xref: /nrf52832-nimble/rt-thread/libcpu/c-sky/common/csi_simd.h (revision 104654410c56c573564690304ae786df310c91fc)
1 /*
2  * Copyright (C) 2017 C-SKY Microsystems Co., Ltd. All rights reserved.
3  *
4  * Licensed under the Apache License, Version 2.0 (the "License");
5  * you may not use this file except in compliance with the License.
6  * You may obtain a copy of the License at
7  *
8  *   http://www.apache.org/licenses/LICENSE-2.0
9  *
10  * Unless required by applicable law or agreed to in writing, software
11  * distributed under the License is distributed on an "AS IS" BASIS,
12  * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13  * See the License for the specific language governing permissions and
14  * limitations under the License.
15  */
16 
17 /******************************************************************************
18  * @file     csi_simd.h
19  * @brief    CSI Single Instruction Multiple Data (SIMD) Header File for GCC.
20  * @version  V1.0
21  * @date     02. June 2017
22  ******************************************************************************/
23 
24 #ifndef _CSI_SIMD_H_
25 #define _CSI_SIMD_H_
26 
27 /**
28   \brief   Halfword packing instruction. Combines bits[15:0] of val1 with bits[31:16]
29            of val2 levitated with the val3.
30   \details Combine a halfword from one register with a halfword from another register.
31            The second argument can be left-shifted before extraction of the halfword.
32   \param [in]    val1   first 16-bit operands
33   \param [in]    val2   second 16-bit operands
34   \param [in]    val3   value for left-shifting val2. Value range [0..31].
35   \return               the combination of halfwords.
36   \remark
37                  res[15:0]  = val1[15:0]              \n
38                  res[31:16] = val2[31:16] << val3
39  */
__PKHBT(uint32_t val1,uint32_t val2,uint32_t val3)40 __ALWAYS_INLINE uint32_t __PKHBT(uint32_t val1, uint32_t val2, uint32_t val3)
41 {
42     return ((((int32_t)(val1) << 0) & (int32_t)0x0000FFFF) | (((int32_t)(val2) << val3) & (int32_t)0xFFFF0000));
43 }
44 
45 /**
46   \brief   Halfword packing instruction. Combines bits[31:16] of val1 with bits[15:0]
47            of val2 right-shifted with the val3.
48   \details Combine a halfword from one register with a halfword from another register.
49            The second argument can be right-shifted before extraction of the halfword.
50   \param [in]    val1   first 16-bit operands
51   \param [in]    val2   second 16-bit operands
52   \param [in]    val3   value for right-shifting val2. Value range [1..32].
53   \return               the combination of halfwords.
54   \remark
55                  res[15:0]  = val2[15:0] >> val3        \n
56                  res[31:16] = val1[31:16]
57  */
__PKHTB(uint32_t val1,uint32_t val2,uint32_t val3)58 __ALWAYS_INLINE uint32_t __PKHTB(uint32_t val1, uint32_t val2, uint32_t val3)
59 {
60     return ((((int32_t)(val1) << 0) & (int32_t)0xFFFF0000) | (((int32_t)(val2) >> val3) & (int32_t)0x0000FFFF));
61 }
62 
63 /**
64   \brief   Dual 16-bit signed saturate.
65   \details This function saturates a signed value.
66   \param [in]    x   two signed 16-bit values to be saturated.
67   \param [in]    y   bit position for saturation, an integral constant expression in the range 1 to 16.
68   \return        the sum of the absolute differences of the following bytes, added to the accumulation value:\n
69                  the signed saturation of the low halfword in val1, saturated to the bit position specified in
70                  val2 and returned in the low halfword of the return value.\n
71                  the signed saturation of the high halfword in val1, saturated to the bit position specified in
72                  val2 and returned in the high halfword of the return value.
73  */
__SSAT16(int32_t x,const uint32_t y)74 __ALWAYS_INLINE uint32_t __SSAT16(int32_t x, const uint32_t y)
75 {
76     int32_t r = 0, s = 0;
77 
78     r = __SSAT((((int32_t)x << 16) >> 16), y) & (int32_t)0x0000FFFF;
79     s = __SSAT((((int32_t)x) >> 16), y) & (int32_t)0x0000FFFF;
80 
81     return ((uint32_t)((s << 16) | (r)));
82 }
83 
84 /**
85   \brief   Dual 16-bit unsigned saturate.
86   \details This function enables you to saturate two signed 16-bit values to a selected unsigned range.
87   \param [in]    x   two signed 16-bit values to be saturated.
88   \param [in]    y   bit position for saturation, an integral constant expression in the range 1 to 16.
89   \return        the saturation of the two signed 16-bit values, as non-negative values:
90                  the saturation of the low halfword in val1, saturated to the bit position specified in
91                  val2 and returned in the low halfword of the return value.\n
92                  the saturation of the high halfword in val1, saturated to the bit position specified in
93                  val2 and returned in the high halfword of the return value.
94  */
__USAT16(uint32_t x,const uint32_t y)95 __ALWAYS_INLINE uint32_t __USAT16(uint32_t x, const uint32_t y)
96 {
97     int32_t r = 0, s = 0;
98 
99     r = __IUSAT(((x << 16) >> 16), y) & 0x0000FFFF;
100     s = __IUSAT(((x) >> 16), y) & 0x0000FFFF;
101 
102     return ((s << 16) | (r));
103 }
104 
105 /**
106   \brief   Quad 8-bit saturating addition.
107   \details This function enables you to perform four 8-bit integer additions,
108            saturating the results to the 8-bit signed integer range -2^7 <= x <= 2^7 - 1.
109   \param [in]    x   first four 8-bit summands.
110   \param [in]    y   second four 8-bit summands.
111   \return        the saturated addition of the first byte of each operand in the first byte of the return value.\n
112                  the saturated addition of the second byte of each operand in the second byte of the return value.\n
113                  the saturated addition of the third byte of each operand in the third byte of the return value.\n
114                  the saturated addition of the fourth byte of each operand in the fourth byte of the return value.\n
115                  The returned results are saturated to the 8-bit signed integer range -2^7 <= x <= 2^7 - 1.
116   \remark
117                  res[7:0]   = val1[7:0]   + val2[7:0]        \n
118                  res[15:8]  = val1[15:8]  + val2[15:8]       \n
119                  res[23:16] = val1[23:16] + val2[23:16]      \n
120                  res[31:24] = val1[31:24] + val2[31:24]
121  */
__QADD8(uint32_t x,uint32_t y)122 __ALWAYS_INLINE uint32_t __QADD8(uint32_t x, uint32_t y)
123 {
124     int32_t r, s, t, u;
125 
126     r = __SSAT(((((int32_t)x << 24) >> 24) + (((int32_t)y << 24) >> 24)), 8) & (int32_t)0x000000FF;
127     s = __SSAT(((((int32_t)x << 16) >> 24) + (((int32_t)y << 16) >> 24)), 8) & (int32_t)0x000000FF;
128     t = __SSAT(((((int32_t)x <<  8) >> 24) + (((int32_t)y <<  8) >> 24)), 8) & (int32_t)0x000000FF;
129     u = __SSAT(((((int32_t)x) >> 24) + (((int32_t)y) >> 24)), 8) & (int32_t)0x000000FF;
130 
131     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
132 }
133 
134 /**
135   \brief   Quad 8-bit unsigned saturating addition.
136   \details This function enables you to perform four unsigned 8-bit integer additions,
137            saturating the results to the 8-bit unsigned integer range 0 < x < 2^8 - 1.
138   \param [in]    x   first four 8-bit summands.
139   \param [in]    y   second four 8-bit summands.
140   \return        the saturated addition of the first byte of each operand in the first byte of the return value.\n
141                  the saturated addition of the second byte of each operand in the second byte of the return value.\n
142                  the saturated addition of the third byte of each operand in the third byte of the return value.\n
143                  the saturated addition of the fourth byte of each operand in the fourth byte of the return value.\n
144                  The returned results are saturated to the 8-bit signed integer range 0 <= x <= 2^8 - 1.
145   \remark
146                  res[7:0]   = val1[7:0]   + val2[7:0]        \n
147                  res[15:8]  = val1[15:8]  + val2[15:8]       \n
148                  res[23:16] = val1[23:16] + val2[23:16]      \n
149                  res[31:24] = val1[31:24] + val2[31:24]
150  */
__UQADD8(uint32_t x,uint32_t y)151 __ALWAYS_INLINE uint32_t __UQADD8(uint32_t x, uint32_t y)
152 {
153     int32_t r, s, t, u;
154 
155     r = __IUSAT((((x << 24) >> 24) + ((y << 24) >> 24)), 8) & 0x000000FF;
156     s = __IUSAT((((x << 16) >> 24) + ((y << 16) >> 24)), 8) & 0x000000FF;
157     t = __IUSAT((((x <<  8) >> 24) + ((y <<  8) >> 24)), 8) & 0x000000FF;
158     u = __IUSAT((((x) >> 24) + ((y) >> 24)), 8) & 0x000000FF;
159 
160     return ((u << 24) | (t << 16) | (s <<  8) | (r));
161 }
162 
163 /**
164   \brief   Quad 8-bit signed addition.
165   \details This function performs four 8-bit signed integer additions.
166   \param [in]    x  first four 8-bit summands.
167   \param [in]    y  second four 8-bit summands.
168   \return        the addition of the first bytes from each operand, in the first byte of the return value.\n
169                  the addition of the second bytes of each operand, in the second byte of the return value.\n
170                  the addition of the third bytes of each operand, in the third byte of the return value.\n
171                  the addition of the fourth bytes of each operand, in the fourth byte of the return value.
172   \remark
173                  res[7:0]   = val1[7:0]   + val2[7:0]        \n
174                  res[15:8]  = val1[15:8]  + val2[15:8]       \n
175                  res[23:16] = val1[23:16] + val2[23:16]      \n
176                  res[31:24] = val1[31:24] + val2[31:24]
177  */
__SADD8(uint32_t x,uint32_t y)178 __ALWAYS_INLINE uint32_t __SADD8(uint32_t x, uint32_t y)
179 {
180     int32_t r, s, t, u;
181 
182     r = ((((int32_t)x << 24) >> 24) + (((int32_t)y << 24) >> 24)) & (int32_t)0x000000FF;
183     s = ((((int32_t)x << 16) >> 24) + (((int32_t)y << 16) >> 24)) & (int32_t)0x000000FF;
184     t = ((((int32_t)x <<  8) >> 24) + (((int32_t)y <<  8) >> 24)) & (int32_t)0x000000FF;
185     u = ((((int32_t)x) >> 24) + (((int32_t)y) >> 24)) & (int32_t)0x000000FF;
186 
187     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
188 }
189 
190 /**
191   \brief   Quad 8-bit unsigned addition.
192   \details This function performs four unsigned 8-bit integer additions.
193   \param [in]    x  first four 8-bit summands.
194   \param [in]    y  second four 8-bit summands.
195   \return        the addition of the first bytes from each operand, in the first byte of the return value.\n
196                  the addition of the second bytes of each operand, in the second byte of the return value.\n
197                  the addition of the third bytes of each operand, in the third byte of the return value.\n
198                  the addition of the fourth bytes of each operand, in the fourth byte of the return value.
199   \remark
200                  res[7:0]   = val1[7:0]   + val2[7:0]        \n
201                  res[15:8]  = val1[15:8]  + val2[15:8]       \n
202                  res[23:16] = val1[23:16] + val2[23:16]      \n
203                  res[31:24] = val1[31:24] + val2[31:24]
204  */
__UADD8(uint32_t x,uint32_t y)205 __ALWAYS_INLINE uint32_t __UADD8(uint32_t x, uint32_t y)
206 {
207     int32_t r, s, t, u;
208 
209     r = (((x << 24) >> 24) + ((y << 24) >> 24)) & 0x000000FF;
210     s = (((x << 16) >> 24) + ((y << 16) >> 24)) & 0x000000FF;
211     t = (((x <<  8) >> 24) + ((y <<  8) >> 24)) & 0x000000FF;
212     u = (((x) >> 24) + ((y) >> 24)) & 0x000000FF;
213 
214     return ((u << 24) | (t << 16) | (s <<  8) | (r));
215 }
216 
217 /**
218   \brief   Quad 8-bit saturating subtract.
219   \details This function enables you to perform four 8-bit integer subtractions,
220            saturating the results to the 8-bit signed integer range -2^7 <= x <= 2^7 - 1.
221   \param [in]    x   first four 8-bit summands.
222   \param [in]    y   second four 8-bit summands.
223   \return        the subtraction of the first byte of each operand in the first byte of the return value.\n
224                  the subtraction of the second byte of each operand in the second byte of the return value.\n
225                  the subtraction of the third byte of each operand in the third byte of the return value.\n
226                  the subtraction of the fourth byte of each operand in the fourth byte of the return value.\n
227                  The returned results are saturated to the 8-bit signed integer range -2^7 <= x <= 2^7 - 1.
228   \remark
229                  res[7:0]   = val1[7:0]   - val2[7:0]        \n
230                  res[15:8]  = val1[15:8]  - val2[15:8]       \n
231                  res[23:16] = val1[23:16] - val2[23:16]      \n
232                  res[31:24] = val1[31:24] - val2[31:24]
233  */
__QSUB8(uint32_t x,uint32_t y)234 __ALWAYS_INLINE uint32_t __QSUB8(uint32_t x, uint32_t y)
235 {
236     int32_t r, s, t, u;
237 
238     r = __SSAT(((((int32_t)x << 24) >> 24) - (((int32_t)y << 24) >> 24)), 8) & (int32_t)0x000000FF;
239     s = __SSAT(((((int32_t)x << 16) >> 24) - (((int32_t)y << 16) >> 24)), 8) & (int32_t)0x000000FF;
240     t = __SSAT(((((int32_t)x <<  8) >> 24) - (((int32_t)y <<  8) >> 24)), 8) & (int32_t)0x000000FF;
241     u = __SSAT(((((int32_t)x) >> 24) - (((int32_t)y) >> 24)), 8) & (int32_t)0x000000FF;
242 
243     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
244 }
245 
246 /**
247   \brief   Quad 8-bit unsigned saturating subtraction.
248   \details This function enables you to perform four unsigned 8-bit integer subtractions,
249            saturating the results to the 8-bit unsigned integer range 0 < x < 2^8 - 1.
250   \param [in]    x   first four 8-bit summands.
251   \param [in]    y   second four 8-bit summands.
252   \return        the subtraction of the first byte of each operand in the first byte of the return value.\n
253                  the subtraction of the second byte of each operand in the second byte of the return value.\n
254                  the subtraction of the third byte of each operand in the third byte of the return value.\n
255                  the subtraction of the fourth byte of each operand in the fourth byte of the return value.\n
256                  The returned results are saturated to the 8-bit unsigned integer range 0 <= x <= 2^8 - 1.
257   \remark
258                  res[7:0]   = val1[7:0]   - val2[7:0]        \n
259                  res[15:8]  = val1[15:8]  - val2[15:8]       \n
260                  res[23:16] = val1[23:16] - val2[23:16]      \n
261                  res[31:24] = val1[31:24] - val2[31:24]
262  */
__UQSUB8(uint32_t x,uint32_t y)263 __ALWAYS_INLINE uint32_t __UQSUB8(uint32_t x, uint32_t y)
264 {
265     int32_t r, s, t, u;
266 
267     r = __IUSAT((((x << 24) >> 24) - ((y << 24) >> 24)), 8) & 0x000000FF;
268     s = __IUSAT((((x << 16) >> 24) - ((y << 16) >> 24)), 8) & 0x000000FF;
269     t = __IUSAT((((x <<  8) >> 24) - ((y <<  8) >> 24)), 8) & 0x000000FF;
270     u = __IUSAT((((x) >> 24) - ((y) >> 24)), 8) & 0x000000FF;
271 
272     return ((u << 24) | (t << 16) | (s <<  8) | (r));
273 }
274 
275 /**
276   \brief   Quad 8-bit signed subtraction.
277   \details This function enables you to perform four 8-bit signed integer subtractions.
278   \param [in]    x  first four 8-bit operands of each subtraction.
279   \param [in]    y  second four 8-bit operands of each subtraction.
280   \return        the subtraction of the first bytes from each operand, in the first byte of the return value.\n
281                  the subtraction of the second bytes of each operand, in the second byte of the return value.\n
282                  the subtraction of the third bytes of each operand, in the third byte of the return value.\n
283                  the subtraction of the fourth bytes of each operand, in the fourth byte of the return value.
284   \remark
285                  res[7:0]   = val1[7:0]   - val2[7:0]        \n
286                  res[15:8]  = val1[15:8]  - val2[15:8]       \n
287                  res[23:16] = val1[23:16] - val2[23:16]      \n
288                  res[31:24] = val1[31:24] - val2[31:24]
289  */
__SSUB8(uint32_t x,uint32_t y)290 __ALWAYS_INLINE uint32_t __SSUB8(uint32_t x, uint32_t y)
291 {
292     int32_t r, s, t, u;
293 
294     r = ((((int32_t)x << 24) >> 24) - (((int32_t)y << 24) >> 24)) & (int32_t)0x000000FF;
295     s = ((((int32_t)x << 16) >> 24) - (((int32_t)y << 16) >> 24)) & (int32_t)0x000000FF;
296     t = ((((int32_t)x <<  8) >> 24) - (((int32_t)y <<  8) >> 24)) & (int32_t)0x000000FF;
297     u = ((((int32_t)x) >> 24) - (((int32_t)y) >> 24)) & (int32_t)0x000000FF;
298 
299     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
300 }
301 
302 /**
303   \brief   Quad 8-bit unsigned subtract.
304   \details This function enables you to perform four 8-bit unsigned integer subtractions.
305   \param [in]    x  first four 8-bit operands of each subtraction.
306   \param [in]    y  second four 8-bit operands of each subtraction.
307   \return        the subtraction of the first bytes from each operand, in the first byte of the return value.\n
308                  the subtraction of the second bytes of each operand, in the second byte of the return value.\n
309                  the subtraction of the third bytes of each operand, in the third byte of the return value.\n
310                  the subtraction of the fourth bytes of each operand, in the fourth byte of the return value.
311   \remark
312                  res[7:0]   = val1[7:0]   - val2[7:0]        \n
313                  res[15:8]  = val1[15:8]  - val2[15:8]       \n
314                  res[23:16] = val1[23:16] - val2[23:16]      \n
315                  res[31:24] = val1[31:24] - val2[31:24]
316  */
__USUB8(uint32_t x,uint32_t y)317 __ALWAYS_INLINE uint32_t __USUB8(uint32_t x, uint32_t y)
318 {
319     int32_t r, s, t, u;
320 
321     r = (((x << 24) >> 24) - ((y << 24) >> 24)) & 0x000000FF;
322     s = (((x << 16) >> 24) - ((y << 16) >> 24)) & 0x000000FF;
323     t = (((x <<  8) >> 24) - ((y <<  8) >> 24)) & 0x000000FF;
324     u = (((x) >> 24) - ((y) >> 24)) & 0x000000FF;
325 
326     return ((u << 24) | (t << 16) | (s <<  8) | (r));
327 }
328 
329 /**
330   \brief   Unsigned sum of quad 8-bit unsigned absolute difference.
331   \details This function enables you to perform four unsigned 8-bit subtractions, and add the absolute values
332            of the differences together, returning the result as a single unsigned integer.
333   \param [in]    x  first four 8-bit operands of each subtraction.
334   \param [in]    y  second four 8-bit operands of each subtraction.
335   \return        the subtraction of the first bytes from each operand, in the first byte of the return value.\n
336                  the subtraction of the second bytes of each operand, in the second byte of the return value.\n
337                  the subtraction of the third bytes of each operand, in the third byte of the return value.\n
338                  the subtraction of the fourth bytes of each operand, in the fourth byte of the return value.\n
339                  The sum is returned as a single unsigned integer.
340   \remark
341                  absdiff1   = val1[7:0]   - val2[7:0]        \n
342                  absdiff2   = val1[15:8]  - val2[15:8]       \n
343                  absdiff3   = val1[23:16] - val2[23:16]      \n
344                  absdiff4   = val1[31:24] - val2[31:24]      \n
345                  res[31:0]  = absdiff1 + absdiff2 + absdiff3 + absdiff4
346  */
__USAD8(uint32_t x,uint32_t y)347 __ALWAYS_INLINE uint32_t __USAD8(uint32_t x, uint32_t y)
348 {
349     int32_t r, s, t, u;
350 
351     r = (((x << 24) >> 24) - ((y << 24) >> 24)) & 0x000000FF;
352     s = (((x << 16) >> 24) - ((y << 16) >> 24)) & 0x000000FF;
353     t = (((x <<  8) >> 24) - ((y <<  8) >> 24)) & 0x000000FF;
354     u = (((x) >> 24) - ((y) >> 24)) & 0x000000FF;
355 
356     return (u + t + s + r);
357 }
358 
359 /**
360   \brief   Unsigned sum of quad 8-bit unsigned absolute difference with 32-bit accumulate.
361   \details This function enables you to perform four unsigned 8-bit subtractions, and add the absolute values
362            of the differences to a 32-bit accumulate operand.
363   \param [in]    x  first four 8-bit operands of each subtraction.
364   \param [in]    y  second four 8-bit operands of each subtraction.
365   \param [in]  sum  accumulation value.
366   \return        the sum of the absolute differences of the following bytes, added to the accumulation value:
367                  the subtraction of the first bytes from each operand, in the first byte of the return value.\n
368                  the subtraction of the second bytes of each operand, in the second byte of the return value.\n
369                  the subtraction of the third bytes of each operand, in the third byte of the return value.\n
370                  the subtraction of the fourth bytes of each operand, in the fourth byte of the return value.
371   \remark
372                  absdiff1 = val1[7:0]   - val2[7:0]        \n
373                  absdiff2 = val1[15:8]  - val2[15:8]       \n
374                  absdiff3 = val1[23:16] - val2[23:16]      \n
375                  absdiff4 = val1[31:24] - val2[31:24]      \n
376                  sum = absdiff1 + absdiff2 + absdiff3 + absdiff4 \n
377                  res[31:0] = sum[31:0] + val3[31:0]
378  */
__USADA8(uint32_t x,uint32_t y,uint32_t sum)379 __ALWAYS_INLINE uint32_t __USADA8(uint32_t x, uint32_t y, uint32_t sum)
380 {
381     int32_t r, s, t, u;
382 
383     r = (abs(((x << 24) >> 24) - ((y << 24) >> 24))) & 0x000000FF;
384     s = (abs(((x << 16) >> 24) - ((y << 16) >> 24))) & 0x000000FF;
385     t = (abs(((x <<  8) >> 24) - ((y <<  8) >> 24))) & 0x000000FF;
386     u = (abs(((x) >> 24) - ((y) >> 24))) & 0x000000FF;
387 
388     return (u + t + s + r + sum);
389 }
390 
391 /**
392   \brief   Dual 16-bit saturating addition.
393   \details This function enables you to perform two 16-bit integer arithmetic additions in parallel,
394            saturating the results to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
395   \param [in]    x   first two 16-bit summands.
396   \param [in]    y   second two 16-bit summands.
397   \return        the saturated addition of the low halfwords, in the low halfword of the return value.\n
398                  the saturated addition of the high halfwords, in the high halfword of the return value.\n
399                  The returned results are saturated to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
400   \remark
401                  res[15:0]  = val1[15:0]  + val2[15:0]        \n
402                  res[31:16] = val1[31:16] + val2[31:16]
403  */
__QADD16(uint32_t x,uint32_t y)404 __ALWAYS_INLINE uint32_t __QADD16(uint32_t x, uint32_t y)
405 {
406     int32_t r = 0, s = 0;
407 
408     r = __SSAT(((((int32_t)x << 16) >> 16) + (((int32_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
409     s = __SSAT(((((int32_t)x) >> 16) + (((int32_t)y) >> 16)), 16) & (int32_t)0x0000FFFF;
410 
411     return ((uint32_t)((s << 16) | (r)));
412 }
413 
414 /**
415   \brief   Dual 16-bit unsigned saturating addition.
416   \details This function enables you to perform two unsigned 16-bit integer additions, saturating
417            the results to the 16-bit unsigned integer range 0 < x < 2^16 - 1.
418   \param [in]    x   first two 16-bit summands.
419   \param [in]    y   second two 16-bit summands.
420   \return        the saturated addition of the low halfwords, in the low halfword of the return value.\n
421                  the saturated addition of the high halfwords, in the high halfword of the return value.\n
422                  The results are saturated to the 16-bit unsigned integer range 0 < x < 2^16 - 1.
423   \remark
424                  res[15:0]  = val1[15:0]  + val2[15:0]        \n
425                  res[31:16] = val1[31:16] + val2[31:16]
426  */
__UQADD16(uint32_t x,uint32_t y)427 __ALWAYS_INLINE uint32_t __UQADD16(uint32_t x, uint32_t y)
428 {
429     int32_t r = 0, s = 0;
430 
431     r = __IUSAT((((x << 16) >> 16) + ((y << 16) >> 16)), 16) & 0x0000FFFF;
432     s = __IUSAT((((x) >> 16) + ((y) >> 16)), 16) & 0x0000FFFF;
433 
434     return ((s << 16) | (r));
435 }
436 
437 /**
438   \brief   Dual 16-bit signed addition.
439   \details This function enables you to perform two 16-bit signed integer additions.
440   \param [in]    x   first two 16-bit summands.
441   \param [in]    y   second two 16-bit summands.
442   \return        the addition of the low halfwords in the low halfword of the return value.\n
443                  the addition of the high halfwords in the high halfword of the return value.
444   \remark
445                  res[15:0]  = val1[15:0]  + val2[15:0]        \n
446                  res[31:16] = val1[31:16] + val2[31:16]
447  */
__SADD16(uint32_t x,uint32_t y)448 __ALWAYS_INLINE uint32_t __SADD16(uint32_t x, uint32_t y)
449 {
450     int32_t r = 0, s = 0;
451 
452     r = ((((int32_t)x << 16) >> 16) + (((int32_t)y << 16) >> 16)) & (int32_t)0x0000FFFF;
453     s = ((((int32_t)x) >> 16) + (((int32_t)y) >> 16)) & (int32_t)0x0000FFFF;
454 
455     return ((uint32_t)((s << 16) | (r)));
456 }
457 
458 /**
459   \brief   Dual 16-bit unsigned addition
460   \details This function enables you to perform two 16-bit unsigned integer additions.
461   \param [in]    x   first two 16-bit summands for each addition.
462   \param [in]    y   second two 16-bit summands for each addition.
463   \return        the addition of the low halfwords in the low halfword of the return value.\n
464                  the addition of the high halfwords in the high halfword of the return value.
465   \remark
466                  res[15:0]  = val1[15:0]  + val2[15:0]        \n
467                  res[31:16] = val1[31:16] + val2[31:16]
468  */
__UADD16(uint32_t x,uint32_t y)469 __ALWAYS_INLINE uint32_t __UADD16(uint32_t x, uint32_t y)
470 {
471     int32_t r = 0, s = 0;
472 
473     r = (((x << 16) >> 16) + ((y << 16) >> 16)) & 0x0000FFFF;
474     s = (((x) >> 16) + ((y) >> 16)) & 0x0000FFFF;
475 
476     return ((s << 16) | (r));
477 }
478 
479 
480 /**
481   \brief   Dual 16-bit signed addition with halved results.
482   \details This function enables you to perform two signed 16-bit integer additions, halving the results.
483   \param [in]    x   first two 16-bit summands.
484   \param [in]    y   second two 16-bit summands.
485   \return        the halved addition of the low halfwords, in the low halfword of the return value.\n
486                  the halved addition of the high halfwords, in the high halfword of the return value.
487   \remark
488                  res[15:0]  = (val1[15:0]  + val2[15:0]) >> 1        \n
489                  res[31:16] = (val1[31:16] + val2[31:16]) >> 1
490  */
__SHADD16(uint32_t x,uint32_t y)491 __ALWAYS_INLINE uint32_t __SHADD16(uint32_t x, uint32_t y)
492 {
493     int32_t r, s;
494 
495     r = (((((int32_t)x << 16) >> 16) + (((int32_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
496     s = (((((int32_t)x) >> 16) + (((int32_t)y) >> 16)) >> 1) & (int32_t)0x0000FFFF;
497 
498     return ((uint32_t)((s << 16) | (r)));
499 }
500 
501 /**
502   \brief   Dual 16-bit unsigned addition with halved results.
503   \details This function enables you to perform two unsigned 16-bit integer additions, halving the results.
504   \param [in]    x   first two 16-bit summands.
505   \param [in]    y   second two 16-bit summands.
506   \return        the halved addition of the low halfwords, in the low halfword of the return value.\n
507                  the halved addition of the high halfwords, in the high halfword of the return value.
508   \remark
509                  res[15:0]  = (val1[15:0]  + val2[15:0]) >> 1        \n
510                  res[31:16] = (val1[31:16] + val2[31:16]) >> 1
511  */
__UHADD16(uint32_t x,uint32_t y)512 __ALWAYS_INLINE uint32_t __UHADD16(uint32_t x, uint32_t y)
513 {
514     int32_t r, s;
515 
516     r = ((((x << 16) >> 16) + ((y << 16) >> 16)) >> 1) & 0x0000FFFF;
517     s = ((((x) >> 16) + ((y) >> 16)) >> 1) & 0x0000FFFF;
518 
519     return ((s << 16) | (r));
520 }
521 
522 /**
523   \brief   Quad 8-bit signed addition with halved results.
524   \details This function enables you to perform four signed 8-bit integer additions, halving the results.
525   \param [in]    x   first four 8-bit summands.
526   \param [in]    y   second four 8-bit summands.
527   \return        the halved addition of the first bytes from each operand, in the first byte of the return value.\n
528                  the halved addition of the second bytes from each operand, in the second byte of the return value.\n
529                  the halved addition of the third bytes from each operand, in the third byte of the return value.\n
530                  the halved addition of the fourth bytes from each operand, in the fourth byte of the return value.
531   \remark
532                  res[7:0]   = (val1[7:0]   + val2[7:0]  ) >> 1    \n
533                  res[15:8]  = (val1[15:8]  + val2[15:8] ) >> 1    \n
534                  res[23:16] = (val1[23:16] + val2[23:16]) >> 1    \n
535                  res[31:24] = (val1[31:24] + val2[31:24]) >> 1
536  */
__SHADD8(uint32_t x,uint32_t y)537 __ALWAYS_INLINE uint32_t __SHADD8(uint32_t x, uint32_t y)
538 {
539     int32_t r, s, t, u;
540 
541     r = (((((int32_t)x << 24) >> 24) + (((int32_t)y << 24) >> 24)) >> 1) & (int32_t)0x000000FF;
542     s = (((((int32_t)x << 16) >> 24) + (((int32_t)y << 16) >> 24)) >> 1) & (int32_t)0x000000FF;
543     t = (((((int32_t)x <<  8) >> 24) + (((int32_t)y <<  8) >> 24)) >> 1) & (int32_t)0x000000FF;
544     u = (((((int32_t)x) >> 24) + (((int32_t)y) >> 24)) >> 1) & (int32_t)0x000000FF;
545 
546     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
547 }
548 
549 /**
550   \brief   Quad 8-bit unsigned addition with halved results.
551   \details This function enables you to perform four unsigned 8-bit integer additions, halving the results.
552   \param [in]    x   first four 8-bit summands.
553   \param [in]    y   second four 8-bit summands.
554   \return        the halved addition of the first bytes from each operand, in the first byte of the return value.\n
555                  the halved addition of the second bytes from each operand, in the second byte of the return value.\n
556                  the halved addition of the third bytes from each operand, in the third byte of the return value.\n
557                  the halved addition of the fourth bytes from each operand, in the fourth byte of the return value.
558   \remark
559                  res[7:0]   = (val1[7:0]   + val2[7:0]  ) >> 1    \n
560                  res[15:8]  = (val1[15:8]  + val2[15:8] ) >> 1    \n
561                  res[23:16] = (val1[23:16] + val2[23:16]) >> 1    \n
562                  res[31:24] = (val1[31:24] + val2[31:24]) >> 1
563  */
__UHADD8(uint32_t x,uint32_t y)564 __ALWAYS_INLINE uint32_t __UHADD8(uint32_t x, uint32_t y)
565 {
566     int32_t r, s, t, u;
567 
568     r = ((((x << 24) >> 24) + ((y << 24) >> 24)) >> 1) & 0x000000FF;
569     s = ((((x << 16) >> 24) + ((y << 16) >> 24)) >> 1) & 0x000000FF;
570     t = ((((x <<  8) >> 24) + ((y <<  8) >> 24)) >> 1) & 0x000000FF;
571     u = ((((x) >> 24) + ((y) >> 24)) >> 1) & 0x000000FF;
572 
573     return ((u << 24) | (t << 16) | (s <<  8) | (r));
574 }
575 
576 /**
577   \brief   Dual 16-bit saturating subtract.
578   \details This function enables you to perform two 16-bit integer subtractions in parallel,
579            saturating the results to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
580   \param [in]    x   first two 16-bit summands.
581   \param [in]    y   second two 16-bit summands.
582   \return        the saturated subtraction of the low halfwords, in the low halfword of the return value.\n
583                  the saturated subtraction of the high halfwords, in the high halfword of the return value.\n
584                  The returned results are saturated to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
585   \remark
586                  res[15:0]  = val1[15:0]  - val2[15:0]        \n
587                  res[31:16] = val1[31:16] - val2[31:16]
588  */
__QSUB16(uint32_t x,uint32_t y)589 __ALWAYS_INLINE uint32_t __QSUB16(uint32_t x, uint32_t y)
590 {
591     int32_t r, s;
592 
593     r = __SSAT(((((int32_t)x << 16) >> 16) - (((int32_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
594     s = __SSAT(((((int32_t)x) >> 16) - (((int32_t)y) >> 16)), 16) & (int32_t)0x0000FFFF;
595 
596     return ((uint32_t)((s << 16) | (r)));
597 }
598 
599 /**
600   \brief   Dual 16-bit unsigned saturating subtraction.
601   \details This function enables you to perform two unsigned 16-bit integer subtractions,
602            saturating the results to the 16-bit unsigned integer range 0 < x < 2^16 - 1.
603   \param [in]    x   first two 16-bit operands for each subtraction.
604   \param [in]    y   second two 16-bit operands for each subtraction.
605   \return        the saturated subtraction of the low halfwords, in the low halfword of the return value.\n
606                  the saturated subtraction of the high halfwords, in the high halfword of the return value.\n
607                  The returned results are saturated to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
608   \remark
609                  res[15:0]  = val1[15:0]  - val2[15:0]        \n
610                  res[31:16] = val1[31:16] - val2[31:16]
611  */
__UQSUB16(uint32_t x,uint32_t y)612 __ALWAYS_INLINE uint32_t __UQSUB16(uint32_t x, uint32_t y)
613 {
614     int32_t r, s;
615 
616     r = __IUSAT((((x << 16) >> 16) - ((y << 16) >> 16)), 16) & 0x0000FFFF;
617     s = __IUSAT((((x) >> 16) - ((y) >> 16)), 16) & 0x0000FFFF;
618 
619     return ((s << 16) | (r));
620 }
621 
622 /**
623   \brief   Dual 16-bit signed subtraction.
624   \details This function enables you to perform two 16-bit signed integer subtractions.
625   \param [in]    x   first two 16-bit operands of each subtraction.
626   \param [in]    y   second two 16-bit operands of each subtraction.
627   \return        the subtraction of the low halfword in the second operand from the low
628                  halfword in the first operand, in the low halfword of the return value. \n
629                  the subtraction of the high halfword in the second operand from the high
630                  halfword in the first operand, in the high halfword of the return value.
631   \remark
632                  res[15:0]  = val1[15:0]  - val2[15:0]        \n
633                  res[31:16] = val1[31:16] - val2[31:16]
634  */
__SSUB16(uint32_t x,uint32_t y)635 __ALWAYS_INLINE uint32_t __SSUB16(uint32_t x, uint32_t y)
636 {
637     int32_t r, s;
638 
639     r = ((((int32_t)x << 16) >> 16) - (((int32_t)y << 16) >> 16)) & (int32_t)0x0000FFFF;
640     s = ((((int32_t)x) >> 16) - (((int32_t)y) >> 16)) & (int32_t)0x0000FFFF;
641 
642     return ((uint32_t)((s << 16) | (r)));
643 }
644 
645 /**
646   \brief   Dual 16-bit unsigned subtract.
647   \details This function enables you to perform two 16-bit unsigned integer subtractions.
648   \param [in]    x   first two 16-bit operands of each subtraction.
649   \param [in]    y   second two 16-bit operands of each subtraction.
650   \return        the subtraction of the low halfword in the second operand from the low
651                  halfword in the first operand, in the low halfword of the return value. \n
652                  the subtraction of the high halfword in the second operand from the high
653                  halfword in the first operand, in the high halfword of the return value.
654   \remark
655                  res[15:0]  = val1[15:0]  - val2[15:0]        \n
656                  res[31:16] = val1[31:16] - val2[31:16]
657  */
__USUB16(uint32_t x,uint32_t y)658 __ALWAYS_INLINE uint32_t __USUB16(uint32_t x, uint32_t y)
659 {
660     int32_t r, s;
661 
662     r = (((x << 16) >> 16) - ((y << 16) >> 16)) & 0x0000FFFF;
663     s = (((x) >> 16) - ((y) >> 16)) & 0x0000FFFF;
664 
665     return ((s << 16) | (r));
666 }
667 
668 /**
669   \brief   Dual 16-bit signed subtraction with halved results.
670   \details This function enables you to perform two signed 16-bit integer subtractions, halving the results.
671   \param [in]    x   first two 16-bit summands.
672   \param [in]    y   second two 16-bit summands.
673   \return        the halved subtraction of the low halfwords, in the low halfword of the return value.\n
674                  the halved subtraction of the high halfwords, in the high halfword of the return value.
675   \remark
676                  res[15:0]  = (val1[15:0]  - val2[15:0]) >> 1        \n
677                  res[31:16] = (val1[31:16] - val2[31:16]) >> 1
678  */
__SHSUB16(uint32_t x,uint32_t y)679 __ALWAYS_INLINE uint32_t __SHSUB16(uint32_t x, uint32_t y)
680 {
681     int32_t r, s;
682 
683     r = (((((int32_t)x << 16) >> 16) - (((int32_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
684     s = (((((int32_t)x) >> 16) - (((int32_t)y) >> 16)) >> 1) & (int32_t)0x0000FFFF;
685 
686     return ((uint32_t)((s << 16) | (r)));
687 }
688 
689 /**
690   \brief   Dual 16-bit unsigned subtraction with halved results.
691   \details This function enables you to perform two unsigned 16-bit integer subtractions, halving the results.
692   \param [in]    x   first two 16-bit summands.
693   \param [in]    y   second two 16-bit summands.
694   \return        the halved subtraction of the low halfwords, in the low halfword of the return value.\n
695                  the halved subtraction of the high halfwords, in the high halfword of the return value.
696   \remark
697                  res[15:0]  = (val1[15:0]  - val2[15:0]) >> 1        \n
698                  res[31:16] = (val1[31:16] - val2[31:16]) >> 1
699  */
__UHSUB16(uint32_t x,uint32_t y)700 __ALWAYS_INLINE uint32_t __UHSUB16(uint32_t x, uint32_t y)
701 {
702     int32_t r, s;
703 
704     r = ((((x << 16) >> 16) - ((y << 16) >> 16)) >> 1) & 0x0000FFFF;
705     s = ((((x) >> 16) - ((y) >> 16)) >> 1) & 0x0000FFFF;
706 
707     return ((s << 16) | (r));
708 }
709 
710 /**
711   \brief   Quad 8-bit signed addition with halved results.
712   \details This function enables you to perform four signed 8-bit integer subtractions, halving the results.
713   \param [in]    x   first four 8-bit summands.
714   \param [in]    y   second four 8-bit summands.
715   \return        the halved subtraction of the first bytes from each operand, in the first byte of the return value.\n
716                  the halved subtraction of the second bytes from each operand, in the second byte of the return value.\n
717                  the halved subtraction of the third bytes from each operand, in the third byte of the return value.\n
718                  the halved subtraction of the fourth bytes from each operand, in the fourth byte of the return value.
719   \remark
720                  res[7:0]   = (val1[7:0]   - val2[7:0]  ) >> 1    \n
721                  res[15:8]  = (val1[15:8]  - val2[15:8] ) >> 1    \n
722                  res[23:16] = (val1[23:16] - val2[23:16]) >> 1    \n
723                  res[31:24] = (val1[31:24] - val2[31:24]) >> 1
724  */
__SHSUB8(uint32_t x,uint32_t y)725 __ALWAYS_INLINE uint32_t __SHSUB8(uint32_t x, uint32_t y)
726 {
727     int32_t r, s, t, u;
728 
729     r = (((((int32_t)x << 24) >> 24) - (((int32_t)y << 24) >> 24)) >> 1) & (int32_t)0x000000FF;
730     s = (((((int32_t)x << 16) >> 24) - (((int32_t)y << 16) >> 24)) >> 1) & (int32_t)0x000000FF;
731     t = (((((int32_t)x <<  8) >> 24) - (((int32_t)y <<  8) >> 24)) >> 1) & (int32_t)0x000000FF;
732     u = (((((int32_t)x) >> 24) - (((int32_t)y) >> 24)) >> 1) & (int32_t)0x000000FF;
733 
734     return ((uint32_t)((u << 24) | (t << 16) | (s <<  8) | (r)));
735 }
736 
737 /**
738   \brief   Quad 8-bit unsigned subtraction with halved results.
739   \details This function enables you to perform four unsigned 8-bit integer subtractions, halving the results.
740   \param [in]    x   first four 8-bit summands.
741   \param [in]    y   second four 8-bit summands.
742   \return        the halved subtraction of the first bytes from each operand, in the first byte of the return value.\n
743                  the halved subtraction of the second bytes from each operand, in the second byte of the return value.\n
744                  the halved subtraction of the third bytes from each operand, in the third byte of the return value.\n
745                  the halved subtraction of the fourth bytes from each operand, in the fourth byte of the return value.
746   \remark
747                  res[7:0]   = (val1[7:0]   - val2[7:0]  ) >> 1    \n
748                  res[15:8]  = (val1[15:8]  - val2[15:8] ) >> 1    \n
749                  res[23:16] = (val1[23:16] - val2[23:16]) >> 1    \n
750                  res[31:24] = (val1[31:24] - val2[31:24]) >> 1
751  */
__UHSUB8(uint32_t x,uint32_t y)752 __ALWAYS_INLINE uint32_t __UHSUB8(uint32_t x, uint32_t y)
753 {
754     int32_t r, s, t, u;
755 
756     r = ((((x << 24) >> 24) - ((y << 24) >> 24)) >> 1) & 0x000000FF;
757     s = ((((x << 16) >> 24) - ((y << 16) >> 24)) >> 1) & 0x000000FF;
758     t = ((((x <<  8) >> 24) - ((y <<  8) >> 24)) >> 1) & 0x000000FF;
759     u = ((((x) >> 24) - ((y) >> 24)) >> 1) & 0x000000FF;
760 
761     return ((u << 24) | (t << 16) | (s <<  8) | (r));
762 }
763 
764 /**
765   \brief   Dual 16-bit add and subtract with exchange.
766   \details This function enables you to exchange the halfwords of the one operand,
767            then add the high halfwords and subtract the low halfwords,
768            saturating the results to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
769   \param [in]    x   first operand for the subtraction in the low halfword,
770                      and the first operand for the addition in the high halfword.
771   \param [in]    y   second operand for the subtraction in the high halfword,
772                      and the second operand for the addition in the low halfword.
773   \return        the saturated subtraction of the high halfword in the second operand from the
774                  low halfword in the first operand, in the low halfword of the return value.\n
775                  the saturated addition of the high halfword in the first operand and the
776                  low halfword in the second operand, in the high halfword of the return value.\n
777                  The returned results are saturated to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
778   \remark
779                  res[15:0]  = val1[15:0]  - val2[31:16]        \n
780                  res[31:16] = val1[31:16] + val2[15:0]
781  */
__QASX(uint32_t x,uint32_t y)782 __ALWAYS_INLINE uint32_t __QASX(uint32_t x, uint32_t y)
783 {
784     int32_t r, s;
785 
786     r = __SSAT(((((int32_t)x << 16) >> 16) - (((int32_t)y) >> 16)), 16) & (int32_t)0x0000FFFF;
787     s = __SSAT(((((int32_t)x) >> 16) + (((int32_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
788 
789     return ((uint32_t)((s << 16) | (r)));
790 }
791 
792 /**
793   \brief   Dual 16-bit unsigned saturating addition and subtraction with exchange.
794   \details This function enables you to exchange the halfwords of the second operand and
795            perform one unsigned 16-bit integer addition and one unsigned 16-bit subtraction,
796            saturating the results to the 16-bit unsigned integer range 0 <= x <= 2^16 - 1.
797   \param [in]    x   first operand for the subtraction in the low halfword,
798                      and the first operand for the addition in the high halfword.
799   \param [in]    y   second operand for the subtraction in the high halfword,
800                      and the second operand for the addition in the low halfword.
801   \return        the saturated subtraction of the high halfword in the second operand from the
802                  low halfword in the first operand, in the low halfword of the return value.\n
803                  the saturated addition of the high halfword in the first operand and the
804                  low halfword in the second operand, in the high halfword of the return value.\n
805                  The returned results are saturated to the 16-bit unsigned integer range 0 <= x <= 2^16 - 1.
806   \remark
807                  res[15:0]  = val1[15:0]  - val2[31:16]        \n
808                  res[31:16] = val1[31:16] + val2[15:0]
809  */
__UQASX(uint32_t x,uint32_t y)810 __ALWAYS_INLINE uint32_t __UQASX(uint32_t x, uint32_t y)
811 {
812     int32_t r, s;
813 
814     r = __IUSAT((((x << 16) >> 16) - ((y) >> 16)), 16) & 0x0000FFFF;
815     s = __IUSAT((((x) >> 16) + ((y << 16) >> 16)), 16) & 0x0000FFFF;
816 
817     return ((s << 16) | (r));
818 }
819 
820 /**
821   \brief   Dual 16-bit addition and subtraction with exchange.
822   \details It enables you to exchange the halfwords of the second operand, add the high halfwords
823            and subtract the low halfwords.
824   \param [in]    x   first operand for the subtraction in the low halfword,
825                      and the first operand for the addition in the high halfword.
826   \param [in]    y   second operand for the subtraction in the high halfword,
827                      and the second operand for the addition in the low halfword.
828   \return        the subtraction of the high halfword in the second operand from the
829                  low halfword in the first operand, in the low halfword of the return value.\n
830                  the addition of the high halfword in the first operand and the
831                  low halfword in the second operand, in the high halfword of the return value.
832   \remark
833                  res[15:0]  = val1[15:0]  - val2[31:16]        \n
834                  res[31:16] = val1[31:16] + val2[15:0]
835  */
__SASX(uint32_t x,uint32_t y)836 __ALWAYS_INLINE uint32_t __SASX(uint32_t x, uint32_t y)
837 {
838     int32_t r, s;
839 
840     r = ((((int32_t)x << 16) >> 16) - (((int32_t)y) >> 16)) & (int32_t)0x0000FFFF;
841     s = ((((int32_t)x) >> 16) + (((int32_t)y << 16) >> 16)) & (int32_t)0x0000FFFF;
842 
843     return ((uint32_t)((s << 16) | (r)));
844 }
845 
846 /**
847   \brief   Dual 16-bit unsigned addition and subtraction with exchange.
848   \details This function enables you to exchange the two halfwords of the second operand,
849            add the high halfwords and subtract the low halfwords.
850   \param [in]    x   first operand for the subtraction in the low halfword,
851                      and the first operand for the addition in the high halfword.
852   \param [in]    y   second operand for the subtraction in the high halfword,
853                      and the second operand for the addition in the low halfword.
854   \return        the subtraction of the high halfword in the second operand from the
855                  low halfword in the first operand, in the low halfword of the return value.\n
856                  the addition of the high halfword in the first operand and the
857                  low halfword in the second operand, in the high halfword of the return value.
858   \remark
859                  res[15:0]  = val1[15:0]  - val2[31:16]        \n
860                  res[31:16] = val1[31:16] + val2[15:0]
861  */
__UASX(uint32_t x,uint32_t y)862 __ALWAYS_INLINE uint32_t __UASX(uint32_t x, uint32_t y)
863 {
864     int32_t r, s;
865 
866     r = (((x << 16) >> 16) - ((y) >> 16)) & 0x0000FFFF;
867     s = (((x) >> 16) + ((y << 16) >> 16)) & 0x0000FFFF;
868 
869     return ((s << 16) | (r));
870 }
871 
872 /**
873   \brief   Dual 16-bit signed addition and subtraction with halved results.
874   \details This function enables you to exchange the two halfwords of one operand, perform one
875            signed 16-bit integer addition and one signed 16-bit subtraction, and halve the results.
876   \param [in]    x   first 16-bit operands.
877   \param [in]    y   second 16-bit operands.
878   \return        the halved subtraction of the high halfword in the second operand from the
879                  low halfword in the first operand, in the low halfword of the return value.\n
880                  the halved addition of the low halfword in the second operand from the high
881                  halfword in the first operand, in the high halfword of the return value.
882   \remark
883                  res[15:0]  = (val1[15:0]  - val2[31:16]) >> 1        \n
884                  res[31:16] = (val1[31:16] + val2[15:0]) >> 1
885  */
__SHASX(uint32_t x,uint32_t y)886 __ALWAYS_INLINE uint32_t __SHASX(uint32_t x, uint32_t y)
887 {
888     int32_t r, s;
889 
890     r = (((((int32_t)x << 16) >> 16) - (((int32_t)y) >> 16)) >> 1) & (int32_t)0x0000FFFF;
891     s = (((((int32_t)x) >> 16) + (((int32_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
892 
893     return ((uint32_t)((s << 16) | (r)));
894 }
895 
896 /**
897   \brief   Dual 16-bit unsigned addition and subtraction with halved results and exchange.
898   \details This function enables you to exchange the halfwords of the second operand,
899            add the high halfwords and subtract the low halfwords, halving the results.
900   \param [in]    x   first operand for the subtraction in the low halfword, and
901                      the first operand for the addition in the high halfword.
902   \param [in]    y   second operand for the subtraction in the high halfword, and
903                      the second operand for the addition in the low halfword.
904   \return        the halved subtraction of the high halfword in the second operand from the
905                  low halfword in the first operand, in the low halfword of the return value.\n
906                  the halved addition of the low halfword in the second operand from the high
907                  halfword in the first operand, in the high halfword of the return value.
908   \remark
909                  res[15:0]  = (val1[15:0]  - val2[31:16]) >> 1        \n
910                  res[31:16] = (val1[31:16] + val2[15:0]) >> 1
911  */
__UHASX(uint32_t x,uint32_t y)912 __ALWAYS_INLINE uint32_t __UHASX(uint32_t x, uint32_t y)
913 {
914     int32_t r, s;
915 
916     r = ((((x << 16) >> 16) - ((y) >> 16)) >> 1) & 0x0000FFFF;
917     s = ((((x) >> 16) + ((y << 16) >> 16)) >> 1) & 0x0000FFFF;
918 
919     return ((s << 16) | (r));
920 }
921 
922 /**
923   \brief   Dual 16-bit subtract and add with exchange.
924   \details This function enables you to exchange the halfwords of one operand,
925            then subtract the high halfwords and add the low halfwords,
926            saturating the results to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
927   \param [in]    x   first operand for the addition in the low halfword,
928                      and the first operand for the subtraction in the high halfword.
929   \param [in]    y   second operand for the addition in the high halfword,
930                      and the second operand for the subtraction in the low halfword.
931   \return        the saturated addition of the low halfword of the first operand and the high
932                  halfword of the second operand, in the low halfword of the return value.\n
933                  the saturated subtraction of the low halfword of the second operand from the
934                  high halfword of the first operand, in the high halfword of the return value.\n
935                  The returned results are saturated to the 16-bit signed integer range -2^15 <= x <= 2^15 - 1.
936   \remark
937                  res[15:0]  = val1[15:0]  + val2[31:16]        \n
938                  res[31:16] = val1[31:16] - val2[15:0]
939  */
__QSAX(uint32_t x,uint32_t y)940 __ALWAYS_INLINE uint32_t __QSAX(uint32_t x, uint32_t y)
941 {
942     int32_t r, s;
943 
944     r = __SSAT(((((int32_t)x << 16) >> 16) + (((int32_t)y) >> 16)), 16) & (int32_t)0x0000FFFF;
945     s = __SSAT(((((int32_t)x) >> 16) - (((int32_t)y << 16) >> 16)), 16) & (int32_t)0x0000FFFF;
946 
947     return ((uint32_t)((s << 16) | (r)));
948 }
949 
950 /**
951   \brief   Dual 16-bit unsigned saturating subtraction and addition with exchange.
952   \details This function enables you to exchange the halfwords of the second operand and perform
953            one unsigned 16-bit integer subtraction and one unsigned 16-bit addition, saturating
954            the results to the 16-bit unsigned integer range 0 <= x <= 2^16 - 1.
955   \param [in]    x   first operand for the addition in the low halfword,
956                      and the first operand for the subtraction in the high halfword.
957   \param [in]    y   second operand for the addition in the high halfword,
958                      and the second operand for the subtraction in the low halfword.
959   \return        the saturated addition of the low halfword of the first operand and the high
960                  halfword of the second operand, in the low halfword of the return value.\n
961                  the saturated subtraction of the low halfword of the second operand from the
962                  high halfword of the first operand, in the high halfword of the return value.\n
963                  The returned results are saturated to the 16-bit unsigned integer range 0 <= x <= 2^16 - 1.
964   \remark
965                  res[15:0]  = val1[15:0]  + val2[31:16]        \n
966                  res[31:16] = val1[31:16] - val2[15:0]
967  */
__UQSAX(uint32_t x,uint32_t y)968 __ALWAYS_INLINE uint32_t __UQSAX(uint32_t x, uint32_t y)
969 {
970     int32_t r, s;
971 
972     r = __IUSAT((((x << 16) >> 16) + ((y) >> 16)), 16) & 0x0000FFFF;
973     s = __IUSAT((((x) >> 16) - ((y << 16) >> 16)), 16) & 0x0000FFFF;
974 
975     return ((s << 16) | (r));
976 }
977 
978 /**
979   \brief   Dual 16-bit unsigned subtract and add with exchange.
980   \details This function enables you to exchange the halfwords of the second operand,
981            subtract the high halfwords and add the low halfwords.
982   \param [in]    x   first operand for the addition in the low halfword,
983                      and the first operand for the subtraction in the high halfword.
984   \param [in]    y   second operand for the addition in the high halfword,
985                      and the second operand for the subtraction in the low halfword.
986   \return        the addition of the low halfword of the first operand and the high
987                  halfword of the second operand, in the low halfword of the return value.\n
988                  the subtraction of the low halfword of the second operand from the
989                  high halfword of the first operand, in the high halfword of the return value.\n
990   \remark
991                  res[15:0]  = val1[15:0]  + val2[31:16]        \n
992                  res[31:16] = val1[31:16] - val2[15:0]
993  */
__USAX(uint32_t x,uint32_t y)994 __ALWAYS_INLINE uint32_t __USAX(uint32_t x, uint32_t y)
995 {
996     int32_t r, s;
997 
998     r = (((x << 16) >> 16) + ((y) >> 16)) & 0x0000FFFF;
999     s = (((x) >> 16) - ((y << 16) >> 16)) & 0x0000FFFF;
1000 
1001     return ((s << 16) | (r));
1002 }
1003 
1004 /**
1005   \brief   Dual 16-bit signed subtraction and addition with exchange.
1006   \details This function enables you to exchange the two halfwords of one operand and perform one
1007            16-bit integer subtraction and one 16-bit addition.
1008   \param [in]    x   first operand for the addition in the low halfword, and the first operand
1009                      for the subtraction in the high halfword.
1010   \param [in]    y   second operand for the addition in the high halfword, and the second
1011                      operand for the subtraction in the low halfword.
1012   \return        the addition of the low halfword of the first operand and the high
1013                  halfword of the second operand, in the low halfword of the return value.\n
1014                  the subtraction of the low halfword of the second operand from the
1015                  high halfword of the first operand, in the high halfword of the return value.\n
1016   \remark
1017                  res[15:0]  = val1[15:0]  + val2[31:16]        \n
1018                  res[31:16] = val1[31:16] - val2[15:0]
1019  */
__SSAX(uint32_t x,uint32_t y)1020 __ALWAYS_INLINE uint32_t __SSAX(uint32_t x, uint32_t y)
1021 {
1022     int32_t r, s;
1023 
1024     r = ((((int32_t)x << 16) >> 16) + (((int32_t)y) >> 16)) & (int32_t)0x0000FFFF;
1025     s = ((((int32_t)x) >> 16) - (((int32_t)y << 16) >> 16)) & (int32_t)0x0000FFFF;
1026 
1027     return ((uint32_t)((s << 16) | (r)));
1028 }
1029 
1030 
1031 /**
1032   \brief   Dual 16-bit signed subtraction and addition with halved results.
1033   \details This function enables you to exchange the two halfwords of one operand, perform one signed
1034            16-bit integer subtraction and one signed 16-bit addition, and halve the results.
1035   \param [in]    x   first 16-bit operands.
1036   \param [in]    y   second 16-bit operands.
1037   \return        the halved addition of the low halfword in the first operand and the
1038                  high halfword in the second operand, in the low halfword of the return value.\n
1039                  the halved subtraction of the low halfword in the second operand from the
1040                  high halfword in the first operand, in the high halfword of the return value.
1041   \remark
1042                  res[15:0]  = (val1[15:0]  + val2[31:16]) >> 1        \n
1043                  res[31:16] = (val1[31:16] - val2[15:0]) >> 1
1044  */
__SHSAX(uint32_t x,uint32_t y)1045 __ALWAYS_INLINE uint32_t __SHSAX(uint32_t x, uint32_t y)
1046 {
1047     int32_t r, s;
1048 
1049     r = (((((int32_t)x << 16) >> 16) + (((int32_t)y) >> 16)) >> 1) & (int32_t)0x0000FFFF;
1050     s = (((((int32_t)x) >> 16) - (((int32_t)y << 16) >> 16)) >> 1) & (int32_t)0x0000FFFF;
1051 
1052     return ((uint32_t)((s << 16) | (r)));
1053 }
1054 
1055 /**
1056   \brief   Dual 16-bit unsigned subtraction and addition with halved results and exchange.
1057   \details This function enables you to exchange the halfwords of the second operand,
1058            subtract the high halfwords and add the low halfwords, halving the results.
1059   \param [in]    x   first operand for the addition in the low halfword, and
1060                      the first operand for the subtraction in the high halfword.
1061   \param [in]    y   second operand for the addition in the high halfword, and
1062                      the second operand for the subtraction in the low halfword.
1063   \return        the halved addition of the low halfword in the first operand and the
1064                  high halfword in the second operand, in the low halfword of the return value.\n
1065                  the halved subtraction of the low halfword in the second operand from the
1066                  high halfword in the first operand, in the high halfword of the return value.
1067   \remark
1068                  res[15:0]  = (val1[15:0]  + val2[31:16]) >> 1        \n
1069                  res[31:16] = (val1[31:16] - val2[15:0]) >> 1
1070  */
__UHSAX(uint32_t x,uint32_t y)1071 __ALWAYS_INLINE uint32_t __UHSAX(uint32_t x, uint32_t y)
1072 {
1073     int32_t r, s;
1074 
1075     r = ((((x << 16) >> 16) + ((y) >> 16)) >> 1) & 0x0000FFFF;
1076     s = ((((x) >> 16) - ((y << 16) >> 16)) >> 1) & 0x0000FFFF;
1077 
1078     return ((s << 16) | (r));
1079 }
1080 
1081 /**
1082   \brief   Dual 16-bit signed multiply with exchange returning difference.
1083   \details This function enables you to perform two 16-bit signed multiplications, subtracting
1084            one of the products from the other. The halfwords of the second operand are exchanged
1085            before performing the arithmetic. This produces top * bottom and bottom * top multiplication.
1086   \param [in]    x   first 16-bit operands for each multiplication.
1087   \param [in]    y   second 16-bit operands for each multiplication.
1088   \return        the difference of the products of the two 16-bit signed multiplications.
1089   \remark
1090                  p1 = val1[15:0]  * val2[31:16]       \n
1091                  p2 = val1[31:16] * val2[15:0]        \n
1092                  res[31:0] = p1 - p2
1093  */
__SMUSDX(uint32_t x,uint32_t y)1094 __ALWAYS_INLINE uint32_t __SMUSDX(uint32_t x, uint32_t y)
1095 {
1096     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) -
1097                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16))));
1098 }
1099 
1100 /**
1101   \brief   Sum of dual 16-bit signed multiply with exchange.
1102   \details This function enables you to perform two 16-bit signed multiplications with exchanged
1103            halfwords of the second operand, adding the products together.
1104   \param [in]    x   first 16-bit operands for each multiplication.
1105   \param [in]    y   second 16-bit operands for each multiplication.
1106   \return        the sum of the products of the two 16-bit signed multiplications with exchanged halfwords of the second operand.
1107   \remark
1108                  p1 = val1[15:0]  * val2[31:16]       \n
1109                  p2 = val1[31:16] * val2[15:0]        \n
1110                  res[31:0] = p1 + p2
1111  */
__SMUADX(uint32_t x,uint32_t y)1112 __ALWAYS_INLINE uint32_t __SMUADX(uint32_t x, uint32_t y)
1113 {
1114     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) +
1115                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16))));
1116 }
1117 
1118 
1119 /**
1120   \brief   Saturating add.
1121   \details This function enables you to obtain the saturating add of two integers.
1122   \param [in]    x   first summand of the saturating add operation.
1123   \param [in]    y   second summand of the saturating add operation.
1124   \return        the saturating addition of val1 and val2.
1125   \remark
1126                  res[31:0] = SAT(val1 + SAT(val2))
1127  */
__QADD(int32_t x,int32_t y)1128 __ALWAYS_INLINE int32_t __QADD(int32_t x, int32_t y)
1129 {
1130     int32_t result;
1131 
1132     if (y >= 0)
1133     {
1134         if (x + y >= x)
1135         {
1136             result = x + y;
1137         }
1138         else
1139         {
1140             result = 0x7FFFFFFF;
1141         }
1142     }
1143     else
1144     {
1145         if (x + y < x)
1146         {
1147             result = x + y;
1148         }
1149         else
1150         {
1151             result = 0x80000000;
1152         }
1153     }
1154 
1155     return result;
1156 }
1157 
1158 /**
1159   \brief   Saturating subtract.
1160   \details This function enables you to obtain the saturating add of two integers.
1161   \param [in]    x   first summand of the saturating add operation.
1162   \param [in]    y   second summand of the saturating add operation.
1163   \return        the saturating addition of val1 and val2.
1164   \remark
1165                  res[31:0] = SAT(val1 + SAT(val2))
1166  */
__QSUB(int32_t x,int32_t y)1167 __ALWAYS_INLINE int32_t __QSUB(int32_t x, int32_t y)
1168 {
1169     int64_t tmp;
1170     int32_t result;
1171 
1172     tmp = (int64_t)x - (int64_t)y;
1173 
1174     if (tmp > 0x7fffffff)
1175     {
1176         tmp = 0x7fffffff;
1177     }
1178     else if (tmp < (-2147483647 - 1))
1179     {
1180         tmp = -2147483647 - 1;
1181     }
1182 
1183     result = tmp;
1184     return result;
1185 }
1186 
1187 /**
1188   \brief   Dual 16-bit signed multiply with single 32-bit accumulator.
1189   \details This function enables you to perform two signed 16-bit multiplications,
1190            adding both results to a 32-bit accumulate operand.
1191   \param [in]    x   first 16-bit operands for each multiplication.
1192   \param [in]    y   second 16-bit operands for each multiplication.
1193   \param [in]  sum   accumulate value.
1194   \return        the product of each multiplication added to the accumulate value, as a 32-bit integer.
1195   \remark
1196                  p1 = val1[15:0]  * val2[15:0]      \n
1197                  p2 = val1[31:16] * val2[31:16]     \n
1198                  res[31:0] = p1 + p2 + val3[31:0]
1199  */
__SMLAD(uint32_t x,uint32_t y,uint32_t sum)1200 __ALWAYS_INLINE uint32_t __SMLAD(uint32_t x, uint32_t y, uint32_t sum)
1201 {
1202     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) +
1203                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16)) +
1204                        (((int32_t)sum))));
1205 }
1206 
1207 /**
1208   \brief   Pre-exchanged dual 16-bit signed multiply with single 32-bit accumulator.
1209   \details This function enables you to perform two signed 16-bit multiplications with exchanged
1210            halfwords of the second operand, adding both results to a 32-bit accumulate operand.
1211   \param [in]    x   first 16-bit operands for each multiplication.
1212   \param [in]    y   second 16-bit operands for each multiplication.
1213   \param [in]  sum   accumulate value.
1214   \return        the product of each multiplication with exchanged halfwords of the second
1215                  operand added to the accumulate value, as a 32-bit integer.
1216   \remark
1217                  p1 = val1[15:0]  * val2[31:16]     \n
1218                  p2 = val1[31:16] * val2[15:0]      \n
1219                  res[31:0] = p1 + p2 + val3[31:0]
1220  */
__SMLADX(uint32_t x,uint32_t y,uint32_t sum)1221 __ALWAYS_INLINE uint32_t __SMLADX(uint32_t x, uint32_t y, uint32_t sum)
1222 {
1223     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) +
1224                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16)) +
1225                        (((int32_t)sum))));
1226 }
1227 
1228 /**
1229   \brief   Dual 16-bit signed multiply with exchange subtract with 32-bit accumulate.
1230   \details This function enables you to perform two 16-bit signed multiplications, take the
1231            difference of the products, subtracting the high halfword product from the low
1232            halfword product, and add the difference to a 32-bit accumulate operand.
1233   \param [in]    x   first 16-bit operands for each multiplication.
1234   \param [in]    y   second 16-bit operands for each multiplication.
1235   \param [in]  sum   accumulate value.
1236   \return        the difference of the product of each multiplication, added to the accumulate value.
1237   \remark
1238                  p1 = val1[15:0]  * val2[15:0]       \n
1239                  p2 = val1[31:16] * val2[31:16]      \n
1240                  res[31:0] = p1 - p2 + val3[31:0]
1241  */
__SMLSD(uint32_t x,uint32_t y,uint32_t sum)1242 __ALWAYS_INLINE uint32_t __SMLSD(uint32_t x, uint32_t y, uint32_t sum)
1243 {
1244     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) -
1245                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16)) +
1246                        (((int32_t)sum))));
1247 }
1248 
1249 /**
1250   \brief   Dual 16-bit signed multiply with exchange subtract with 32-bit accumulate.
1251   \details This function enables you to exchange the halfwords in the second operand, then perform two 16-bit
1252            signed multiplications. The difference of the products is added to a 32-bit accumulate operand.
1253   \param [in]    x   first 16-bit operands for each multiplication.
1254   \param [in]    y   second 16-bit operands for each multiplication.
1255   \param [in]  sum   accumulate value.
1256   \return        the difference of the product of each multiplication, added to the accumulate value.
1257   \remark
1258                  p1 = val1[15:0]  * val2[31:16]     \n
1259                  p2 = val1[31:16] * val2[15:0]      \n
1260                  res[31:0] = p1 - p2 + val3[31:0]
1261  */
__SMLSDX(uint32_t x,uint32_t y,uint32_t sum)1262 __ALWAYS_INLINE uint32_t __SMLSDX(uint32_t x, uint32_t y, uint32_t sum)
1263 {
1264     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) -
1265                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16)) +
1266                        (((int32_t)sum))));
1267 }
1268 
1269 /**
1270   \brief   Dual 16-bit signed multiply with single 64-bit accumulator.
1271   \details This function enables you to perform two signed 16-bit multiplications, adding both results
1272            to a 64-bit accumulate operand. Overflow is only possible as a result of the 64-bit addition.
1273            This overflow is not detected if it occurs. Instead, the result wraps around modulo2^64.
1274   \param [in]    x   first 16-bit operands for each multiplication.
1275   \param [in]    y   second 16-bit operands for each multiplication.
1276   \param [in]  sum   accumulate value.
1277   \return        the product of each multiplication added to the accumulate value.
1278   \remark
1279                  p1 = val1[15:0]  * val2[15:0]      \n
1280                  p2 = val1[31:16] * val2[31:16]     \n
1281                  sum = p1 + p2 + val3[63:32][31:0]  \n
1282                  res[63:32] = sum[63:32]            \n
1283                  res[31:0]  = sum[31:0]
1284  */
__SMLALD(uint32_t x,uint32_t y,uint64_t sum)1285 __ALWAYS_INLINE uint64_t __SMLALD(uint32_t x, uint32_t y, uint64_t sum)
1286 {
1287     return ((uint64_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) +
1288                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16)) +
1289                        (((uint64_t)sum))));
1290 }
1291 
1292 /**
1293   \brief   Dual 16-bit signed multiply with exchange with single 64-bit accumulator.
1294   \details This function enables you to exchange the halfwords of the second operand, and perform two
1295            signed 16-bit multiplications, adding both results to a 64-bit accumulate operand. Overflow
1296            is only possible as a result of the 64-bit addition. This overflow is not detected if it occurs.
1297            Instead, the result wraps around modulo2^64.
1298   \param [in]    x   first 16-bit operands for each multiplication.
1299   \param [in]    y   second 16-bit operands for each multiplication.
1300   \param [in]  sum   accumulate value.
1301   \return        the product of each multiplication added to the accumulate value.
1302   \remark
1303                  p1 = val1[15:0]  * val2[31:16]     \n
1304                  p2 = val1[31:16] * val2[15:0]      \n
1305                  sum = p1 + p2 + val3[63:32][31:0]  \n
1306                  res[63:32] = sum[63:32]            \n
1307                  res[31:0]  = sum[31:0]
1308  */
__SMLALDX(uint32_t x,uint32_t y,uint64_t sum)1309 __ALWAYS_INLINE uint64_t __SMLALDX(uint32_t x, uint32_t y, uint64_t sum)
1310 {
1311     return ((uint64_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) +
1312                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16)) +
1313                        (((uint64_t)sum))));
1314 }
1315 
1316 /**
1317   \brief   dual 16-bit signed multiply subtract with 64-bit accumulate.
1318   \details This function It enables you to perform two 16-bit signed multiplications, take the difference
1319            of the products, subtracting the high halfword product from the low halfword product, and add the
1320            difference to a 64-bit accumulate operand. Overflow cannot occur during the multiplications or the
1321            subtraction. Overflow can occur as a result of the 64-bit addition, and this overflow is not
1322            detected. Instead, the result wraps round to modulo2^64.
1323   \param [in]    x   first 16-bit operands for each multiplication.
1324   \param [in]    y   second 16-bit operands for each multiplication.
1325   \param [in]  sum   accumulate value.
1326   \return        the difference of the product of each multiplication, added to the accumulate value.
1327   \remark
1328                  p1 = val1[15:0]  * val2[15:0]      \n
1329                  p2 = val1[31:16] * val2[31:16]     \n
1330                  res[63:0] = p1 - p2 + val3[63:0]
1331  */
__SMLSLD(uint32_t x,uint32_t y,uint64_t sum)1332 __ALWAYS_INLINE uint64_t __SMLSLD(uint32_t x, uint32_t y, uint64_t sum)
1333 {
1334     return ((uint64_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) -
1335                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16)) +
1336                        (((uint64_t)sum))));
1337 }
1338 
1339 /**
1340   \brief   Dual 16-bit signed multiply with exchange subtract with 64-bit accumulate.
1341   \details This function enables you to exchange the halfwords of the second operand, perform two 16-bit multiplications,
1342            adding the difference of the products to a 64-bit accumulate operand. Overflow cannot occur during the
1343            multiplications or the subtraction. Overflow can occur as a result of the 64-bit addition, and this overflow
1344            is not detected. Instead, the result wraps round to modulo2^64.
1345   \param [in]    x   first 16-bit operands for each multiplication.
1346   \param [in]    y   second 16-bit operands for each multiplication.
1347   \param [in]  sum   accumulate value.
1348   \return        the difference of the product of each multiplication, added to the accumulate value.
1349   \remark
1350                  p1 = val1[15:0]  * val2[31:16]      \n
1351                  p2 = val1[31:16] * val2[15:0]       \n
1352                  res[63:0] = p1 - p2 + val3[63:0]
1353  */
__SMLSLDX(uint32_t x,uint32_t y,uint64_t sum)1354 __ALWAYS_INLINE uint64_t __SMLSLDX(uint32_t x, uint32_t y, uint64_t sum)
1355 {
1356     return ((uint64_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y) >> 16)) -
1357                        ((((int32_t)x) >> 16) * (((int32_t)y << 16) >> 16)) +
1358                        (((uint64_t)sum))));
1359 }
1360 
1361 /**
1362   \brief   32-bit signed multiply with 32-bit truncated accumulator.
1363   \details This function enables you to perform a signed 32-bit multiplications, adding the most
1364            significant 32 bits of the 64-bit result to a 32-bit accumulate operand.
1365   \param [in]    x   first operand for multiplication.
1366   \param [in]    y   second operand for multiplication.
1367   \param [in]  sum   accumulate value.
1368   \return        the product of multiplication (most significant 32 bits) is added to the accumulate value, as a 32-bit integer.
1369   \remark
1370                  p = val1 * val2      \n
1371                  res[31:0] = p[61:32] + val3[31:0]
1372  */
__SMMLA(int32_t x,int32_t y,int32_t sum)1373 __ALWAYS_INLINE uint32_t __SMMLA(int32_t x, int32_t y, int32_t sum)
1374 {
1375     return (uint32_t)((int32_t)((int64_t)((int64_t)x * (int64_t)y) >> 32) + sum);
1376 }
1377 
1378 /**
1379   \brief   Sum of dual 16-bit signed multiply.
1380   \details This function enables you to perform two 16-bit signed multiplications, adding the products together.
1381   \param [in]    x   first 16-bit operands for each multiplication.
1382   \param [in]    y   second 16-bit operands for each multiplication.
1383   \return        the sum of the products of the two 16-bit signed multiplications.
1384   \remark
1385                  p1 = val1[15:0]  * val2[15:0]      \n
1386                  p2 = val1[31:16] * val2[31:16]     \n
1387                  res[31:0] = p1 + p2
1388  */
__SMUAD(uint32_t x,uint32_t y)1389 __ALWAYS_INLINE uint32_t __SMUAD(uint32_t x, uint32_t y)
1390 {
1391     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) +
1392                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16))));
1393 }
1394 
1395 /**
1396   \brief   Dual 16-bit signed multiply returning difference.
1397   \details This function enables you to perform two 16-bit signed multiplications, taking the difference
1398            of the products by subtracting the high halfword product from the low halfword product.
1399   \param [in]    x   first 16-bit operands for each multiplication.
1400   \param [in]    y   second 16-bit operands for each multiplication.
1401   \return        the difference of the products of the two 16-bit signed multiplications.
1402   \remark
1403                  p1 = val1[15:0]  * val2[15:0]      \n
1404                  p2 = val1[31:16] * val2[31:16]     \n
1405                  res[31:0] = p1 - p2
1406  */
__SMUSD(uint32_t x,uint32_t y)1407 __ALWAYS_INLINE uint32_t __SMUSD(uint32_t x, uint32_t y)
1408 {
1409     return ((uint32_t)(((((int32_t)x << 16) >> 16) * (((int32_t)y << 16) >> 16)) -
1410                        ((((int32_t)x) >> 16) * (((int32_t)y) >> 16))));
1411 }
1412 
1413 /**
1414   \brief   Dual extracted 8-bit to 16-bit signed addition.
1415   \details This function enables you to extract two 8-bit values from the second operand (at bit positions
1416            [7:0] and [23:16]), sign-extend them to 16-bits each, and add the results to the first operand.
1417   \param [in]    x   values added to the sign-extended to 16-bit values.
1418   \param [in]    y   two 8-bit values to be extracted and sign-extended.
1419   \return        the addition of val1 and val2, where the 8-bit values in val2[7:0] and
1420                  val2[23:16] have been extracted and sign-extended prior to the addition.
1421   \remark
1422                  res[15:0]  = val1[15:0] + SignExtended(val2[7:0])      \n
1423                  res[31:16] = val1[31:16] + SignExtended(val2[23:16])
1424  */
__SXTAB16(uint32_t x,uint32_t y)1425 __ALWAYS_INLINE uint32_t __SXTAB16(uint32_t x, uint32_t y)
1426 {
1427     return ((uint32_t)((((((int32_t)y << 24) >> 24) + (((int32_t)x << 16) >> 16)) & (int32_t)0x0000FFFF) |
1428                        (((((int32_t)y <<  8) >>  8)  + (((int32_t)x >> 16) << 16)) & (int32_t)0xFFFF0000)));
1429 }
1430 
1431 /**
1432   \brief   Extracted 16-bit to 32-bit unsigned addition.
1433   \details This function enables you to extract two 8-bit values from one operand, zero-extend
1434            them to 16 bits each, and add the results to two 16-bit values from another operand.
1435   \param [in]    x   values added to the zero-extended to 16-bit values.
1436   \param [in]    y   two 8-bit values to be extracted and zero-extended.
1437   \return        the addition of val1 and val2, where the 8-bit values in val2[7:0] and
1438                  val2[23:16] have been extracted and zero-extended prior to the addition.
1439   \remark
1440                  res[15:0]  = ZeroExt(val2[7:0]   to 16 bits) + val1[15:0]      \n
1441                  res[31:16] = ZeroExt(val2[31:16] to 16 bits) + val1[31:16]
1442  */
__UXTAB16(uint32_t x,uint32_t y)1443 __ALWAYS_INLINE uint32_t __UXTAB16(uint32_t x, uint32_t y)
1444 {
1445     return ((uint32_t)(((((y << 24) >> 24) + ((x << 16) >> 16)) & 0x0000FFFF) |
1446                        ((((y <<  8) >>  8) + ((x >> 16) << 16)) & 0xFFFF0000)));
1447 }
1448 
1449 /**
1450   \brief   Dual extract 8-bits and sign extend each to 16-bits.
1451   \details This function enables you to extract two 8-bit values from an operand and sign-extend them to 16 bits each.
1452   \param [in]    x   two 8-bit values in val[7:0] and val[23:16] to be sign-extended.
1453   \return        the 8-bit values sign-extended to 16-bit values.\n
1454                  sign-extended value of val[7:0] in the low halfword of the return value.\n
1455                  sign-extended value of val[23:16] in the high halfword of the return value.
1456   \remark
1457                  res[15:0]  = SignExtended(val[7:0])       \n
1458                  res[31:16] = SignExtended(val[23:16])
1459  */
__SXTB16(uint32_t x)1460 __ALWAYS_INLINE uint32_t __SXTB16(uint32_t x)
1461 {
1462     return ((uint32_t)(((((int32_t)x << 24) >> 24) & (int32_t)0x0000FFFF) |
1463                        ((((int32_t)x <<  8) >>  8) & (int32_t)0xFFFF0000)));
1464 }
1465 
1466 /**
1467   \brief   Dual extract 8-bits and zero-extend to 16-bits.
1468   \details This function enables you to extract two 8-bit values from an operand and zero-extend them to 16 bits each.
1469   \param [in]    x   two 8-bit values in val[7:0] and val[23:16] to be zero-extended.
1470   \return        the 8-bit values sign-extended to 16-bit values.\n
1471                  sign-extended value of val[7:0] in the low halfword of the return value.\n
1472                  sign-extended value of val[23:16] in the high halfword of the return value.
1473   \remark
1474                  res[15:0]  = SignExtended(val[7:0])       \n
1475                  res[31:16] = SignExtended(val[23:16])
1476  */
__UXTB16(uint32_t x)1477 __ALWAYS_INLINE uint32_t __UXTB16(uint32_t x)
1478 {
1479     return ((uint32_t)((((x << 24) >> 24) & 0x0000FFFF) |
1480                        (((x <<  8) >>  8) & 0xFFFF0000)));
1481 }
1482 
1483 #endif /* _CSI_SIMD_H_ */
1484