0
|
1 /* PowerPC AltiVec include file.
|
|
2 Copyright (C) 2002, 2003, 2004, 2005, 2008, 2009 Free Software Foundation, Inc.
|
|
3 Contributed by Aldy Hernandez (aldyh@redhat.com).
|
|
4 Rewritten by Paolo Bonzini (bonzini@gnu.org).
|
|
5
|
|
6 This file is part of GCC.
|
|
7
|
|
8 GCC is free software; you can redistribute it and/or modify it
|
|
9 under the terms of the GNU General Public License as published
|
|
10 by the Free Software Foundation; either version 3, or (at your
|
|
11 option) any later version.
|
|
12
|
|
13 GCC is distributed in the hope that it will be useful, but WITHOUT
|
|
14 ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
|
|
15 or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
|
|
16 License for more details.
|
|
17
|
|
18 Under Section 7 of GPL version 3, you are granted additional
|
|
19 permissions described in the GCC Runtime Library Exception, version
|
|
20 3.1, as published by the Free Software Foundation.
|
|
21
|
|
22 You should have received a copy of the GNU General Public License and
|
|
23 a copy of the GCC Runtime Library Exception along with this program;
|
|
24 see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
|
25 <http://www.gnu.org/licenses/>. */
|
|
26
|
|
27 /* Implemented to conform to the specification included in the AltiVec
|
|
28 Technology Programming Interface Manual (ALTIVECPIM/D 6/1999 Rev 0). */
|
|
29
|
|
30 #ifndef _ALTIVEC_H
|
|
31 #define _ALTIVEC_H 1
|
|
32
|
|
33 #if !defined(__VEC__) || !defined(__ALTIVEC__)
|
|
34 #error Use the "-maltivec" flag to enable PowerPC AltiVec support
|
|
35 #endif
|
|
36
|
|
37 /* If __APPLE_ALTIVEC__ is defined, the compiler supports 'vector',
|
|
38 'pixel' and 'bool' as context-sensitive AltiVec keywords (in
|
|
39 non-AltiVec contexts, they revert to their original meanings,
|
|
40 if any), so we do not need to define them as macros. */
|
|
41
|
|
42 #if !defined(__APPLE_ALTIVEC__)
|
|
43 /* You are allowed to undef these for C++ compatibility. */
|
|
44 #define vector __vector
|
|
45 #define pixel __pixel
|
|
46 #define bool __bool
|
|
47 #endif
|
|
48
|
|
49 /* Condition register codes for AltiVec predicates. */
|
|
50
|
|
51 #define __CR6_EQ 0
|
|
52 #define __CR6_EQ_REV 1
|
|
53 #define __CR6_LT 2
|
|
54 #define __CR6_LT_REV 3
|
|
55
|
|
56 /* Synonyms. */
|
|
57 #define vec_vaddcuw vec_addc
|
|
58 #define vec_vand vec_and
|
|
59 #define vec_vandc vec_andc
|
|
60 #define vec_vrfip vec_ceil
|
|
61 #define vec_vcmpbfp vec_cmpb
|
|
62 #define vec_vcmpgefp vec_cmpge
|
|
63 #define vec_vctsxs vec_cts
|
|
64 #define vec_vctuxs vec_ctu
|
|
65 #define vec_vexptefp vec_expte
|
|
66 #define vec_vrfim vec_floor
|
|
67 #define vec_lvx vec_ld
|
|
68 #define vec_lvxl vec_ldl
|
|
69 #define vec_vlogefp vec_loge
|
|
70 #define vec_vmaddfp vec_madd
|
|
71 #define vec_vmhaddshs vec_madds
|
|
72 #define vec_vmladduhm vec_mladd
|
|
73 #define vec_vmhraddshs vec_mradds
|
|
74 #define vec_vnmsubfp vec_nmsub
|
|
75 #define vec_vnor vec_nor
|
|
76 #define vec_vor vec_or
|
|
77 #define vec_vpkpx vec_packpx
|
|
78 #define vec_vperm vec_perm
|
|
79 #define vec_vrefp vec_re
|
|
80 #define vec_vrfin vec_round
|
|
81 #define vec_vrsqrtefp vec_rsqrte
|
|
82 #define vec_vsel vec_sel
|
|
83 #define vec_vsldoi vec_sld
|
|
84 #define vec_vsl vec_sll
|
|
85 #define vec_vslo vec_slo
|
|
86 #define vec_vspltisb vec_splat_s8
|
|
87 #define vec_vspltish vec_splat_s16
|
|
88 #define vec_vspltisw vec_splat_s32
|
|
89 #define vec_vsr vec_srl
|
|
90 #define vec_vsro vec_sro
|
|
91 #define vec_stvx vec_st
|
|
92 #define vec_stvxl vec_stl
|
|
93 #define vec_vsubcuw vec_subc
|
|
94 #define vec_vsum2sws vec_sum2s
|
|
95 #define vec_vsumsws vec_sums
|
|
96 #define vec_vrfiz vec_trunc
|
|
97 #define vec_vxor vec_xor
|
|
98
|
|
99 /* Functions that are resolved by the backend to one of the
|
|
100 typed builtins. */
|
|
101 #define vec_vaddfp __builtin_vec_vaddfp
|
|
102 #define vec_addc __builtin_vec_addc
|
|
103 #define vec_vaddsws __builtin_vec_vaddsws
|
|
104 #define vec_vaddshs __builtin_vec_vaddshs
|
|
105 #define vec_vaddsbs __builtin_vec_vaddsbs
|
|
106 #define vec_vavgsw __builtin_vec_vavgsw
|
|
107 #define vec_vavguw __builtin_vec_vavguw
|
|
108 #define vec_vavgsh __builtin_vec_vavgsh
|
|
109 #define vec_vavguh __builtin_vec_vavguh
|
|
110 #define vec_vavgsb __builtin_vec_vavgsb
|
|
111 #define vec_vavgub __builtin_vec_vavgub
|
|
112 #define vec_ceil __builtin_vec_ceil
|
|
113 #define vec_cmpb __builtin_vec_cmpb
|
|
114 #define vec_vcmpeqfp __builtin_vec_vcmpeqfp
|
|
115 #define vec_cmpge __builtin_vec_cmpge
|
|
116 #define vec_vcmpgtfp __builtin_vec_vcmpgtfp
|
|
117 #define vec_vcmpgtsw __builtin_vec_vcmpgtsw
|
|
118 #define vec_vcmpgtuw __builtin_vec_vcmpgtuw
|
|
119 #define vec_vcmpgtsh __builtin_vec_vcmpgtsh
|
|
120 #define vec_vcmpgtuh __builtin_vec_vcmpgtuh
|
|
121 #define vec_vcmpgtsb __builtin_vec_vcmpgtsb
|
|
122 #define vec_vcmpgtub __builtin_vec_vcmpgtub
|
|
123 #define vec_vcfsx __builtin_vec_vcfsx
|
|
124 #define vec_vcfux __builtin_vec_vcfux
|
|
125 #define vec_cts __builtin_vec_cts
|
|
126 #define vec_ctu __builtin_vec_ctu
|
|
127 #define vec_expte __builtin_vec_expte
|
|
128 #define vec_floor __builtin_vec_floor
|
|
129 #define vec_loge __builtin_vec_loge
|
|
130 #define vec_madd __builtin_vec_madd
|
|
131 #define vec_madds __builtin_vec_madds
|
|
132 #define vec_mtvscr __builtin_vec_mtvscr
|
|
133 #define vec_vmaxfp __builtin_vec_vmaxfp
|
|
134 #define vec_vmaxsw __builtin_vec_vmaxsw
|
|
135 #define vec_vmaxsh __builtin_vec_vmaxsh
|
|
136 #define vec_vmaxsb __builtin_vec_vmaxsb
|
|
137 #define vec_vminfp __builtin_vec_vminfp
|
|
138 #define vec_vminsw __builtin_vec_vminsw
|
|
139 #define vec_vminsh __builtin_vec_vminsh
|
|
140 #define vec_vminsb __builtin_vec_vminsb
|
|
141 #define vec_mradds __builtin_vec_mradds
|
|
142 #define vec_vmsumshm __builtin_vec_vmsumshm
|
|
143 #define vec_vmsumuhm __builtin_vec_vmsumuhm
|
|
144 #define vec_vmsummbm __builtin_vec_vmsummbm
|
|
145 #define vec_vmsumubm __builtin_vec_vmsumubm
|
|
146 #define vec_vmsumshs __builtin_vec_vmsumshs
|
|
147 #define vec_vmsumuhs __builtin_vec_vmsumuhs
|
|
148 #define vec_vmulesb __builtin_vec_vmulesb
|
|
149 #define vec_vmulesh __builtin_vec_vmulesh
|
|
150 #define vec_vmuleuh __builtin_vec_vmuleuh
|
|
151 #define vec_vmuleub __builtin_vec_vmuleub
|
|
152 #define vec_vmulosh __builtin_vec_vmulosh
|
|
153 #define vec_vmulouh __builtin_vec_vmulouh
|
|
154 #define vec_vmulosb __builtin_vec_vmulosb
|
|
155 #define vec_vmuloub __builtin_vec_vmuloub
|
|
156 #define vec_nmsub __builtin_vec_nmsub
|
|
157 #define vec_packpx __builtin_vec_packpx
|
|
158 #define vec_vpkswss __builtin_vec_vpkswss
|
|
159 #define vec_vpkuwus __builtin_vec_vpkuwus
|
|
160 #define vec_vpkshss __builtin_vec_vpkshss
|
|
161 #define vec_vpkuhus __builtin_vec_vpkuhus
|
|
162 #define vec_vpkswus __builtin_vec_vpkswus
|
|
163 #define vec_vpkshus __builtin_vec_vpkshus
|
|
164 #define vec_re __builtin_vec_re
|
|
165 #define vec_round __builtin_vec_round
|
|
166 #define vec_rsqrte __builtin_vec_rsqrte
|
|
167 #define vec_vsubfp __builtin_vec_vsubfp
|
|
168 #define vec_subc __builtin_vec_subc
|
|
169 #define vec_vsubsws __builtin_vec_vsubsws
|
|
170 #define vec_vsubshs __builtin_vec_vsubshs
|
|
171 #define vec_vsubsbs __builtin_vec_vsubsbs
|
|
172 #define vec_sum4s __builtin_vec_sum4s
|
|
173 #define vec_vsum4shs __builtin_vec_vsum4shs
|
|
174 #define vec_vsum4sbs __builtin_vec_vsum4sbs
|
|
175 #define vec_vsum4ubs __builtin_vec_vsum4ubs
|
|
176 #define vec_sum2s __builtin_vec_sum2s
|
|
177 #define vec_sums __builtin_vec_sums
|
|
178 #define vec_trunc __builtin_vec_trunc
|
|
179 #define vec_vupkhpx __builtin_vec_vupkhpx
|
|
180 #define vec_vupkhsh __builtin_vec_vupkhsh
|
|
181 #define vec_vupkhsb __builtin_vec_vupkhsb
|
|
182 #define vec_vupklpx __builtin_vec_vupklpx
|
|
183 #define vec_vupklsh __builtin_vec_vupklsh
|
|
184 #define vec_vupklsb __builtin_vec_vupklsb
|
|
185 #define vec_abs __builtin_vec_abs
|
|
186 #define vec_abss __builtin_vec_abss
|
|
187 #define vec_add __builtin_vec_add
|
|
188 #define vec_adds __builtin_vec_adds
|
|
189 #define vec_and __builtin_vec_and
|
|
190 #define vec_andc __builtin_vec_andc
|
|
191 #define vec_avg __builtin_vec_avg
|
|
192 #define vec_cmpeq __builtin_vec_cmpeq
|
|
193 #define vec_cmpgt __builtin_vec_cmpgt
|
|
194 #define vec_ctf __builtin_vec_ctf
|
|
195 #define vec_dst __builtin_vec_dst
|
|
196 #define vec_dstst __builtin_vec_dstst
|
|
197 #define vec_dststt __builtin_vec_dststt
|
|
198 #define vec_dstt __builtin_vec_dstt
|
|
199 #define vec_ld __builtin_vec_ld
|
|
200 #define vec_lde __builtin_vec_lde
|
|
201 #define vec_ldl __builtin_vec_ldl
|
|
202 #define vec_lvebx __builtin_vec_lvebx
|
|
203 #define vec_lvehx __builtin_vec_lvehx
|
|
204 #define vec_lvewx __builtin_vec_lvewx
|
|
205 /* Cell only intrinsics. */
|
|
206 #ifdef __PPU__
|
|
207 #define vec_lvlx __builtin_vec_lvlx
|
|
208 #define vec_lvlxl __builtin_vec_lvlxl
|
|
209 #define vec_lvrx __builtin_vec_lvrx
|
|
210 #define vec_lvrxl __builtin_vec_lvrxl
|
|
211 #endif
|
|
212 #define vec_lvsl __builtin_vec_lvsl
|
|
213 #define vec_lvsr __builtin_vec_lvsr
|
|
214 #define vec_max __builtin_vec_max
|
|
215 #define vec_mergeh __builtin_vec_mergeh
|
|
216 #define vec_mergel __builtin_vec_mergel
|
|
217 #define vec_min __builtin_vec_min
|
|
218 #define vec_mladd __builtin_vec_mladd
|
|
219 #define vec_msum __builtin_vec_msum
|
|
220 #define vec_msums __builtin_vec_msums
|
|
221 #define vec_mule __builtin_vec_mule
|
|
222 #define vec_mulo __builtin_vec_mulo
|
|
223 #define vec_nor __builtin_vec_nor
|
|
224 #define vec_or __builtin_vec_or
|
|
225 #define vec_pack __builtin_vec_pack
|
|
226 #define vec_packs __builtin_vec_packs
|
|
227 #define vec_packsu __builtin_vec_packsu
|
|
228 #define vec_perm __builtin_vec_perm
|
|
229 #define vec_rl __builtin_vec_rl
|
|
230 #define vec_sel __builtin_vec_sel
|
|
231 #define vec_sl __builtin_vec_sl
|
|
232 #define vec_sld __builtin_vec_sld
|
|
233 #define vec_sll __builtin_vec_sll
|
|
234 #define vec_slo __builtin_vec_slo
|
|
235 #define vec_splat __builtin_vec_splat
|
|
236 #define vec_sr __builtin_vec_sr
|
|
237 #define vec_sra __builtin_vec_sra
|
|
238 #define vec_srl __builtin_vec_srl
|
|
239 #define vec_sro __builtin_vec_sro
|
|
240 #define vec_st __builtin_vec_st
|
|
241 #define vec_ste __builtin_vec_ste
|
|
242 #define vec_stl __builtin_vec_stl
|
|
243 #define vec_stvebx __builtin_vec_stvebx
|
|
244 #define vec_stvehx __builtin_vec_stvehx
|
|
245 #define vec_stvewx __builtin_vec_stvewx
|
|
246 /* Cell only intrinsics. */
|
|
247 #ifdef __PPU__
|
|
248 #define vec_stvlx __builtin_vec_stvlx
|
|
249 #define vec_stvlxl __builtin_vec_stvlxl
|
|
250 #define vec_stvrx __builtin_vec_stvrx
|
|
251 #define vec_stvrxl __builtin_vec_stvrxl
|
|
252 #endif
|
|
253 #define vec_sub __builtin_vec_sub
|
|
254 #define vec_subs __builtin_vec_subs
|
|
255 #define vec_sum __builtin_vec_sum
|
|
256 #define vec_unpackh __builtin_vec_unpackh
|
|
257 #define vec_unpackl __builtin_vec_unpackl
|
|
258 #define vec_vaddubm __builtin_vec_vaddubm
|
|
259 #define vec_vaddubs __builtin_vec_vaddubs
|
|
260 #define vec_vadduhm __builtin_vec_vadduhm
|
|
261 #define vec_vadduhs __builtin_vec_vadduhs
|
|
262 #define vec_vadduwm __builtin_vec_vadduwm
|
|
263 #define vec_vadduws __builtin_vec_vadduws
|
|
264 #define vec_vcmpequb __builtin_vec_vcmpequb
|
|
265 #define vec_vcmpequh __builtin_vec_vcmpequh
|
|
266 #define vec_vcmpequw __builtin_vec_vcmpequw
|
|
267 #define vec_vmaxub __builtin_vec_vmaxub
|
|
268 #define vec_vmaxuh __builtin_vec_vmaxuh
|
|
269 #define vec_vmaxuw __builtin_vec_vmaxuw
|
|
270 #define vec_vminub __builtin_vec_vminub
|
|
271 #define vec_vminuh __builtin_vec_vminuh
|
|
272 #define vec_vminuw __builtin_vec_vminuw
|
|
273 #define vec_vmrghb __builtin_vec_vmrghb
|
|
274 #define vec_vmrghh __builtin_vec_vmrghh
|
|
275 #define vec_vmrghw __builtin_vec_vmrghw
|
|
276 #define vec_vmrglb __builtin_vec_vmrglb
|
|
277 #define vec_vmrglh __builtin_vec_vmrglh
|
|
278 #define vec_vmrglw __builtin_vec_vmrglw
|
|
279 #define vec_vpkuhum __builtin_vec_vpkuhum
|
|
280 #define vec_vpkuwum __builtin_vec_vpkuwum
|
|
281 #define vec_vrlb __builtin_vec_vrlb
|
|
282 #define vec_vrlh __builtin_vec_vrlh
|
|
283 #define vec_vrlw __builtin_vec_vrlw
|
|
284 #define vec_vslb __builtin_vec_vslb
|
|
285 #define vec_vslh __builtin_vec_vslh
|
|
286 #define vec_vslw __builtin_vec_vslw
|
|
287 #define vec_vspltb __builtin_vec_vspltb
|
|
288 #define vec_vsplth __builtin_vec_vsplth
|
|
289 #define vec_vspltw __builtin_vec_vspltw
|
|
290 #define vec_vsrab __builtin_vec_vsrab
|
|
291 #define vec_vsrah __builtin_vec_vsrah
|
|
292 #define vec_vsraw __builtin_vec_vsraw
|
|
293 #define vec_vsrb __builtin_vec_vsrb
|
|
294 #define vec_vsrh __builtin_vec_vsrh
|
|
295 #define vec_vsrw __builtin_vec_vsrw
|
|
296 #define vec_vsububs __builtin_vec_vsububs
|
|
297 #define vec_vsububm __builtin_vec_vsububm
|
|
298 #define vec_vsubuhm __builtin_vec_vsubuhm
|
|
299 #define vec_vsubuhs __builtin_vec_vsubuhs
|
|
300 #define vec_vsubuwm __builtin_vec_vsubuwm
|
|
301 #define vec_vsubuws __builtin_vec_vsubuws
|
|
302 #define vec_xor __builtin_vec_xor
|
|
303
|
|
304 #define vec_extract __builtin_vec_extract
|
|
305 #define vec_insert __builtin_vec_insert
|
|
306 #define vec_splats __builtin_vec_splats
|
|
307 #define vec_promote __builtin_vec_promote
|
|
308
|
|
309 /* Predicates.
|
|
310 For C++, we use templates in order to allow non-parenthesized arguments.
|
|
311 For C, instead, we use macros since non-parenthesized arguments were
|
|
312 not allowed even in older GCC implementation of AltiVec.
|
|
313
|
|
314 In the future, we may add more magic to the back-end, so that no
|
|
315 one- or two-argument macros are used. */
|
|
316
|
|
317 #ifdef __cplusplus__
|
|
318 #define __altivec_unary_pred(NAME, CALL) \
|
|
319 template <class T> int NAME (T a1) { return CALL; }
|
|
320
|
|
321 #define __altivec_scalar_pred(NAME, CALL) \
|
|
322 template <class T, class U> int NAME (T a1, U a2) { return CALL; }
|
|
323
|
|
324 /* Given the vec_step of a type, return the corresponding bool type. */
|
|
325 template <int STEP> class __altivec_bool_ret { };
|
|
326 template <> class __altivec_bool_ret <4> {
|
|
327 typedef __vector __bool int __ret;
|
|
328 };
|
|
329 template <> class __altivec_bool_ret <8> {
|
|
330 typedef __vector __bool short __ret;
|
|
331 };
|
|
332 template <> class __altivec_bool_ret <16> {
|
|
333 typedef __vector __bool char __ret;
|
|
334 };
|
|
335
|
|
336 /* Be very liberal in the pairs we accept. Mistakes such as passing
|
|
337 a `vector char' and `vector short' will be caught by the middle-end,
|
|
338 while any attempt to detect them here would produce hard to understand
|
|
339 error messages involving the implementation details of AltiVec. */
|
|
340 #define __altivec_binary_pred(NAME, CALL) \
|
|
341 template <class T, class U> \
|
|
342 typename __altivec_bool_ret <vec_step (T)>::__ret \
|
|
343 NAME (T a1, U a2) \
|
|
344 { \
|
|
345 return CALL; \
|
|
346 }
|
|
347
|
|
348 __altivec_binary_pred(vec_cmplt,
|
|
349 __builtin_vec_cmpgt (a2, a1))
|
|
350 __altivec_binary_pred(vec_cmple,
|
|
351 __builtin_vec_cmpge (a2, a1))
|
|
352
|
|
353 __altivec_scalar_pred(vec_all_in,
|
|
354 __builtin_altivec_vcmpbfp_p (__CR6_EQ, a1, a2))
|
|
355 __altivec_scalar_pred(vec_any_out,
|
|
356 __builtin_altivec_vcmpbfp_p (__CR6_EQ_REV, a1, a2))
|
|
357
|
|
358 __altivec_unary_pred(vec_all_nan,
|
|
359 __builtin_altivec_vcmpeqfp_p (__CR6_EQ, a1, a1))
|
|
360 __altivec_unary_pred(vec_any_nan,
|
|
361 __builtin_altivec_vcmpeqfp_p (__CR6_LT_REV, a1, a1))
|
|
362
|
|
363 __altivec_unary_pred(vec_all_numeric,
|
|
364 __builtin_altivec_vcmpeqfp_p (__CR6_LT, a1, a1))
|
|
365 __altivec_unary_pred(vec_any_numeric,
|
|
366 __builtin_altivec_vcmpeqfp_p (__CR6_EQ_REV, a1, a1))
|
|
367
|
|
368 __altivec_scalar_pred(vec_all_eq,
|
|
369 __builtin_vec_vcmpeq_p (__CR6_LT, a1, a2))
|
|
370 __altivec_scalar_pred(vec_all_ne,
|
|
371 __builtin_vec_vcmpeq_p (__CR6_EQ, a1, a2))
|
|
372 __altivec_scalar_pred(vec_any_eq,
|
|
373 __builtin_vec_vcmpeq_p (__CR6_EQ_REV, a1, a2))
|
|
374 __altivec_scalar_pred(vec_any_ne,
|
|
375 __builtin_vec_vcmpeq_p (__CR6_LT_REV, a1, a2))
|
|
376
|
|
377 __altivec_scalar_pred(vec_all_gt,
|
|
378 __builtin_vec_vcmpgt_p (__CR6_LT, a1, a2))
|
|
379 __altivec_scalar_pred(vec_all_lt,
|
|
380 __builtin_vec_vcmpgt_p (__CR6_LT, a2, a1))
|
|
381 __altivec_scalar_pred(vec_any_gt,
|
|
382 __builtin_vec_vcmpgt_p (__CR6_EQ_REV, a1, a2))
|
|
383 __altivec_scalar_pred(vec_any_lt,
|
|
384 __builtin_vec_vcmpgt_p (__CR6_EQ_REV, a2, a1))
|
|
385
|
|
386 __altivec_scalar_pred(vec_all_ngt,
|
|
387 __builtin_altivec_vcmpgtfp_p (__CR6_EQ, a1, a2))
|
|
388 __altivec_scalar_pred(vec_all_nlt,
|
|
389 __builtin_altivec_vcmpgtfp_p (__CR6_EQ, a2, a1))
|
|
390 __altivec_scalar_pred(vec_any_ngt,
|
|
391 __builtin_altivec_vcmpgtfp_p (__CR6_LT_REV, a1, a2))
|
|
392 __altivec_scalar_pred(vec_any_nlt,
|
|
393 __builtin_altivec_vcmpgtfp_p (__CR6_LT_REV, a2, a1))
|
|
394
|
|
395 /* __builtin_vec_vcmpge_p is vcmpgefp for floating-point vector types,
|
|
396 while for integer types it is converted to __builtin_vec_vcmpgt_p,
|
|
397 with inverted args and condition code. */
|
|
398 __altivec_scalar_pred(vec_all_le,
|
|
399 __builtin_vec_vcmpge_p (__CR6_LT, a2, a1))
|
|
400 __altivec_scalar_pred(vec_all_ge,
|
|
401 __builtin_vec_vcmpge_p (__CR6_LT, a1, a2))
|
|
402 __altivec_scalar_pred(vec_any_le,
|
|
403 __builtin_vec_vcmpge_p (__CR6_EQ_REV, a2, a1))
|
|
404 __altivec_scalar_pred(vec_any_ge,
|
|
405 __builtin_vec_vcmpge_p (__CR6_EQ_REV, a1, a2))
|
|
406
|
|
407 __altivec_scalar_pred(vec_all_nge,
|
|
408 __builtin_altivec_vcmpgefp_p (__CR6_EQ, a1, a2))
|
|
409 __altivec_scalar_pred(vec_all_nle,
|
|
410 __builtin_altivec_vcmpgefp_p (__CR6_EQ, a2, a1))
|
|
411 __altivec_scalar_pred(vec_any_nge,
|
|
412 __builtin_altivec_vcmpgefp_p (__CR6_LT_REV, a1, a2))
|
|
413 __altivec_scalar_pred(vec_any_nle,
|
|
414 __builtin_altivec_vcmpgefp_p (__CR6_LT_REV, a2, a1))
|
|
415
|
|
416 #undef __altivec_scalar_pred
|
|
417 #undef __altivec_unary_pred
|
|
418 #undef __altivec_binary_pred
|
|
419 #else
|
|
420 #define vec_cmplt(a1, a2) __builtin_vec_cmpgt ((a2), (a1))
|
|
421 #define vec_cmple(a1, a2) __builtin_vec_cmpge ((a2), (a1))
|
|
422
|
|
423 #define vec_all_in(a1, a2) __builtin_altivec_vcmpbfp_p (__CR6_EQ, (a1), (a2))
|
|
424 #define vec_any_out(a1, a2) __builtin_altivec_vcmpbfp_p (__CR6_EQ_REV, (a1), (a2))
|
|
425
|
|
426 #define vec_all_nan(a1) __builtin_altivec_vcmpeqfp_p (__CR6_EQ, (a1), (a1))
|
|
427 #define vec_any_nan(a1) __builtin_altivec_vcmpeqfp_p (__CR6_LT_REV, (a1), (a1))
|
|
428
|
|
429 #define vec_all_numeric(a1) __builtin_altivec_vcmpeqfp_p (__CR6_LT, (a1), (a1))
|
|
430 #define vec_any_numeric(a1) __builtin_altivec_vcmpeqfp_p (__CR6_EQ_REV, (a1), (a1))
|
|
431
|
|
432 #define vec_all_eq(a1, a2) __builtin_vec_vcmpeq_p (__CR6_LT, (a1), (a2))
|
|
433 #define vec_all_ne(a1, a2) __builtin_vec_vcmpeq_p (__CR6_EQ, (a1), (a2))
|
|
434 #define vec_any_eq(a1, a2) __builtin_vec_vcmpeq_p (__CR6_EQ_REV, (a1), (a2))
|
|
435 #define vec_any_ne(a1, a2) __builtin_vec_vcmpeq_p (__CR6_LT_REV, (a1), (a2))
|
|
436
|
|
437 #define vec_all_gt(a1, a2) __builtin_vec_vcmpgt_p (__CR6_LT, (a1), (a2))
|
|
438 #define vec_all_lt(a1, a2) __builtin_vec_vcmpgt_p (__CR6_LT, (a2), (a1))
|
|
439 #define vec_any_gt(a1, a2) __builtin_vec_vcmpgt_p (__CR6_EQ_REV, (a1), (a2))
|
|
440 #define vec_any_lt(a1, a2) __builtin_vec_vcmpgt_p (__CR6_EQ_REV, (a2), (a1))
|
|
441
|
|
442 #define vec_all_ngt(a1, a2) __builtin_altivec_vcmpgtfp_p (__CR6_EQ, (a1), (a2))
|
|
443 #define vec_all_nlt(a1, a2) __builtin_altivec_vcmpgtfp_p (__CR6_EQ, (a2), (a1))
|
|
444 #define vec_any_ngt(a1, a2) __builtin_altivec_vcmpgtfp_p (__CR6_LT_REV, (a1), (a2))
|
|
445 #define vec_any_nlt(a1, a2) __builtin_altivec_vcmpgtfp_p (__CR6_LT_REV, (a2), (a1))
|
|
446
|
|
447 /* __builtin_vec_vcmpge_p is vcmpgefp for floating-point vector types,
|
|
448 while for integer types it is converted to __builtin_vec_vcmpgt_p,
|
|
449 with inverted args and condition code. */
|
|
450 #define vec_all_le(a1, a2) __builtin_vec_vcmpge_p (__CR6_LT, (a2), (a1))
|
|
451 #define vec_all_ge(a1, a2) __builtin_vec_vcmpge_p (__CR6_LT, (a1), (a2))
|
|
452 #define vec_any_le(a1, a2) __builtin_vec_vcmpge_p (__CR6_EQ_REV, (a2), (a1))
|
|
453 #define vec_any_ge(a1, a2) __builtin_vec_vcmpge_p (__CR6_EQ_REV, (a1), (a2))
|
|
454
|
|
455 #define vec_all_nge(a1, a2) __builtin_altivec_vcmpgefp_p (__CR6_EQ, (a1), (a2))
|
|
456 #define vec_all_nle(a1, a2) __builtin_altivec_vcmpgefp_p (__CR6_EQ, (a2), (a1))
|
|
457 #define vec_any_nge(a1, a2) __builtin_altivec_vcmpgefp_p (__CR6_LT_REV, (a1), (a2))
|
|
458 #define vec_any_nle(a1, a2) __builtin_altivec_vcmpgefp_p (__CR6_LT_REV, (a2), (a1))
|
|
459 #endif
|
|
460
|
|
461 /* These do not accept vectors, so they do not have a __builtin_vec_*
|
|
462 counterpart. */
|
|
463 #define vec_dss(x) __builtin_altivec_dss((x))
|
|
464 #define vec_dssall() __builtin_altivec_dssall ()
|
|
465 #define vec_mfvscr() ((__vector unsigned short) __builtin_altivec_mfvscr ())
|
|
466 #define vec_splat_s8(x) __builtin_altivec_vspltisb ((x))
|
|
467 #define vec_splat_s16(x) __builtin_altivec_vspltish ((x))
|
|
468 #define vec_splat_s32(x) __builtin_altivec_vspltisw ((x))
|
|
469 #define vec_splat_u8(x) ((__vector unsigned char) vec_splat_s8 ((x)))
|
|
470 #define vec_splat_u16(x) ((__vector unsigned short) vec_splat_s16 ((x)))
|
|
471 #define vec_splat_u32(x) ((__vector unsigned int) vec_splat_s32 ((x)))
|
|
472
|
|
473 /* This also accepts a type for its parameter, so it is not enough
|
|
474 to #define vec_step to __builtin_vec_step. */
|
|
475 #define vec_step(x) __builtin_vec_step (* (__typeof__ (x) *) 0)
|
|
476
|
|
477 #endif /* _ALTIVEC_H */
|