1 /* ###################### CMSIS Support for Cortex-M4 SIMD Instructions ####################### */
4 \defgroup intrinsic_SIMD_gr Intrinsic Functions for SIMD Instructions [only Cortex-M4 and Cortex-M7]
5 \brief Access to dedicated SIMD instructions.
9 <b>Single Instruction Multiple Data (SIMD)</b> extensions are provided <b>only for Cortex-M4 and Cortex-M7 cores</b>
10 to simplify development of application software. SIMD extensions increase the processing capability
11 without materially increasing the power consumption. The SIMD extensions are completely transparent
12 to the operating system (OS), allowing existing OS ports to be used.
16 - Simultaneous computation of 2x16-bit or 4x8-bit operands
17 - Fractional arithmetic
18 - User definable saturation modes (arbitrary word-width)
19 - Dual 16x16 multiply-add/subtract 32x32 fractional MAC
20 - Simultaneous 8/16-bit select operations
21 - Performance up to 3.2 GOPS at 800MHz
22 - Performance is achieved with a "near zero" increase in power consumption on a typical implementation
26 \b Addition: Add two values using SIMD function
29 uint32_t add_halfwords(uint32_t val1, uint32_t val2)
31 return __SADD16(val1, val2);
37 \b Subtraction: Subtract two values using SIMD function
40 uint32_t sub_halfwords(uint32_t val1, uint32_t val2)
42 return __SSUB16(val1, val2);
47 \b Multiplication: Performing a multiplication using SIMD function
50 uint32_t dual_mul_add_products(uint32_t val1, uint32_t val2)
52 return __SMUAD(val1, val2);
60 /**************************************************************************************************/
62 \brief GE setting quad 8-bit signed addition
63 \details This function performs four 8-bit signed integer additions.
64 The GE bits of the APSR are set according to the results of the additions.
65 \param val1 first four 8-bit summands.
66 \param val2 second four 8-bit summands.
69 \li the addition of the first bytes from each operand, in the first byte of the return value.
70 \li the addition of the second bytes of each operand, in the second byte of the return value.
71 \li the addition of the third bytes of each operand, in the third byte of the return value.
72 \li the addition of the fourth bytes of each operand, in the fourth byte of the return value.
74 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
75 the results of the operation.
77 If \em res is the return value, then:
78 \li if res[7:0] \>= 0 then APSR.GE[0] = 1 else 0
79 \li if res[15:8] \>= 0 then APSR.GE[1] = 1 else 0
80 \li if res[23:16] \>= 0 then APSR.GE[2] = 1 else 0
81 \li if res[31:24] \>= 0 then APSR.GE[3] = 1 else 0
85 res[7:0] = val1[7:0] + val2[7:0]
86 res[15:8] = val1[15:8] + val2[15:8]
87 res[23:16] = val1[23:16] + val2[23:16]
88 res[31:24] = val1[31:24] + val2[31:24]
91 uint32_t __SADD8(uint32_t val1, uint32_t val2);
94 /**************************************************************************************************/
95 /** \ingroup Intrinsic_SIMD_gr
96 \brief Q setting quad 8-bit saturating addition
97 \details This function enables you to perform four 8-bit integer additions, saturating the results to
98 the 8-bit signed integer range -2<sup>7</sup> \<= x \<= 2<sup>7</sup> - 1.
99 \param val1 first four 8-bit summands.
100 \param val2 second four 8-bit summands.
103 \li the saturated addition of the first byte of each operand in the first byte of the return value.
104 \li the saturated addition of the second byte of each operand in the second byte of the return value.
105 \li the saturated addition of the third byte of each operand in the third byte of the return value.
106 \li the saturated addition of the fourth byte of each operand in the fourth byte of the return value.
108 The returned results are saturated to the 16-bit signed integer range -2<sup>7</sup> \<= x \<= 2<sup>7</sup> - 1.
112 res[7:0] = val1[7:0] + val2[7:0]
113 res[15:8] = val1[15:8] + val2[15:8]
114 res[23:16] = val1[23:16] + val2[23:16]
115 res[31:24] = val1[31:24] + val2[31:24]
118 uint32_t __QADD8(uint32_t val1, uint32_t val2);
121 /**************************************************************************************************/
123 \brief Quad 8-bit signed addition with halved results
124 \details This function enables you to perform four signed 8-bit integer additions, halving the results.
125 \param val1 first four 8-bit summands.
126 \param val2 second four 8-bit summands.
129 \li the halved addition of the first bytes from each operand, in the first byte of the return value.
130 \li the halved addition of the second bytes from each operand, in the second byte of the return value.
131 \li the halved addition of the third bytes from each operand, in the third byte of the return value.
132 \li the halved addition of the fourth bytes from each operand, in the fourth byte of the return value.
136 res[7:0] = val1[7:0] + val2[7:0] >> 1
137 res[15:8] = val1[15:8] + val2[15:8] >> 1
138 res[23:16] = val1[23:16] + val2[23:16] >> 1
139 res[31:24] = val1[31:24] + val2[31:24] >> 1
142 uint32_t __SHADD8(uint32_t val1, uint32_t val2);
145 /**************************************************************************************************/
146 /** \ingroup Intrinsic_SIMD_gr
147 \brief GE setting quad 8-bit unsigned addition
149 \details This function enables you to perform four unsigned 8-bit integer additions.
150 The GE bits of the APSR are set according to the results.
152 \param val1 first four 8-bit summands for each addition.
153 \param val2 second four 8-bit summands for each addition.
156 \li the halved addition of the first bytes from each operand, in the first byte of the return value.
157 \li the halved addition of the second bytes from each operand, in the second byte of the return value.
158 \li the halved addition of the third bytes from each operand, in the third byte of the return value.
159 \li the halved addition of the fourth bytes from each operand, in the fourth byte of the return value.
162 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on the results of the operation.
165 If \em res is the return value, then:
166 \li if res[7:0] \>= 0x100 then APSR.GE[0] = 1 else 0
167 \li if res[15:8] \>= 0x100 then APSR.GE[1] = 1 else 0
168 \li if res[23:16] \>= 0x100 then APSR.GE[2] = 1 else 0
169 \li if res[31:24] \>= 0x100 then APSR.GE[3] = 1 else 0
173 res[7:0] = val1[7:0] + val2[7:0]
174 res[15:8] = val1[15:8] + val2[15:8]
175 res[23:16] = val1[23:16] + val2[23:16]
176 res[31:24] = val1[31:24] + val2[31:24]
179 uint32_t __UADD8(uint32_t val1, uint32_t val2);
182 /**************************************************************************************************/
184 \brief Quad 8-bit unsigned saturating addition
186 \details This function enables you to perform four unsigned 8-bit integer additions, saturating the
187 results to the 8-bit unsigned integer range 0 \< x \< 2<sup>8</sup> - 1.
189 \param val1 first four 8-bit summands.
190 \param val2 second four 8-bit summands.
193 \li the halved addition of the first bytes in each operand, in the first byte of the return value.
194 \li the halved addition of the second bytes in each operand, in the second byte of the return value.
195 \li the halved addition of the third bytes in each operand, in the third byte of the return value.
196 \li the halved addition of the fourth bytes in each operand, in the fourth byte of the return value.
199 The results are saturated to the 8-bit unsigned integer range 0 \< x \< 2<sup>8</sup> - 1.
203 res[7:0] = val1[7:0] + val2[7:0]
204 res[15:8] = val1[15:8] + val2[15:8]
205 res[23:16] = val1[23:16] + val2[23:16]
206 res[31:24] = val1[31:24] + val2[31:24]
209 uint32_t __UQADD8(uint32_t val1, uint32_t val2);
212 /**************************************************************************************************/
214 \brief Quad 8-bit unsigned addition with halved results
216 \details This function enables you to perform four unsigned 8-bit integer additions, halving the results.
218 \param val1 first four 8-bit summands.
219 \param val2 second four 8-bit summands.
222 \li the halved addition of the first bytes in each operand, in the first byte of the return value.
223 \li the halved addition of the second bytes in each operand, in the second byte of the return value.
224 \li the halved addition of the third bytes in each operand, in the third byte of the return value.
225 \li the halved addition of the fourth bytes in each operand, in the fourth byte of the return value.
229 res[7:0] = val1[7:0] + val2[7:0] >> 1
230 res[15:8] = val1[15:8] + val2[15:8] >> 1
231 res[23:16] = val1[23:16] + val2[23:16] >> 1
232 res[31:24] = val1[31:24] + val2[31:24] >> 1
235 uint32_t __UHADD8(uint32_t val1, uint32_t val2);
238 /**************************************************************************************************/
240 \brief GE setting quad 8-bit signed subtraction
242 \details This function enables you to perform four 8-bit signed integer subtractions.<br>
243 The GE bits in the APSR are set according to the results.
245 \param val1 first four 8-bit operands of each subtraction.
246 \param val2 second four 8-bit operands of each subtraction.
249 \li the subtraction of the first byte in the second operand from the first byte in the
250 first operand, in the first bytes of the return value.
251 \li the subtraction of the second byte in the second operand from the second byte in
252 the first operand, in the second byte of the return value.
253 \li the subtraction of the third byte in the second operand from the third byte in the
254 first operand, in the third byte of the return value.
255 \li the subtraction of the fourth byte in the second operand from the fourth byte in
256 the first operand, in the fourth byte of the return value.
258 \par Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
259 the results of the operation.
262 If \em res is the return value, then:
263 \li if res[8:0] \>= 0 then APSR.GE[0] = 1 else 0
264 \li if res[15:8] \>= 0 then APSR.GE[1] = 1 else 0
265 \li if res[23:16] \>= 0 then APSR.GE[2] = 1 else 0
266 \li if res[31:24] \>= 0 then APSR.GE[3] = 1 else 0
271 res[7:0] = val1[7:0] - val2[7:0]
272 res[15:8] = val1[15:8] - val2[15:8]
273 res[23:16] = val1[23:16] - val2[23:16]
274 res[31:24] = val1[31:24] - val2[31:24]
277 uint32_t __SSUB8(uint32_t val1, uint32_t val2);
280 /**************************************************************************************************/
282 \brief Q setting quad 8-bit saturating subtract
284 \details This function enables you to perform four 8-bit integer subtractions, saturating the results
285 to the 8-bit signed integer range -2<sup>7</sup> \<= x \<= 2<sup>7</sup> - 1.
287 \param val1 first four 8-bit operands.
288 \param val2 second four 8-bit operands.
291 \li the subtraction of the first byte in the second operand from the first byte in the
292 first operand, in the first bytes of the return value.
293 \li the subtraction of the second byte in the second operand from the second byte in
294 the first operand, in the second byte of the return value.
295 \li the subtraction of the third byte in the second operand from the third byte in the
296 first operand, in the third byte of the return value.
297 \li the subtraction of the fourth byte in the second operand from the fourth byte in
298 the first operand, in the fourth byte of the return value.
301 The returned results are saturated to the 8-bit signed integer range -2<sup>7</sup> \<= x \<= 2<sup>7</sup> - 1.
306 res[7:0] = val1[7:0] - val2[7:0]
307 res[15:8] = val1[15:8] - val2[15:8]
308 res[23:16] = val1[23:16] - val2[23:16]
309 res[31:24] = val1[31:24] - val2[31:24]
312 uint32_t __QSUB8(uint32_t val1, uint32_t val2);
315 /**************************************************************************************************/
317 \brief Quad 8-bit signed subtraction with halved results
319 \details This function enables you to perform four signed 8-bit integer subtractions, halving the
322 \param val1 first four 8-bit operands.
323 \param val2 second four 8-bit operands.
326 \li the halved subtraction of the first byte in the second operand from the first byte in the
327 first operand, in the first bytes of the return value.
328 \li the halved subtraction of the second byte in the second operand from the second byte in
329 the first operand, in the second byte of the return value.
330 \li the halved subtraction of the third byte in the second operand from the third byte in the
331 first operand, in the third byte of the return value.
332 \li the halved subtraction of the fourth byte in the second operand from the fourth byte in
333 the first operand, in the fourth byte of the return value.
337 res[7:0] = val1[7:0] - val2[7:0] >> 1
338 res[15:8] = val1[15:8] - val2[15:8] >> 1
339 res[23:16] = val1[23:16] - val2[23:16] >> 1
340 res[31:24] = val1[31:24] - val2[31:24] >> 1
343 uint32_t __SHSUB8(uint32_t val1, uint32_t val2);
346 /**************************************************************************************************/
348 \brief GE setting quad 8-bit unsigned subtract
350 \details This function enables you to perform four 8-bit unsigned integer subtractions.
351 The GE bits in the APSR are set according to the results.
353 \param val1 first four 8-bit operands.
354 \param val2 second four 8-bit operands.
357 \li the subtraction of the first byte in the second operand from the first byte in the
358 first operand, in the first bytes of the return value.
359 \li the subtraction of the second byte in the second operand from the second byte in
360 the first operand, in the second byte of the return value.
361 \li the subtraction of the third byte in the second operand from the third byte in the
362 first operand, in the third byte of the return value.
363 \li the subtraction of the fourth byte in the second operand from the fourth byte in
364 the first operand, in the fourth byte of the return value.
367 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
368 the results of the operation.
371 If \em res is the return value, then:
372 \li if res[8:0] \>= 0 then APSR.GE[0] = 1 else 0
373 \li if res[15:8] \>= 0 then APSR.GE[1] = 1 else 0
374 \li if res[23:16] \>= 0 then APSR.GE[2] = 1 else 0
375 \li if res[31:24] \>= 0 then APSR.GE[3] = 1 else 0
380 res[7:0] = val1[7:0] - val2[7:0]
381 res[15:8] = val1[15:8] - val2[15:8]
382 res[23:16] = val1[23:16] - val2[23:16]
383 res[31:24] = val1[31:24] - val2[31:24]
386 uint32_t __USUB8(uint32_t val1, uint32_t val2);
389 /**************************************************************************************************/
391 \brief Quad 8-bit unsigned saturating subtraction
393 \details This function enables you to perform four unsigned 8-bit integer subtractions, saturating
394 the results to the 8-bit unsigned integer range 0 \< x \< 2<sup>8</sup> - 1.
396 \param val1 first four 8-bit operands.
397 \param val2 second four 8-bit operands.
400 \li the subtraction of the first byte in the second operand from the first byte in the
401 first operand, in the first bytes of the return value.
402 \li the subtraction of the second byte in the second operand from the second byte in
403 the first operand, in the second byte of the return value.
404 \li the subtraction of the third byte in the second operand from the third byte in the
405 first operand, in the third byte of the return value.
406 \li the subtraction of the fourth byte in the second operand from the fourth byte in
407 the first operand, in the fourth byte of the return value.
410 The results are saturated to the 8-bit unsigned integer range 0 \< x \< 2<sup>8</sup> - 1.
415 res[7:0] = val1[7:0] - val2[7:0]
416 res[15:8] = val1[15:8] - val2[15:8]
417 res[23:16] = val1[23:16] - val2[23:16]
418 res[31:24] = val1[31:24] - val2[31:24]
421 uint32_t __UQSUB8(uint32_t val1, uint32_t val2);
424 /**************************************************************************************************/
426 \brief Quad 8-bit unsigned subtraction with halved results
428 \details This function enables you to perform four unsigned 8-bit integer subtractions, halving the
431 \param val1 first four 8-bit operands.
432 \param val2 second four 8-bit operands.
435 \li the halved subtraction of the first byte in the second operand from the first byte in the
436 first operand, in the first bytes of the return value.
437 \li the halved subtraction of the second byte in the second operand from the second byte in
438 the first operand, in the second byte of the return value.
439 \li the halved subtraction of the third byte in the second operand from the third byte in the
440 first operand, in the third byte of the return value.
441 \li the halved subtraction of the fourth byte in the second operand from the fourth byte in
442 the first operand, in the fourth byte of the return value.
446 res[7:0] = val1[7:0] - val2[7:0] >> 1
447 res[15:8] = val1[15:8] - val2[15:8] >> 1
448 res[23:16] = val1[23:16] - val2[23:16] >> 1
449 res[31:24] = val1[31:24] - val2[31:24] >> 1
452 uint32_t __UHSUB8(uint32_t val1, uint32_t val2);
455 /**************************************************************************************************/
457 \brief GE setting dual 16-bit signed addition
459 \details This function enables you to perform two 16-bit signed integer additions.<br>
460 The GE bits in the APSR are set according to the results of the additions.
462 \param val1 first two 16-bit summands.
463 \param val2 second two 16-bit summands.
466 \li the addition of the low halfwords in the low halfword of the return value.
467 \li the addition of the high halfwords in the high halfword of the return value.
470 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
471 the results of the operation.
473 If \em res is the return value, then:
474 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
475 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
480 res[15:0] = val1[15:0] + val2[15:0]
481 res[31:16] = val1[31:16] + val2[31:16]
484 uint32_t __SADD16(uint32_t val1, uint32_t val2);
487 /**************************************************************************************************/
489 \brief Q setting dual 16-bit saturating addition
491 \details This function enables you to perform two 16-bit integer arithmetic additions in parallel,
492 saturating the results to the 16-bit signed integer range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
494 \param val1 first two 16-bit summands.
495 \param val2 second two 16-bit summands.
498 \li the saturated addition of the low halfwords, in the low halfword of the return value.
499 \li the saturated addition of the high halfwords, in the high halfword of the return value.
502 The returned results are saturated to the 16-bit signed integer
503 range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1
507 res[15:0] = val1[15:0] + val2[15:0]
508 res[31:16] = val1[31:16] + val2[31:16]
511 uint32_t __QADD16(uint32_t val1, uint32_t val2);
514 /**************************************************************************************************/
516 \brief Dual 16-bit signed addition with halved results
518 \details This function enables you to perform two signed 16-bit integer additions, halving the
521 \param val1 first two 16-bit summands.
522 \param val2 second two 16-bit summands.
525 \li the halved addition of the low halfwords, in the low halfword of the return value.
526 \li the halved addition of the high halfwords, in the high halfword of the return value.
530 res[15:0] = val1[15:0] + val2[15:0] >> 1
531 res[31:16] = val1[31:16] + val2[31:16] >> 1
534 uint32_t __SHADD16(uint32_t val1, uint32_t val2);
537 /**************************************************************************************************/
539 \brief GE setting dual 16-bit unsigned addition
541 \details This function enables you to perform two 16-bit unsigned integer additions.<br>
542 The GE bits in the APSR are set according to the results.
544 \param val1 first two 16-bit summands for each addition.
545 \param val2 second two 16-bit summands for each addition.
548 \li the addition of the low halfwords in each operand, in the low halfword of the
550 \li the addition of the high halfwords in each operand, in the high halfword of the
554 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
555 the results of the operation.
557 If \em res is the return value, then:
558 \li if res[15:0] \>= 0x10000 then APSR.GE[0] = 11 else 00
559 \li if res[31:16] \>= 0x10000 then APSR.GE[1] = 11 else 00
563 res[15:0] = val1[15:0] + val2[15:0]
564 res[31:16] = val1[31:16] + val2[31:16]
567 uint32_t __UADD16(uint32_t val1, uint32_t val2);
570 /**************************************************************************************************/
572 \brief Dual 16-bit unsigned saturating addition
574 \details This function enables you to perform two unsigned 16-bit integer additions, saturating the
575 results to the 16-bit unsigned integer range 0 \< x \< 2<sup>16</sup> - 1.
577 \param val1 first two 16-bit summands.
578 \param val2 second two 16-bit summands.
581 \li the addition of the low halfword in the first operand and the low halfword in the
582 second operand, in the low halfword of the return value.
583 \li the addition of the high halfword in the first operand and the high halfword in the
584 second operand, in the high halfword of the return value.
587 The results are saturated to the 16-bit unsigned integer
588 range 0 \< x \< 2<sup>16</sup> - 1.
592 res[15:0] = val1[15:0] + val2[15:0]
593 res[31:16] = val1[31:16] + val2[31:16]
596 uint32_t __UQADD16(uint32_t val1, uint32_t val2);
599 /**************************************************************************************************/
601 \brief Dual 16-bit unsigned addition with halved results
603 \details This function enables you to perform two unsigned 16-bit integer additions, halving the
606 \param val1 first two 16-bit summands.
607 \param val2 second two 16-bit summands.
610 \li the halved addition of the low halfwords in each operand, in the low halfword of
612 \li the halved addition of the high halfwords in each operand, in the high halfword
617 res[15:0] = val1[15:0] + val2[15:0] >> 1
618 res[31:16] = val1[31:16] + val2[31:16] >> 1
621 uint32_t __UHADD16(uint32_t val1, uint32_t val2);
624 /**************************************************************************************************/
626 \brief GE setting dual 16-bit signed subtraction
628 \details This function enables you to perform two 16-bit signed integer subtractions.<br>
629 The GE bits in the APSR are set according to the results.
631 \param val1 first two 16-bit operands of each subtraction.
632 \param val2 second two 16-bit operands of each subtraction.
635 \li the subtraction of the low halfword in the second operand from the low halfword
636 in the first operand, in the low halfword of the return value.
637 \li the subtraction of the high halfword in the second operand from the high halfword
638 in the first operand, in the high halfword of the return value.
641 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
642 the results of the operation.
644 If \li res is the return value, then:
645 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
646 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
651 res[15:0] = val1[15:0] - val2[15:0]
652 res[31:16] = val1[31:16] - val2[31:16]
655 uint32_t __SSUB16(uint32_t val1, uint32_t val2);
658 /**************************************************************************************************/
660 \brief Q setting dual 16-bit saturating subtract
662 \details This function enables you to perform two 16-bit integer subtractions, saturating the
663 results to the 16-bit signed integer range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
665 \param val1 first two 16-bit operands.
666 \param val2 second two 16-bit operands.
669 \li the saturated subtraction of the low halfword in the second operand from the low
670 halfword in the first operand, in the low halfword of the returned result.
671 \li the saturated subtraction of the high halfword in the second operand from the high
672 halfword in the first operand, in the high halfword of the returned result.
675 The returned results are saturated to the 16-bit signed integer
676 range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
680 res[15:0] = val1[15:0] - val2[15:0]
681 res[31:16] = val1[31:16] - val2[31:16]
684 uint32_t __QSUB16(uint32_t val1, uint32_t val2);
687 /**************************************************************************************************/
689 \brief Dual 16-bit signed subtraction with halved results
691 \details This function enables you to perform two signed 16-bit integer subtractions, halving the
694 \param val1 first two 16-bit operands.
695 \param val2 second two 16-bit operands.
698 \li the halved subtraction of the low halfword in the second operand from the low
699 halfword in the first operand, in the low halfword of the returned result.
700 \li the halved subtraction of the high halfword in the second operand from the high
701 halfword in the first operand, in the high halfword of the returned result.
706 res[15:0] = val1[15:0] - val2[15:0] >> 1
707 res[31:16] = val1[31:16] - val2[31:16] >> 1
710 uint32_t __SHSUB16(uint32_t val1, uint32_t val2);
713 /**************************************************************************************************/
715 \brief GE setting dual 16-bit unsigned subtract
717 \details This function enables you to perform two 16-bit unsigned integer subtractions.<br>
718 The GE bits in the APSR are set according to the results.
720 \param val1 first two 16-bit operands.
721 \param val2 second two 16-bit operands.
724 \li the subtraction of the low halfword in the second operand from the low halfword
725 in the first operand, in the low halfword of the return value.
726 \li the subtraction of the high halfword in the second operand from the high halfword
727 in the first operand, in the high halfword of the return value.
730 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
731 the results of the operation.
734 If \em res is the return value, then:
735 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
736 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
740 res[15:0] = val1[15:0] - val2[15:0]
741 res[31:16] = val1[31:16] - val2[31:16]
744 uint32_t __USUB16(uint32_t val1, uint32_t val2);
747 /**************************************************************************************************/
749 \brief Dual 16-bit unsigned saturating subtraction
751 \details This function enables you to perform two unsigned 16-bit integer subtractions, saturating
752 the results to the 16-bit unsigned integer range 0 \< x \< 2<sup>16</sup> - 1.
754 \param val1 first two 16-bit operands for each subtraction.
755 \param val2 second two 16-bit operands for each subtraction.
758 \li the subtraction of the low halfword in the second operand from the low halfword
759 in the first operand, in the low halfword of the return value.
760 \li the subtraction of the high halfword in the second operand from the high halfword
761 in the first operand, in the high halfword of the return value.
764 The results are saturated to the 16-bit unsigned integer range 0 \< x \< 2<sup>16</sup> - 1.
769 res[15:0] = val1[15:0] - val2[15:0]
770 res[31:16] = val1[31:16] - val2[31:16]
773 uint32_t __UQSUB16(uint32_t val1, uint32_t val2);
776 /**************************************************************************************************/
778 \brief Dual 16-bit unsigned subtraction with halved results
780 \details This function enables you to perform two unsigned 16-bit integer subtractions, halving
783 \param val1 first two 16-bit operands.
784 \param val2 second two 16-bit operands.
787 \li the halved subtraction of the low halfword in the second operand from the low halfword
788 in the first operand, in the low halfword of the return value.
789 \li the halved subtraction of the high halfword in the second operand from the high halfword
790 in the first operand, in the high halfword of the return value.
795 res[15:0] = val1[15:0] - val2[15:0] >> 1
796 res[31:16] = val1[31:16] - val2[31:16] >> 1
799 uint32_t __UHSUB16(uint32_t val1, uint32_t val2);
802 /**************************************************************************************************/
804 \brief GE setting dual 16-bit addition and subtraction with exchange
806 \details This function inserts an SASX instruction into the instruction stream generated by the
807 compiler. It enables you to exchange the halfwords of the second operand, add the high
808 halfwords and subtract the low halfwords.<br>
809 The GE bits in the APRS are set according to the results.
811 \param val1 first operand for the subtraction in the low halfword, and the
812 first operand for the addition in the high halfword.
813 \param val2 second operand for the subtraction in the high halfword, and the
814 second operand for the addition in the low halfword.
817 \li the subtraction of the high halfword in the second operand from the low halfword
818 in the first operand, in the low halfword of the return value.
819 \li the addition of the high halfword in the first operand and the low halfword in the
820 second operand, in the high halfword of the return value.
823 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
824 the results of the operation.
826 If \em res is the return value, then:
827 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
828 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
832 res[15:0] = val1[15:0] - val2[31:16]
833 res[31:16] = val1[31:16] + val2[15:0]
836 uint32_t __SASX(uint32_t val1, uint32_t val2);
839 /**************************************************************************************************/
841 \brief Q setting dual 16-bit add and subtract with exchange
843 \details This function enables you to exchange the halfwords of the one operand, then add the high
844 halfwords and subtract the low halfwords, saturating the results to the 16-bit signed
845 integer range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
847 \param val1 first operand for the subtraction in the low halfword, and the
848 first operand for the addition in the high halfword.
849 \param val2 second operand for the subtraction in the high halfword, and the
850 second operand for the addition in the low halfword.
853 \li the saturated subtraction of the high halfword in the second operand from the low
854 halfword in the first operand, in the low halfword of the return value.
855 \li the saturated addition of the high halfword in the first operand and the low
856 halfword in the second operand, in the high halfword of the return value.
859 The returned results are saturated to the 16-bit signed integer
860 range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
864 res[15:0] = val1[15:0] - val2[31:16]
865 res[31:16] = val1[31:16] + val2[15:0]
868 uint32_t __QASX(uint32_t val1, uint32_t val2);
871 /**************************************************************************************************/
873 \brief Dual 16-bit signed addition and subtraction with halved results
875 \details This function enables you to exchange the two halfwords of one operand, perform one
876 signed 16-bit integer addition and one signed 16-bit subtraction, and halve the results.
878 \param val1 first 16-bit operands.
879 \param val2 second 16-bit operands.
882 \li the halved subtraction of the high halfword in the second operand from the low
883 halfword in the first operand, in the low halfword of the return value.
884 \li the halved subtraction of the low halfword in the second operand from the high
885 halfword in the first operand, in the high halfword of the return value.
889 res[15:0] = (val1[15:0] - val2[31:16]) >> 1
890 res[31:16] = (val1[31:16] - val2[15:0] ) >> 1
893 uint32_t __SHASX(uint32_t val1, uint32_t val2);
896 /**************************************************************************************************/
898 \brief GE setting dual 16-bit unsigned addition and subtraction with exchange
900 \details This function enables you to exchange the two halfwords of the second operand, add the
901 high halfwords and subtract the low halfwords.<br>
902 The GE bits in the APSR are set according to the results.
904 \param val1 first operand for the subtraction in the low halfword, and the
905 first operand for the addition in the high halfword.
906 \param val2 second operand for the subtraction in the high halfword and the
907 second operand for the addition in the low halfword.
910 \li the subtraction of the high halfword in the second operand from the low halfword
911 in the first operand, in the low halfword of the return value.
912 \li the addition of the high halfword in the first operand and the low halfword in the
913 second operand, in the high halfword of the return value.
916 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
917 the results of the operation.
919 \par If \em res is the return value, then:
920 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
921 \li if res[31:16] \>= 0x10000 then APSR.GE[3:2] = 11 else 00
925 res[15:0] = val1[15:0] - val2[31:16]
926 res[31:16] = val1[31:16] + val2[15:0]
929 uint32_t __UASX(uint32_t val1, uint32_t val2);
932 /**************************************************************************************************/
934 \brief Dual 16-bit unsigned saturating addition and subtraction with exchange
936 \details This function enables you to exchange the halfwords of the second operand and perform
937 one unsigned 16-bit integer addition and one unsigned 16-bit subtraction, saturating the
938 results to the 16-bit unsigned integer range 0 \<= x \<= 2<sup>16</sup> - 1.
940 \param val1 first two 16-bit operands.
941 \param val2 second two 16-bit operands.
944 \li the subtraction of the high halfword in the second operand from the low halfword
945 in the first operand, in the low halfword of the return value.
946 \li the subtraction of the low halfword in the second operand from the high halfword
947 in the first operand, in the high halfword of the return value.
950 The results are saturated to the 16-bit unsigned integer
951 range 0 \<= x \<= 2<sup>16</sup> - 1.
955 res[15:0] = val1[15:0] - val2[31:16]
956 res[31:16] = val1[31:16] + val2[15:0]
959 uint32_t __UQASX(uint32_t val1, uint32_t val2);
962 /**************************************************************************************************/
964 \brief Dual 16-bit unsigned addition and subtraction with halved results and exchange
966 \details This function enables you to exchange the halfwords of the second operand, add the high
967 halfwords and subtract the low halfwords, halving the results.
969 \param val1 first operand for the subtraction in the low halfword, and the
970 first operand for the addition in the high halfword.
971 \param val2 second operand for the subtraction in the high halfword, and the
972 second operand for the addition in the low halfword.
975 \li the halved subtraction of the high halfword in the second operand from the low
976 halfword in the first operand.
977 \li the halved addition of the high halfword in the first operand and the low halfword
978 in the second operand.
983 res[15:0] = (val1[15:0] - val2[31:16]) >> 1
984 res[31:16] = (val1[31:16] + val2[15:0] ) >> 1
987 uint32_t __UHASX(uint32_t val1, uint32_t val2);
990 /**************************************************************************************************/
992 \brief GE setting dual 16-bit signed subtraction and addition with exchange
994 \details This function enables you to exchange the two halfwords of one operand and perform one
995 16-bit integer subtraction and one 16-bit addition.<br>
996 The GE bits in the APSR are set according to the results.
998 \param val1 first operand for the addition in the low halfword, and the first
999 operand for the subtraction in the high halfword.
1000 \param val2 second operand for the addition in the high halfword, and the
1001 second operand for the subtraction in the low halfword.
1004 \li the addition of the low halfword in the first operand and the high halfword in the
1005 second operand, in the low halfword of the return value.
1006 \li the subtraction of the low halfword in the second operand from the high halfword
1007 in the first operand, in the high halfword of the return value.
1009 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
1010 the results of the operation.
1012 If \em res is the return value, then:
1013 \li if res[15:0] \>= 0 then APSR.GE[1:0] = 11 else 00
1014 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
1018 res[15:0] = val1[15:0] + val2[31:16]
1019 res[31:16] = val1[31:16] - val2[15:0]
1022 uint32_t __SSAX(uint32_t val1, uint32_t val2);
1025 /**************************************************************************************************/
1027 \brief Q setting dual 16-bit subtract and add with exchange
1029 \details This function enables you to exchange the halfwords of one operand, then subtract the
1030 high halfwords and add the low halfwords, saturating the results to the 16-bit signed
1031 integer range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
1033 \param val1 first operand for the addition in the low halfword, and the first
1034 operand for the subtraction in the high halfword.
1035 \param val2 second operand for the addition in the high halfword, and the
1036 second operand for the subtraction in the low halfword.
1039 \li the saturated addition of the low halfword of the first operand and the high
1040 halfword of the second operand, in the low halfword of the return value.
1041 \li the saturated subtraction of the low halfword of the second operand from the high
1042 halfword of the first operand, in the high halfword of the return value.
1044 The returned results are saturated to the 16-bit signed integer
1045 range -2<sup>15</sup> \<= x \<= 2<sup>15</sup> - 1.
1049 res[15:0] = val1[15:0] + val2[31:16]
1050 res[31:16] = val1[31:16] - val2[15:0]
1053 uint32_t __QSAX(uint32_t val1, uint32_t val2);
1056 /**************************************************************************************************/
1058 \brief Dual 16-bit signed subtraction and addition with halved results
1060 \details This function enables you to exchange the two halfwords of one operand, perform one
1061 signed 16-bit integer subtraction and one signed 16-bit addition, and halve the results.
1063 \param val1 first 16-bit operands.
1064 \param val2 second 16-bit operands.
1067 \li the halved addition of the low halfword in the first operand and the high halfword
1068 in the second operand, in the low halfword of the return value.
1069 \li the halved subtraction of the low halfword in the second operand from the high
1070 halfword in the first operand, in the high halfword of the return value.
1074 res[15:0] = (val1[15:0] + val2[31:16]) >> 1
1075 res[31:16] = (val1[31:16] - val2[15:0] ) >> 1
1078 uint32_t __SHSAX(uint32_t val1, uint32_t val2);
1081 /**************************************************************************************************/
1083 \brief GE setting dual 16-bit unsigned subtract and add with exchange
1085 \details This function enables you to exchange the halfwords of the second operand, subtract the
1086 high halfwords and add the low halfwords.<br>
1087 The GE bits in the APSR are set according to the results.
1089 \param val1 first operand for the addition in the low halfword, and the first
1090 operand for the subtraction in the high halfword.
1091 \param val2 second operand for the addition in the high halfword, and the
1092 second operand for the subtraction in the low halfword.
1095 \li the addition of the low halfword in the first operand and the high halfword in the
1096 second operand, in the low halfword of the return value.
1097 \li the subtraction of the low halfword in the second operand from the high halfword
1098 in the first operand, in the high halfword of the return value.
1100 Each bit in APSR.GE is set or cleared for each byte in the return value, depending on
1101 the results of the operation.
1103 If \em res is the return value, then:
1104 \li if res[15:0] \>= 0x10000 then APSR.GE[1:0] = 11 else 00
1105 \li if res[31:16] \>= 0 then APSR.GE[3:2] = 11 else 00
1109 res[15:0] = val1[15:0] + val2[31:16]
1110 res[31:16] = val1[31:16] - val2[15:0]
1113 uint32_t __USAX(uint32_t val1, uint32_t val2);
1116 /**************************************************************************************************/
1118 \brief Dual 16-bit unsigned saturating subtraction and addition with exchange
1120 \details This function enables you to exchange the halfwords of the second operand and perform
1121 one unsigned 16-bit integer subtraction and one unsigned 16-bit addition, saturating the
1122 results to the 16-bit unsigned integer range 0 \<= x \<= 2<sup>16</sup> - 1.
1124 \param val1 first 16-bit operand for the addition in the low halfword, and the
1125 first 16-bit operand for the subtraction in the high halfword.
1126 \param val2 second 16-bit halfword for the addition in the high halfword,
1127 and the second 16-bit halfword for the subtraction in the low halfword.
1130 \li the addition of the low halfword in the first operand and the high halfword in the
1131 second operand, in the low halfword of the return value.
1132 \li the subtraction of the low halfword in the second operand from the high halfword
1133 in the first operand, in the high halfword of the return value.
1135 The results are saturated to the 16-bit unsigned integer
1136 range 0 \<= x \<= 2<sup>16</sup> - 1.
1140 res[15:0] = val1[15:0] + val2[31:16]
1141 res[31:16] = val1[31:16] - val2[15:0]
1144 uint32_t __UQSAX(uint32_t val1, uint32_t val2);
1147 /**************************************************************************************************/
1149 \brief Dual 16-bit unsigned subtraction and addition with halved results and exchange
1151 \details This function enables you to exchange the halfwords of the second operand, subtract the
1152 high halfwords and add the low halfwords, halving the results.
1154 \param val1 first operand for the addition in the low halfword, and the first
1155 operand for the subtraction in the high halfword.
1156 \param val2 second operand for the addition in the high halfword, and the
1157 second operand for the subtraction in the low halfword.
1160 \li the halved addition of the high halfword in the second operand and the low
1161 halfword in the first operand, in the low halfword of the return value.
1162 \li the halved subtraction of the low halfword in the second operand from the high
1163 halfword in the first operand, in the high halfword of the return value.
1167 res[15:0] = (val1[15:0] + val2[31:16]) >> 1
1168 res[31:16] = (val1[31:16] - val2[15:0] ) >> 1
1171 uint32_t __UHSAX(uint32_t val1, uint32_t val2);
1174 /**************************************************************************************************/
1176 \brief Unsigned sum of quad 8-bit unsigned absolute difference
1178 \details This function enables you to perform four unsigned 8-bit subtractions, and add the
1179 absolute values of the differences together, returning the result as a single unsigned
1182 \param val1 first four 8-bit operands for the subtractions.
1183 \param val2 second four 8-bit operands for the subtractions.
1186 \li the subtraction of the first byte in the second operand from the first byte in the
1188 \li the subtraction of the second byte in the second operand from the second byte in
1190 \li the subtraction of the third byte in the second operand from the third byte in the
1192 \li the subtraction of the fourth byte in the second operand from the fourth byte in
1195 The sum is returned as a single unsigned integer.
1200 absdiff1 = val1[7:0] - val2[7:0]
1201 absdiff2 = val1[15:8] - val2[15:8]
1202 absdiff3 = val1[23:16] - val2[23:16]
1203 absdiff4 = val1[31:24] - val2[31:24]
1204 res[31:0] = absdiff1 + absdiff2 + absdiff3 + absdiff4
1207 uint32_t __USAD8(uint32_t val1, uint32_t val2);
1210 /**************************************************************************************************/
1212 \brief Unsigned sum of quad 8-bit unsigned absolute difference with 32-bit accumulate
1214 \details This function enables you to perform four unsigned 8-bit subtractions, and add the
1215 absolute values of the differences to a 32-bit accumulate operand.
1217 \param val1 first four 8-bit operands for the subtractions.
1218 \param val2 second four 8-bit operands for the subtractions.
1219 \param val3 accumulation value.
1222 the sum of the absolute differences of the following
1223 bytes, added to the accumulation value:
1224 \li the subtraction of the first byte in the second operand from the first byte in the
1226 \li the subtraction of the second byte in the second operand from the second byte in
1228 \li the subtraction of the third byte in the second operand from the third byte in the
1230 \li the subtraction of the fourth byte in the second operand from the fourth byte in
1236 absdiff1 = val1[7:0] - val2[7:0]
1237 absdiff2 = val1[15:8] - val2[15:8]
1238 absdiff3 = val1[23:16] - val2[23:16]
1239 absdiff4 = val1[31:24] - val2[31:24]
1240 sum = absdiff1 + absdiff2 + absdiff3 + absdiff4
1241 res[31:0] = sum[31:0] + val3[31:0]
1244 uint32_t __USADA8(uint32_t val1, uint32_t val2, uint32_t val3);
1247 /**************************************************************************************************/
1249 \brief Q setting dual 16-bit saturate
1251 \details This function enables you to saturate two signed 16-bit values to a selected signed range.<br>
1252 The Q bit is set if either operation saturates.
1254 \param val1 two signed 16-bit values to be saturated.
1255 \param val2 bit position for saturation, an integral constant expression in the
1260 the sum of the absolute differences of the following
1261 bytes, added to the accumulation value:
1262 \li the signed saturation of the low halfword in \em val1, saturated to the bit position
1263 specified in \em val2 and returned in the low halfword of the return value.
1264 \li the signed saturation of the high halfword in <i>val1</i>, saturated to the bit position
1265 specified in <i>val2</i> and returned in the high halfword of the return value.
1270 Saturate halfwords in val1 to the signed range specified by the bit position in val2
1273 uint32_t __SSAT16(uint32_t val1, const uint32_t val2);
1276 /**************************************************************************************************/
1278 \brief Q setting dual 16-bit unsigned saturate
1280 \details This function enables you to saturate two signed 16-bit values to a selected unsigned
1282 The Q bit is set if either operation saturates.
1284 \param val1 two 16-bit values that are to be saturated.
1285 \param val2 bit position for saturation, and must be an integral constant
1286 expression in the range 0 to 15.
1290 the saturation of the two signed 16-bit values, as non-negative values.
1291 \li the saturation of the low halfword in \em val1, saturated to the bit position
1292 specified in \em val2 and returned in the low halfword of the return value.
1293 \li the saturation of the high halfword in \em val1, saturated to the bit position
1294 specified in \em val2 and returned in the high halfword of the return value.
1299 Saturate halfwords in val1 to the unsigned range specified by the bit position in val2
1302 uint32_t __USAT16(uint32_t val1, const uint32_t val2);
1305 /**************************************************************************************************/
1307 \brief Dual extract 8-bits and zero-extend to 16-bits
1309 \details This function enables you to extract two 8-bit values from an operand and zero-extend
1310 them to 16 bits each.
1312 \param val two 8-bit values in val[7:0] and val[23:16] to be sign-extended.
1316 the 8-bit values zero-extended to 16-bit values.
1317 \li zero-extended value of val[7:0] in the low halfword of the return value.
1318 \li zero-extended value of val[23:16] in the high halfword of the return value.
1323 res[15:0] = ZeroExtended(val[7:0] )
1324 res[31:16] = ZeroExtended(val[23:16])
1327 uint32_t __UXTB16(uint32_t val);
1330 /**************************************************************************************************/
1332 \brief Extracted 16-bit to 32-bit unsigned addition
1334 \details This function enables you to extract two 8-bit values from one operand, zero-extend them
1335 to 16 bits each, and add the results to two 16-bit values from another operand.
1337 \param val1 value added to the zero-extended to 16-bit values.
1338 \param val2 two 8-bit values to be extracted and zero-extended.
1342 the 8-bit values in \em val2, zero-extended to 16-bit values
1343 and added to \em val1.
1348 res[15:0] = ZeroExt(val2[7:0] to 16 bits) + val1[15:0]
1349 res[31:16] = ZeroExt(val2[31:16] to 16 bits) + val1[31:16]
1352 uint32_t __UXTAB16(uint32_t val1, uint32_t val2);
1355 /**************************************************************************************************/
1357 \brief Dual extract 8-bits and sign extend each to 16-bits
1359 \details This function enables you to extract two 8-bit values from an operand and sign-extend
1360 them to 16 bits each.
1362 \param val two 8-bit values in val[7:0] and val[23:16] to be sign-extended.
1367 the 8-bit values sign-extended to 16-bit values.
1368 \li sign-extended value of val[7:0] in the low halfword of the return value.
1369 \li sign-extended value of val[23:16] in the high halfword of the return value.
1374 res[15:0] = SignExtended(val[7:0]
1375 res[31:16] = SignExtended(val[23:16]
1378 uint32_t __SXTB16(uint32_t val);
1381 /**************************************************************************************************/
1383 \brief Rotate right, dual extract 8-bits and sign extend each to 16-bits
1385 \details This function enables you to rotate an operand by 8/16/24 bit, extract two 8-bit values and sign-extend
1386 them to 16 bits each.
1388 \param val two 8-bit values in val[7:0] and val[23:16] to be sign-extended.
1389 \param rotate number of bits to rotate val. Only 8,16 and 24 are accepted
1393 the 8-bit values sign-extended to 16-bit values.
1394 \li sign-extended value of val[7:0] in the low halfword of the return value.
1395 \li sign-extended value of val[23:16] in the high halfword of the return value.
1400 val = Rotate(val, rotate)
1401 res[15:0] = SignExtended(val[7:0])
1402 res[31:16] = SignExtended(val[23:16])
1405 uint32_t __SXTB16_RORn(uint32_t val, uint32_r rotate);
1408 /**************************************************************************************************/
1410 \brief Dual extracted 8-bit to 16-bit signed addition
1412 \details This function enables you to extract two 8-bit values from the second operand (at bit
1413 positions [7:0] and [23:16]), sign-extend them to 16-bits each, and add the results to the
1416 \param val1 values added to the zero-extended to 16-bit values.
1417 \param val2 two 8-bit values to be extracted and zero-extended.
1422 the addition of \em val1 and \em val2, where the 8-bit values in
1423 val2[7:0] and val2[23:16] have been extracted and sign-extended prior to the addition.
1428 res[15:0] = val1[15:0] + SignExtended(val2[7:0])
1429 res[31:16] = val1[31:16] + SignExtended(val2[23:16])
1432 uint32_t __SXTAB16(uint32_t val1, uint32_t val2);
1435 /**************************************************************************************************/
1437 \brief Q setting sum of dual 16-bit signed multiply
1439 \details This function enables you to perform two 16-bit signed multiplications, adding the
1440 products together.<br>
1441 The Q bit is set if the addition overflows.
1443 \param val1 first 16-bit operands for each multiplication.
1444 \param val2 second 16-bit operands for each multiplication.
1449 the sum of the products of the two 16-bit signed multiplications.
1454 p1 = val1[15:0] * val2[15:0]
1455 p2 = val1[31:16] * val2[31:16]
1459 uint32_t __SMUAD(uint32_t val1, uint32_t val2);
1462 /**************************************************************************************************/
1464 \brief Q setting sum of dual 16-bit signed multiply with exchange
1466 \details This function enables you to perform two 16-bit signed multiplications with exchanged
1467 halfwords of the second operand, adding the products together.<br>
1468 The Q bit is set if the addition overflows.
1470 \param val1 first 16-bit operands for each multiplication.
1471 \param val2 second 16-bit operands for each multiplication.
1476 the sum of the products of the two 16-bit signed multiplications with exchanged
1477 halfwords of the second operand.
1482 p1 = val1[15:0] * val2[31:16]
1483 p2 = val1[31:16] * val2[15:0]
1487 uint32_t __SMUADX(uint32_t val1, uint32_t val2);
1490 /**************************************************************************************************/
1492 \brief 32-bit signed multiply with 32-bit truncated accumulator.
1494 \details This function enables you to perform a signed 32-bit multiplications, adding the most significant 32 bits
1495 of the 64-bit result to a 32-bit accumulate operand.<br>
1497 \param val1 first operand for multiplication.
1498 \param val2 second operand for multiplication.
1499 \param val3 accumulate value.
1502 \returns the product of multiplication (most significant 32 bits) is added to the accumulate
1503 value, as a 32-bit integer.
1508 res[31:0] = p[61:32] + val3[31:0]
1511 uint32_t __SMMLA (int32_t val1, int32_t val2, int32_t val3);
1514 /**************************************************************************************************/
1516 \brief Q setting dual 16-bit signed multiply with single 32-bit accumulator
1518 \details This function enables you to perform two signed 16-bit multiplications, adding both
1519 results to a 32-bit accumulate operand.<br>
1520 The Q bit is set if the addition overflows. Overflow cannot occur during the multiplications.
1522 \param val1 first 16-bit operands for each multiplication.
1523 \param val2 second 16-bit operands for each multiplication.
1524 \param val3 accumulate value.
1528 the product of each multiplication added to the accumulate
1529 value, as a 32-bit integer.
1534 p1 = val1[15:0] * val2[15:0]
1535 p2 = val1[31:16] * val2[31:16]
1536 res[31:0] = p1 + p2 + val3[31:0]
1539 uint32_t __SMLAD(uint32_t val1, uint32_t val2, uint32_t val3);
1542 /**************************************************************************************************/
1544 \brief Q setting pre-exchanged dual 16-bit signed multiply with single 32-bit accumulator
1546 \details This function enables you to perform two signed 16-bit multiplications with exchanged
1547 halfwords of the second operand, adding both results to a 32-bit accumulate operand.<br>
1548 The Q bit is set if the addition overflows. Overflow cannot occur during the multiplications.
1550 \param val1 first 16-bit operands for each multiplication.
1551 \param val2 second 16-bit operands for each multiplication.
1552 \param val3 accumulate value.
1556 the product of each multiplication with exchanged
1557 halfwords of the second operand added to the accumulate value, as a 32-bit integer.
1562 p1 = val1[15:0] * val2[31:16]
1563 p2 = val1[31:16] * val2[15:0]
1564 res[31:0] = p1 + p2 + val3[31:0]
1567 uint32_t __SMLADX(uint32_t val1, uint32_t val2, uint32_t val3);
1570 /**************************************************************************************************/
1572 \brief Dual 16-bit signed multiply with single 64-bit accumulator
1574 \details This function enables you to perform two signed 16-bit multiplications, adding both
1575 results to a 64-bit accumulate operand. Overflow is only possible as a result of the 64-bit
1576 addition. This overflow is not detected if it occurs. Instead, the result wraps around
1577 modulo2<sup>64</sup>.
1579 \param val1 first 16-bit operands for each multiplication.
1580 \param val2 second 16-bit operands for each multiplication.
1581 \param val3 accumulate value.
1585 the product of each multiplication added to the accumulate value.
1590 p1 = val1[15:0] * val2[15:0]
1591 p2 = val1[31:16] * val2[31:16]
1592 sum = p1 + p2 + val3[63:32][31:0]
1593 res[63:32] = sum[63:32]
1594 res[31:0] = sum[31:0]
1597 uint64_t __SMLALD(uint32_t val1, uint32_t val2, uint64_t val3);
1600 /**************************************************************************************************/
1602 \brief Dual 16-bit signed multiply with exchange with single 64-bit accumulator
1604 \details This function enables you to exchange the halfwords of the second operand, and perform
1605 two signed 16-bit multiplications, adding both results to a 64-bit accumulate operand.
1606 Overflow is only possible as a result of the 64-bit addition. This overflow is not detected
1607 if it occurs. Instead, the result wraps around modulo2<sup>64</sup>.
1609 \param val1 first 16-bit operands for each multiplication.
1610 \param val2 second 16-bit operands for each multiplication.
1611 \param val3 accumulate value.
1615 the product of each multiplication added to the accumulate value.
1620 p1 = val1[15:0] * val2[31:16]
1621 p2 = val1[31:16] * val2[15:0]
1622 sum = p1 + p2 + val3[63:32][31:0]
1623 res[63:32] = sum[63:32]
1624 res[31:0] = sum[31:0]
1627 unsigned long long __SMLALDX(uint32_t val1, uint32_t val2, unsigned long long val3);
1630 /**************************************************************************************************/
1632 \brief Dual 16-bit signed multiply returning difference
1634 \details This function enables you to perform two 16-bit signed multiplications, taking the
1635 difference of the products by subtracting the high halfword product from the low
1638 \param val1 first 16-bit operands for each multiplication.
1639 \param val2 second 16-bit operands for each multiplication.
1643 the difference of the products of the two 16-bit signed multiplications.
1648 p1 = val1[15:0] * val2[15:0]
1649 p2 = val1[31:16] * val2[31:16]
1653 uint32_t __SMUSD(uint32_t val1, uint32_t val2);
1656 /**************************************************************************************************/
1658 \brief Dual 16-bit signed multiply with exchange returning difference
1660 \details This function enables you to perform two 16-bit signed multiplications, subtracting one
1661 of the products from the other. The halfwords of the second operand are exchanged
1662 before performing the arithmetic. This produces top * bottom and bottom * top
1665 \param val1 first 16-bit operands for each multiplication.
1666 \param val2 second 16-bit operands for each multiplication.
1670 the difference of the products of the two 16-bit signed multiplications.
1675 p1 = val1[15:0] * val2[31:16]
1676 p2 = val1[31:16] * val2[15:0]
1680 uint32_t __SMUSDX(uint32_t val1, uint32_t val2);
1683 /**************************************************************************************************/
1685 \brief Q setting dual 16-bit signed multiply subtract with 32-bit accumulate
1687 \details This function enables you to perform two 16-bit signed multiplications, take the
1688 difference of the products, subtracting the high halfword product from the low halfword
1689 product, and add the difference to a 32-bit accumulate operand.<br>
1690 The Q bit is set if the accumulation overflows. Overflow cannot occur during the multiplications or the
1693 \param val1 first 16-bit operands for each multiplication.
1694 \param val2 second 16-bit operands for each multiplication.
1695 \param val3 accumulate value.
1699 the difference of the product of each multiplication, added
1700 to the accumulate value.
1705 p1 = val1[15:0] * val2[15:0]
1706 p2 = val1[31:16] * val2[31:16]
1707 res[31:0] = p1 - p2 + val3[31:0]
1710 uint32_t __SMLSD(uint32_t val1, uint32_t val2, uint32_t val3);
1713 /**************************************************************************************************/
1715 \brief Q setting dual 16-bit signed multiply with exchange subtract with 32-bit accumulate
1717 \details This function enables you to exchange the halfwords in the second operand, then perform
1718 two 16-bit signed multiplications. The difference of the products is added to a 32-bit
1719 accumulate operand.<br>
1720 The Q bit is set if the addition overflows. Overflow cannot occur during the multiplications or the subtraction.
1722 \param val1 first 16-bit operands for each multiplication.
1723 \param val2 second 16-bit operands for each multiplication.
1724 \param val3 accumulate value.
1728 the difference of the product of each multiplication, added
1729 to the accumulate value.
1734 p1 = val1[15:0] * val2[31:16]
1735 p2 = val1[31:16] * val2[15:0]
1736 res[31:0] = p1 - p2 + val3[31:0]
1739 uint32_t __SMLSDX(uint32_t val1, uint32_t val2, uint32_t val3);
1742 /**************************************************************************************************/
1744 \brief Q setting dual 16-bit signed multiply subtract with 64-bit accumulate
1746 \details This function It enables you to perform two 16-bit signed multiplications, take the
1747 difference of the products, subtracting the high halfword product from the low halfword
1748 product, and add the difference to a 64-bit accumulate operand. Overflow cannot occur
1749 during the multiplications or the subtraction. Overflow can occur as a result of the 64-bit
1750 addition, and this overflow is not detected. Instead, the result wraps round to
1751 modulo2<sup>64</sup>.
1753 \param val1 first 16-bit operands for each multiplication.
1754 \param val2 second 16-bit operands for each multiplication.
1755 \param val3 accumulate value.
1759 the difference of the product of each multiplication,
1760 added to the accumulate value.
1765 p1 = val1[15:0] * val2[15:0]
1766 p2 = val1[31:16] * val2[31:16]
1767 res[63:0] = p1 - p2 + val3[63:0]
1770 uint64_t __SMLSLD(uint32_t val1, uint32_t val2, uint64_t val3);
1773 /**************************************************************************************************/
1775 \brief Q setting dual 16-bit signed multiply with exchange subtract with 64-bit accumulate
1777 \details This function enables you to exchange the halfwords of the second operand, perform two
1778 16-bit multiplications, adding the difference of the products to a 64-bit accumulate
1779 operand. Overflow cannot occur during the multiplications or the subtraction. Overflow
1780 can occur as a result of the 64-bit addition, and this overflow is not detected. Instead,
1781 the result wraps round to modulo2<sup>64</sup>.
1783 \param val1 first 16-bit operands for each multiplication.
1784 \param val2 second 16-bit operands for each multiplication.
1785 \param val3 accumulate value.
1789 the difference of the product of each multiplication,
1790 added to the accumulate value.
1795 p1 = val1[15:0] * val2[31:16]
1796 p2 = val1[31:16] * val2[15:0]
1797 res[63:0] = p1 - p2 + val3[63:0]
1800 unsigned long long __SMLSLDX(uint32_t val1, uint32_t val2, unsigned long long val3);
1803 /**************************************************************************************************/
1805 \brief Select bytes based on GE bits
1807 \details This function inserts a SEL instruction into the instruction stream generated by the
1808 compiler. It enables you to select bytes from the input parameters, whereby the bytes
1809 that are selected depend upon the results of previous SIMD instruction function. The
1810 results of previous SIMD instruction function are represented by the Greater than or
1811 Equal flags in the Application Program Status Register (APSR).
1812 The __SEL function works equally well on both halfword and byte operand function
1813 results. This is because halfword operand operations set two (duplicate) GE bits per
1816 \param val1 four selectable 8-bit values.
1817 \param val2 four selectable 8-bit values.
1821 The function selects bytes from the input parameters and returns them in the
1822 return value, res, according to the following criteria:
1823 \li if APSR.GE[0] == 1 then res[7:0] = val1[7:0] else res[7:0] = val2[7:0]
1824 \li if APSR.GE[1] == 1 then res[15:8] = val1[15:8] else res[15:8] = val2[15:8]
1825 \li if APSR.GE[2] == 1 then res[23:16] = val1[23:16] else res[23:16] = val2[23:16]
1826 \li if APSR.GE[3] == 1 then res[31;24] = val1[31:24] else res = val2[31:24]
1829 uint32_t __SEL(uint32_t val1, uint32_t val2);
1832 /**************************************************************************************************/
1834 \brief Q setting saturating add
1836 \details This function enables you to obtain the saturating add of two integers.<br>
1837 The Q bit is set if the operation saturates.
1839 \param val1 first summand of the saturating add operation.
1840 \param val2 second summand of the saturating add operation.
1844 the saturating addition of val1 and val2.
1848 res[31:0] = SAT(val1 + SAT(val2))
1851 uint32_t __QADD(uint32_t val1, uint32_t val2);
1854 /**************************************************************************************************/
1856 \brief Q setting saturating subtract
1858 \details This function enables you to obtain the saturating subtraction of two integers.<br>
1859 The Q bit is set if the operation saturates.
1861 \param val1 minuend of the saturating subtraction operation.
1862 \param val2 subtrahend of the saturating subtraction operation.
1866 the saturating subtraction of val1 and val2.
1870 res[31:0] = SAT(val1 - SAT(val2))
1873 uint32_t __QSUB(uint32_t val1, uint32_t val2);
1876 /**************************************************************************************************/
1878 \brief Halfword packing instruction. Combines bits[15:0] of <i>val1</i>
1879 with bits[31:16] of <i>val2</i> levitated with the <i>val3</i>.
1881 \details Combine a halfword from one register with a halfword from another register.
1882 The second argument can be left-shifted before extraction of the halfword. The registers
1883 PC and SP are not allowed as arguments. This instruction does not change the flags.
1885 \param val1 first 16-bit operands
1886 \param val2 second 16-bit operands
1887 \param val3 value for left-shifting <i>val2</i>. Value range [0..31].
1891 the combination of halfwords.
1895 res[15:0] = val1[15:0]
1896 res[31:16] = val2[31:16]<<val3
1899 uint32_t __PKHBT(uint32_t val1, uint32_t val2, uint32_t val3);
1902 /**************************************************************************************************/
1904 \brief Halfword packing instruction. Combines bits[31:16] of <i>val1</i>
1905 with bits[15:0] of <i>val2</i> right-shifted with the <i>val3</i>.
1907 \details Combines a halfword from one register with a halfword from another register.
1908 The second argument can be right-shifted before extraction of the halfword. The registers
1909 PC and SP are not allowed as arguments. This instruction does not change the flags.
1911 \param val1 second 16-bit operands
1912 \param val2 first 16-bit operands
1913 \param val3 value for right-shifting <i>val2</i>. Value range [1..32].
1917 the combination of halfwords.
1921 res[15:0] = val2[15:0]>>val3
1922 res[31:16] = val1[31:16]
1925 uint32_t __PKHTB(uint32_t val1, uint32_t val2, uint32_t val3);
1927 /** @} */ /* end group intrinsic_SIMD_gr */