| blob | 58ef5f2adb78f46468bafef01c76878f93386fa5 |
1 .file "acoshf.s"
4 // Copyright (c) 2000 - 2003, Intel Corporation
5 // All rights reserved.
6 //
7 // Contributed 2000 by the Intel Numerics Group, Intel Corporation
8 //
9 // Redistribution and use in source and binary forms, with or without
10 // modification, are permitted provided that the following conditions are
11 // met:
12 //
13 // * Redistributions of source code must retain the above copyright
14 // notice, this list of conditions and the following disclaimer.
15 //
16 // * Redistributions in binary form must reproduce the above copyright
17 // notice, this list of conditions and the following disclaimer in the
18 // documentation and/or other materials provided with the distribution.
19 //
20 // * The name of Intel Corporation may not be used to endorse or promote
21 // products derived from this software without specific prior written
22 // permission.
24 // THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
25 // "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
26 // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
27 // A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS
28 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
29 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
30 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
31 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
32 // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING
33 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
34 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
35 //
36 // Intel Corporation is the author of this code, and requests that all
37 // problem reports or change requests be submitted to it directly at
38 // http://www.intel.com/software/products/opensource/libraries/num.htm.
39 //
40 // ==============================================================
41 // History
42 // ==============================================================
43 // 03/28/01 Initial version
44 // 04/19/01 Improved speed of the paths #1,2,3,4,5
45 // 05/20/02 Cleaned up namespace and sf0 syntax
46 // 02/06/03 Reordered header: .section, .global, .proc, .align
47 // 05/14/03 Improved performance, set denormal flag for unorms >= 1.0
48 //
49 // API
50 // ==============================================================
51 // float acoshf(float)
52 //
53 // Overview of operation
54 // ==============================================================
55 //
56 // There are 7 paths:
57 // 1. x = 1.0
58 // Return acoshf(x) = 0.0
59 // 2. 1.0 < x < 1.000499725341796875(0x3FF0020C00000000)
60 // Return acoshf(x) = sqrt(x-1) * Pol4(x),
61 // where Pol4(x) = (x*C2 + C1)*(x-1) + C0
62 //
63 // 3. 1.000499725341796875(0x3FF0020C00000000) <= x < 2^51
64 // Return acoshf(x) = log(x + sqrt(x^2 -1.0))
65 // To compute x + sqrt(x^2 -1.0) modified Newton Raphson method is used
66 // (2 iterations)
67 // Algorithm description for log function see below.
68 //
69 // 4. 2^51 <= x < +INF
70 // Return acoshf(x) = log(2*x)
71 // Algorithm description for log function see below.
72 //
73 // 5. x = +INF
74 // Return acoshf(x) = +INF
75 //
76 // 6. x = [S,Q]NaN
77 // Return acoshf(x) = QNaN
78 //
79 // 7. x < 1.0
80 // It's domain error. Error handler with tag = 137 is called
81 //
82 //==============================================================
83 // Algorithm Description for log(x) function
84 // Below we are using the fact that inequality x - 1.0 > 2^(-6) is always
85 // true for this acosh implementation
86 //
87 // Consider x = 2^N 1.f1 f2 f3 f4...f63
88 // Log(x) = log(frcpa(x) x/frcpa(x))
89 // = log(1/frcpa(x)) + log(frcpa(x) x)
90 // = -log(frcpa(x)) + log(frcpa(x) x)
91 //
92 // frcpa(x) = 2^-N frcpa((1.f1 f2 ... f63)
93 //
94 // -log(frcpa(x)) = -log(C)
95 // = -log(2^-N) - log(frcpa(1.f1 f2 ... f63))
96 //
97 // -log(frcpa(x)) = -log(C)
98 // = +Nlog2 - log(frcpa(1.f1 f2 ... f63))
99 //
100 // -log(frcpa(x)) = -log(C)
101 // = +Nlog2 + log(frcpa(1.f1 f2 ... f63))
102 //
103 // Log(x) = log(1/frcpa(x)) + log(frcpa(x) x)
104 //
105 // Log(x) = +Nlog2 + log(1./frcpa(1.f1 f2 ... f63)) + log(frcpa(x) x)
106 // Log(x) = +Nlog2 - log(/frcpa(1.f1 f2 ... f63)) + log(frcpa(x) x)
107 // Log(x) = +Nlog2 + T + log(frcpa(x) x)
108 //
109 // Log(x) = +Nlog2 + T + log(C x)
110 //
111 // Cx = 1 + r
112 //
113 // Log(x) = +Nlog2 + T + log(1+r)
114 // Log(x) = +Nlog2 + T + Series( r - r^2/2 + r^3/3 - r^4/4 ....)
115 //
116 // 1.f1 f2 ... f8 has 256 entries.
117 // They are 1 + k/2^8, k = 0 ... 255
118 // These 256 values are the table entries.
119 //
120 // Implementation
121 //==============================================================
122 // C = frcpa(x)
123 // r = C * x - 1
124 //
125 // Form rseries = r + P1*r^2 + P2*r^3 + P3*r^4
126 //
127 // x = f * 2*n where f is 1.f_1f_2f_3....f_63
128 // Nfloat = float(n) where n is the true unbiased exponent
129 // pre-index = f_1f_2....f_8
130 // index = pre_index * 8
131 // get the dxt table entry at index + offset = T
132 //
133 // result = (T + Nfloat * log(2)) + rseries
134 //
135 // The T table is calculated as follows
136 // Form x_k = 1 + k/2^8 where k goes from 0... 255
137 // y_k = frcpa(x_k)
138 // log(1/y_k) in quad and round to double
139 //
141 // Registers used
142 //==============================================================
143 // Floating Point registers used:
144 // f8, input
145 // f9 -> f15, f32 -> f62
146 //
147 // General registers used:
148 // r14 -> r27, r32 -> r39
149 //
150 // Predicate registers used:
151 // p6 -> p15
152 //
153 // p6 to filter out case when x = [Q,S]NaN
154 // p7,p8 to filter out case when x < 1.0
155 //
156 // p10 to select path #1
157 // p11 to filter out case when x = +INF
158 // p12 used in the frcpa
159 // p13 to select path #4
160 // p14,p15 to select path #2
162 // Assembly macros
163 //==============================================================
164 log_GR_exp_17_ones = r14
165 log_GR_signexp_f8 = r15
166 log_table_address2 = r16
167 log_GR_exp_16_ones = r17
168 log_GR_exp_f8 = r18
169 log_GR_true_exp_f8 = r19
170 log_GR_significand_f8 = r20
171 log_GR_index = r21
172 log_GR_comp2 = r22
173 acosh_GR_f8 = r23
174 log_GR_comp = r24
175 acosh_GR_f8_sig = r25
176 log_table_address3 = r26
177 NR_table_address = r27
179 GR_SAVE_B0 = r33
180 GR_SAVE_GP = r34
181 GR_SAVE_PFS = r35
183 GR_Parameter_X = r36
184 GR_Parameter_Y = r37
185 GR_Parameter_RESULT = r38
186 acosh_GR_tag = r39
188 //==============================================================
189 log_y = f9
190 NR1 = f10
191 NR2 = f11
192 log_y_rs = f12
193 log_y_rs_iter = f13
194 log_y_rs_iter1 = f14
195 log_NORM_f8 = f15
196 log_w = f32
197 acosh_comp = f34
198 acosh_comp2 = f33
199 log_P3 = f35
200 log_P2 = f36
201 log_P1 = f37
202 log2 = f38
203 log_C0 = f39
204 log_C1 = f40
205 log_C2 = f41
206 acosh_w_rs = f42
207 log_C = f43
208 log_arg = f44
209 acosh_w_iter1 = f45
210 acosh_w_iter2 = f46
211 log_int_Nfloat = f47
212 log_r = f48
213 log_rsq = f49
214 log_rp_p4 = f50
215 log_rp_p32 = f51
216 log_rcube = f52
217 log_rp_p10 = f53
218 log_rp_p2 = f54
219 log_Nfloat = f55
220 log_T = f56
221 log_r2P_r = f57
222 log_T_plus_Nlog2 = f58
223 acosh_w_sqrt = f59
224 acosh_w_1 = f60
225 log_arg_early = f61
226 log_y_rs_iter2 = f62
229 // Data tables
230 //==============================================================
232 RODATA
233 .align 16
235 LOCAL_OBJECT_START(log_table_1)
236 data8 0xbfd0001008f39d59 // p3
237 data8 0x3fd5556073e0c45a // p2
238 data8 0xbfdffffffffaea15 // p1
239 data8 0x3FE62E42FEFA39EF // log2
240 LOCAL_OBJECT_END(log_table_1)
242 LOCAL_OBJECT_START(log_table_2)
244 data8 0x3FE0000000000000 // 0.5
245 data8 0x4008000000000000 // 3.0
246 data8 0xD92CBAD213719F11, 0x00003FF9 // C2 3FF9D92CBAD213719F11
247 data8 0x93D38EBF2EC9B073, 0x0000BFFC // C1 BFFC93D38EBF2EC9B073
248 data8 0xB504F333F9DA0E32, 0x00003FFF // C0 3FFFB504F333F9DA0E32
249 LOCAL_OBJECT_END(log_table_2)
251 LOCAL_OBJECT_START(log_table_3)
252 data8 0x3F60040155D5889E //log(1/frcpa(1+ 0/256)
253 data8 0x3F78121214586B54 //log(1/frcpa(1+ 1/256)
254 data8 0x3F841929F96832F0 //log(1/frcpa(1+ 2/256)
255 data8 0x3F8C317384C75F06 //log(1/frcpa(1+ 3/256)
256 data8 0x3F91A6B91AC73386 //log(1/frcpa(1+ 4/256)
257 data8 0x3F95BA9A5D9AC039 //log(1/frcpa(1+ 5/256)
258 data8 0x3F99D2A8074325F4 //log(1/frcpa(1+ 6/256)
259 data8 0x3F9D6B2725979802 //log(1/frcpa(1+ 7/256)
260 data8 0x3FA0C58FA19DFAAA //log(1/frcpa(1+ 8/256)
261 data8 0x3FA2954C78CBCE1B //log(1/frcpa(1+ 9/256)
262 data8 0x3FA4A94D2DA96C56 //log(1/frcpa(1+ 10/256)
263 data8 0x3FA67C94F2D4BB58 //log(1/frcpa(1+ 11/256)
264 data8 0x3FA85188B630F068 //log(1/frcpa(1+ 12/256)
265 data8 0x3FAA6B8ABE73AF4C //log(1/frcpa(1+ 13/256)
266 data8 0x3FAC441E06F72A9E //log(1/frcpa(1+ 14/256)
267 data8 0x3FAE1E6713606D07 //log(1/frcpa(1+ 15/256)
268 data8 0x3FAFFA6911AB9301 //log(1/frcpa(1+ 16/256)
269 data8 0x3FB0EC139C5DA601 //log(1/frcpa(1+ 17/256)
270 data8 0x3FB1DBD2643D190B //log(1/frcpa(1+ 18/256)
271 data8 0x3FB2CC7284FE5F1C //log(1/frcpa(1+ 19/256)
272 data8 0x3FB3BDF5A7D1EE64 //log(1/frcpa(1+ 20/256)
273 data8 0x3FB4B05D7AA012E0 //log(1/frcpa(1+ 21/256)
274 data8 0x3FB580DB7CEB5702 //log(1/frcpa(1+ 22/256)
275 data8 0x3FB674F089365A7A //log(1/frcpa(1+ 23/256)
276 data8 0x3FB769EF2C6B568D //log(1/frcpa(1+ 24/256)
277 data8 0x3FB85FD927506A48 //log(1/frcpa(1+ 25/256)
278 data8 0x3FB9335E5D594989 //log(1/frcpa(1+ 26/256)
279 data8 0x3FBA2B0220C8E5F5 //log(1/frcpa(1+ 27/256)
280 data8 0x3FBB0004AC1A86AC //log(1/frcpa(1+ 28/256)
281 data8 0x3FBBF968769FCA11 //log(1/frcpa(1+ 29/256)
282 data8 0x3FBCCFEDBFEE13A8 //log(1/frcpa(1+ 30/256)
283 data8 0x3FBDA727638446A2 //log(1/frcpa(1+ 31/256)
284 data8 0x3FBEA3257FE10F7A //log(1/frcpa(1+ 32/256)
285 data8 0x3FBF7BE9FEDBFDE6 //log(1/frcpa(1+ 33/256)
286 data8 0x3FC02AB352FF25F4 //log(1/frcpa(1+ 34/256)
287 data8 0x3FC097CE579D204D //log(1/frcpa(1+ 35/256)
288 data8 0x3FC1178E8227E47C //log(1/frcpa(1+ 36/256)
289 data8 0x3FC185747DBECF34 //log(1/frcpa(1+ 37/256)
290 data8 0x3FC1F3B925F25D41 //log(1/frcpa(1+ 38/256)
291 data8 0x3FC2625D1E6DDF57 //log(1/frcpa(1+ 39/256)
292 data8 0x3FC2D1610C86813A //log(1/frcpa(1+ 40/256)
293 data8 0x3FC340C59741142E //log(1/frcpa(1+ 41/256)
294 data8 0x3FC3B08B6757F2A9 //log(1/frcpa(1+ 42/256)
295 data8 0x3FC40DFB08378003 //log(1/frcpa(1+ 43/256)
296 data8 0x3FC47E74E8CA5F7C //log(1/frcpa(1+ 44/256)
297 data8 0x3FC4EF51F6466DE4 //log(1/frcpa(1+ 45/256)
298 data8 0x3FC56092E02BA516 //log(1/frcpa(1+ 46/256)
299 data8 0x3FC5D23857CD74D5 //log(1/frcpa(1+ 47/256)
300 data8 0x3FC6313A37335D76 //log(1/frcpa(1+ 48/256)
301 data8 0x3FC6A399DABBD383 //log(1/frcpa(1+ 49/256)
302 data8 0x3FC70337DD3CE41B //log(1/frcpa(1+ 50/256)
303 data8 0x3FC77654128F6127 //log(1/frcpa(1+ 51/256)
304 data8 0x3FC7E9D82A0B022D //log(1/frcpa(1+ 52/256)
305 data8 0x3FC84A6B759F512F //log(1/frcpa(1+ 53/256)
306 data8 0x3FC8AB47D5F5A310 //log(1/frcpa(1+ 54/256)
307 data8 0x3FC91FE49096581B //log(1/frcpa(1+ 55/256)
308 data8 0x3FC981634011AA75 //log(1/frcpa(1+ 56/256)
309 data8 0x3FC9F6C407089664 //log(1/frcpa(1+ 57/256)
310 data8 0x3FCA58E729348F43 //log(1/frcpa(1+ 58/256)
311 data8 0x3FCABB55C31693AD //log(1/frcpa(1+ 59/256)
312 data8 0x3FCB1E104919EFD0 //log(1/frcpa(1+ 60/256)
313 data8 0x3FCB94EE93E367CB //log(1/frcpa(1+ 61/256)
314 data8 0x3FCBF851C067555F //log(1/frcpa(1+ 62/256)
315 data8 0x3FCC5C0254BF23A6 //log(1/frcpa(1+ 63/256)
316 data8 0x3FCCC000C9DB3C52 //log(1/frcpa(1+ 64/256)
317 data8 0x3FCD244D99C85674 //log(1/frcpa(1+ 65/256)
318 data8 0x3FCD88E93FB2F450 //log(1/frcpa(1+ 66/256)
319 data8 0x3FCDEDD437EAEF01 //log(1/frcpa(1+ 67/256)
320 data8 0x3FCE530EFFE71012 //log(1/frcpa(1+ 68/256)
321 data8 0x3FCEB89A1648B971 //log(1/frcpa(1+ 69/256)
322 data8 0x3FCF1E75FADF9BDE //log(1/frcpa(1+ 70/256)
323 data8 0x3FCF84A32EAD7C35 //log(1/frcpa(1+ 71/256)
324 data8 0x3FCFEB2233EA07CD //log(1/frcpa(1+ 72/256)
325 data8 0x3FD028F9C7035C1C //log(1/frcpa(1+ 73/256)
326 data8 0x3FD05C8BE0D9635A //log(1/frcpa(1+ 74/256)
327 data8 0x3FD085EB8F8AE797 //log(1/frcpa(1+ 75/256)
328 data8 0x3FD0B9C8E32D1911 //log(1/frcpa(1+ 76/256)
329 data8 0x3FD0EDD060B78081 //log(1/frcpa(1+ 77/256)
330 data8 0x3FD122024CF0063F //log(1/frcpa(1+ 78/256)
331 data8 0x3FD14BE2927AECD4 //log(1/frcpa(1+ 79/256)
332 data8 0x3FD180618EF18ADF //log(1/frcpa(1+ 80/256)
333 data8 0x3FD1B50BBE2FC63B //log(1/frcpa(1+ 81/256)
334 data8 0x3FD1DF4CC7CF242D //log(1/frcpa(1+ 82/256)
335 data8 0x3FD214456D0EB8D4 //log(1/frcpa(1+ 83/256)
336 data8 0x3FD23EC5991EBA49 //log(1/frcpa(1+ 84/256)
337 data8 0x3FD2740D9F870AFB //log(1/frcpa(1+ 85/256)
338 data8 0x3FD29ECDABCDFA04 //log(1/frcpa(1+ 86/256)
339 data8 0x3FD2D46602ADCCEE //log(1/frcpa(1+ 87/256)
340 data8 0x3FD2FF66B04EA9D4 //log(1/frcpa(1+ 88/256)
341 data8 0x3FD335504B355A37 //log(1/frcpa(1+ 89/256)
342 data8 0x3FD360925EC44F5D //log(1/frcpa(1+ 90/256)
343 data8 0x3FD38BF1C3337E75 //log(1/frcpa(1+ 91/256)
344 data8 0x3FD3C25277333184 //log(1/frcpa(1+ 92/256)
345 data8 0x3FD3EDF463C1683E //log(1/frcpa(1+ 93/256)
346 data8 0x3FD419B423D5E8C7 //log(1/frcpa(1+ 94/256)
347 data8 0x3FD44591E0539F49 //log(1/frcpa(1+ 95/256)
348 data8 0x3FD47C9175B6F0AD //log(1/frcpa(1+ 96/256)
349 data8 0x3FD4A8B341552B09 //log(1/frcpa(1+ 97/256)
350 data8 0x3FD4D4F3908901A0 //log(1/frcpa(1+ 98/256)
351 data8 0x3FD501528DA1F968 //log(1/frcpa(1+ 99/256)
352 data8 0x3FD52DD06347D4F6 //log(1/frcpa(1+ 100/256)
353 data8 0x3FD55A6D3C7B8A8A //log(1/frcpa(1+ 101/256)
354 data8 0x3FD5925D2B112A59 //log(1/frcpa(1+ 102/256)
355 data8 0x3FD5BF406B543DB2 //log(1/frcpa(1+ 103/256)
356 data8 0x3FD5EC433D5C35AE //log(1/frcpa(1+ 104/256)
357 data8 0x3FD61965CDB02C1F //log(1/frcpa(1+ 105/256)
358 data8 0x3FD646A84935B2A2 //log(1/frcpa(1+ 106/256)
359 data8 0x3FD6740ADD31DE94 //log(1/frcpa(1+ 107/256)
360 data8 0x3FD6A18DB74A58C5 //log(1/frcpa(1+ 108/256)
361 data8 0x3FD6CF31058670EC //log(1/frcpa(1+ 109/256)
362 data8 0x3FD6F180E852F0BA //log(1/frcpa(1+ 110/256)
363 data8 0x3FD71F5D71B894F0 //log(1/frcpa(1+ 111/256)
364 data8 0x3FD74D5AEFD66D5C //log(1/frcpa(1+ 112/256)
365 data8 0x3FD77B79922BD37E //log(1/frcpa(1+ 113/256)
366 data8 0x3FD7A9B9889F19E2 //log(1/frcpa(1+ 114/256)
367 data8 0x3FD7D81B037EB6A6 //log(1/frcpa(1+ 115/256)
368 data8 0x3FD8069E33827231 //log(1/frcpa(1+ 116/256)
369 data8 0x3FD82996D3EF8BCB //log(1/frcpa(1+ 117/256)
370 data8 0x3FD85855776DCBFB //log(1/frcpa(1+ 118/256)
371 data8 0x3FD8873658327CCF //log(1/frcpa(1+ 119/256)
372 data8 0x3FD8AA75973AB8CF //log(1/frcpa(1+ 120/256)
373 data8 0x3FD8D992DC8824E5 //log(1/frcpa(1+ 121/256)
374 data8 0x3FD908D2EA7D9512 //log(1/frcpa(1+ 122/256)
375 data8 0x3FD92C59E79C0E56 //log(1/frcpa(1+ 123/256)
376 data8 0x3FD95BD750EE3ED3 //log(1/frcpa(1+ 124/256)
377 data8 0x3FD98B7811A3EE5B //log(1/frcpa(1+ 125/256)
378 data8 0x3FD9AF47F33D406C //log(1/frcpa(1+ 126/256)
379 data8 0x3FD9DF270C1914A8 //log(1/frcpa(1+ 127/256)
380 data8 0x3FDA0325ED14FDA4 //log(1/frcpa(1+ 128/256)
381 data8 0x3FDA33440224FA79 //log(1/frcpa(1+ 129/256)
382 data8 0x3FDA57725E80C383 //log(1/frcpa(1+ 130/256)
383 data8 0x3FDA87D0165DD199 //log(1/frcpa(1+ 131/256)
384 data8 0x3FDAAC2E6C03F896 //log(1/frcpa(1+ 132/256)
385 data8 0x3FDADCCC6FDF6A81 //log(1/frcpa(1+ 133/256)
386 data8 0x3FDB015B3EB1E790 //log(1/frcpa(1+ 134/256)
387 data8 0x3FDB323A3A635948 //log(1/frcpa(1+ 135/256)
388 data8 0x3FDB56FA04462909 //log(1/frcpa(1+ 136/256)
389 data8 0x3FDB881AA659BC93 //log(1/frcpa(1+ 137/256)
390 data8 0x3FDBAD0BEF3DB165 //log(1/frcpa(1+ 138/256)
391 data8 0x3FDBD21297781C2F //log(1/frcpa(1+ 139/256)
392 data8 0x3FDC039236F08819 //log(1/frcpa(1+ 140/256)
393 data8 0x3FDC28CB1E4D32FD //log(1/frcpa(1+ 141/256)
394 data8 0x3FDC4E19B84723C2 //log(1/frcpa(1+ 142/256)
395 data8 0x3FDC7FF9C74554C9 //log(1/frcpa(1+ 143/256)
396 data8 0x3FDCA57B64E9DB05 //log(1/frcpa(1+ 144/256)
397 data8 0x3FDCCB130A5CEBB0 //log(1/frcpa(1+ 145/256)
398 data8 0x3FDCF0C0D18F326F //log(1/frcpa(1+ 146/256)
399 data8 0x3FDD232075B5A201 //log(1/frcpa(1+ 147/256)
400 data8 0x3FDD490246DEFA6B //log(1/frcpa(1+ 148/256)
401 data8 0x3FDD6EFA918D25CD //log(1/frcpa(1+ 149/256)
402 data8 0x3FDD9509707AE52F //log(1/frcpa(1+ 150/256)
403 data8 0x3FDDBB2EFE92C554 //log(1/frcpa(1+ 151/256)
404 data8 0x3FDDEE2F3445E4AF //log(1/frcpa(1+ 152/256)
405 data8 0x3FDE148A1A2726CE //log(1/frcpa(1+ 153/256)
406 data8 0x3FDE3AFC0A49FF40 //log(1/frcpa(1+ 154/256)
407 data8 0x3FDE6185206D516E //log(1/frcpa(1+ 155/256)
408 data8 0x3FDE882578823D52 //log(1/frcpa(1+ 156/256)
409 data8 0x3FDEAEDD2EAC990C //log(1/frcpa(1+ 157/256)
410 data8 0x3FDED5AC5F436BE3 //log(1/frcpa(1+ 158/256)
411 data8 0x3FDEFC9326D16AB9 //log(1/frcpa(1+ 159/256)
412 data8 0x3FDF2391A2157600 //log(1/frcpa(1+ 160/256)
413 data8 0x3FDF4AA7EE03192D //log(1/frcpa(1+ 161/256)
414 data8 0x3FDF71D627C30BB0 //log(1/frcpa(1+ 162/256)
415 data8 0x3FDF991C6CB3B379 //log(1/frcpa(1+ 163/256)
416 data8 0x3FDFC07ADA69A910 //log(1/frcpa(1+ 164/256)
417 data8 0x3FDFE7F18EB03D3E //log(1/frcpa(1+ 165/256)
418 data8 0x3FE007C053C5002E //log(1/frcpa(1+ 166/256)
419 data8 0x3FE01B942198A5A1 //log(1/frcpa(1+ 167/256)
420 data8 0x3FE02F74400C64EB //log(1/frcpa(1+ 168/256)
421 data8 0x3FE04360BE7603AD //log(1/frcpa(1+ 169/256)
422 data8 0x3FE05759AC47FE34 //log(1/frcpa(1+ 170/256)
423 data8 0x3FE06B5F1911CF52 //log(1/frcpa(1+ 171/256)
424 data8 0x3FE078BF0533C568 //log(1/frcpa(1+ 172/256)
425 data8 0x3FE08CD9687E7B0E //log(1/frcpa(1+ 173/256)
426 data8 0x3FE0A10074CF9019 //log(1/frcpa(1+ 174/256)
427 data8 0x3FE0B5343A234477 //log(1/frcpa(1+ 175/256)
428 data8 0x3FE0C974C89431CE //log(1/frcpa(1+ 176/256)
429 data8 0x3FE0DDC2305B9886 //log(1/frcpa(1+ 177/256)
430 data8 0x3FE0EB524BAFC918 //log(1/frcpa(1+ 178/256)
431 data8 0x3FE0FFB54213A476 //log(1/frcpa(1+ 179/256)
432 data8 0x3FE114253DA97D9F //log(1/frcpa(1+ 180/256)
433 data8 0x3FE128A24F1D9AFF //log(1/frcpa(1+ 181/256)
434 data8 0x3FE1365252BF0865 //log(1/frcpa(1+ 182/256)
435 data8 0x3FE14AE558B4A92D //log(1/frcpa(1+ 183/256)
436 data8 0x3FE15F85A19C765B //log(1/frcpa(1+ 184/256)
437 data8 0x3FE16D4D38C119FA //log(1/frcpa(1+ 185/256)
438 data8 0x3FE18203C20DD133 //log(1/frcpa(1+ 186/256)
439 data8 0x3FE196C7BC4B1F3B //log(1/frcpa(1+ 187/256)
440 data8 0x3FE1A4A738B7A33C //log(1/frcpa(1+ 188/256)
441 data8 0x3FE1B981C0C9653D //log(1/frcpa(1+ 189/256)
442 data8 0x3FE1CE69E8BB106B //log(1/frcpa(1+ 190/256)
443 data8 0x3FE1DC619DE06944 //log(1/frcpa(1+ 191/256)
444 data8 0x3FE1F160A2AD0DA4 //log(1/frcpa(1+ 192/256)
445 data8 0x3FE2066D7740737E //log(1/frcpa(1+ 193/256)
446 data8 0x3FE2147DBA47A394 //log(1/frcpa(1+ 194/256)
447 data8 0x3FE229A1BC5EBAC3 //log(1/frcpa(1+ 195/256)
448 data8 0x3FE237C1841A502E //log(1/frcpa(1+ 196/256)
449 data8 0x3FE24CFCE6F80D9A //log(1/frcpa(1+ 197/256)
450 data8 0x3FE25B2C55CD5762 //log(1/frcpa(1+ 198/256)
451 data8 0x3FE2707F4D5F7C41 //log(1/frcpa(1+ 199/256)
452 data8 0x3FE285E0842CA384 //log(1/frcpa(1+ 200/256)
453 data8 0x3FE294294708B773 //log(1/frcpa(1+ 201/256)
454 data8 0x3FE2A9A2670AFF0C //log(1/frcpa(1+ 202/256)
455 data8 0x3FE2B7FB2C8D1CC1 //log(1/frcpa(1+ 203/256)
456 data8 0x3FE2C65A6395F5F5 //log(1/frcpa(1+ 204/256)
457 data8 0x3FE2DBF557B0DF43 //log(1/frcpa(1+ 205/256)
458 data8 0x3FE2EA64C3F97655 //log(1/frcpa(1+ 206/256)
459 data8 0x3FE3001823684D73 //log(1/frcpa(1+ 207/256)
460 data8 0x3FE30E97E9A8B5CD //log(1/frcpa(1+ 208/256)
461 data8 0x3FE32463EBDD34EA //log(1/frcpa(1+ 209/256)
462 data8 0x3FE332F4314AD796 //log(1/frcpa(1+ 210/256)
463 data8 0x3FE348D90E7464D0 //log(1/frcpa(1+ 211/256)
464 data8 0x3FE35779F8C43D6E //log(1/frcpa(1+ 212/256)
465 data8 0x3FE36621961A6A99 //log(1/frcpa(1+ 213/256)
466 data8 0x3FE37C299F3C366A //log(1/frcpa(1+ 214/256)
467 data8 0x3FE38AE2171976E7 //log(1/frcpa(1+ 215/256)
468 data8 0x3FE399A157A603E7 //log(1/frcpa(1+ 216/256)
469 data8 0x3FE3AFCCFE77B9D1 //log(1/frcpa(1+ 217/256)
470 data8 0x3FE3BE9D503533B5 //log(1/frcpa(1+ 218/256)
471 data8 0x3FE3CD7480B4A8A3 //log(1/frcpa(1+ 219/256)
472 data8 0x3FE3E3C43918F76C //log(1/frcpa(1+ 220/256)
473 data8 0x3FE3F2ACB27ED6C7 //log(1/frcpa(1+ 221/256)
474 data8 0x3FE4019C2125CA93 //log(1/frcpa(1+ 222/256)
475 data8 0x3FE4181061389722 //log(1/frcpa(1+ 223/256)
476 data8 0x3FE42711518DF545 //log(1/frcpa(1+ 224/256)
477 data8 0x3FE436194E12B6BF //log(1/frcpa(1+ 225/256)
478 data8 0x3FE445285D68EA69 //log(1/frcpa(1+ 226/256)
479 data8 0x3FE45BCC464C893A //log(1/frcpa(1+ 227/256)
480 data8 0x3FE46AED21F117FC //log(1/frcpa(1+ 228/256)
481 data8 0x3FE47A1527E8A2D3 //log(1/frcpa(1+ 229/256)
482 data8 0x3FE489445EFFFCCC //log(1/frcpa(1+ 230/256)
483 data8 0x3FE4A018BCB69835 //log(1/frcpa(1+ 231/256)
484 data8 0x3FE4AF5A0C9D65D7 //log(1/frcpa(1+ 232/256)
485 data8 0x3FE4BEA2A5BDBE87 //log(1/frcpa(1+ 233/256)
486 data8 0x3FE4CDF28F10AC46 //log(1/frcpa(1+ 234/256)
487 data8 0x3FE4DD49CF994058 //log(1/frcpa(1+ 235/256)
488 data8 0x3FE4ECA86E64A684 //log(1/frcpa(1+ 236/256)
489 data8 0x3FE503C43CD8EB68 //log(1/frcpa(1+ 237/256)
490 data8 0x3FE513356667FC57 //log(1/frcpa(1+ 238/256)
491 data8 0x3FE522AE0738A3D8 //log(1/frcpa(1+ 239/256)
492 data8 0x3FE5322E26867857 //log(1/frcpa(1+ 240/256)
493 data8 0x3FE541B5CB979809 //log(1/frcpa(1+ 241/256)
494 data8 0x3FE55144FDBCBD62 //log(1/frcpa(1+ 242/256)
495 data8 0x3FE560DBC45153C7 //log(1/frcpa(1+ 243/256)
496 data8 0x3FE5707A26BB8C66 //log(1/frcpa(1+ 244/256)
497 data8 0x3FE587F60ED5B900 //log(1/frcpa(1+ 245/256)
498 data8 0x3FE597A7977C8F31 //log(1/frcpa(1+ 246/256)
499 data8 0x3FE5A760D634BB8B //log(1/frcpa(1+ 247/256)
500 data8 0x3FE5B721D295F10F //log(1/frcpa(1+ 248/256)
501 data8 0x3FE5C6EA94431EF9 //log(1/frcpa(1+ 249/256)
502 data8 0x3FE5D6BB22EA86F6 //log(1/frcpa(1+ 250/256)
503 data8 0x3FE5E6938645D390 //log(1/frcpa(1+ 251/256)
504 data8 0x3FE5F673C61A2ED2 //log(1/frcpa(1+ 252/256)
505 data8 0x3FE6065BEA385926 //log(1/frcpa(1+ 253/256)
506 data8 0x3FE6164BFA7CC06B //log(1/frcpa(1+ 254/256)
507 data8 0x3FE62643FECF9743 //log(1/frcpa(1+ 255/256)
508 LOCAL_OBJECT_END(log_table_3)
511 .section .text
512 GLOBAL_LIBM_ENTRY(acoshf)
514 { .mfi
515 getf.exp acosh_GR_f8 = f8
516 fclass.m p6,p0 = f8, 0xc3 // Test for x = NaN
517 mov log_GR_comp2 = 0x10032
518 }
519 { .mfi
520 addl NR_table_address = @ltoff(log_table_1), gp
521 fms.s1 log_y = f8, f8, f1 // y = x^2-1
522 nop.i 0
523 }
524 ;;
526 { .mfi
527 getf.sig acosh_GR_f8_sig = f8
528 fclass.m p11,p0 = f8, 0x21 // Test for x=+inf
529 mov log_GR_exp_17_ones = 0x1ffff
530 }
531 { .mfi
532 ld8 NR_table_address = [NR_table_address]
533 fms.s1 log_w = f8,f1,f1 // w = x - 1
534 nop.i 0
535 }
536 ;;
538 { .mfi
539 nop.m 0
540 fcmp.lt.s1 p7,p8 = f8, f1 // Test for x<1.0
541 addl log_GR_comp = 0x10020C,r0 // Upper 21 bits of signif of 1.0005
542 }
543 { .mfb
544 mov log_GR_exp_16_ones = 0xffff //BIAS
545 (p6) fma.s.s0 f8 = f8,f1,f0 // quietize nan result if x=nan
546 (p6) br.ret.spnt b0 // Exit for x=nan
547 }
548 ;;
550 { .mfb
551 //get second table address
552 adds log_table_address2 = 0x20, NR_table_address
553 fcmp.eq.s1 p10,p0 = f8, f1 // Test for x=+1.0
554 (p11) br.ret.spnt b0 // Exit for x=+inf
555 }
556 ;;
558 { .mfi
559 ldfpd NR1,NR2 = [log_table_address2],16
560 frsqrta.s1 log_y_rs,p0 = log_y // z=1/sqrt(y)
561 nop.i 0
562 }
563 { .mfb
564 nop.m 0
565 fma.s1 log_arg = f8,f1,f8
566 (p7) br.cond.spnt ACOSH_LESS_ONE // Branch if path 7, x < 1.0
567 }
568 ;;
570 { .mfi
571 ldfe log_C2 = [log_table_address2],16
572 (p8) fcmp.eq.s0 p6,p0 = f8, f0 // Dummy op sets denorm flag if unorm>=1.0
573 nop.i 0
574 }
575 { .mfb
576 (p8) cmp.le.unc p13,p0 = log_GR_comp2,acosh_GR_f8
577 nop.f 0
578 (p13) br.cond.spnt LOG_COMMON1 // Branch if path 4, x >= 2^51
579 }
580 ;;
582 { .mfi
583 ldfe log_C1 = [log_table_address2],16
584 (p10) fmerge.s f8 = f0, f0 // Return 0 if x=1.0
585 shr.u acosh_GR_f8_sig = acosh_GR_f8_sig,43
586 }
587 { .mib
588 cmp.eq p14,p0 = log_GR_exp_16_ones,acosh_GR_f8
589 nop.i 0
590 (p10) br.ret.spnt b0 // Exit for x=1.0
591 }
592 ;;
594 { .mfi
595 ldfe log_C0 = [log_table_address2],16
596 frsqrta.s1 acosh_w_rs,p0 = log_w // t=1/sqrt(w)
597 nop.i 0
598 }
599 { .mfb
600 (p14) cmp.lt.unc p15,p0 = acosh_GR_f8_sig,log_GR_comp
601 nop.f 0
602 (p15) br.cond.spnt ACOSH_NEAR_ONE // Branch if path 2, 1.0 < x < 1.0005
603 }
604 ;;
606 // Here is main path, 1.0005 <= x < 2^51
607 /////////////// The first iteration //////////////////////////////////
608 { .mfi
609 ldfpd log_P3,log_P2 = [NR_table_address],16
610 fma.s1 log_y_rs_iter = log_y_rs,log_y,f0 // y*z
611 nop.i 0
612 }
613 ;;
615 { .mfi
616 ldfpd log_P1,log2 = [NR_table_address],16
617 fnma.s1 log_y_rs_iter2 = log_y_rs_iter,log_y_rs,NR2 // 3-(y*z)*z
618 nop.i 0
619 }
620 { .mfi
621 nop.m 0
622 fma.s1 log_y_rs_iter1 = log_y_rs,NR1,f0 // 0.5*z
623 nop.i 0
624 }
625 ;;
627 { .mfi
628 nop.m 0
629 // (0.5*z)*(3-(y*z)*z)
630 fma.s1 log_y_rs_iter = log_y_rs_iter1,log_y_rs_iter2,f0
631 nop.i 0
632 }
633 { .mfi
634 nop.m 0
635 // (0.5*z)*(3-(y*z)*z)
636 fma.s1 log_arg_early = log_y_rs_iter1,log_y_rs_iter2,f0
637 nop.i 0
638 }
639 ;;
641 /////////////////////////// The second iteration /////////////////////////////
642 { .mfi
643 nop.m 0
644 fma.s1 log_y_rs = log_y_rs_iter,log_y,f0
645 nop.i 0
646 }
647 { .mfi
648 nop.m 0
649 fma.s1 log_y_rs_iter1 = log_y_rs_iter,NR1,f0
650 nop.i 0
651 }
652 ;;
654 { .mfi
655 nop.m 0
656 fma.s1 log_arg_early = log_arg_early,log_y,f8
657 nop.i 0
658 }
659 ;;
661 { .mfi
662 nop.m 0
663 fnma.s1 log_y_rs = log_y_rs,log_y_rs_iter,NR2
664 nop.i 0
665 }
666 { .mfi
667 nop.m 0
668 fma.s1 log_y_rs_iter1 = log_y_rs_iter1,log_y,f0
669 nop.i 0
670 }
671 ;;
673 { .mfi
674 nop.m 0
675 frcpa.s1 log_C,p0 = f1,log_arg_early
676 nop.i 0
677 }
678 ;;
680 { .mfi
681 getf.exp log_GR_signexp_f8 = log_arg_early
682 nop.f 0
683 nop.i 0
684 }
685 ;;
687 { .mfi
688 getf.sig log_GR_significand_f8 = log_arg_early
689 fma.s1 log_arg = log_y_rs_iter1,log_y_rs,f8 // (0.5*z)*(3-(y*z)*z)
690 adds log_table_address3 = 0x40, NR_table_address
691 }
692 ;;
694 ///////////////////////////////// The end NR iterations /////////////////////
696 { .mmi
697 //significant bit destruction
698 and log_GR_exp_f8 = log_GR_signexp_f8, log_GR_exp_17_ones
699 ;;
700 //BIAS subtraction
701 sub log_GR_true_exp_f8 = log_GR_exp_f8, log_GR_exp_16_ones
702 nop.i 0
703 }
704 ;;
706 { .mfi
707 setf.sig log_int_Nfloat = log_GR_true_exp_f8
708 fms.s1 log_r = log_C,log_arg,f1 // C = frcpa(x); r = C * x - 1
709 extr.u log_GR_index = log_GR_significand_f8,55,8 //Extract 8 bits
710 }
711 ;;
713 { .mmi
714 //pre-index*8 + index
715 shladd log_table_address3 = log_GR_index,3,log_table_address3
716 ;;
717 ldfd log_T = [log_table_address3]
718 nop.i 0
719 }
720 ;;
722 { .mfi
723 nop.m 0
724 fma.s1 log_rsq = log_r, log_r, f0 //r^2
725 nop.i 0
726 }
727 { .mfi
728 nop.m 0
729 fma.s1 log_rp_p32 = log_P3, log_r, log_P2 //P3*r + P2
730 nop.i 0
731 }
732 ;;
734 { .mfi
735 nop.m 0
736 fma.s1 log_rp_p10 = log_P1, log_r, f1 //P1*r + 1.0
737 nop.i 0
738 }
739 ;;
741 { .mfi
742 nop.m 0
743 //convert N to the floating-point format log_Nfloat
744 fcvt.xf log_Nfloat = log_int_Nfloat
745 nop.i 0
746 }
747 ;;
749 { .mfi
750 nop.m 0
751 //(P3*r + P2)*r^2 + P1*r + 1.0
752 fma.s1 log_rp_p2 = log_rp_p32, log_rsq, log_rp_p10
753 nop.i 0
754 }
755 ;;
757 { .mfi
758 nop.m 0
759 fma.s1 log_T_plus_Nlog2 = log_Nfloat,log2,log_T //N*log2 + T
760 nop.i 0
761 }
762 ;;
764 { .mfb
765 nop.m 0
766 fma.s.s0 f8 = log_rp_p2,log_r,log_T_plus_Nlog2
767 br.ret.sptk b0 // Exit main path, path 3: 1.0005 <= x < 2^51
768 }
769 ;;
771 // Here if path 2, 1.0 < x < 1.0005
772 ACOSH_NEAR_ONE:
773 // The first NR iteration
774 { .mfi
775 nop.m 0
776 fma.s1 acosh_w_iter1 = acosh_w_rs,log_w,f0 //t*w
777 nop.i 0
778 }
779 { .mfi
780 nop.m 0
781 fma.s1 acosh_w_1 = f8,log_C2,log_C1 //x*C2 + C1
782 nop.i 0
783 }
784 ;;
786 { .mfi
787 nop.m 0
788 fma.s1 acosh_w_iter2 = acosh_w_rs,NR1,f0 //t*0.5
789 nop.i 0
790 }
791 { .mfi
792 nop.m 0
793 fnma.s1 acosh_w_iter1 = acosh_w_iter1,acosh_w_rs,NR2 //3-t*t*w
794 nop.i 0
795 }
796 ;;
798 { .mfi
799 nop.m 0
800 //(3-t*t*w)*t*0.5
801 fma.s1 acosh_w_iter2 = acosh_w_iter2,acosh_w_iter1,f0
802 nop.i 0
803 }
804 { .mfi
805 nop.m 0
806 fma.s1 acosh_w_1 = acosh_w_1,log_w,log_C0 //(x*C2 + C1)*(x-1) + C0
807 nop.i 0
808 }
809 ;;
811 // The second NR iteration
812 { .mfi
813 nop.m 0
814 fma.s1 acosh_w_rs = acosh_w_iter2,log_w,f0 //t*w
815 nop.i 0
816 }
817 ;;
819 { .mfi
820 nop.m 0
821 fnma.s1 acosh_w_iter1 = acosh_w_iter2,acosh_w_rs,NR2
822 nop.i 0
823 }
824 { .mfi
825 nop.m 0
826 fma.s1 acosh_w_iter2 = acosh_w_iter2,NR1,f0
827 nop.i 0
828 }
829 ;;
831 { .mfi
832 nop.m 0
833 fma.s1 acosh_w_iter2 = acosh_w_iter2,acosh_w_iter1,f0
834 nop.i 0
835 }
836 ;;
838 { .mfi
839 nop.m 0
840 fma.s1 acosh_w_sqrt = acosh_w_iter2,log_w,f0
841 nop.i 0
842 }
843 ;;
845 { .mfb
846 nop.m 0
847 fma.s.s0 f8 = acosh_w_1,acosh_w_sqrt,f0
848 br.ret.sptk b0 // Exit path 2, 1.0 < x < 1.0005
849 }
850 ;;
852 // Here if path 4, x >= 2^51
853 LOG_COMMON1:
854 { .mfi
855 ldfpd log_P3,log_P2 = [NR_table_address],16
856 frcpa.s1 log_C,p0 = f1,log_arg
857 nop.i 0
858 }
859 ;;
861 { .mmi
862 getf.exp log_GR_signexp_f8 = log_arg
863 ldfpd log_P1,log2 = [NR_table_address],16
864 nop.i 0
865 }
866 ;;
868 { .mmi
869 getf.sig log_GR_significand_f8 = log_arg
870 nop.m 0
871 nop.i 0
872 }
873 ;;
875 { .mfi
876 adds log_table_address3 = 0x40, NR_table_address
877 nop.f 0
878 //significant bit destruction
879 and log_GR_exp_f8 = log_GR_signexp_f8, log_GR_exp_17_ones
880 }
881 ;;
883 { .mmf
884 nop.m 0
885 //BIAS subtraction
886 sub log_GR_true_exp_f8 = log_GR_exp_f8, log_GR_exp_16_ones
887 fms.s1 log_r = log_C,log_arg,f1 // C = frcpa(x); r = C * x - 1
888 }
889 ;;
891 { .mfi
892 setf.sig log_int_Nfloat = log_GR_true_exp_f8
893 nop.f 0
894 extr.u log_GR_index = log_GR_significand_f8,55,8 //Extract 8 bits
895 }
896 ;;
898 { .mmi
899 //pre-index*8 + index
900 shladd log_table_address3 = log_GR_index,3,log_table_address3
901 ;;
902 ldfd log_T = [log_table_address3]
903 nop.i 0
904 }
905 ;;
907 { .mfi
908 nop.m 0
909 fma.s1 log_rsq = log_r, log_r, f0 //r^2
910 nop.i 0
911 }
912 { .mfi
913 nop.m 0
914 fma.s1 log_rp_p32 = log_P3, log_r, log_P2 //P3*r + P2
915 nop.i 0
916 }
917 ;;
919 { .mfi
920 nop.m 0
921 fma.s1 log_rp_p10 = log_P1, log_r, f1 //P1*r + 1.0
922 nop.i 0
923 }
924 ;;
926 { .mfi
927 nop.m 0
928 //convert N to the floating-point format log_Nfloat
929 fcvt.xf log_Nfloat = log_int_Nfloat
930 nop.i 0
931 }
932 { .mfi
933 nop.m 0
934 fma.s1 log_rp_p2 = log_rp_p32, log_rsq, log_rp_p10
935 nop.i 0
936 }
937 ;;
939 { .mfi
940 nop.m 0
941 fma.s1 log_T_plus_Nlog2 = log_Nfloat,log2,log_T //N*log2 + T
942 nop.i 0
943 }
944 ;;
946 { .mfb
947 nop.m 0
948 fma.s.s0 f8 = log_rp_p2,log_r,log_T_plus_Nlog2
949 br.ret.sptk b0 // Exit path 4, x >= 2^51
950 }
951 ;;
953 // Here if path 7, x < 1.0
954 ACOSH_LESS_ONE:
955 { .mfi
956 alloc r32 = ar.pfs,1,3,4,0
957 fmerge.s f10 = f8,f8
958 nop.i 0
959 }
960 ;;
962 { .mfb
963 mov acosh_GR_tag = 137
964 frcpa.s0 f8,p0 = f0,f0
965 br.cond.sptk __libm_error_region
966 }
967 ;;
969 GLOBAL_LIBM_END(acoshf)
972 LOCAL_LIBM_ENTRY(__libm_error_region)
973 .prologue
975 { .mfi
976 add GR_Parameter_Y=-32,sp // Parameter 2 value
977 nop.f 0
978 .save ar.pfs,GR_SAVE_PFS
979 mov GR_SAVE_PFS=ar.pfs // Save ar.pfs
980 }
981 { .mfi
982 .fframe 64
983 add sp=-64,sp // Create new stack
984 nop.f 0
985 mov GR_SAVE_GP=gp // Save gp
986 };;
988 { .mmi
989 stfs [GR_Parameter_Y] = f1,16 // STORE Parameter 2 on stack
990 add GR_Parameter_X = 16,sp // Parameter 1 address
991 .save b0, GR_SAVE_B0
992 mov GR_SAVE_B0=b0 // Save b0
993 };;
995 .body
996 { .mib
997 stfs [GR_Parameter_X] = f10 // STORE Parameter 1 on stack
998 add GR_Parameter_RESULT = 0,GR_Parameter_Y // Parameter 3 address
999 nop.b 0
1000 }
1001 { .mib
1002 stfs [GR_Parameter_Y] = f8 // STORE Parameter 3 on stack
1003 add GR_Parameter_Y = -16,GR_Parameter_Y
1004 br.call.sptk b0=__libm_error_support# // Call error handling function
1005 };;
1007 { .mmi
1008 add GR_Parameter_RESULT = 48,sp
1009 nop.m 0
1010 nop.i 0
1011 };;
1013 { .mmi
1014 ldfs f8 = [GR_Parameter_RESULT] // Get return result off stack
1015 .restore sp
1016 add sp = 64,sp // Restore stack pointer
1017 mov b0 = GR_SAVE_B0 // Restore return address
1018 };;
1020 { .mib
1021 mov gp = GR_SAVE_GP // Restore gp
1022 mov ar.pfs = GR_SAVE_PFS // Restore ar.pfs
1023 br.ret.sptk b0 // Return
1024 };;
1026 LOCAL_LIBM_END(__libm_error_region)
1029 .type __libm_error_support#,@function
1030 .global __libm_error_support#