\t" \
217  "punpcklhw "#t5", "#m4", "#zero" \n\t" \
218  "pshufh "#t5", "#t5", "#shift" \n\t" \
219  "or "#t3", "#t3", "#t5" \n\t" \
220  "punpckhhw "#t4", "#m3", "#zero" \n\t" \
221  "punpckhhw "#t5", "#m4", "#zero" \n\t" \
222  "pshufh "#t5", "#t5", "#shift" \n\t" \
223  "or "#t4", "#t4", "#t5" \n\t" \
224  "punpcklwd "#m1", "#t1", "#t3" \n\t" \
225  "punpckhwd "#m2", "#t1", "#t3" \n\t" \
226  "punpcklwd "#m3", "#t2", "#t4" \n\t" \
227  "punpckhwd "#m4", "#t2", "#t4" \n\t"
228 
229 
230 #define PSRAH_4_MMI(fp1, fp2, fp3, fp4, shift) \
231  "psrah "#fp1", "#fp1", "#shift" \n\t" \
232  "psrah "#fp2", "#fp2", "#shift" \n\t" \
233  "psrah "#fp3", "#fp3", "#shift" \n\t" \
234  "psrah "#fp4", "#fp4", "#shift" \n\t"
235 
236 #define PSRAH_8_MMI(fp1, fp2, fp3, fp4, fp5, fp6, fp7, fp8, shift) \
237  PSRAH_4_MMI(fp1, fp2, fp3, fp4, shift) \
238  PSRAH_4_MMI(fp5, fp6, fp7, fp8, shift)
239 
240 
241 #endif /* AVUTILS_MIPS_MMIUTILS_H */
MIPS assembly defines from sys/asm.h but rewritten for use with C inline assembly (rather than from w...