]> git.ipfire.org Git - thirdparty/glibc.git/blob - sysdeps/ia64/fpu/e_asinl.S
Remove "Contributed by" lines
[thirdparty/glibc.git] / sysdeps / ia64 / fpu / e_asinl.S
1 .file "asinl.s"
2
3
4 // Copyright (c) 2001 - 2003, Intel Corporation
5 // All rights reserved.
6 //
7 //
8 // Redistribution and use in source and binary forms, with or without
9 // modification, are permitted provided that the following conditions are
10 // met:
11 //
12 // * Redistributions of source code must retain the above copyright
13 // notice, this list of conditions and the following disclaimer.
14 //
15 // * Redistributions in binary form must reproduce the above copyright
16 // notice, this list of conditions and the following disclaimer in the
17 // documentation and/or other materials provided with the distribution.
18 //
19 // * The name of Intel Corporation may not be used to endorse or promote
20 // products derived from this software without specific prior written
21 // permission.
22
23 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
24 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
25 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
26 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
27 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
28 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
29 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
30 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
31 // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
32 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
33 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 //
35 // Intel Corporation is the author of this code, and requests that all
36 // problem reports or change requests be submitted to it directly at
37 // http://www.intel.com/software/products/opensource/libraries/num.htm.
38 //
39 // History
40 //==============================================================
41 // 08/28/01 New version
42 // 05/20/02 Cleaned up namespace and sf0 syntax
43 // 02/06/03 Reordered header: .section, .global, .proc, .align
44 //
45 // API
46 //==============================================================
47 // long double asinl(long double)
48 //
49 // Overview of operation
50 //==============================================================
51 // Background
52 //
53 // Implementation
54 //
55 // For |s| in [2^{-4}, sqrt(2)/2]:
56 // Let t= 2^k*1.b1 b2..b6 1, where s= 2^k*1.b1 b2.. b52
57 // asin(s)= asin(t)+asin(r), where r= s*sqrt(1-t^2)-t*sqrt(1-s^2), i.e.
58 // r= (s-t)*sqrt(1-t^2)-t*sqrt(1-t^2)*(sqrt((1-s^2)/(1-t^2))-1)
59 // asin(r)-r evaluated as 9-degree polynomial (c3*r^3+c5*r^5+c7*r^7+c9*r^9)
60 // The 64-bit significands of sqrt(1-t^2), 1/(1-t^2) are read from the table,
61 // along with the high and low parts of asin(t) (stored as two double precision
62 // values)
63 //
64 // |s| in (sqrt(2)/2, sqrt(255/256)):
65 // Let t= 2^k*1.b1 b2..b6 1, where (1-s^2)*frsqrta(1-s^2)= 2^k*1.b1 b2..b6..
66 // asin(|s|)= pi/2-asin(t)+asin(r), r= s*t-sqrt(1-s^2)*sqrt(1-t^2)
67 // To minimize accumulated errors, r is computed as
68 // r= (t*s)_s-t^2*y*z+z*y*(t^2-1+s^2)_s+z*y*(1-s^2)_s*x+z'*y*(1-s^2)*PS29+
69 // +(t*s-(t*s)_s)+z*y*((t^2-1-(t^2-1+s^2)_s)+s^2)+z*y*(1-s^2-(1-s^2)_s)+
70 // +ez*z'*y*(1-s^2)*(1-x),
71 // where y= frsqrta(1-s^2), z= (sqrt(1-t^2))_s (rounded to 24 significant bits)
72 // z'= sqrt(1-t^2), x= ((1-s^2)*y^2-1)/2
73 //
74 // |s|<2^{-4}: evaluate as 17-degree polynomial
75 // (or simply return s, if|s|<2^{-64})
76 //
77 // |s| in [sqrt(255/256), 1): asin(|s|)= pi/2-asin(sqrt(1-s^2))
78 // use 17-degree polynomial for asin(sqrt(1-s^2)),
79 // 9-degree polynomial to evaluate sqrt(1-s^2)
80 // High order term is (pi/2)_high-(y*(1-s^2))_high
81 //
82
83
84
85 // Registers used
86 //==============================================================
87 // f6-f15, f32-f36
88 // r2-r3, r23-r23
89 // p6, p7, p8, p12
90 //
91
92
93 GR_SAVE_B0= r33
94 GR_SAVE_PFS= r34
95 GR_SAVE_GP= r35 // This reg. can safely be used
96 GR_SAVE_SP= r36
97
98 GR_Parameter_X= r37
99 GR_Parameter_Y= r38
100 GR_Parameter_RESULT= r39
101 GR_Parameter_TAG= r40
102
103 FR_X= f10
104 FR_Y= f1
105 FR_RESULT= f8
106
107
108
109 RODATA
110
111 .align 16
112
113
114
115 LOCAL_OBJECT_START(T_table)
116
117 // stores 64-bit significand of 1/(1-t^2), 64-bit significand of sqrt(1-t^2),
118 // asin(t)_high (double precision), asin(t)_low (double precision)
119
120 data8 0x80828692b71c4391, 0xff7ddcec2d87e879
121 data8 0x3fb022bc0ae531a0, 0x3c9f599c7bb42af6
122 data8 0x80869f0163d0b082, 0xff79cad2247914d3
123 data8 0x3fb062dd26afc320, 0x3ca4eff21bd49c5c
124 data8 0x808ac7d5a8690705, 0xff75a89ed6b626b9
125 data8 0x3fb0a2ff4a1821e0, 0x3cb7e33b58f164cc
126 data8 0x808f0112ad8ad2e0, 0xff7176517c2cc0cb
127 data8 0x3fb0e32279319d80, 0x3caee31546582c43
128 data8 0x80934abba8a1da0a, 0xff6d33e949b1ed31
129 data8 0x3fb12346b8101da0, 0x3cb8bfe463d087cd
130 data8 0x8097a4d3dbe63d8f, 0xff68e16571015c63
131 data8 0x3fb1636c0ac824e0, 0x3c8870a7c5a3556f
132 data8 0x809c0f5e9662b3dd, 0xff647ec520bca0f0
133 data8 0x3fb1a392756ed280, 0x3c964f1a927461ae
134 data8 0x80a08a5f33fadc66, 0xff600c07846a6830
135 data8 0x3fb1e3b9fc19e580, 0x3c69eb3576d56332
136 data8 0x80a515d91d71acd4, 0xff5b892bc475affa
137 data8 0x3fb223e2a2dfbe80, 0x3c6a4e19fd972fb6
138 data8 0x80a9b1cfc86ff7cd, 0xff56f631062cf93d
139 data8 0x3fb2640c6dd76260, 0x3c62041160e0849e
140 data8 0x80ae5e46b78b0d68, 0xff5253166bc17794
141 data8 0x3fb2a43761187c80, 0x3cac61651af678c0
142 data8 0x80b31b417a4b756b, 0xff4d9fdb14463dc8
143 data8 0x3fb2e46380bb6160, 0x3cb06ef23eeba7a1
144 data8 0x80b7e8c3ad33c369, 0xff48dc7e1baf6738
145 data8 0x3fb32490d0d910c0, 0x3caa05f480b300d5
146 data8 0x80bcc6d0f9c784d6, 0xff4408fe9ad13e37
147 data8 0x3fb364bf558b3820, 0x3cb01e7e403aaab9
148 data8 0x80c1b56d1692492d, 0xff3f255ba75f5f4e
149 data8 0x3fb3a4ef12ec3540, 0x3cb4fe8fcdf5f5f1
150 data8 0x80c6b49bc72ec446, 0xff3a319453ebd961
151 data8 0x3fb3e5200d171880, 0x3caf2dc089b2b7e2
152 data8 0x80cbc460dc4e0ae8, 0xff352da7afe64ac6
153 data8 0x3fb425524827a720, 0x3cb75a855e7c6053
154 data8 0x80d0e4c033bee9c4, 0xff301994c79afb32
155 data8 0x3fb46585c83a5e00, 0x3cb3264981c019ab
156 data8 0x80d615bdb87556db, 0xff2af55aa431f291
157 data8 0x3fb4a5ba916c73c0, 0x3c994251d94427b5
158 data8 0x80db575d6291fd8a, 0xff25c0f84bae0cb9
159 data8 0x3fb4e5f0a7dbdb20, 0x3cbee2fcc4c786cb
160 data8 0x80e0a9a33769e535, 0xff207c6cc0ec09fd
161 data8 0x3fb526280fa74620, 0x3c940656e5549b91
162 data8 0x80e60c93498e32cd, 0xff1b27b703a19c98
163 data8 0x3fb56660ccee2740, 0x3ca7082374d7b2cd
164 data8 0x80eb8031b8d4052d, 0xff15c2d6105c72f8
165 data8 0x3fb5a69ae3d0b520, 0x3c7c4d46e09ac68a
166 data8 0x80f10482b25c6c8a, 0xff104dc8e0813ed4
167 data8 0x3fb5e6d6586fec20, 0x3c9aa84ffd9b4958
168 data8 0x80f6998a709c7cfb, 0xff0ac88e6a4ab926
169 data8 0x3fb627132eed9140, 0x3cbced2cbbbe7d16
170 data8 0x80fc3f4d3b657c44, 0xff053325a0c8a2ec
171 data8 0x3fb667516b6c34c0, 0x3c6489c5fc68595a
172 data8 0x8101f5cf67ed2af8, 0xfeff8d8d73dec2bb
173 data8 0x3fb6a791120f33a0, 0x3cbe12acf159dfad
174 data8 0x8107bd1558d6291f, 0xfef9d7c4d043df29
175 data8 0x3fb6e7d226fabba0, 0x3ca386d099cd0dc7
176 data8 0x810d95237e38766a, 0xfef411ca9f80b5f7
177 data8 0x3fb72814ae53cc20, 0x3cb9f35731e71dd6
178 data8 0x81137dfe55aa0e29, 0xfeee3b9dc7eef009
179 data8 0x3fb76858ac403a00, 0x3c74df3dd959141a
180 data8 0x811977aa6a479f0f, 0xfee8553d2cb8122c
181 data8 0x3fb7a89e24e6b0e0, 0x3ca6034406ee42bc
182 data8 0x811f822c54bd5ef8, 0xfee25ea7add46a91
183 data8 0x3fb7e8e51c6eb6a0, 0x3cb82f8f78e68ed7
184 data8 0x81259d88bb4ffac1, 0xfedc57dc2809fb1d
185 data8 0x3fb8292d9700ad60, 0x3cbebb73c0e653f9
186 data8 0x812bc9c451e5a257, 0xfed640d974eb6068
187 data8 0x3fb8697798c5d620, 0x3ca2feee76a9701b
188 data8 0x813206e3da0f3124, 0xfed0199e6ad6b585
189 data8 0x3fb8a9c325e852e0, 0x3cb9e88f2f4d0efe
190 data8 0x813854ec231172f9, 0xfec9e229dcf4747d
191 data8 0x3fb8ea1042932a00, 0x3ca5ff40d81f66fd
192 data8 0x813eb3e209ee858f, 0xfec39a7a9b36538b
193 data8 0x3fb92a5ef2f247c0, 0x3cb5e3bece4d6b07
194 data8 0x814523ca796f56ce, 0xfebd428f72561efe
195 data8 0x3fb96aaf3b3281a0, 0x3cb7b9e499436d7c
196 data8 0x814ba4aa6a2d3ff9, 0xfeb6da672bd48fe4
197 data8 0x3fb9ab011f819860, 0x3cb9168143cc1a7f
198 data8 0x81523686e29bbdd7, 0xfeb062008df81f50
199 data8 0x3fb9eb54a40e3ac0, 0x3cb6e544197eb1e1
200 data8 0x8158d964f7124614, 0xfea9d95a5bcbd65a
201 data8 0x3fba2ba9cd080800, 0x3ca9a717be8f7446
202 data8 0x815f8d49c9d639e4, 0xfea34073551e1ac8
203 data8 0x3fba6c009e9f9260, 0x3c741e989a60938a
204 data8 0x8166523a8b24f626, 0xfe9c974a367f785c
205 data8 0x3fbaac591d0661a0, 0x3cb2c1290107e57d
206 data8 0x816d283c793e0114, 0xfe95ddddb94166cb
207 data8 0x3fbaecb34c6ef600, 0x3c9c7d5fbaec405d
208 data8 0x81740f54e06d55bd, 0xfe8f142c93750c50
209 data8 0x3fbb2d0f310cca00, 0x3cbc09479a9cbcfb
210 data8 0x817b07891b15cd5e, 0xfe883a3577e9fceb
211 data8 0x3fbb6d6ccf1455e0, 0x3cb9450bff4ee307
212 data8 0x818210de91bba6c8, 0xfe814ff7162cf62f
213 data8 0x3fbbadcc2abb1180, 0x3c9227fda12a8d24
214 data8 0x81892b5abb0f2bf9, 0xfe7a55701a8697b1
215 data8 0x3fbbee2d48377700, 0x3cb6fad72acfe356
216 data8 0x819057031bf7760e, 0xfe734a9f2dfa1810
217 data8 0x3fbc2e902bc10600, 0x3cb4465b588d16ad
218 data8 0x819793dd479d4fbe, 0xfe6c2f82f643f68b
219 data8 0x3fbc6ef4d9904580, 0x3c8b9ac54823960d
220 data8 0x819ee1eedf76367a, 0xfe65041a15d8a92c
221 data8 0x3fbcaf5b55dec6a0, 0x3ca2b8d28a954db2
222 data8 0x81a6413d934f7a66, 0xfe5dc8632be3477f
223 data8 0x3fbcefc3a4e727a0, 0x3c9380da83713ab4
224 data8 0x81adb1cf21597d4b, 0xfe567c5cd44431d5
225 data8 0x3fbd302dcae51600, 0x3ca995b83421756a
226 data8 0x81b533a9563310b8, 0xfe4f2005a78fb50f
227 data8 0x3fbd7099cc155180, 0x3caefa2f7a817d5f
228 data8 0x81bcc6d20cf4f373, 0xfe47b35c3b0caaeb
229 data8 0x3fbdb107acb5ae80, 0x3cb455fc372dd026
230 data8 0x81c46b4f2f3d6e68, 0xfe40365f20b316d6
231 data8 0x3fbdf177710518c0, 0x3cbee3dcc5b01434
232 data8 0x81cc2126b53c1144, 0xfe38a90ce72abf36
233 data8 0x3fbe31e91d439620, 0x3cb3e131c950aebd
234 data8 0x81d3e85ea5bd8ee2, 0xfe310b6419c9c33a
235 data8 0x3fbe725cb5b24900, 0x3c01d3fac6029027
236 data8 0x81dbc0fd1637b9c1, 0xfe295d6340932d15
237 data8 0x3fbeb2d23e937300, 0x3c6304cc44aeedd1
238 data8 0x81e3ab082ad5a0a4, 0xfe219f08e03580b3
239 data8 0x3fbef349bc2a77e0, 0x3cac1d2d6abe9c72
240 data8 0x81eba6861683cb97, 0xfe19d0537a0946e2
241 data8 0x3fbf33c332bbe020, 0x3ca0909dba4e96ca
242 data8 0x81f3b37d1afc9979, 0xfe11f1418c0f94e2
243 data8 0x3fbf743ea68d5b60, 0x3c937fc12a2a779a
244 data8 0x81fbd1f388d4be45, 0xfe0a01d190f09063
245 data8 0x3fbfb4bc1be5c340, 0x3cbf51a504b55813
246 data8 0x820401efbf87e248, 0xfe020201fff9efea
247 data8 0x3fbff53b970d1e80, 0x3ca625444b260078
248 data8 0x82106ad2ffdca049, 0xfdf5e3940a49135e
249 data8 0x3fc02aff52065460, 0x3c9125d113e22a57
250 data8 0x8221343d6ea1d3e2, 0xfde581a45429b0a0
251 data8 0x3fc06b84f8e03220, 0x3caccf362295894b
252 data8 0x82324434adbf99c2, 0xfdd4de1a001fb775
253 data8 0x3fc0ac0ed1fe7240, 0x3cc22f676096b0af
254 data8 0x82439aee8d0c7747, 0xfdc3f8e8269d1f03
255 data8 0x3fc0ec9cee9e4820, 0x3cca147e2886a628
256 data8 0x825538a1d0fcb2f0, 0xfdb2d201a9b1ba66
257 data8 0x3fc12d2f6006f0a0, 0x3cc72b36633bc2d4
258 data8 0x82671d86345c5cee, 0xfda1695934d723e7
259 data8 0x3fc16dc63789de60, 0x3cb11f9c47c7b83f
260 data8 0x827949d46a121770, 0xfd8fbee13cbbb823
261 data8 0x3fc1ae618682e620, 0x3cce1b59020cef8e
262 data8 0x828bbdc61eeab9ba, 0xfd7dd28bff0c9f34
263 data8 0x3fc1ef015e586c40, 0x3cafec043e0225ee
264 data8 0x829e7995fb6de9e1, 0xfd6ba44b823ee1ca
265 data8 0x3fc22fa5d07b90c0, 0x3cba905409caf8e3
266 data8 0x82b17d7fa5bbc982, 0xfd5934119557883a
267 data8 0x3fc2704eee685da0, 0x3cb5ef21838a823e
268 data8 0x82c4c9bfc373d276, 0xfd4681cfcfb2c161
269 data8 0x3fc2b0fcc9a5f3e0, 0x3ccc7952c5e0e312
270 data8 0x82d85e93fba50136, 0xfd338d7790ca0f41
271 data8 0x3fc2f1af73c6ba00, 0x3cbecf5f977d1ca9
272 data8 0x82ec3c3af8c76b32, 0xfd2056f9fff97727
273 data8 0x3fc33266fe6889a0, 0x3c9d329c022ebdb5
274 data8 0x830062f46abf6022, 0xfd0cde480c43b327
275 data8 0x3fc373237b34de60, 0x3cc95806d4928adb
276 data8 0x8314d30108ea35f0, 0xfcf923526c1562b2
277 data8 0x3fc3b3e4fbe10520, 0x3cbc299fe7223d54
278 data8 0x83298ca29434df97, 0xfce526099d0737ed
279 data8 0x3fc3f4ab922e4a60, 0x3cb59d8bb8fdbccc
280 data8 0x833e901bd93c7009, 0xfcd0e65de39f1f7c
281 data8 0x3fc435774fea2a60, 0x3c9ec18b43340914
282 data8 0x8353ddb0b278aad8, 0xfcbc643f4b106055
283 data8 0x3fc4764846ee80a0, 0x3cb90402efd87ed6
284 data8 0x836975a60a70c52e, 0xfca79f9da4fab13a
285 data8 0x3fc4b71e8921b860, 0xbc58f23449ed6365
286 data8 0x837f5841ddfa7a46, 0xfc92986889284148
287 data8 0x3fc4f7fa2876fca0, 0xbc6294812bf43acd
288 data8 0x839585cb3e839773, 0xfc7d4e8f554ab12f
289 data8 0x3fc538db36ee6960, 0x3cb910b773d4c578
290 data8 0x83abfe8a5466246f, 0xfc67c2012cb6fa68
291 data8 0x3fc579c1c6953cc0, 0x3cc5ede909fc47fc
292 data8 0x83c2c2c861474d91, 0xfc51f2acf82041d5
293 data8 0x3fc5baade9860880, 0x3cac63cdfc3588e5
294 data8 0x83d9d2cfc2813637, 0xfc3be08165519325
295 data8 0x3fc5fb9fb1e8e3a0, 0x3cbf7c8466578c29
296 data8 0x83f12eebf397daac, 0xfc258b6ce6e6822f
297 data8 0x3fc63c9731f39d40, 0x3cb6d2a7ffca3e9e
298 data8 0x8408d76990b9296e, 0xfc0ef35db402af94
299 data8 0x3fc67d947be9eec0, 0x3cb1980da09e6566
300 data8 0x8420cc9659487cd7, 0xfbf81841c8082dc4
301 data8 0x3fc6be97a21daf00, 0x3cc2ac8330e59aa5
302 data8 0x84390ec132759ecb, 0xfbe0fa06e24cc390
303 data8 0x3fc6ffa0b6ef05e0, 0x3ccc1a030fee56c4
304 data8 0x84519e3a29df811a, 0xfbc9989a85ce0954
305 data8 0x3fc740afcccca000, 0x3cc19692a5301ca6
306 data8 0x846a7b527842d61b, 0xfbb1f3e9f8e45dc4
307 data8 0x3fc781c4f633e2c0, 0x3cc0e98f3868a508
308 data8 0x8483a65c8434b5f0, 0xfb9a0be244f4af45
309 data8 0x3fc7c2e045b12140, 0x3cb2a8d309754420
310 data8 0x849d1fabe4e97dd7, 0xfb81e070362116d1
311 data8 0x3fc80401cddfd120, 0x3ca7a44544aa4ce6
312 data8 0x84b6e795650817ea, 0xfb6971805af8411e
313 data8 0x3fc84529a16ac020, 0x3c9e3b709c7d6f94
314 data8 0x84d0fe6f0589da92, 0xfb50beff0423a2f5
315 data8 0x3fc88657d30c49e0, 0x3cc60d65a7f0a278
316 data8 0x84eb649000a73014, 0xfb37c8d84414755c
317 data8 0x3fc8c78c758e8e80, 0x3cc94b2ee984c2b7
318 data8 0x85061a50ccd13781, 0xfb1e8ef7eeaf764b
319 data8 0x3fc908c79bcba900, 0x3cc8540ae794a2fe
320 data8 0x8521200b1fb8916e, 0xfb05114998f76a83
321 data8 0x3fc94a0958ade6c0, 0x3ca127f49839fa9c
322 data8 0x853c7619f1618bf6, 0xfaeb4fb898b65d19
323 data8 0x3fc98b51bf2ffee0, 0x3c8c9ba7a803909a
324 data8 0x85581cd97f45e274, 0xfad14a3004259931
325 data8 0x3fc9cca0e25d4ac0, 0x3cba458e91d3bf54
326 data8 0x857414a74f8446b4, 0xfab7009ab1945a54
327 data8 0x3fca0df6d551fe80, 0x3cc78ea1d329d2b2
328 data8 0x85905de2341dea46, 0xfa9c72e3370d2fbc
329 data8 0x3fca4f53ab3b6200, 0x3ccf60dca86d57ef
330 data8 0x85acf8ea4e423ff8, 0xfa81a0f3e9fa0ee9
331 data8 0x3fca90b777580aa0, 0x3ca4c4e2ec8a867e
332 data8 0x85c9e62111a92e7d, 0xfa668ab6dec711b1
333 data8 0x3fcad2224cf814e0, 0x3c303de5980d071c
334 data8 0x85e725e947fbee97, 0xfa4b3015e883dbfe
335 data8 0x3fcb13943f7d5f80, 0x3cc29d4eefa5cb1e
336 data8 0x8604b8a7144cd054, 0xfa2f90fa9883a543
337 data8 0x3fcb550d625bc6a0, 0x3c9e01a746152daf
338 data8 0x86229ebff69e2415, 0xfa13ad4e3dfbe1c1
339 data8 0x3fcb968dc9195ea0, 0x3ccc091bd73ae518
340 data8 0x8640d89acf78858c, 0xf9f784f9e5a1877b
341 data8 0x3fcbd815874eb160, 0x3cb5f4b89875e187
342 data8 0x865f669fe390c7f5, 0xf9db17e65944eacf
343 data8 0x3fcc19a4b0a6f9c0, 0x3cc5c0bc2b0bbf14
344 data8 0x867e4938df7dc45f, 0xf9be65fc1f6c2e6e
345 data8 0x3fcc5b3b58e061e0, 0x3cc1ca70df8f57e7
346 data8 0x869d80d0db7e4c0c, 0xf9a16f237aec427a
347 data8 0x3fcc9cd993cc4040, 0x3cbae93acc85eccf
348 data8 0x86bd0dd45f4f8265, 0xf98433446a806e70
349 data8 0x3fccde7f754f5660, 0x3cb22f70e64568d0
350 data8 0x86dcf0b16613e37a, 0xf966b246a8606170
351 data8 0x3fcd202d11620fa0, 0x3c962030e5d4c849
352 data8 0x86fd29d7624b3d5d, 0xf948ec11a9d4c45b
353 data8 0x3fcd61e27c10c0a0, 0x3cc7083c91d59217
354 data8 0x871db9b741dbe44a, 0xf92ae08c9eca4941
355 data8 0x3fcda39fc97be7c0, 0x3cc9258579e57211
356 data8 0x873ea0c3722d6af2, 0xf90c8f9e71633363
357 data8 0x3fcde5650dd86d60, 0x3ca4755a9ea582a9
358 data8 0x875fdf6fe45529e8, 0xf8edf92dc5875319
359 data8 0x3fce27325d6fe520, 0x3cbc1e2b6c1954f9
360 data8 0x878176321154e2bc, 0xf8cf1d20f87270b8
361 data8 0x3fce6907cca0d060, 0x3cb6ca4804750830
362 data8 0x87a36580fe6bccf5, 0xf8affb5e20412199
363 data8 0x3fceaae56fdee040, 0x3cad6b310d6fd46c
364 data8 0x87c5add5417a5cb9, 0xf89093cb0b7c0233
365 data8 0x3fceeccb5bb33900, 0x3cc16e99cedadb20
366 data8 0x87e84fa9057914ca, 0xf870e64d40a15036
367 data8 0x3fcf2eb9a4bcb600, 0x3cc75ee47c8b09e9
368 data8 0x880b4b780f02b709, 0xf850f2c9fdacdf78
369 data8 0x3fcf70b05fb02e20, 0x3cad6350d379f41a
370 data8 0x882ea1bfc0f228ac, 0xf830b926379e6465
371 data8 0x3fcfb2afa158b8a0, 0x3cce0ccd9f829985
372 data8 0x885252ff21146108, 0xf810394699fe0e8e
373 data8 0x3fcff4b77e97f3e0, 0x3c9b30faa7a4c703
374 data8 0x88765fb6dceebbb3, 0xf7ef730f865f6df0
375 data8 0x3fd01b6406332540, 0x3cdc5772c9e0b9bd
376 data8 0x88ad1f69be2cc730, 0xf7bdc59bc9cfbd97
377 data8 0x3fd04cf8ad203480, 0x3caeef44fe21a74a
378 data8 0x88f763f70ae2245e, 0xf77a91c868a9c54e
379 data8 0x3fd08f23ce0162a0, 0x3cd6290ab3fe5889
380 data8 0x89431fc7bc0c2910, 0xf73642973c91298e
381 data8 0x3fd0d1610f0c1ec0, 0x3cc67401a01f08cf
382 data8 0x8990573407c7738e, 0xf6f0d71d1d7a2dd6
383 data8 0x3fd113b0c65d88c0, 0x3cc7aa4020fe546f
384 data8 0x89df0eb108594653, 0xf6aa4e6a05cfdef2
385 data8 0x3fd156134ada6fe0, 0x3cc87369da09600c
386 data8 0x8a2f4ad16e0ed78a, 0xf662a78900c35249
387 data8 0x3fd19888f43427a0, 0x3cc62b220f38e49c
388 data8 0x8a811046373e0819, 0xf619e180181d97cc
389 data8 0x3fd1db121aed7720, 0x3ca3ede7490b52f4
390 data8 0x8ad463df6ea0fa2c, 0xf5cffb504190f9a2
391 data8 0x3fd21daf185fa360, 0x3caafad98c1d6c1b
392 data8 0x8b294a8cf0488daf, 0xf584f3f54b8604e6
393 data8 0x3fd2606046bf95a0, 0x3cdb2d704eeb08fa
394 data8 0x8b7fc95f35647757, 0xf538ca65c960b582
395 data8 0x3fd2a32601231ec0, 0x3cc661619fa2f126
396 data8 0x8bd7e588272276f8, 0xf4eb7d92ff39fccb
397 data8 0x3fd2e600a3865760, 0x3c8a2a36a99aca4a
398 data8 0x8c31a45bf8e9255e, 0xf49d0c68cd09b689
399 data8 0x3fd328f08ad12000, 0x3cb9efaf1d7ab552
400 data8 0x8c8d0b520a35eb18, 0xf44d75cd993cfad2
401 data8 0x3fd36bf614dcc040, 0x3ccacbb590bef70d
402 data8 0x8cea2005d068f23d, 0xf3fcb8a23ab4942b
403 data8 0x3fd3af11a079a6c0, 0x3cd9775872cf037d
404 data8 0x8d48e837c8cd5027, 0xf3aad3c1e2273908
405 data8 0x3fd3f2438d754b40, 0x3ca03304f667109a
406 data8 0x8da969ce732f3ac7, 0xf357c60202e2fd7e
407 data8 0x3fd4358c3ca032e0, 0x3caecf2504ff1a9d
408 data8 0x8e0baad75555e361, 0xf3038e323ae9463a
409 data8 0x3fd478ec0fd419c0, 0x3cc64bdc3d703971
410 data8 0x8e6fb18807ba877e, 0xf2ae2b1c3a6057f7
411 data8 0x3fd4bc6369fa40e0, 0x3cbb7122ec245cf2
412 data8 0x8ed5843f4bda74d5, 0xf2579b83aa556f0c
413 data8 0x3fd4fff2af11e2c0, 0x3c9cfa2dc792d394
414 data8 0x8f3d29862c861fef, 0xf1ffde2612ca1909
415 data8 0x3fd5439a4436d000, 0x3cc38d46d310526b
416 data8 0x8fa6a81128940b2d, 0xf1a6f1bac0075669
417 data8 0x3fd5875a8fa83520, 0x3cd8bf59b8153f8a
418 data8 0x901206c1686317a6, 0xf14cd4f2a730d480
419 data8 0x3fd5cb33f8cf8ac0, 0x3c9502b5c4d0e431
420 data8 0x907f4ca5fe9cf739, 0xf0f186784a125726
421 data8 0x3fd60f26e847b120, 0x3cc8a1a5e0acaa33
422 data8 0x90ee80fd34aeda5e, 0xf09504ef9a212f18
423 data8 0x3fd65333c7e43aa0, 0x3cae5b029cb1f26e
424 data8 0x915fab35e37421c6, 0xf0374ef5daab5c45
425 data8 0x3fd6975b02b8e360, 0x3cd5aa1c280c45e6
426 data8 0x91d2d2f0d894d73c, 0xefd86321822dbb51
427 data8 0x3fd6db9d05213b20, 0x3cbecf2c093ccd8b
428 data8 0x9248000249200009, 0xef7840021aca5a72
429 data8 0x3fd71ffa3cc87fc0, 0x3cb8d273f08d00d9
430 data8 0x92bf3a7351f081d2, 0xef16e42021d7cbd5
431 data8 0x3fd7647318b1ad20, 0x3cbce099d79cdc46
432 data8 0x93388a8386725713, 0xeeb44dfce6820283
433 data8 0x3fd7a908093fc1e0, 0x3ccb033ec17a30d9
434 data8 0x93b3f8aa8e653812, 0xee507c126774fa45
435 data8 0x3fd7edb9803e3c20, 0x3cc10aedb48671eb
436 data8 0x94318d99d341ade4, 0xedeb6cd32f891afb
437 data8 0x3fd83287f0e9cf80, 0x3c994c0c1505cd2a
438 data8 0x94b1523e3dedc630, 0xed851eaa3168f43c
439 data8 0x3fd87773cff956e0, 0x3cda3b7bce6a6b16
440 data8 0x95334fc20577563f, 0xed1d8ffaa2279669
441 data8 0x3fd8bc7d93a70440, 0x3cd4922edc792ce2
442 data8 0x95b78f8e8f92f274, 0xecb4bf1fd2be72da
443 data8 0x3fd901a5b3b9cf40, 0x3cd3fea1b00f9d0d
444 data8 0x963e1b4e63a87c3f, 0xec4aaa6d08694cc1
445 data8 0x3fd946eca98f2700, 0x3cdba4032d968ff1
446 data8 0x96c6fcef314074fc, 0xebdf502d53d65fea
447 data8 0x3fd98c52f024e800, 0x3cbe7be1ab8c95c9
448 data8 0x97523ea3eab028b2, 0xeb72aea36720793e
449 data8 0x3fd9d1d904239860, 0x3cd72d08a6a22b70
450 data8 0x97dfeae6f4ee4a9a, 0xeb04c4096a884e94
451 data8 0x3fda177f63e8ef00, 0x3cd818c3c1ebfac7
452 data8 0x98700c7c6d85d119, 0xea958e90cfe1efd7
453 data8 0x3fda5d468f92a540, 0x3cdf45fbfaa080fe
454 data8 0x9902ae7487a9caa1, 0xea250c6224aab21a
455 data8 0x3fdaa32f090998e0, 0x3cd715a9353cede4
456 data8 0x9997dc2e017a9550, 0xe9b33b9ce2bb7638
457 data8 0x3fdae939540d3f00, 0x3cc545c014943439
458 data8 0x9a2fa158b29b649b, 0xe9401a573f8aa706
459 data8 0x3fdb2f65f63f6c60, 0x3cd4a63c2f2ca8e2
460 data8 0x9aca09f835466186, 0xe8cba69df9f0bf35
461 data8 0x3fdb75b5773075e0, 0x3cda310ce1b217ec
462 data8 0x9b672266ab1e0136, 0xe855de74266193d4
463 data8 0x3fdbbc28606babc0, 0x3cdc84b75cca6c44
464 data8 0x9c06f7579f0b7bd5, 0xe7debfd2f98c060b
465 data8 0x3fdc02bf3d843420, 0x3cd225d967ffb922
466 data8 0x9ca995db058cabdc, 0xe76648a991511c6e
467 data8 0x3fdc497a9c224780, 0x3cde08101c5b825b
468 data8 0x9d4f0b605ce71e88, 0xe6ec76dcbc02d9a7
469 data8 0x3fdc905b0c10d420, 0x3cb1abbaa3edf120
470 data8 0x9df765b9eecad5e6, 0xe6714846bdda7318
471 data8 0x3fdcd7611f4b8a00, 0x3cbf6217ae80aadf
472 data8 0x9ea2b320350540fe, 0xe5f4bab71494cd6b
473 data8 0x3fdd1e8d6a0d56c0, 0x3cb726e048cc235c
474 data8 0x9f51023562fc5676, 0xe576cbf239235ecb
475 data8 0x3fdd65e082df5260, 0x3cd9e66872bd5250
476 data8 0xa002620915c2a2f6, 0xe4f779b15f5ec5a7
477 data8 0x3fddad5b02a82420, 0x3c89743b0b57534b
478 data8 0xa0b6e21c2caf9992, 0xe476c1a233a7873e
479 data8 0x3fddf4fd84bbe160, 0x3cbf7adea9ee3338
480 data8 0xa16e9264cc83a6b2, 0xe3f4a16696608191
481 data8 0x3fde3cc8a6ec6ee0, 0x3cce46f5a51f49c6
482 data8 0xa22983528f3d8d49, 0xe3711694552da8a8
483 data8 0x3fde84bd099a6600, 0x3cdc78f6490a2d31
484 data8 0xa2e7c5d2e2e69460, 0xe2ec1eb4e1e0a5fb
485 data8 0x3fdeccdb4fc685c0, 0x3cdd3aedb56a4825
486 data8 0xa3a96b5599bd2532, 0xe265b74506fbe1c9
487 data8 0x3fdf15241f23b3e0, 0x3cd440f3c6d65f65
488 data8 0xa46e85d1ae49d7de, 0xe1ddddb499b3606f
489 data8 0x3fdf5d98202994a0, 0x3cd6c44bd3fb745a
490 data8 0xa53727ca3e11b99e, 0xe1548f662951b00d
491 data8 0x3fdfa637fe27bf60, 0x3ca8ad1cd33054dd
492 data8 0xa6036453bdc20186, 0xe0c9c9aeabe5e481
493 data8 0x3fdfef0467599580, 0x3cc0f1ac0685d78a
494 data8 0xa6d34f1969dda338, 0xe03d89d5281e4f81
495 data8 0x3fe01bff067d6220, 0x3cc0731e8a9ef057
496 data8 0xa7a6fc62f7246ff3, 0xdfafcd125c323f54
497 data8 0x3fe04092d1ae3b40, 0x3ccabda24b59906d
498 data8 0xa87e811a861df9b9, 0xdf20909061bb9760
499 data8 0x3fe0653df0fd9fc0, 0x3ce94c8dcc722278
500 data8 0xa959f2d2dd687200, 0xde8fd16a4e5f88bd
501 data8 0x3fe08a00c1cae320, 0x3ce6b888bb60a274
502 data8 0xaa3967cdeea58bda, 0xddfd8cabd1240d22
503 data8 0x3fe0aedba3221c00, 0x3ced5941cd486e46
504 data8 0xab904fd587263c84, 0xdd1f4472e1cf64ed
505 data8 0x3fe0e651e85229c0, 0x3cdb6701042299b1
506 data8 0xad686d44dd5a74bb, 0xdbf173e1f6b46e92
507 data8 0x3fe1309cbf4cdb20, 0x3cbf1be7bb3f0ec5
508 data8 0xaf524e15640ebee4, 0xdabd54896f1029f6
509 data8 0x3fe17b4ee1641300, 0x3ce81dd055b792f1
510 data8 0xb14eca24ef7db3fa, 0xd982cb9ae2f47e41
511 data8 0x3fe1c66b9ffd6660, 0x3cd98ea31eb5ddc7
512 data8 0xb35ec807669920ce, 0xd841bd1b8291d0b6
513 data8 0x3fe211f66db3a5a0, 0x3ca480c35a27b4a2
514 data8 0xb5833e4755e04dd1, 0xd6fa0bd3150b6930
515 data8 0x3fe25df2e05b6c40, 0x3ca4bc324287a351
516 data8 0xb7bd34c8000b7bd3, 0xd5ab9939a7d23aa1
517 data8 0x3fe2aa64b32f7780, 0x3cba67314933077c
518 data8 0xba0dc64d126cc135, 0xd4564563ce924481
519 data8 0x3fe2f74fc9289ac0, 0x3cec1a1dc0efc5ec
520 data8 0xbc76222cbbfa74a6, 0xd2f9eeed501125a8
521 data8 0x3fe344b82f859ac0, 0x3ceeef218de413ac
522 data8 0xbef78e31985291a9, 0xd19672e2182f78be
523 data8 0x3fe392a22087b7e0, 0x3cd2619ba201204c
524 data8 0xc19368b2b0629572, 0xd02baca5427e436a
525 data8 0x3fe3e11206694520, 0x3cb5d0b3143fe689
526 data8 0xc44b2ae8c6733e51, 0xceb975d60b6eae5d
527 data8 0x3fe4300c7e945020, 0x3cbd367143da6582
528 data8 0xc7206b894212dfef, 0xcd3fa6326ff0ac9a
529 data8 0x3fe47f965d201d60, 0x3ce797c7a4ec1d63
530 data8 0xca14e1b0622de526, 0xcbbe13773c3c5338
531 data8 0x3fe4cfb4b09d1a20, 0x3cedfadb5347143c
532 data8 0xcd2a6825eae65f82, 0xca34913d425a5ae9
533 data8 0x3fe5206cc637e000, 0x3ce2798b38e54193
534 data8 0xd06301095e1351ee, 0xc8a2f0d3679c08c0
535 data8 0x3fe571c42e3d0be0, 0x3ccd7cb9c6c2ca68
536 data8 0xd3c0d9f50057adda, 0xc70901152d59d16b
537 data8 0x3fe5c3c0c108f940, 0x3ceb6c13563180ab
538 data8 0xd74650a98cc14789, 0xc5668e3d4cbf8828
539 data8 0x3fe61668a46ffa80, 0x3caa9092e9e3c0e5
540 data8 0xdaf5f8579dcc8f8f, 0xc3bb61b3eed42d02
541 data8 0x3fe669c251ad69e0, 0x3cccf896ef3b4fee
542 data8 0xded29f9f9a6171b4, 0xc20741d7f8e8e8af
543 data8 0x3fe6bdd49bea05c0, 0x3cdc6b29937c575d
544 data8 0xe2df5765854ccdb0, 0xc049f1c2d1b8014b
545 data8 0x3fe712a6b76c6e80, 0x3ce1ddc6f2922321
546 data8 0xe71f7a9b94fcb4c3, 0xbe833105ec291e91
547 data8 0x3fe76840418978a0, 0x3ccda46e85432c3d
548 data8 0xeb96b72d3374b91e, 0xbcb2bb61493b28b3
549 data8 0x3fe7bea9496d5a40, 0x3ce37b42ec6e17d3
550 data8 0xf049183c3f53c39b, 0xbad848720223d3a8
551 data8 0x3fe815ea59dab0a0, 0x3cb03ad41bfc415b
552 data8 0xf53b11ec7f415f15, 0xb8f38b57c53c9c48
553 data8 0x3fe86e0c84010760, 0x3cc03bfcfb17fe1f
554 data8 0xfa718f05adbf2c33, 0xb70432500286b185
555 data8 0x3fe8c7196b9225c0, 0x3ced99fcc6866ba9
556 data8 0xfff200c3f5489608, 0xb509e6454dca33cc
557 data8 0x3fe9211b54441080, 0x3cb789cb53515688
558 // The following table entries are not used
559 //data8 0x82e138a0fac48700, 0xb3044a513a8e6132
560 //data8 0x3fe97c1d30f5b7c0, 0x3ce1eb765612d1d0
561 //data8 0x85f4cc7fc670d021, 0xb0f2fb2ea6cbbc88
562 //data8 0x3fe9d82ab4b5fde0, 0x3ced3fe6f27e8039
563 //data8 0x89377c1387d5b908, 0xaed58e9a09014d5c
564 //data8 0x3fea355065f87fa0, 0x3cbef481d25f5b58
565 //data8 0x8cad7a2c98dec333, 0xacab929ce114d451
566 //data8 0x3fea939bb451e2a0, 0x3c8e92b4fbf4560f
567 //data8 0x905b7dfc99583025, 0xaa748cc0dbbbc0ec
568 //data8 0x3feaf31b11270220, 0x3cdced8c61bd7bd5
569 //data8 0x9446d8191f80dd42, 0xa82ff92687235baf
570 //data8 0x3feb53de0bcffc20, 0x3cbe1722fb47509e
571 //data8 0x98758ba086e4000a, 0xa5dd497a9c184f58
572 //data8 0x3febb5f571cb0560, 0x3ce0c7774329a613
573 //data8 0x9cee6c7bf18e4e24, 0xa37be3c3cd1de51b
574 //data8 0x3fec197373bc7be0, 0x3ce08ebdb55c3177
575 //data8 0xa1b944000a1b9440, 0xa10b2101b4f27e03
576 //data8 0x3fec7e6bd023da60, 0x3ce5fc5fd4995959
577 //data8 0xa6defd8ba04d3e38, 0x9e8a4b93cad088ec
578 //data8 0x3fece4f404e29b20, 0x3cea3413401132b5
579 //data8 0xac69dd408a10c62d, 0x9bf89d5d17ddae8c
580 //data8 0x3fed4d2388f63600, 0x3cd5a7fb0d1d4276
581 //data8 0xb265c39cbd80f97a, 0x99553d969fec7beb
582 //data8 0x3fedb714101e0a00, 0x3cdbda21f01193f2
583 //data8 0xb8e081a16ae4ae73, 0x969f3e3ed2a0516c
584 //data8 0x3fee22e1da97bb00, 0x3ce7231177f85f71
585 //data8 0xbfea427678945732, 0x93d5990f9ee787af
586 //data8 0x3fee90ac13b18220, 0x3ce3c8a5453363a5
587 //data8 0xc79611399b8c90c5, 0x90f72bde80febc31
588 //data8 0x3fef009542b712e0, 0x3ce218fd79e8cb56
589 //data8 0xcffa8425040624d7, 0x8e02b4418574ebed
590 //data8 0x3fef72c3d2c57520, 0x3cd32a717f82203f
591 //data8 0xd93299cddcf9cf23, 0x8af6ca48e9c44024
592 //data8 0x3fefe762b77744c0, 0x3ce53478a6bbcf94
593 //data8 0xe35eda760af69ad9, 0x87d1da0d7f45678b
594 //data8 0x3ff02f511b223c00, 0x3ced6e11782c28fc
595 //data8 0xeea6d733421da0a6, 0x84921bbe64ae029a
596 //data8 0x3ff06c5c6f8ce9c0, 0x3ce71fc71c1ffc02
597 //data8 0xfb3b2c73fc6195cc, 0x813589ba3a5651b6
598 //data8 0x3ff0aaf2613700a0, 0x3cf2a72d2fd94ef3
599 //data8 0x84ac1fcec4203245, 0xfb73a828893df19e
600 //data8 0x3ff0eb367c3fd600, 0x3cf8054c158610de
601 //data8 0x8ca50621110c60e6, 0xf438a14c158d867c
602 //data8 0x3ff12d51caa6b580, 0x3ce6bce9748739b6
603 //data8 0x95b8c2062d6f8161, 0xecb3ccdd37b369da
604 //data8 0x3ff1717418520340, 0x3ca5c2732533177c
605 //data8 0xa0262917caab4ad1, 0xe4dde4ddc81fd119
606 //data8 0x3ff1b7d59dd40ba0, 0x3cc4c7c98e870ff5
607 //data8 0xac402c688b72f3f4, 0xdcae469be46d4c8d
608 //data8 0x3ff200b93cc5a540, 0x3c8dd6dc1bfe865a
609 //data8 0xba76968b9eabd9ab, 0xd41a8f3df1115f7f
610 //data8 0x3ff24c6f8f6affa0, 0x3cf1acb6d2a7eff7
611 //data8 0xcb63c87c23a71dc5, 0xcb161074c17f54ec
612 //data8 0x3ff29b5b338b7c80, 0x3ce9b5845f6ec746
613 //data8 0xdfe323b8653af367, 0xc19107d99ab27e42
614 //data8 0x3ff2edf6fac7f5a0, 0x3cf77f961925fa02
615 //data8 0xf93746caaba3e1f1, 0xb777744a9df03bff
616 //data8 0x3ff344df237486c0, 0x3cf6ddf5f6ddda43
617 //data8 0x8ca77052f6c340f0, 0xacaf476f13806648
618 //data8 0x3ff3a0dfa4bb4ae0, 0x3cfee01bbd761bff
619 //data8 0xa1a48604a81d5c62, 0xa11575d30c0aae50
620 //data8 0x3ff4030b73c55360, 0x3cf1cf0e0324d37c
621 //data8 0xbe45074b05579024, 0x9478e362a07dd287
622 //data8 0x3ff46ce4c738c4e0, 0x3ce3179555367d12
623 //data8 0xe7a08b5693d214ec, 0x8690e3575b8a7c3b
624 //data8 0x3ff4e0a887c40a80, 0x3cfbd5d46bfefe69
625 //data8 0x94503d69396d91c7, 0xedd2ce885ff04028
626 //data8 0x3ff561ebd9c18cc0, 0x3cf331bd176b233b
627 //data8 0xced1d96c5bb209e6, 0xc965278083808702
628 //data8 0x3ff5f71d7ff42c80, 0x3ce3301cc0b5a48c
629 //data8 0xabac2cee0fc24e20, 0x9c4eb1136094cbbd
630 //data8 0x3ff6ae4c63222720, 0x3cf5ff46874ee51e
631 //data8 0x8040201008040201, 0xb4d7ac4d9acb1bf4
632 //data8 0x3ff7b7d33b928c40, 0x3cfacdee584023bb
633 LOCAL_OBJECT_END(T_table)
634
635
636
637 .align 16
638
639 LOCAL_OBJECT_START(poly_coeffs)
640 // C_3
641 data8 0xaaaaaaaaaaaaaaab, 0x0000000000003ffc
642 // C_5
643 data8 0x999999999999999a, 0x0000000000003ffb
644 // C_7, C_9
645 data8 0x3fa6db6db6db6db7, 0x3f9f1c71c71c71c8
646 // pi/2 (low, high)
647 data8 0x3C91A62633145C07, 0x3FF921FB54442D18
648 // C_11, C_13
649 data8 0x3f96e8ba2e8ba2e9, 0x3f91c4ec4ec4ec4e
650 // C_15, C_17
651 data8 0x3f8c99999999999a, 0x3f87a87878787223
652 LOCAL_OBJECT_END(poly_coeffs)
653
654
655 R_DBL_S = r21
656 R_EXP0 = r22
657 R_EXP = r15
658 R_SGNMASK = r23
659 R_TMP = r24
660 R_TMP2 = r25
661 R_INDEX = r26
662 R_TMP3 = r27
663 R_TMP03 = r27
664 R_TMP4 = r28
665 R_TMP5 = r23
666 R_TMP6 = r22
667 R_TMP7 = r21
668 R_T = r29
669 R_BIAS = r20
670
671 F_T = f6
672 F_1S2 = f7
673 F_1S2_S = f9
674 F_INV_1T2 = f10
675 F_SQRT_1T2 = f11
676 F_S2T2 = f12
677 F_X = f13
678 F_D = f14
679 F_2M64 = f15
680
681 F_CS2 = f32
682 F_CS3 = f33
683 F_CS4 = f34
684 F_CS5 = f35
685 F_CS6 = f36
686 F_CS7 = f37
687 F_CS8 = f38
688 F_CS9 = f39
689 F_S23 = f40
690 F_S45 = f41
691 F_S67 = f42
692 F_S89 = f43
693 F_S25 = f44
694 F_S69 = f45
695 F_S29 = f46
696 F_X2 = f47
697 F_X4 = f48
698 F_TSQRT = f49
699 F_DTX = f50
700 F_R = f51
701 F_R2 = f52
702 F_R3 = f53
703 F_R4 = f54
704
705 F_C3 = f55
706 F_C5 = f56
707 F_C7 = f57
708 F_C9 = f58
709 F_P79 = f59
710 F_P35 = f60
711 F_P39 = f61
712
713 F_ATHI = f62
714 F_ATLO = f63
715
716 F_T1 = f64
717 F_Y = f65
718 F_Y2 = f66
719 F_ANDMASK = f67
720 F_ORMASK = f68
721 F_S = f69
722 F_05 = f70
723 F_SQRT_1S2 = f71
724 F_DS = f72
725 F_Z = f73
726 F_1T2 = f74
727 F_DZ = f75
728 F_ZE = f76
729 F_YZ = f77
730 F_Y1S2 = f78
731 F_Y1S2X = f79
732 F_1X = f80
733 F_ST = f81
734 F_1T2_ST = f82
735 F_TSS = f83
736 F_Y1S2X2 = f84
737 F_DZ_TERM = f85
738 F_DTS = f86
739 F_DS2X = f87
740 F_T2 = f88
741 F_ZY1S2S = f89
742 F_Y1S2_1X = f90
743 F_TS = f91
744 F_PI2_LO = f92
745 F_PI2_HI = f93
746 F_S19 = f94
747 F_INV1T2_2 = f95
748 F_CORR = f96
749 F_DZ0 = f97
750
751 F_C11 = f98
752 F_C13 = f99
753 F_C15 = f100
754 F_C17 = f101
755 F_P1113 = f102
756 F_P1517 = f103
757 F_P1117 = f104
758 F_P317 = f105
759 F_R8 = f106
760 F_HI = f107
761 F_1S2_HI = f108
762 F_DS2 = f109
763 F_Y2_2 = f110
764 F_S2 = f111
765 F_S_DS2 = f112
766 F_S_1S2S = f113
767 F_XL = f114
768 F_2M128 = f115
769
770
771 .section .text
772 GLOBAL_LIBM_ENTRY(asinl)
773
774 {.mfi
775 // get exponent, mantissa (rounded to double precision) of s
776 getf.d R_DBL_S = f8
777 // 1-s^2
778 fnma.s1 F_1S2 = f8, f8, f1
779 // r2 = pointer to T_table
780 addl r2 = @ltoff(T_table), gp
781 }
782
783 {.mfi
784 // sign mask
785 mov R_SGNMASK = 0x20000
786 nop.f 0
787 // bias-63-1
788 mov R_TMP03 = 0xffff-64;;
789 }
790
791
792 {.mfi
793 // get exponent of s
794 getf.exp R_EXP = f8
795 nop.f 0
796 // R_TMP4 = 2^45
797 shl R_TMP4 = R_SGNMASK, 45-17
798 }
799
800 {.mlx
801 // load bias-4
802 mov R_TMP = 0xffff-4
803 // load RU(sqrt(2)/2) to integer register (in double format, shifted left by 1)
804 movl R_TMP2 = 0x7fcd413cccfe779a;;
805 }
806
807
808 {.mfi
809 // load 2^{-64} in FP register
810 setf.exp F_2M64 = R_TMP03
811 nop.f 0
812 // index = (0x7-exponent)|b1 b2.. b6
813 extr.u R_INDEX = R_DBL_S, 46, 9
814 }
815
816 {.mfi
817 // get t = sign|exponent|b1 b2.. b6 1 x.. x
818 or R_T = R_DBL_S, R_TMP4
819 nop.f 0
820 // R_TMP4 = 2^45-1
821 sub R_TMP4 = R_TMP4, r0, 1;;
822 }
823
824
825 {.mfi
826 // get t = sign|exponent|b1 b2.. b6 1 0.. 0
827 andcm R_T = R_T, R_TMP4
828 nop.f 0
829 // eliminate sign from R_DBL_S (shift left by 1)
830 shl R_TMP3 = R_DBL_S, 1
831 }
832
833 {.mfi
834 // R_BIAS = 3*2^6
835 mov R_BIAS = 0xc0
836 nop.f 0
837 // eliminate sign from R_EXP
838 andcm R_EXP0 = R_EXP, R_SGNMASK;;
839 }
840
841
842
843 {.mfi
844 // load start address for T_table
845 ld8 r2 = [r2]
846 nop.f 0
847 // p8 = 1 if |s|> = sqrt(2)/2
848 cmp.geu p8, p0 = R_TMP3, R_TMP2
849 }
850
851 {.mlx
852 // p7 = 1 if |s|<2^{-4} (exponent of s<bias-4)
853 cmp.lt p7, p0 = R_EXP0, R_TMP
854 // sqrt coefficient cs8 = -33*13/128
855 movl R_TMP2 = 0xc0568000;;
856 }
857
858
859
860 {.mbb
861 // load t in FP register
862 setf.d F_T = R_T
863 // if |s|<2^{-4}, take alternate path
864 (p7) br.cond.spnt SMALL_S
865 // if |s|> = sqrt(2)/2, take alternate path
866 (p8) br.cond.sptk LARGE_S
867 }
868
869 {.mlx
870 // index = (4-exponent)|b1 b2.. b6
871 sub R_INDEX = R_INDEX, R_BIAS
872 // sqrt coefficient cs9 = 55*13/128
873 movl R_TMP = 0x40b2c000;;
874 }
875
876
877 {.mfi
878 // sqrt coefficient cs8 = -33*13/128
879 setf.s F_CS8 = R_TMP2
880 nop.f 0
881 // shift R_INDEX by 5
882 shl R_INDEX = R_INDEX, 5
883 }
884
885 {.mfi
886 // sqrt coefficient cs3 = 0.5 (set exponent = bias-1)
887 mov R_TMP4 = 0xffff - 1
888 nop.f 0
889 // sqrt coefficient cs6 = -21/16
890 mov R_TMP6 = 0xbfa8;;
891 }
892
893
894 {.mlx
895 // table index
896 add r2 = r2, R_INDEX
897 // sqrt coefficient cs7 = 33/16
898 movl R_TMP2 = 0x40040000;;
899 }
900
901
902 {.mmi
903 // load cs9 = 55*13/128
904 setf.s F_CS9 = R_TMP
905 // sqrt coefficient cs5 = 7/8
906 mov R_TMP3 = 0x3f60
907 // sqrt coefficient cs6 = 21/16
908 shl R_TMP6 = R_TMP6, 16;;
909 }
910
911
912 {.mmi
913 // load significand of 1/(1-t^2)
914 ldf8 F_INV_1T2 = [r2], 8
915 // sqrt coefficient cs7 = 33/16
916 setf.s F_CS7 = R_TMP2
917 // sqrt coefficient cs4 = -5/8
918 mov R_TMP5 = 0xbf20;;
919 }
920
921
922 {.mmi
923 // load significand of sqrt(1-t^2)
924 ldf8 F_SQRT_1T2 = [r2], 8
925 // sqrt coefficient cs6 = 21/16
926 setf.s F_CS6 = R_TMP6
927 // sqrt coefficient cs5 = 7/8
928 shl R_TMP3 = R_TMP3, 16;;
929 }
930
931
932 {.mmi
933 // sqrt coefficient cs3 = 0.5 (set exponent = bias-1)
934 setf.exp F_CS3 = R_TMP4
935 // r3 = pointer to polynomial coefficients
936 addl r3 = @ltoff(poly_coeffs), gp
937 // sqrt coefficient cs4 = -5/8
938 shl R_TMP5 = R_TMP5, 16;;
939 }
940
941
942 {.mfi
943 // sqrt coefficient cs5 = 7/8
944 setf.s F_CS5 = R_TMP3
945 // d = s-t
946 fms.s1 F_D = f8, f1, F_T
947 // set p6 = 1 if s<0, p11 = 1 if s> = 0
948 cmp.ge p6, p11 = R_EXP, R_DBL_S
949 }
950
951 {.mfi
952 // r3 = load start address to polynomial coefficients
953 ld8 r3 = [r3]
954 // s+t
955 fma.s1 F_S2T2 = f8, f1, F_T
956 nop.i 0;;
957 }
958
959
960 {.mfi
961 // sqrt coefficient cs4 = -5/8
962 setf.s F_CS4 = R_TMP5
963 // s^2-t^2
964 fma.s1 F_S2T2 = F_S2T2, F_D, f0
965 nop.i 0;;
966 }
967
968
969 {.mfi
970 // load C3
971 ldfe F_C3 = [r3], 16
972 // 0.5/(1-t^2) = 2^{-64}*(2^63/(1-t^2))
973 fma.s1 F_INV_1T2 = F_INV_1T2, F_2M64, f0
974 nop.i 0;;
975 }
976
977 {.mfi
978 // load C_5
979 ldfe F_C5 = [r3], 16
980 // set correct exponent for sqrt(1-t^2)
981 fma.s1 F_SQRT_1T2 = F_SQRT_1T2, F_2M64, f0
982 nop.i 0;;
983 }
984
985
986 {.mfi
987 // load C_7, C_9
988 ldfpd F_C7, F_C9 = [r3]
989 // x = -(s^2-t^2)/(1-t^2)/2
990 fnma.s1 F_X = F_INV_1T2, F_S2T2, f0
991 nop.i 0;;
992 }
993
994
995 {.mfi
996 // load asin(t)_high, asin(t)_low
997 ldfpd F_ATHI, F_ATLO = [r2]
998 // t*sqrt(1-t^2)
999 fma.s1 F_TSQRT = F_T, F_SQRT_1T2, f0
1000 nop.i 0;;
1001 }
1002
1003
1004 {.mfi
1005 nop.m 0
1006 // cs9*x+cs8
1007 fma.s1 F_S89 = F_CS9, F_X, F_CS8
1008 nop.i 0
1009 }
1010
1011 {.mfi
1012 nop.m 0
1013 // cs7*x+cs6
1014 fma.s1 F_S67 = F_CS7, F_X, F_CS6
1015 nop.i 0;;
1016 }
1017
1018 {.mfi
1019 nop.m 0
1020 // cs5*x+cs4
1021 fma.s1 F_S45 = F_CS5, F_X, F_CS4
1022 nop.i 0
1023 }
1024
1025 {.mfi
1026 nop.m 0
1027 // x*x
1028 fma.s1 F_X2 = F_X, F_X, f0
1029 nop.i 0;;
1030 }
1031
1032
1033 {.mfi
1034 nop.m 0
1035 // (s-t)-t*x
1036 fnma.s1 F_DTX = F_T, F_X, F_D
1037 nop.i 0
1038 }
1039
1040 {.mfi
1041 nop.m 0
1042 // cs3*x+cs2 (cs2 = -0.5 = -cs3)
1043 fms.s1 F_S23 = F_CS3, F_X, F_CS3
1044 nop.i 0;;
1045 }
1046
1047
1048 {.mfi
1049 nop.m 0
1050 // cs9*x^3+cs8*x^2+cs7*x+cs6
1051 fma.s1 F_S69 = F_S89, F_X2, F_S67
1052 nop.i 0
1053 }
1054
1055 {.mfi
1056 nop.m 0
1057 // x^4
1058 fma.s1 F_X4 = F_X2, F_X2, f0
1059 nop.i 0;;
1060 }
1061
1062
1063 {.mfi
1064 nop.m 0
1065 // t*sqrt(1-t^2)*x^2
1066 fma.s1 F_TSQRT = F_TSQRT, F_X2, f0
1067 nop.i 0
1068 }
1069
1070 {.mfi
1071 nop.m 0
1072 // cs5*x^3+cs4*x^2+cs3*x+cs2
1073 fma.s1 F_S25 = F_S45, F_X2, F_S23
1074 nop.i 0;;
1075 }
1076
1077
1078 {.mfi
1079 nop.m 0
1080 // ((s-t)-t*x)*sqrt(1-t^2)
1081 fma.s1 F_DTX = F_DTX, F_SQRT_1T2, f0
1082 nop.i 0;;
1083 }
1084
1085
1086 {.mfi
1087 nop.m 0
1088 // if sign is negative, negate table values: asin(t)_low
1089 (p6) fnma.s1 F_ATLO = F_ATLO, f1, f0
1090 nop.i 0
1091 }
1092
1093 {.mfi
1094 nop.m 0
1095 // PS29 = cs9*x^7+..+cs5*x^3+cs4*x^2+cs3*x+cs2
1096 fma.s1 F_S29 = F_S69, F_X4, F_S25
1097 nop.i 0;;
1098 }
1099
1100
1101 {.mfi
1102 nop.m 0
1103 // if sign is negative, negate table values: asin(t)_high
1104 (p6) fnma.s1 F_ATHI = F_ATHI, f1, f0
1105 nop.i 0
1106 }
1107
1108 {.mfi
1109 nop.m 0
1110 // R = ((s-t)-t*x)*sqrt(1-t^2)-t*sqrt(1-t^2)*x^2*PS29
1111 fnma.s1 F_R = F_S29, F_TSQRT, F_DTX
1112 nop.i 0;;
1113 }
1114
1115
1116 {.mfi
1117 nop.m 0
1118 // R^2
1119 fma.s1 F_R2 = F_R, F_R, f0
1120 nop.i 0;;
1121 }
1122
1123
1124 {.mfi
1125 nop.m 0
1126 // c7+c9*R^2
1127 fma.s1 F_P79 = F_C9, F_R2, F_C7
1128 nop.i 0
1129 }
1130
1131 {.mfi
1132 nop.m 0
1133 // c3+c5*R^2
1134 fma.s1 F_P35 = F_C5, F_R2, F_C3
1135 nop.i 0;;
1136 }
1137
1138 {.mfi
1139 nop.m 0
1140 // R^3
1141 fma.s1 F_R4 = F_R2, F_R2, f0
1142 nop.i 0;;
1143 }
1144
1145 {.mfi
1146 nop.m 0
1147 // R^3
1148 fma.s1 F_R3 = F_R2, F_R, f0
1149 nop.i 0;;
1150 }
1151
1152
1153
1154 {.mfi
1155 nop.m 0
1156 // c3+c5*R^2+c7*R^4+c9*R^6
1157 fma.s1 F_P39 = F_P79, F_R4, F_P35
1158 nop.i 0;;
1159 }
1160
1161
1162 {.mfi
1163 nop.m 0
1164 // asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1165 fma.s1 F_P39 = F_P39, F_R3, F_ATLO
1166 nop.i 0;;
1167 }
1168
1169
1170 {.mfi
1171 nop.m 0
1172 // R+asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1173 fma.s1 F_P39 = F_P39, f1, F_R
1174 nop.i 0;;
1175 }
1176
1177
1178 {.mfb
1179 nop.m 0
1180 // result = asin(t)_high+R+asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1181 fma.s0 f8 = F_ATHI, f1, F_P39
1182 // return
1183 br.ret.sptk b0;;
1184 }
1185
1186
1187
1188
1189 LARGE_S:
1190
1191 {.mfi
1192 // bias-1
1193 mov R_TMP3 = 0xffff - 1
1194 // y ~ 1/sqrt(1-s^2)
1195 frsqrta.s1 F_Y, p7 = F_1S2
1196 // c9 = 55*13*17/128
1197 mov R_TMP4 = 0x10af7b
1198 }
1199
1200 {.mlx
1201 // c8 = -33*13*15/128
1202 mov R_TMP5 = 0x184923
1203 movl R_TMP2 = 0xff00000000000000;;
1204 }
1205
1206 {.mfi
1207 // set p6 = 1 if s<0, p11 = 1 if s>0
1208 cmp.ge p6, p11 = R_EXP, R_DBL_S
1209 // 1-s^2
1210 fnma.s1 F_1S2 = f8, f8, f1
1211 // set p9 = 1
1212 cmp.eq p9, p0 = r0, r0;;
1213 }
1214
1215
1216 {.mfi
1217 // load 0.5
1218 setf.exp F_05 = R_TMP3
1219 // (1-s^2) rounded to single precision
1220 fnma.s.s1 F_1S2_S = f8, f8, f1
1221 // c9 = 55*13*17/128
1222 shl R_TMP4 = R_TMP4, 10
1223 }
1224
1225 {.mlx
1226 // AND mask for getting t ~ sqrt(1-s^2)
1227 setf.sig F_ANDMASK = R_TMP2
1228 // OR mask
1229 movl R_TMP2 = 0x0100000000000000;;
1230 }
1231
1232
1233 {.mfi
1234 nop.m 0
1235 // (s^2)_s
1236 fma.s.s1 F_S2 = f8, f8, f0
1237 nop.i 0;;
1238 }
1239
1240
1241 {.mmi
1242 // c9 = 55*13*17/128
1243 setf.s F_CS9 = R_TMP4
1244 // c7 = 33*13/16
1245 mov R_TMP4 = 0x41d68
1246 // c8 = -33*13*15/128
1247 shl R_TMP5 = R_TMP5, 11;;
1248 }
1249
1250
1251 {.mfi
1252 setf.sig F_ORMASK = R_TMP2
1253 // y^2
1254 fma.s1 F_Y2 = F_Y, F_Y, f0
1255 // c7 = 33*13/16
1256 shl R_TMP4 = R_TMP4, 12
1257 }
1258
1259 {.mfi
1260 // c6 = -33*7/16
1261 mov R_TMP6 = 0xc1670
1262 // y' ~ sqrt(1-s^2)
1263 fma.s1 F_T1 = F_Y, F_1S2, f0
1264 // c5 = 63/8
1265 mov R_TMP7 = 0x40fc;;
1266 }
1267
1268
1269 {.mlx
1270 // load c8 = -33*13*15/128
1271 setf.s F_CS8 = R_TMP5
1272 // c4 = -35/8
1273 movl R_TMP5 = 0xc08c0000;;
1274 }
1275
1276 {.mfi
1277 // r3 = pointer to polynomial coefficients
1278 addl r3 = @ltoff(poly_coeffs), gp
1279 // 1-(1-s^2)_s
1280 fnma.s1 F_DS = F_1S2_S, f1, f1
1281 // p9 = 0 if p7 = 1 (p9 = 1 for special cases only)
1282 (p7) cmp.ne p9, p0 = r0, r0
1283 }
1284
1285 {.mlx
1286 // load c7 = 33*13/16
1287 setf.s F_CS7 = R_TMP4
1288 // c3 = 5/2
1289 movl R_TMP4 = 0x40200000;;
1290 }
1291
1292
1293 {.mfi
1294 nop.m 0
1295 // 1-(s^2)_s
1296 fnma.s1 F_S_1S2S = F_S2, f1, f1
1297 nop.i 0
1298 }
1299
1300 {.mlx
1301 // load c4 = -35/8
1302 setf.s F_CS4 = R_TMP5
1303 // c2 = -3/2
1304 movl R_TMP5 = 0xbfc00000;;
1305 }
1306
1307
1308 {.mfi
1309 // load c3 = 5/2
1310 setf.s F_CS3 = R_TMP4
1311 // x = (1-s^2)_s*y^2-1
1312 fms.s1 F_X = F_1S2_S, F_Y2, f1
1313 // c6 = -33*7/16
1314 shl R_TMP6 = R_TMP6, 12
1315 }
1316
1317 {.mfi
1318 nop.m 0
1319 // y^2/2
1320 fma.s1 F_Y2_2 = F_Y2, F_05, f0
1321 nop.i 0;;
1322 }
1323
1324
1325 {.mfi
1326 // load c6 = -33*7/16
1327 setf.s F_CS6 = R_TMP6
1328 // eliminate lower bits from y'
1329 fand F_T = F_T1, F_ANDMASK
1330 // c5 = 63/8
1331 shl R_TMP7 = R_TMP7, 16
1332 }
1333
1334 {.mfb
1335 // r3 = load start address to polynomial coefficients
1336 ld8 r3 = [r3]
1337 // 1-(1-s^2)_s-s^2
1338 fnma.s1 F_DS = f8, f8, F_DS
1339 // p9 = 1 if s is a special input (NaN, or |s|> = 1)
1340 (p9) br.cond.spnt ASINL_SPECIAL_CASES;;
1341 }
1342
1343 {.mmf
1344 // get exponent, significand of y' (in single prec.)
1345 getf.s R_TMP = F_T1
1346 // load c3 = -3/2
1347 setf.s F_CS2 = R_TMP5
1348 // y*(1-s^2)
1349 fma.s1 F_Y1S2 = F_Y, F_1S2, f0;;
1350 }
1351
1352
1353 {.mfi
1354 nop.m 0
1355 // x' = (y^2/2)*(1-(s^2)_s)-0.5
1356 fms.s1 F_XL = F_Y2_2, F_S_1S2S, F_05
1357 nop.i 0
1358 }
1359
1360 {.mfi
1361 nop.m 0
1362 // s^2-(s^2)_s
1363 fms.s1 F_S_DS2 = f8, f8, F_S2
1364 nop.i 0;;
1365 }
1366
1367
1368 {.mfi
1369 nop.m 0
1370 // if s<0, set s = -s
1371 (p6) fnma.s1 f8 = f8, f1, f0
1372 nop.i 0;;
1373 }
1374
1375 {.mfi
1376 // load c5 = 63/8
1377 setf.s F_CS5 = R_TMP7
1378 // x = (1-s^2)_s*y^2-1+(1-(1-s^2)_s-s^2)*y^2
1379 fma.s1 F_X = F_DS, F_Y2, F_X
1380 // for t = 2^k*1.b1 b2.., get 7-k|b1.. b6
1381 extr.u R_INDEX = R_TMP, 17, 9;;
1382 }
1383
1384
1385 {.mmi
1386 // index = (4-exponent)|b1 b2.. b6
1387 sub R_INDEX = R_INDEX, R_BIAS
1388 nop.m 0
1389 // get exponent of y
1390 shr.u R_TMP2 = R_TMP, 23;;
1391 }
1392
1393 {.mmi
1394 // load C3
1395 ldfe F_C3 = [r3], 16
1396 // set p8 = 1 if y'<2^{-4}
1397 cmp.gt p8, p0 = 0x7b, R_TMP2
1398 // shift R_INDEX by 5
1399 shl R_INDEX = R_INDEX, 5;;
1400 }
1401
1402
1403 {.mfb
1404 // get table index for sqrt(1-t^2)
1405 add r2 = r2, R_INDEX
1406 // get t = 2^k*1.b1 b2.. b7 1
1407 for F_T = F_T, F_ORMASK
1408 (p8) br.cond.spnt VERY_LARGE_INPUT;;
1409 }
1410
1411
1412
1413 {.mmf
1414 // load C5
1415 ldfe F_C5 = [r3], 16
1416 // load 1/(1-t^2)
1417 ldfp8 F_INV_1T2, F_SQRT_1T2 = [r2], 16
1418 // x = ((1-s^2)*y^2-1)/2
1419 fma.s1 F_X = F_X, F_05, f0;;
1420 }
1421
1422
1423
1424 {.mmf
1425 nop.m 0
1426 // C7, C9
1427 ldfpd F_C7, F_C9 = [r3], 16
1428 // set correct exponent for t
1429 fmerge.se F_T = F_T1, F_T;;
1430 }
1431
1432
1433
1434 {.mfi
1435 // pi/2 (low, high)
1436 ldfpd F_PI2_LO, F_PI2_HI = [r3]
1437 // c9*x+c8
1438 fma.s1 F_S89 = F_X, F_CS9, F_CS8
1439 nop.i 0
1440 }
1441
1442 {.mfi
1443 nop.m 0
1444 // x^2
1445 fma.s1 F_X2 = F_X, F_X, f0
1446 nop.i 0;;
1447 }
1448
1449
1450 {.mfi
1451 nop.m 0
1452 // y*(1-s^2)*x
1453 fma.s1 F_Y1S2X = F_Y1S2, F_X, f0
1454 nop.i 0
1455 }
1456
1457 {.mfi
1458 nop.m 0
1459 // c7*x+c6
1460 fma.s1 F_S67 = F_X, F_CS7, F_CS6
1461 nop.i 0;;
1462 }
1463
1464
1465 {.mfi
1466 nop.m 0
1467 // 1-x
1468 fnma.s1 F_1X = F_X, f1, f1
1469 nop.i 0
1470 }
1471
1472 {.mfi
1473 nop.m 0
1474 // c3*x+c2
1475 fma.s1 F_S23 = F_X, F_CS3, F_CS2
1476 nop.i 0;;
1477 }
1478
1479
1480 {.mfi
1481 nop.m 0
1482 // 1-t^2
1483 fnma.s1 F_1T2 = F_T, F_T, f1
1484 nop.i 0
1485 }
1486
1487 {.mfi
1488 // load asin(t)_high, asin(t)_low
1489 ldfpd F_ATHI, F_ATLO = [r2]
1490 // c5*x+c4
1491 fma.s1 F_S45 = F_X, F_CS5, F_CS4
1492 nop.i 0;;
1493 }
1494
1495
1496
1497 {.mfi
1498 nop.m 0
1499 // t*s
1500 fma.s1 F_TS = F_T, f8, f0
1501 nop.i 0
1502 }
1503
1504 {.mfi
1505 nop.m 0
1506 // 0.5/(1-t^2)
1507 fma.s1 F_INV_1T2 = F_INV_1T2, F_2M64, f0
1508 nop.i 0;;
1509 }
1510
1511 {.mfi
1512 nop.m 0
1513 // z~sqrt(1-t^2), rounded to 24 significant bits
1514 fma.s.s1 F_Z = F_SQRT_1T2, F_2M64, f0
1515 nop.i 0
1516 }
1517
1518 {.mfi
1519 nop.m 0
1520 // sqrt(1-t^2)
1521 fma.s1 F_SQRT_1T2 = F_SQRT_1T2, F_2M64, f0
1522 nop.i 0;;
1523 }
1524
1525
1526 {.mfi
1527 nop.m 0
1528 // y*(1-s^2)*x^2
1529 fma.s1 F_Y1S2X2 = F_Y1S2, F_X2, f0
1530 nop.i 0
1531 }
1532
1533 {.mfi
1534 nop.m 0
1535 // x^4
1536 fma.s1 F_X4 = F_X2, F_X2, f0
1537 nop.i 0;;
1538 }
1539
1540
1541 {.mfi
1542 nop.m 0
1543 // s*t rounded to 24 significant bits
1544 fma.s.s1 F_TSS = F_T, f8, f0
1545 nop.i 0
1546 }
1547
1548 {.mfi
1549 nop.m 0
1550 // c9*x^3+..+c6
1551 fma.s1 F_S69 = F_X2, F_S89, F_S67
1552 nop.i 0;;
1553 }
1554
1555
1556 {.mfi
1557 nop.m 0
1558 // ST = (t^2-1+s^2) rounded to 24 significant bits
1559 fms.s.s1 F_ST = f8, f8, F_1T2
1560 nop.i 0
1561 }
1562
1563 {.mfi
1564 nop.m 0
1565 // c5*x^3+..+c2
1566 fma.s1 F_S25 = F_X2, F_S45, F_S23
1567 nop.i 0;;
1568 }
1569
1570
1571 {.mfi
1572 nop.m 0
1573 // 0.25/(1-t^2)
1574 fma.s1 F_INV1T2_2 = F_05, F_INV_1T2, f0
1575 nop.i 0
1576 }
1577
1578 {.mfi
1579 nop.m 0
1580 // t*s-sqrt(1-t^2)*(1-s^2)*y
1581 fnma.s1 F_TS = F_Y1S2, F_SQRT_1T2, F_TS
1582 nop.i 0;;
1583 }
1584
1585
1586 {.mfi
1587 nop.m 0
1588 // z*0.5/(1-t^2)
1589 fma.s1 F_ZE = F_INV_1T2, F_SQRT_1T2, f0
1590 nop.i 0
1591 }
1592
1593 {.mfi
1594 nop.m 0
1595 // z^2+t^2-1
1596 fms.s1 F_DZ0 = F_Z, F_Z, F_1T2
1597 nop.i 0;;
1598 }
1599
1600
1601 {.mfi
1602 nop.m 0
1603 // (1-s^2-(1-s^2)_s)*x
1604 fma.s1 F_DS2X = F_X, F_DS, f0
1605 nop.i 0;;
1606 }
1607
1608
1609 {.mfi
1610 nop.m 0
1611 // t*s-(t*s)_s
1612 fms.s1 F_DTS = F_T, f8, F_TSS
1613 nop.i 0
1614 }
1615
1616 {.mfi
1617 nop.m 0
1618 // c9*x^7+..+c2
1619 fma.s1 F_S29 = F_X4, F_S69, F_S25
1620 nop.i 0;;
1621 }
1622
1623
1624 {.mfi
1625 nop.m 0
1626 // y*z
1627 fma.s1 F_YZ = F_Z, F_Y, f0
1628 nop.i 0
1629 }
1630
1631 {.mfi
1632 nop.m 0
1633 // t^2
1634 fma.s1 F_T2 = F_T, F_T, f0
1635 nop.i 0;;
1636 }
1637
1638
1639 {.mfi
1640 nop.m 0
1641 // 1-t^2+ST
1642 fma.s1 F_1T2_ST = F_ST, f1, F_1T2
1643 nop.i 0;;
1644 }
1645
1646
1647 {.mfi
1648 nop.m 0
1649 // y*(1-s^2)(1-x)
1650 fma.s1 F_Y1S2_1X = F_Y1S2, F_1X, f0
1651 nop.i 0
1652 }
1653
1654 {.mfi
1655 nop.m 0
1656 // dz ~ sqrt(1-t^2)-z
1657 fma.s1 F_DZ = F_DZ0, F_ZE, f0
1658 nop.i 0;;
1659 }
1660
1661
1662 {.mfi
1663 nop.m 0
1664 // -1+correction for sqrt(1-t^2)-z
1665 fnma.s1 F_CORR = F_INV1T2_2, F_DZ0, f0
1666 nop.i 0;;
1667 }
1668
1669
1670 {.mfi
1671 nop.m 0
1672 // (PS29*x^2+x)*y*(1-s^2)
1673 fma.s1 F_S19 = F_Y1S2X2, F_S29, F_Y1S2X
1674 nop.i 0;;
1675 }
1676
1677
1678 {.mfi
1679 nop.m 0
1680 // z*y*(1-s^2)_s
1681 fma.s1 F_ZY1S2S = F_YZ, F_1S2_S, f0
1682 nop.i 0
1683 }
1684
1685 {.mfi
1686 nop.m 0
1687 // s^2-(1-t^2+ST)
1688 fms.s1 F_1T2_ST = f8, f8, F_1T2_ST
1689 nop.i 0;;
1690 }
1691
1692
1693 {.mfi
1694 nop.m 0
1695 // (t*s-(t*s)_s)+z*y*(1-s^2-(1-s^2)_s)*x
1696 fma.s1 F_DTS = F_YZ, F_DS2X, F_DTS
1697 nop.i 0
1698 }
1699
1700 {.mfi
1701 nop.m 0
1702 // dz*y*(1-s^2)*(1-x)
1703 fma.s1 F_DZ_TERM = F_DZ, F_Y1S2_1X, f0
1704 nop.i 0;;
1705 }
1706
1707
1708 {.mfi
1709 nop.m 0
1710 // R = t*s-sqrt(1-t^2)*(1-s^2)*y+sqrt(1-t^2)*(1-s^2)*y*PS19
1711 // (used for polynomial evaluation)
1712 fma.s1 F_R = F_S19, F_SQRT_1T2, F_TS
1713 nop.i 0;;
1714 }
1715
1716
1717 {.mfi
1718 nop.m 0
1719 // (PS29*x^2)*y*(1-s^2)
1720 fma.s1 F_S29 = F_Y1S2X2, F_S29, f0
1721 nop.i 0
1722 }
1723
1724 {.mfi
1725 nop.m 0
1726 // apply correction to dz*y*(1-s^2)*(1-x)
1727 fma.s1 F_DZ_TERM = F_DZ_TERM, F_CORR, F_DZ_TERM
1728 nop.i 0;;
1729 }
1730
1731
1732 {.mfi
1733 nop.m 0
1734 // R^2
1735 fma.s1 F_R2 = F_R, F_R, f0
1736 nop.i 0;;
1737 }
1738
1739
1740 {.mfi
1741 nop.m 0
1742 // (t*s-(t*s)_s)+z*y*(1-s^2-(1-s^2)_s)*x+dz*y*(1-s^2)*(1-x)
1743 fma.s1 F_DZ_TERM = F_DZ_TERM, f1, F_DTS
1744 nop.i 0;;
1745 }
1746
1747
1748 {.mfi
1749 nop.m 0
1750 // c7+c9*R^2
1751 fma.s1 F_P79 = F_C9, F_R2, F_C7
1752 nop.i 0
1753 }
1754
1755 {.mfi
1756 nop.m 0
1757 // c3+c5*R^2
1758 fma.s1 F_P35 = F_C5, F_R2, F_C3
1759 nop.i 0;;
1760 }
1761
1762 {.mfi
1763 nop.m 0
1764 // asin(t)_low-(pi/2)_low
1765 fms.s1 F_ATLO = F_ATLO, f1, F_PI2_LO
1766 nop.i 0
1767 }
1768
1769 {.mfi
1770 nop.m 0
1771 // R^4
1772 fma.s1 F_R4 = F_R2, F_R2, f0
1773 nop.i 0;;
1774 }
1775
1776 {.mfi
1777 nop.m 0
1778 // R^3
1779 fma.s1 F_R3 = F_R2, F_R, f0
1780 nop.i 0;;
1781 }
1782
1783
1784 {.mfi
1785 nop.m 0
1786 // (t*s)_s-t^2*y*z
1787 fnma.s1 F_TSS = F_T2, F_YZ, F_TSS
1788 nop.i 0
1789 }
1790
1791 {.mfi
1792 nop.m 0
1793 // d(ts)+z*y*d(1-s^2)*x+dz*y*(1-s^2)*(1-x)+z*y*(s^2-1+t^2-ST)
1794 fma.s1 F_DZ_TERM = F_YZ, F_1T2_ST, F_DZ_TERM
1795 nop.i 0;;
1796 }
1797
1798
1799 {.mfi
1800 nop.m 0
1801 // (pi/2)_hi-asin(t)_hi
1802 fms.s1 F_ATHI = F_PI2_HI, f1, F_ATHI
1803 nop.i 0
1804 }
1805
1806 {.mfi
1807 nop.m 0
1808 // c3+c5*R^2+c7*R^4+c9*R^6
1809 fma.s1 F_P39 = F_P79, F_R4, F_P35
1810 nop.i 0;;
1811 }
1812
1813
1814 {.mfi
1815 nop.m 0
1816 // d(ts)+z*y*d(1-s^2)*x+dz*y*(1-s^2)*(1-x)+z*y*(s^2-1+t^2-ST)+
1817 // + sqrt(1-t^2)*y*(1-s^2)*x^2*PS29
1818 fma.s1 F_DZ_TERM = F_SQRT_1T2, F_S29, F_DZ_TERM
1819 nop.i 0;;
1820 }
1821
1822
1823 {.mfi
1824 nop.m 0
1825 // (t*s)_s-t^2*y*z+z*y*ST
1826 fma.s1 F_TSS = F_YZ, F_ST, F_TSS
1827 nop.i 0
1828 }
1829
1830 {.mfi
1831 nop.m 0
1832 // -asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1833 fms.s1 F_P39 = F_P39, F_R3, F_ATLO
1834 nop.i 0;;
1835 }
1836
1837
1838 {.mfi
1839 nop.m 0
1840 // if s<0, change sign of F_ATHI
1841 (p6) fnma.s1 F_ATHI = F_ATHI, f1, f0
1842 nop.i 0
1843 }
1844
1845 {.mfi
1846 nop.m 0
1847 // d(ts)+z*y*d(1-s^2)*x+dz*y*(1-s^2)*(1-x)+z*y*(s^2-1+t^2-ST) +
1848 // + sqrt(1-t^2)*y*(1-s^2)*x^2*PS29 +
1849 // - asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1850 fma.s1 F_DZ_TERM = F_P39, f1, F_DZ_TERM
1851 nop.i 0;;
1852 }
1853
1854
1855 {.mfi
1856 nop.m 0
1857 // d(ts)+z*y*d(1-s^2)*x+dz*y*(1-s^2)*(1-x)+z*y*(s^2-1+t^2-ST) +
1858 // + sqrt(1-t^2)*y*(1-s^2)*x^2*PS29 + z*y*(1-s^2)_s*x +
1859 // - asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6)
1860 fma.s1 F_DZ_TERM = F_ZY1S2S, F_X, F_DZ_TERM
1861 nop.i 0;;
1862 }
1863
1864
1865 {.mfi
1866 nop.m 0
1867 // d(ts)+z*y*d(1-s^2)*x+dz*y*(1-s^2)*(1-x)+z*y*(s^2-1+t^2-ST) +
1868 // + sqrt(1-t^2)*y*(1-s^2)*x^2*PS29 + z*y*(1-s^2)_s*x +
1869 // - asin(t)_low+R^3*(c3+c5*R^2+c7*R^4+c9*R^6) +
1870 // + (t*s)_s-t^2*y*z+z*y*ST
1871 fma.s1 F_DZ_TERM = F_TSS, f1, F_DZ_TERM
1872 nop.i 0;;
1873 }
1874
1875
1876 .pred.rel "mutex", p6, p11
1877 {.mfi
1878 nop.m 0
1879 // result: add high part of pi/2-table value
1880 // s>0 in this case
1881 (p11) fma.s0 f8 = F_DZ_TERM, f1, F_ATHI
1882 nop.i 0
1883 }
1884
1885 {.mfb
1886 nop.m 0
1887 // result: add high part of pi/2-table value
1888 // if s<0
1889 (p6) fnma.s0 f8 = F_DZ_TERM, f1, F_ATHI
1890 br.ret.sptk b0;;
1891 }
1892
1893
1894
1895
1896
1897
1898 SMALL_S:
1899
1900 // use 15-term polynomial approximation
1901
1902 {.mmi
1903 // r3 = pointer to polynomial coefficients
1904 addl r3 = @ltoff(poly_coeffs), gp;;
1905 // load start address for coefficients
1906 ld8 r3 = [r3]
1907 mov R_TMP = 0x3fbf;;
1908 }
1909
1910
1911 {.mmi
1912 add r2 = 64, r3
1913 ldfe F_C3 = [r3], 16
1914 // p7 = 1 if |s|<2^{-64} (exponent of s<bias-64)
1915 cmp.lt p7, p0 = R_EXP0, R_TMP;;
1916 }
1917
1918 {.mmf
1919 ldfe F_C5 = [r3], 16
1920 ldfpd F_C11, F_C13 = [r2], 16
1921 // 2^{-128}
1922 fma.s1 F_2M128 = F_2M64, F_2M64, f0;;
1923 }
1924
1925 {.mmf
1926 ldfpd F_C7, F_C9 = [r3]
1927 ldfpd F_C15, F_C17 = [r2]
1928 // if |s|<2^{-64}, return s+2^{-128}*s
1929 (p7) fma.s0 f8 = f8, F_2M128, f8;;
1930 }
1931
1932
1933
1934 {.mfb
1935 nop.m 0
1936 // s^2
1937 fma.s1 F_R2 = f8, f8, f0
1938 // if |s|<2^{-64}, return s
1939 (p7) br.ret.spnt b0;;
1940 }
1941
1942
1943 {.mfi
1944 nop.m 0
1945 // s^3
1946 fma.s1 F_R3 = f8, F_R2, f0
1947 nop.i 0
1948 }
1949
1950 {.mfi
1951 nop.m 0
1952 // s^4
1953 fma.s1 F_R4 = F_R2, F_R2, f0
1954 nop.i 0;;
1955 }
1956
1957
1958 {.mfi
1959 nop.m 0
1960 // c3+c5*s^2
1961 fma.s1 F_P35 = F_C5, F_R2, F_C3
1962 nop.i 0
1963 }
1964
1965 {.mfi
1966 nop.m 0
1967 // c11+c13*s^2
1968 fma.s1 F_P1113 = F_C13, F_R2, F_C11
1969 nop.i 0;;
1970 }
1971
1972
1973 {.mfi
1974 nop.m 0
1975 // c7+c9*s^2
1976 fma.s1 F_P79 = F_C9, F_R2, F_C7
1977 nop.i 0
1978 }
1979
1980 {.mfi
1981 nop.m 0
1982 // c15+c17*s^2
1983 fma.s1 F_P1517 = F_C17, F_R2, F_C15
1984 nop.i 0;;
1985 }
1986
1987
1988 {.mfi
1989 nop.m 0
1990 // s^8
1991 fma.s1 F_R8 = F_R4, F_R4, f0
1992 nop.i 0;;
1993 }
1994
1995
1996 {.mfi
1997 nop.m 0
1998 // c3+c5*s^2+c7*s^4+c9*s^6
1999 fma.s1 F_P39 = F_P79, F_R4, F_P35
2000 nop.i 0
2001 }
2002
2003 {.mfi
2004 nop.m 0
2005 // c11+c13*s^2+c15*s^4+c17*s^6
2006 fma.s1 F_P1117 = F_P1517, F_R4, F_P1113
2007 nop.i 0;;
2008 }
2009
2010
2011 {.mfi
2012 nop.m 0
2013 // c3+..+c17*s^14
2014 fma.s1 F_P317 = F_R8, F_P1117, F_P39
2015 nop.i 0;;
2016 }
2017
2018
2019 {.mfb
2020 nop.m 0
2021 // result
2022 fma.s0 f8 = F_P317, F_R3, f8
2023 br.ret.sptk b0;;
2024 }
2025
2026
2027 {.mfb
2028 nop.m 0
2029 fma.s0 f8 = F_P317, F_R3, f0//F_P317, F_R3, F_S29
2030 // nop.f 0//fma.s0 f8 = f13, f6, f0
2031 br.ret.sptk b0;;
2032 }
2033
2034
2035
2036
2037
2038 VERY_LARGE_INPUT:
2039
2040 {.mfi
2041 nop.m 0
2042 // s rounded to 24 significant bits
2043 fma.s.s1 F_S = f8, f1, f0
2044 nop.i 0
2045 }
2046
2047 {.mfi
2048 // load C5
2049 ldfe F_C5 = [r3], 16
2050 // x = ((1-(s^2)_s)*y^2-1)/2-(s^2-(s^2)_s)*y^2/2
2051 fnma.s1 F_X = F_S_DS2, F_Y2_2, F_XL
2052 nop.i 0;;
2053 }
2054
2055
2056
2057 {.mmf
2058 nop.m 0
2059 // C7, C9
2060 ldfpd F_C7, F_C9 = [r3], 16
2061 nop.f 0;;
2062 }
2063
2064
2065
2066 {.mfi
2067 // pi/2 (low, high)
2068 ldfpd F_PI2_LO, F_PI2_HI = [r3], 16
2069 // c9*x+c8
2070 fma.s1 F_S89 = F_X, F_CS9, F_CS8
2071 nop.i 0
2072 }
2073
2074 {.mfi
2075 nop.m 0
2076 // x^2
2077 fma.s1 F_X2 = F_X, F_X, f0
2078 nop.i 0;;
2079 }
2080
2081
2082 {.mfi
2083 nop.m 0
2084 // y*(1-s^2)*x
2085 fma.s1 F_Y1S2X = F_Y1S2, F_X, f0
2086 nop.i 0
2087 }
2088
2089 {.mfi
2090 // C11, C13
2091 ldfpd F_C11, F_C13 = [r3], 16
2092 // c7*x+c6
2093 fma.s1 F_S67 = F_X, F_CS7, F_CS6
2094 nop.i 0;;
2095 }
2096
2097
2098 {.mfi
2099 // C15, C17
2100 ldfpd F_C15, F_C17 = [r3], 16
2101 // c3*x+c2
2102 fma.s1 F_S23 = F_X, F_CS3, F_CS2
2103 nop.i 0;;
2104 }
2105
2106
2107 {.mfi
2108 nop.m 0
2109 // c5*x+c4
2110 fma.s1 F_S45 = F_X, F_CS5, F_CS4
2111 nop.i 0;;
2112 }
2113
2114
2115 {.mfi
2116 nop.m 0
2117 // (s_s)^2
2118 fma.s1 F_DS = F_S, F_S, f0
2119 nop.i 0
2120 }
2121
2122 {.mfi
2123 nop.m 0
2124 // 1-(s_s)^2
2125 fnma.s1 F_1S2_S = F_S, F_S, f1
2126 nop.i 0;;
2127 }
2128
2129
2130 {.mfi
2131 nop.m 0
2132 // y*(1-s^2)*x^2
2133 fma.s1 F_Y1S2X2 = F_Y1S2, F_X2, f0
2134 nop.i 0
2135 }
2136
2137 {.mfi
2138 nop.m 0
2139 // x^4
2140 fma.s1 F_X4 = F_X2, F_X2, f0
2141 nop.i 0;;
2142 }
2143
2144
2145 {.mfi
2146 nop.m 0
2147 // c9*x^3+..+c6
2148 fma.s1 F_S69 = F_X2, F_S89, F_S67
2149 nop.i 0;;
2150 }
2151
2152
2153 {.mfi
2154 nop.m 0
2155 // c5*x^3+..+c2
2156 fma.s1 F_S25 = F_X2, F_S45, F_S23
2157 nop.i 0;;
2158 }
2159
2160
2161 {.mfi
2162 nop.m 0
2163 // ((s_s)^2-s^2)
2164 fnma.s1 F_DS = f8, f8, F_DS
2165 nop.i 0
2166 }
2167
2168 {.mfi
2169 nop.m 0
2170 // (pi/2)_high-y*(1-(s_s)^2)
2171 fnma.s1 F_HI = F_Y, F_1S2_S, F_PI2_HI
2172 nop.i 0;;
2173 }
2174
2175
2176 {.mfi
2177 nop.m 0
2178 // c9*x^7+..+c2
2179 fma.s1 F_S29 = F_X4, F_S69, F_S25
2180 nop.i 0;;
2181 }
2182
2183
2184 {.mfi
2185 nop.m 0
2186 // -(y*(1-(s_s)^2))_high
2187 fms.s1 F_1S2_HI = F_HI, f1, F_PI2_HI
2188 nop.i 0;;
2189 }
2190
2191
2192 {.mfi
2193 nop.m 0
2194 // (PS29*x^2+x)*y*(1-s^2)
2195 fma.s1 F_S19 = F_Y1S2X2, F_S29, F_Y1S2X
2196 nop.i 0;;
2197 }
2198
2199
2200 {.mfi
2201 nop.m 0
2202 // y*(1-(s_s)^2)-(y*(1-s^2))_high
2203 fma.s1 F_DS2 = F_Y, F_1S2_S, F_1S2_HI
2204 nop.i 0;;
2205 }
2206
2207
2208
2209 {.mfi
2210 nop.m 0
2211 // R ~ sqrt(1-s^2)
2212 // (used for polynomial evaluation)
2213 fnma.s1 F_R = F_S19, f1, F_Y1S2
2214 nop.i 0;;
2215 }
2216
2217
2218 {.mfi
2219 nop.m 0
2220 // y*(1-s^2)-(y*(1-s^2))_high
2221 fma.s1 F_DS2 = F_Y, F_DS, F_DS2
2222 nop.i 0
2223 }
2224
2225 {.mfi
2226 nop.m 0
2227 // (pi/2)_low+(PS29*x^2)*y*(1-s^2)
2228 fma.s1 F_S29 = F_Y1S2X2, F_S29, F_PI2_LO
2229 nop.i 0;;
2230 }
2231
2232
2233
2234 {.mfi
2235 nop.m 0
2236 // R^2
2237 fma.s1 F_R2 = F_R, F_R, f0
2238 nop.i 0;;
2239 }
2240
2241
2242 {.mfi
2243 nop.m 0
2244 // (pi/2)_low+(PS29*x^2)*y*(1-s^2)-(y*(1-s^2)-(y*(1-s^2))_high)
2245 fms.s1 F_S29 = F_S29, f1, F_DS2
2246 nop.i 0;;
2247 }
2248
2249
2250 {.mfi
2251 nop.m 0
2252 // c7+c9*R^2
2253 fma.s1 F_P79 = F_C9, F_R2, F_C7
2254 nop.i 0
2255 }
2256
2257 {.mfi
2258 nop.m 0
2259 // c3+c5*R^2
2260 fma.s1 F_P35 = F_C5, F_R2, F_C3
2261 nop.i 0;;
2262 }
2263
2264
2265
2266 {.mfi
2267 nop.m 0
2268 // R^4
2269 fma.s1 F_R4 = F_R2, F_R2, f0
2270 nop.i 0
2271 }
2272
2273 {.mfi
2274 nop.m 0
2275 // R^3
2276 fma.s1 F_R3 = F_R2, F_R, f0
2277 nop.i 0;;
2278 }
2279
2280
2281 {.mfi
2282 nop.m 0
2283 // c11+c13*R^2
2284 fma.s1 F_P1113 = F_C13, F_R2, F_C11
2285 nop.i 0
2286 }
2287
2288 {.mfi
2289 nop.m 0
2290 // c15+c17*R^2
2291 fma.s1 F_P1517 = F_C17, F_R2, F_C15
2292 nop.i 0;;
2293 }
2294
2295
2296 {.mfi
2297 nop.m 0
2298 // (pi/2)_low+(PS29*x^2)*y*(1-s^2)-(y*(1-s^2)-(y*(1-s^2))_high)+y*(1-s^2)*x
2299 fma.s1 F_S29 = F_Y1S2, F_X, F_S29
2300 nop.i 0;;
2301 }
2302
2303
2304 {.mfi
2305 nop.m 0
2306 // c11+c13*R^2+c15*R^4+c17*R^6
2307 fma.s1 F_P1117 = F_P1517, F_R4, F_P1113
2308 nop.i 0
2309 }
2310
2311 {.mfi
2312 nop.m 0
2313 // c3+c5*R^2+c7*R^4+c9*R^6
2314 fma.s1 F_P39 = F_P79, F_R4, F_P35
2315 nop.i 0;;
2316 }
2317
2318
2319 {.mfi
2320 nop.m 0
2321 // R^8
2322 fma.s1 F_R8 = F_R4, F_R4, f0
2323 nop.i 0;;
2324 }
2325
2326
2327 {.mfi
2328 nop.m 0
2329 // c3+c5*R^2+c7*R^4+c9*R^6+..+c17*R^14
2330 fma.s1 F_P317 = F_P1117, F_R8, F_P39
2331 nop.i 0;;
2332 }
2333
2334
2335 {.mfi
2336 nop.m 0
2337 // (pi/2)_low-(PS29*x^2)*y*(1-s^2)-(y*(1-s^2)-
2338 // -(y*(1-s^2))_high)+y*(1-s^2)*x - P3, 17
2339 fnma.s1 F_S29 = F_P317, F_R3, F_S29
2340 nop.i 0;;
2341 }
2342
2343 {.mfi
2344 nop.m 0
2345 // set sign
2346 (p6) fnma.s1 F_S29 = F_S29, f1, f0
2347 nop.i 0
2348 }
2349
2350 {.mfi
2351 nop.m 0
2352 (p6) fnma.s1 F_HI = F_HI, f1, f0
2353 nop.i 0;;
2354 }
2355
2356
2357 {.mfb
2358 nop.m 0
2359 // Result:
2360 // (pi/2)_low-(PS29*x^2)*y*(1-s^2)-(y*(1-s^2)-
2361 // -(y*(1-s^2))_high)+y*(1-s^2)*x - P3, 17
2362 // +(pi/2)_high-(y*(1-s^2))_high
2363 fma.s0 f8 = F_S29, f1, F_HI
2364 br.ret.sptk b0;;
2365 }
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375 ASINL_SPECIAL_CASES:
2376
2377 {.mfi
2378 alloc r32 = ar.pfs, 1, 4, 4, 0
2379 // check if the input is a NaN, or unsupported format
2380 // (i.e. not infinity or normal/denormal)
2381 fclass.nm p7, p8 = f8, 0x3f
2382 // pointer to pi/2
2383 add r3 = 48, r3;;
2384 }
2385
2386
2387 {.mfi
2388 // load pi/2
2389 ldfpd F_PI2_HI, F_PI2_LO = [r3]
2390 // get |s|
2391 fmerge.s F_S = f0, f8
2392 nop.i 0
2393 }
2394
2395 {.mfb
2396 nop.m 0
2397 // if NaN, quietize it, and return
2398 (p7) fma.s0 f8 = f8, f1, f0
2399 (p7) br.ret.spnt b0;;
2400 }
2401
2402
2403 {.mfi
2404 nop.m 0
2405 // |s| = 1 ?
2406 fcmp.eq.s0 p9, p0 = F_S, f1
2407 nop.i 0
2408 }
2409
2410 {.mfi
2411 nop.m 0
2412 // load FR_X
2413 fma.s1 FR_X = f8, f1, f0
2414 // load error tag
2415 mov GR_Parameter_TAG = 60;;
2416 }
2417
2418
2419 {.mfb
2420 nop.m 0
2421 // change sign if s = -1
2422 (p6) fnma.s1 F_PI2_HI = F_PI2_HI, f1, f0
2423 nop.b 0
2424 }
2425
2426 {.mfb
2427 nop.m 0
2428 // change sign if s = -1
2429 (p6) fnma.s1 F_PI2_LO = F_PI2_LO, f1, f0
2430 nop.b 0;;
2431 }
2432
2433 {.mfb
2434 nop.m 0
2435 // if s = 1, result is pi/2
2436 (p9) fma.s0 f8 = F_PI2_HI, f1, F_PI2_LO
2437 // return if |s| = 1
2438 (p9) br.ret.sptk b0;;
2439 }
2440
2441
2442 {.mfi
2443 nop.m 0
2444 // get Infinity
2445 frcpa.s1 FR_RESULT, p0 = f1, f0
2446 nop.i 0;;
2447 }
2448
2449
2450 {.mfi
2451 nop.m 0
2452 // return QNaN indefinite (0*Infinity)
2453 fma.s0 FR_RESULT = f0, FR_RESULT, f0
2454 nop.i 0;;
2455 }
2456
2457
2458 GLOBAL_LIBM_END(asinl)
2459 libm_alias_ldouble_other (asin, asin)
2460
2461
2462
2463 LOCAL_LIBM_ENTRY(__libm_error_region)
2464 .prologue
2465 // (1)
2466 { .mfi
2467 add GR_Parameter_Y=-32,sp // Parameter 2 value
2468 nop.f 0
2469 .save ar.pfs,GR_SAVE_PFS
2470 mov GR_SAVE_PFS=ar.pfs // Save ar.pfs
2471 }
2472 { .mfi
2473 .fframe 64
2474 add sp=-64,sp // Create new stack
2475 nop.f 0
2476 mov GR_SAVE_GP=gp // Save gp
2477 };;
2478
2479
2480 // (2)
2481 { .mmi
2482 stfe [GR_Parameter_Y] = f1,16 // Store Parameter 2 on stack
2483 add GR_Parameter_X = 16,sp // Parameter 1 address
2484 .save b0, GR_SAVE_B0
2485 mov GR_SAVE_B0=b0 // Save b0
2486 };;
2487
2488 .body
2489 // (3)
2490 { .mib
2491 stfe [GR_Parameter_X] = FR_X // Store Parameter 1 on stack
2492 add GR_Parameter_RESULT = 0,GR_Parameter_Y
2493 nop.b 0 // Parameter 3 address
2494 }
2495 { .mib
2496 stfe [GR_Parameter_Y] = FR_RESULT // Store Parameter 3 on stack
2497 add GR_Parameter_Y = -16,GR_Parameter_Y
2498 br.call.sptk b0=__libm_error_support# // Call error handling function
2499 };;
2500 { .mmi
2501 nop.m 0
2502 nop.m 0
2503 add GR_Parameter_RESULT = 48,sp
2504 };;
2505
2506 // (4)
2507 { .mmi
2508 ldfe f8 = [GR_Parameter_RESULT] // Get return result off stack
2509 .restore sp
2510 add sp = 64,sp // Restore stack pointer
2511 mov b0 = GR_SAVE_B0 // Restore return address
2512 };;
2513
2514 { .mib
2515 mov gp = GR_SAVE_GP // Restore gp
2516 mov ar.pfs = GR_SAVE_PFS // Restore ar.pfs
2517 br.ret.sptk b0 // Return
2518 };;
2519
2520 LOCAL_LIBM_END(__libm_error_region)
2521
2522 .type __libm_error_support#,@function
2523 .global __libm_error_support#