1/* Function tanf vectorized with SSE4.
2   Copyright (C) 2021-2022 Free Software Foundation, Inc.
3   This file is part of the GNU C Library.
4
5   The GNU C Library is free software; you can redistribute it and/or
6   modify it under the terms of the GNU Lesser General Public
7   License as published by the Free Software Foundation; either
8   version 2.1 of the License, or (at your option) any later version.
9
10   The GNU C Library is distributed in the hope that it will be useful,
11   but WITHOUT ANY WARRANTY; without even the implied warranty of
12   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13   Lesser General Public License for more details.
14
15   You should have received a copy of the GNU Lesser General Public
16   License along with the GNU C Library; if not, see
17   https://www.gnu.org/licenses/.  */
18
19/*
20 * ALGORITHM DESCRIPTION:
21 *
22 *      1) Range reduction to [-Pi/4; +Pi/4] interval
23 *         a) Grab sign from source argument and save it.
24 *         b) Remove sign using AND 0x7fffffff operation
25 *         c) Getting octant Y by 2/Pi multiplication
26 *         d) Add "Right Shifter" (0x4B000000) value
27 *         e) Treat obtained value as integer for destination sign setting.
28 *            Shift first bit of this value to the last (sign) position (S << 31)
29 *         f) Change destination sign if source sign is negative
30 *            using XOR operation.
31 *         g) Subtract "Right Shifter" (0x4B000000) value
32 *         h) Subtract Y*(PI/2) from X argument, where PI/2 divided to 4 parts:
33 *            X = X - Y*PI1 - Y*PI2 - Y*PI3 - Y*PI4;
34 *      2) Rational polynomial approximation ( at [-Pi/4; +Pi/4] interval)
35 *         a) Calculate X^2 = X * X
36 *         b) Calculate 2 polynomials:
37 *            P = X * (P0 + X^2 * P1);
38 *            Q = Q0 + X^2 * (Q1 + x^2 * Q2);
39 *         c) Swap P and Q if first bit of obtained value after
40 *            Right Shifting is set to 1. Using And, Andnot & Or operations.
41 *         d) Divide R = P / Q;
42 *      3) Destination sign setting
43 *         a) Set shifted destination sign using XOR operation:
44 *            R = XOR( R, S );
45 *
46 */
47
48/* Offsets for data table __svml_stan_data_internal
49 */
50#define _sInvPI_uisa			0
51#define _sPI1_uisa			16
52#define _sPI2_uisa			32
53#define _sPI3_uisa			48
54#define _sPI2_ha_uisa			64
55#define _sPI3_ha_uisa			80
56#define Th_tbl_uisa			96
57#define Tl_tbl_uisa			224
58#define _sPC3_uisa			352
59#define _sPC5_uisa			368
60#define _sRangeReductionVal_uisa	384
61#define _sInvPi				400
62#define _sSignMask			416
63#define _sAbsMask			432
64#define _sRangeVal			448
65#define _sRShifter			464
66#define _sOne				480
67#define _sRangeReductionVal		496
68#define _sPI1				512
69#define _sPI2				528
70#define _sPI3				544
71#define _sPI4				560
72#define _sPI1_FMA			576
73#define _sPI2_FMA			592
74#define _sPI3_FMA			608
75#define _sP0				624
76#define _sP1				640
77#define _sQ0				656
78#define _sQ1				672
79#define _sQ2				688
80#define _sTwo				704
81#define _sCoeffs			720
82
83#include <sysdep.h>
84
85	.section .text.sse4, "ax", @progbits
86ENTRY(_ZGVbN4v_tanf_sse4)
87	subq	$232, %rsp
88	cfi_def_cfa_offset(240)
89	movaps	%xmm0, %xmm13
90	movups	_sAbsMask+__svml_stan_data_internal(%rip), %xmm12
91
92	/*
93	 * Legacy Code
94	 * Here HW FMA can be unavailable
95	 */
96	xorl	%eax, %eax
97	movaps	%xmm12, %xmm4
98	pxor	%xmm10, %xmm10
99	movups	_sInvPi+__svml_stan_data_internal(%rip), %xmm2
100	andps	%xmm13, %xmm4
101	mulps	%xmm4, %xmm2
102
103	/* Range reduction */
104	movaps	%xmm4, %xmm1
105
106	/*
107	 *
108	 * Main path (_LA_ and _EP_)
109	 *
110	 * Octant calculation
111	 */
112	movups	_sRShifter+__svml_stan_data_internal(%rip), %xmm3
113
114	/* Large values check */
115	movaps	%xmm4, %xmm11
116	movups	_sPI1+__svml_stan_data_internal(%rip), %xmm5
117	andnps	%xmm13, %xmm12
118	movups	_sPI2+__svml_stan_data_internal(%rip), %xmm6
119	addps	%xmm3, %xmm2
120	cmpnleps _sRangeReductionVal+__svml_stan_data_internal(%rip), %xmm11
121	movaps	%xmm2, %xmm8
122	movups	_sPI3+__svml_stan_data_internal(%rip), %xmm7
123	subps	%xmm3, %xmm8
124	movmskps %xmm11, %edx
125	movups	_sPI4+__svml_stan_data_internal(%rip), %xmm9
126	mulps	%xmm8, %xmm5
127	mulps	%xmm8, %xmm6
128	mulps	%xmm8, %xmm7
129	subps	%xmm5, %xmm1
130	mulps	%xmm8, %xmm9
131	subps	%xmm6, %xmm1
132	movups	_sQ2+__svml_stan_data_internal(%rip), %xmm15
133
134	/* Inversion mask and sign calculation */
135	movaps	%xmm2, %xmm5
136
137	/* Rational approximation */
138	movups	_sP1+__svml_stan_data_internal(%rip), %xmm14
139	pslld	$30, %xmm2
140	cmpneqps %xmm10, %xmm2
141	subps	%xmm7, %xmm1
142
143	/* Exchanged numerator and denominator if necessary */
144	movaps	%xmm2, %xmm0
145	movaps	%xmm2, %xmm10
146	pslld	$31, %xmm5
147	subps	%xmm9, %xmm1
148	movaps	%xmm1, %xmm3
149	pxor	%xmm12, %xmm5
150	mulps	%xmm1, %xmm3
151	mulps	%xmm3, %xmm15
152	mulps	%xmm3, %xmm14
153	addps	_sQ1+__svml_stan_data_internal(%rip), %xmm15
154	addps	_sP0+__svml_stan_data_internal(%rip), %xmm14
155	mulps	%xmm15, %xmm3
156	mulps	%xmm14, %xmm1
157	addps	_sQ0+__svml_stan_data_internal(%rip), %xmm3
158	andnps	%xmm1, %xmm0
159	andps	%xmm3, %xmm10
160	andps	%xmm2, %xmm1
161	andnps	%xmm3, %xmm2
162	orps	%xmm10, %xmm0
163	orps	%xmm2, %xmm1
164
165	/* Division */
166	divps	%xmm1, %xmm0
167
168	/* Sign setting */
169	pxor	%xmm5, %xmm0
170
171	/*
172	 *
173	 * End of main path (_LA_ and _EP_)
174	 */
175
176	testl	%edx, %edx
177
178	/* Go to auxilary branch */
179	jne	L(AUX_BRANCH)
180	# LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm4 xmm11 xmm12 xmm13
181
182	/* Return from auxilary branch
183	 * for out of main path inputs
184	 */
185
186L(AUX_BRANCH_RETURN):
187	testl	%eax, %eax
188
189	/* Go to special inputs processing branch */
190	jne	L(SPECIAL_VALUES_BRANCH)
191	# LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm13
192
193	/* Restore registers
194	 * and exit the function
195	 */
196
197L(EXIT):
198	addq	$232, %rsp
199	cfi_def_cfa_offset(8)
200	ret
201	cfi_def_cfa_offset(240)
202
203	/* Branch to process
204	 * special inputs
205	 */
206
207L(SPECIAL_VALUES_BRANCH):
208	movups	%xmm13, 32(%rsp)
209	movups	%xmm0, 48(%rsp)
210	# LOE rbx rbp r12 r13 r14 r15 eax xmm0
211
212	xorl	%edx, %edx
213	movq	%r12, 16(%rsp)
214	cfi_offset(12, -224)
215	movl	%edx, %r12d
216	movq	%r13, 8(%rsp)
217	cfi_offset(13, -232)
218	movl	%eax, %r13d
219	movq	%r14, (%rsp)
220	cfi_offset(14, -240)
221	# LOE rbx rbp r15 r12d r13d
222
223	/* Range mask
224	 * bits check
225	 */
226
227L(RANGEMASK_CHECK):
228	btl	%r12d, %r13d
229
230	/* Call scalar math function */
231	jc	L(SCALAR_MATH_CALL)
232	# LOE rbx rbp r15 r12d r13d
233
234	/* Special inputs
235	 * processing loop
236	 */
237
238L(SPECIAL_VALUES_LOOP):
239	incl	%r12d
240	cmpl	$4, %r12d
241
242	/* Check bits in range mask */
243	jl	L(RANGEMASK_CHECK)
244	# LOE rbx rbp r15 r12d r13d
245
246	movq	16(%rsp), %r12
247	cfi_restore(12)
248	movq	8(%rsp), %r13
249	cfi_restore(13)
250	movq	(%rsp), %r14
251	cfi_restore(14)
252	movups	48(%rsp), %xmm0
253
254	/* Go to exit */
255	jmp	L(EXIT)
256	cfi_offset(12, -224)
257	cfi_offset(13, -232)
258	cfi_offset(14, -240)
259	# LOE rbx rbp r12 r13 r14 r15 xmm0
260
261	/* Scalar math fucntion call
262	 * to process special input
263	 */
264
265L(SCALAR_MATH_CALL):
266	movl	%r12d, %r14d
267	movss	32(%rsp, %r14, 4), %xmm0
268	call	tanf@PLT
269	# LOE rbx rbp r14 r15 r12d r13d xmm0
270
271	movss	%xmm0, 48(%rsp, %r14, 4)
272
273	/* Process special inputs in loop */
274	jmp	L(SPECIAL_VALUES_LOOP)
275	cfi_restore(12)
276	cfi_restore(13)
277	cfi_restore(14)
278	# LOE rbx rbp r15 r12d r13d
279
280	/* Auxilary branch
281	 * for out of main path inputs
282	 */
283
284L(AUX_BRANCH):
285	movl	$2139095040, %eax
286
287	/*
288	 * Get the (2^a / 2pi) mod 1 values from the table.
289	 * Because doesn't have I-type gather, we need a trivial cast
290	 */
291	lea	__svml_stan_reduction_data_internal(%rip), %r8
292	movups	%xmm13, 64(%rsp)
293
294	/*
295	 * Also get the significand as an integer
296	 * NB: adding in the integer bit is wrong for denorms!
297	 * To make this work for denorms we should do something slightly different
298	 */
299	movl	$8388607, %r9d
300	movups	%xmm12, 80(%rsp)
301	movl	$8388608, %r10d
302	movups	%xmm11, 96(%rsp)
303
304	/*
305	 * Break the P_xxx and m into 16-bit chunks ready for
306	 * the long multiplication via 16x16->32 multiplications
307	 */
308	movl	$65535, %r11d
309	movd	%eax, %xmm3
310	pshufd	$0, %xmm3, %xmm2
311	andps	%xmm2, %xmm13
312	cmpeqps	%xmm2, %xmm13
313	pand	%xmm4, %xmm2
314	psrld	$23, %xmm2
315	movdqa	%xmm2, %xmm12
316	pslld	$1, %xmm12
317	paddd	%xmm2, %xmm12
318	pslld	$2, %xmm12
319	pshufd	$1, %xmm12, %xmm10
320	pshufd	$2, %xmm12, %xmm11
321	pshufd	$3, %xmm12, %xmm14
322	movd	%xmm12, %edx
323	movd	%xmm10, %ecx
324	movd	%xmm11, %esi
325	movd	%r9d, %xmm11
326	movd	%xmm14, %edi
327	movd	4(%rdx, %r8), %xmm6
328	movd	4(%rcx, %r8), %xmm7
329	movd	4(%rsi, %r8), %xmm3
330	movl	$872415232, %r9d
331	movd	4(%rdi, %r8), %xmm5
332	punpckldq %xmm7, %xmm6
333	punpckldq %xmm5, %xmm3
334	movd	8(%rdi, %r8), %xmm10
335	movmskps %xmm13, %eax
336	punpcklqdq %xmm3, %xmm6
337	movd	8(%rdx, %r8), %xmm3
338	movd	8(%rcx, %r8), %xmm2
339	movd	8(%rsi, %r8), %xmm13
340	punpckldq %xmm2, %xmm3
341	punpckldq %xmm10, %xmm13
342	punpcklqdq %xmm13, %xmm3
343	pshufd	$0, %xmm11, %xmm13
344	movdqa	%xmm3, %xmm2
345	movups	%xmm4, 48(%rsp)
346	pand	%xmm4, %xmm13
347	movd	%r10d, %xmm4
348	psrld	$16, %xmm2
349	movd	(%rdx, %r8), %xmm9
350
351	/*
352	 * We want to incorporate the original sign now too.
353	 * Do it here for convenience in getting the right N value,
354	 * though we could wait right to the end if we were prepared
355	 * to modify the sign of N later too.
356	 * So get the appropriate sign mask now (or sooner).
357	 */
358	movl	$-2147483648, %edx
359	movd	(%rcx, %r8), %xmm8
360
361	/*
362	 * Create floating-point high part, implicitly adding integer bit 1
363	 * Incorporate overall sign at this stage too.
364	 */
365	movl	$1065353216, %ecx
366	movd	(%rsi, %r8), %xmm15
367
368	/*
369	 * Now round at the 2^-8 bit position for reduction mod pi/2^7
370	 * instead of the original 2pi (but still with the same 2pi scaling).
371	 * Use a shifter of 2^15 + 2^14.
372	 * The N we get is our final version; it has an offset of
373	 * 2^8 because of the implicit integer bit, and anyway for negative
374	 * starting value it's a 2s complement thing. But we need to mask
375	 * off the exponent part anyway so it's fine.
376	 */
377	movl	$1195376640, %esi
378	movd	(%rdi, %r8), %xmm1
379	movl	$511, %r10d
380	movups	%xmm0, 112(%rsp)
381	movd	%r11d, %xmm0
382	pshufd	$0, %xmm4, %xmm12
383	movdqa	%xmm2, %xmm4
384	punpckldq %xmm8, %xmm9
385	paddd	%xmm12, %xmm13
386	punpckldq %xmm1, %xmm15
387	movdqa	%xmm13, %xmm12
388	pshufd	$0, %xmm0, %xmm8
389	movdqa	%xmm6, %xmm0
390	punpcklqdq %xmm15, %xmm9
391	pand	%xmm8, %xmm13
392	movdqa	%xmm9, %xmm14
393	pand	%xmm8, %xmm9
394	movdqa	%xmm13, %xmm10
395	psrld	$16, %xmm14
396	movdqu	%xmm14, 128(%rsp)
397
398	/* Now do the big multiplication and carry propagation */
399	movdqa	%xmm9, %xmm14
400	psrlq	$32, %xmm10
401	psrlq	$32, %xmm14
402	movdqa	%xmm13, %xmm15
403	movdqa	%xmm10, %xmm7
404	pmuludq	%xmm9, %xmm15
405	psrld	$16, %xmm0
406	pmuludq	%xmm14, %xmm7
407	movdqu	%xmm9, 144(%rsp)
408	psllq	$32, %xmm7
409	movdqu	.FLT_16(%rip), %xmm9
410	pand	%xmm8, %xmm6
411	pand	%xmm9, %xmm15
412	psrld	$16, %xmm12
413	movdqa	%xmm0, %xmm1
414	por	%xmm7, %xmm15
415	movdqa	%xmm13, %xmm7
416	pand	%xmm8, %xmm3
417	movdqu	%xmm0, 160(%rsp)
418	movdqa	%xmm12, %xmm11
419	movdqu	%xmm15, 208(%rsp)
420	psrlq	$32, %xmm1
421	pmuludq	%xmm0, %xmm7
422	movdqa	%xmm6, %xmm5
423	movdqa	%xmm10, %xmm15
424	movdqa	%xmm12, %xmm0
425	movdqu	%xmm14, 176(%rsp)
426	psrlq	$32, %xmm11
427	movdqu	%xmm1, 192(%rsp)
428	psrlq	$32, %xmm5
429	pmuludq	%xmm1, %xmm15
430	movdqa	%xmm13, %xmm1
431	pmuludq	%xmm3, %xmm0
432	pmuludq	%xmm6, %xmm1
433	pmuludq	%xmm12, %xmm6
434	movdqa	%xmm10, %xmm14
435	psrlq	$32, %xmm3
436	pmuludq	%xmm5, %xmm14
437	pand	%xmm9, %xmm1
438	pmuludq	%xmm11, %xmm3
439	pmuludq	%xmm11, %xmm5
440	psllq	$32, %xmm14
441	pand	%xmm9, %xmm0
442	psllq	$32, %xmm3
443	psrlq	$32, %xmm4
444	por	%xmm14, %xmm1
445	por	%xmm3, %xmm0
446	movdqa	%xmm12, %xmm14
447	movdqa	%xmm11, %xmm3
448	pmuludq	%xmm2, %xmm14
449	pand	%xmm9, %xmm7
450	pmuludq	%xmm4, %xmm3
451	pmuludq	%xmm13, %xmm2
452	pmuludq	%xmm10, %xmm4
453	pand	%xmm9, %xmm2
454	psllq	$32, %xmm4
455	psllq	$32, %xmm15
456	pand	%xmm9, %xmm14
457	psllq	$32, %xmm3
458	por	%xmm4, %xmm2
459	por	%xmm15, %xmm7
460	por	%xmm3, %xmm14
461	psrld	$16, %xmm2
462	pand	%xmm9, %xmm6
463	psllq	$32, %xmm5
464	movdqa	%xmm1, %xmm15
465	paddd	%xmm2, %xmm14
466	movdqa	%xmm7, %xmm2
467	por	%xmm5, %xmm6
468	psrld	$16, %xmm1
469	pand	%xmm8, %xmm2
470	paddd	%xmm1, %xmm6
471	movdqu	160(%rsp), %xmm1
472	paddd	%xmm6, %xmm2
473	movdqu	192(%rsp), %xmm6
474	psrld	$16, %xmm7
475	pmuludq	%xmm12, %xmm1
476	pand	%xmm8, %xmm15
477	pmuludq	%xmm11, %xmm6
478	pmuludq	144(%rsp), %xmm12
479	pmuludq	176(%rsp), %xmm11
480	pand	%xmm9, %xmm1
481	psllq	$32, %xmm6
482	por	%xmm6, %xmm1
483	psrld	$16, %xmm0
484	paddd	%xmm7, %xmm1
485	paddd	%xmm14, %xmm15
486	movdqu	128(%rsp), %xmm7
487	paddd	%xmm15, %xmm0
488	pmuludq	%xmm7, %xmm13
489	psrlq	$32, %xmm7
490	pmuludq	%xmm7, %xmm10
491	movdqa	%xmm0, %xmm14
492	pand	%xmm9, %xmm13
493	movdqu	208(%rsp), %xmm5
494	psrld	$16, %xmm14
495	paddd	%xmm2, %xmm14
496	movdqa	%xmm5, %xmm15
497	movdqa	%xmm14, %xmm3
498	pand	%xmm8, %xmm15
499	psrld	$16, %xmm3
500	paddd	%xmm1, %xmm15
501	psllq	$32, %xmm10
502	pand	%xmm9, %xmm12
503	psllq	$32, %xmm11
504	paddd	%xmm15, %xmm3
505	por	%xmm10, %xmm13
506	por	%xmm11, %xmm12
507	psrld	$16, %xmm5
508	movdqa	%xmm3, %xmm4
509	pand	%xmm8, %xmm13
510	paddd	%xmm5, %xmm12
511	psrld	$16, %xmm4
512	paddd	%xmm12, %xmm13
513	paddd	%xmm13, %xmm4
514	pand	%xmm8, %xmm3
515	pslld	$16, %xmm4
516	movd	%edx, %xmm9
517	movups	48(%rsp), %xmm15
518	paddd	%xmm3, %xmm4
519	pshufd	$0, %xmm9, %xmm7
520
521	/* Assemble reduced argument from the pieces */
522	pand	%xmm8, %xmm0
523	movd	%ecx, %xmm8
524	pand	%xmm15, %xmm7
525	pshufd	$0, %xmm8, %xmm1
526	movdqa	%xmm4, %xmm5
527	psrld	$9, %xmm5
528	pxor	%xmm7, %xmm1
529	por	%xmm1, %xmm5
530	movd	%esi, %xmm6
531	pshufd	$0, %xmm6, %xmm3
532	movdqa	%xmm5, %xmm6
533	movl	$262143, %r8d
534
535	/*
536	 * Create floating-point low and medium parts, respectively
537	 * lo_17, ... lo_0, 0, ..., 0
538	 * hi_8, ... hi_0, lo_31, ..., lo_18
539	 * then subtract off the implicitly added integer bits,
540	 * 2^-46 and 2^-23, respectively.
541	 * Put the original sign into all of them at this stage.
542	 */
543	movl	$679477248, %edi
544	movd	%r10d, %xmm13
545	pslld	$16, %xmm14
546	pshufd	$0, %xmm13, %xmm1
547	paddd	%xmm0, %xmm14
548	movd	%r9d, %xmm11
549	pand	%xmm4, %xmm1
550	movd	%r8d, %xmm9
551	movd	%edi, %xmm10
552	pshufd	$0, %xmm9, %xmm8
553	pslld	$14, %xmm1
554	pshufd	$0, %xmm10, %xmm0
555	pand	%xmm14, %xmm8
556	pshufd	$0, %xmm11, %xmm12
557	psrld	$18, %xmm14
558	pxor	%xmm7, %xmm0
559	pxor	%xmm12, %xmm7
560	por	%xmm14, %xmm1
561	pslld	$5, %xmm8
562	por	%xmm7, %xmm1
563
564	/*
565	 * Now multiply those numbers all by 2 pi, reasonably accurately.
566	 * The top part uses 2pi = s2pi_lead + s2pi_trail, where
567	 * s2pi_lead has 12 significant bits.
568	 */
569	movl	$1086918619, %r11d
570
571	/* Split RHi into 12-bit leading and trailing parts. */
572	movl	$-4096, %esi
573	por	%xmm0, %xmm8
574	movl	$1086918656, %edx
575	movl	$-1214941318, %ecx
576
577	/*
578	 * If the magnitude of the input is <= 2^-20, then
579	 * just pass through the input, since no reduction will be needed and
580	 * the main path will only work accurately if the reduced argument is
581	 * about >= 2^-40 (which it is for all large pi multiples)
582	 */
583	movl	$2147483647, %edi
584	addps	%xmm3, %xmm6
585	subps	%xmm7, %xmm1
586	subps	%xmm0, %xmm8
587	movaps	%xmm6, %xmm2
588	movd	%r11d, %xmm14
589	movd	%esi, %xmm4
590	movd	%edx, %xmm7
591	movl	$897581056, %r8d
592	subps	%xmm3, %xmm2
593
594	/* Grab our final N value as an integer, appropriately masked mod 2^8 */
595	movl	$255, %r9d
596	subps	%xmm2, %xmm5
597
598	/* Now add them up into 2 reasonably aligned pieces */
599	movaps	%xmm5, %xmm3
600
601	/*
602	 * The output is _VRES_R (high) + _VRES_E (low), and the integer part is _VRES_IND
603	 * Set sRp2 = _VRES_R^2 and then resume the original code.
604	 * Argument reduction is now finished: x = n * pi/128 + r
605	 * where n = iIndex and r = sR (high) + sE (low).
606	 * But we have n modulo 256, needed for sin/cos with period 2pi
607	 * but we want it modulo 128 since tan has period pi.
608	 */
609	movl	$127, %r10d
610	pshufd	$0, %xmm14, %xmm2
611	addps	%xmm1, %xmm3
612	pshufd	$0, %xmm4, %xmm14
613	movd	%r8d, %xmm4
614	pshufd	$0, %xmm4, %xmm9
615	subps	%xmm3, %xmm5
616	movdqa	%xmm9, %xmm11
617	addps	%xmm5, %xmm1
618	movd	%ecx, %xmm5
619	addps	%xmm1, %xmm8
620	pshufd	$0, %xmm7, %xmm1
621	movdqa	%xmm14, %xmm7
622	andps	%xmm3, %xmm7
623
624	/*
625	 * Do the multiplication as exact top part and "naive" low part.
626	 * This still maintains a similar level of offset and doesn't drop
627	 * the accuracy much below what we already have.
628	 */
629	movdqa	%xmm1, %xmm10
630	pshufd	$0, %xmm5, %xmm5
631	subps	%xmm7, %xmm3
632	mulps	%xmm7, %xmm10
633	mulps	%xmm5, %xmm7
634	mulps	%xmm3, %xmm1
635	mulps	%xmm8, %xmm2
636	mulps	%xmm3, %xmm5
637	addps	%xmm7, %xmm1
638	addps	%xmm5, %xmm2
639	movd	%edi, %xmm8
640	addps	%xmm2, %xmm1
641
642	/*
643	 * Do another stage of compensated summation to get full offset
644	 * between the pieces sRedHi + sRedLo.
645	 * Depending on the later algorithm, we might avoid this stage.
646	 */
647	movaps	%xmm1, %xmm0
648
649	/*  Load constants (not all needed at once)  */
650	lea	_sCoeffs+36+__svml_stan_data_internal(%rip), %rdi
651	pshufd	$0, %xmm8, %xmm8
652	addps	%xmm10, %xmm0
653	andps	%xmm15, %xmm8
654	subps	%xmm0, %xmm10
655	cmpltps	%xmm8, %xmm11
656	cmpleps	%xmm9, %xmm8
657	addps	%xmm10, %xmm1
658	andps	%xmm15, %xmm8
659	movd	%r9d, %xmm15
660	andps	%xmm11, %xmm0
661	andps	%xmm1, %xmm11
662	pshufd	$0, %xmm15, %xmm1
663	movd	%r10d, %xmm15
664	pshufd	$0, %xmm15, %xmm7
665	pand	%xmm1, %xmm6
666	pand	%xmm7, %xmm6
667	orps	%xmm0, %xmm8
668	movaps	%xmm6, %xmm4
669
670	/*
671	 * Simply combine the two parts of the reduced argument
672	 * since we can afford a few ulps in this case.
673	 */
674	addps	%xmm11, %xmm8
675	pslld	$2, %xmm4
676	paddd	%xmm6, %xmm4
677	pslld	$3, %xmm4
678	pshufd	$1, %xmm4, %xmm6
679	pshufd	$2, %xmm4, %xmm5
680	pshufd	$3, %xmm4, %xmm3
681	movd	%xmm4, %r11d
682	movd	%xmm6, %edx
683	movd	%xmm5, %ecx
684	movd	%xmm3, %esi
685	movd	-32(%r11, %rdi), %xmm15
686	movd	-32(%rdx, %rdi), %xmm12
687	movd	-32(%rcx, %rdi), %xmm7
688	movd	-32(%rsi, %rdi), %xmm13
689	punpckldq %xmm12, %xmm15
690	punpckldq %xmm13, %xmm7
691	movd	-28(%rsi, %rdi), %xmm5
692	punpcklqdq %xmm7, %xmm15
693	movd	-28(%r11, %rdi), %xmm7
694	movd	-28(%rdx, %rdi), %xmm6
695	movd	-28(%rcx, %rdi), %xmm4
696	movd	-36(%rcx, %rdi), %xmm9
697	movd	-36(%r11, %rdi), %xmm1
698	movd	-36(%rdx, %rdi), %xmm2
699	movd	-24(%rdx, %rdi), %xmm3
700	movd	-36(%rsi, %rdi), %xmm10
701	punpckldq %xmm6, %xmm7
702	punpckldq %xmm5, %xmm4
703	movd	-24(%r11, %rdi), %xmm6
704	punpckldq %xmm2, %xmm1
705	punpckldq %xmm10, %xmm9
706	punpcklqdq %xmm4, %xmm7
707	movd	-16(%r11, %rdi), %xmm4
708	punpckldq %xmm3, %xmm6
709	movd	-24(%rcx, %rdi), %xmm10
710	movd	-16(%rcx, %rdi), %xmm3
711	movd	-24(%rsi, %rdi), %xmm2
712	movd	-16(%rsi, %rdi), %xmm13
713	movd	-16(%rdx, %rdi), %xmm12
714	punpcklqdq %xmm9, %xmm1
715	movd	-20(%rdx, %rdi), %xmm9
716	punpckldq %xmm2, %xmm10
717	movd	-20(%r11, %rdi), %xmm5
718	movd	-20(%rcx, %rdi), %xmm11
719	movd	-20(%rsi, %rdi), %xmm0
720	punpckldq %xmm12, %xmm4
721	punpckldq %xmm13, %xmm3
722	punpcklqdq %xmm10, %xmm6
723	movd	-12(%rsi, %rdi), %xmm10
724	punpckldq %xmm9, %xmm5
725	punpckldq %xmm0, %xmm11
726	punpcklqdq %xmm3, %xmm4
727	movd	-12(%r11, %rdi), %xmm3
728	movd	-12(%rdx, %rdi), %xmm2
729	movd	-12(%rcx, %rdi), %xmm9
730	punpcklqdq %xmm11, %xmm5
731	punpckldq %xmm2, %xmm3
732	punpckldq %xmm10, %xmm9
733	movd	-8(%rcx, %rdi), %xmm10
734	movd	-8(%r11, %rdi), %xmm2
735	movd	-8(%rdx, %rdi), %xmm0
736	movd	-8(%rsi, %rdi), %xmm11
737	punpckldq %xmm0, %xmm2
738	punpckldq %xmm11, %xmm10
739	movd	-4(%rsi, %rdi), %xmm13
740	punpcklqdq %xmm9, %xmm3
741	punpcklqdq %xmm10, %xmm2
742	movd	-4(%r11, %rdi), %xmm10
743	movd	-4(%rdx, %rdi), %xmm12
744	movd	-4(%rcx, %rdi), %xmm9
745	punpckldq %xmm12, %xmm10
746	punpckldq %xmm13, %xmm9
747	punpcklqdq %xmm9, %xmm10
748
749	/*
750	 *  Compute 2-part reciprocal component
751	 * Construct a separate reduced argument modulo pi near pi/2 multiples.
752	 * i.e. (pi/2 - x) mod pi, simply by subtracting the reduced argument
753	 * from an accurate B_hi + B_lo = (128 - n) pi/128. Force the upper part
754	 * of this reduced argument to half-length to simplify accurate
755	 * reciprocation later on.
756	 */
757	movdqa	%xmm1, %xmm9
758	movd	(%r11, %rdi), %xmm13
759	subps	%xmm8, %xmm9
760	movd	(%rdx, %rdi), %xmm0
761	subps	%xmm9, %xmm1
762	punpckldq %xmm0, %xmm13
763	movdqa	%xmm14, %xmm0
764	andps	%xmm9, %xmm0
765	subps	%xmm8, %xmm1
766	subps	%xmm0, %xmm9
767	movd	(%rcx, %rdi), %xmm12
768	addps	%xmm9, %xmm15
769
770	/*
771	 * Now compute an approximate reciprocal to mix into the computation
772	 * To avoid any danger of nonportability, force it to 12 bits,
773	 * though I suspect it always is anyway on current platforms.
774	 */
775	rcpps	%xmm0, %xmm9
776	addps	%xmm15, %xmm1
777	andps	%xmm14, %xmm9
778	mulps	%xmm9, %xmm0
779
780	/*
781	 * Get a better approximation to  1/sR_hi (not far short of an ulp)
782	 * using a third-order polynomial approximation
783	 */
784	movaps	%xmm9, %xmm14
785	movd	(%rsi, %rdi), %xmm11
786
787	/*
788	 * Now compute the error sEr where sRecip_hi = (1/R_hi) * (1 - sEr)
789	 * so that we can compensate for it.
790	 */
791	movups	_sOne+__svml_stan_data_internal(%rip), %xmm15
792	punpckldq %xmm11, %xmm12
793	movaps	%xmm15, %xmm11
794	punpcklqdq %xmm12, %xmm13
795	subps	%xmm0, %xmm11
796	mulps	%xmm11, %xmm14
797	movups	%xmm11, (%rsp)
798	addps	%xmm9, %xmm14
799	mulps	%xmm11, %xmm11
800	movups	%xmm13, 32(%rsp)
801	movups	%xmm11, 16(%rsp)
802	movups	112(%rsp), %xmm0
803	movups	96(%rsp), %xmm11
804	movups	80(%rsp), %xmm12
805	movups	64(%rsp), %xmm13
806	# LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm1 xmm2 xmm3 xmm4 xmm5 xmm6 xmm7 xmm8 xmm9 xmm10 xmm11 xmm12 xmm13 xmm14 xmm15
807
808	/*
809	 *  Compensated sum of dominant component(s)
810	 * Compute C0_hi + C1_hi * Z + Recip_hi + Recip_lo = H4 (hi) + H9 (lo)
811	 * H1 = C1_hi * Z (exact since C1_hi is 1 bit)
812	 */
813	mulps	%xmm8, %xmm4
814	addps	16(%rsp), %xmm15
815
816	/* Finally, multiplex both parts so they are only used in cotangent path */
817	mulps	%xmm7, %xmm9
818
819	/*
820	 *  Higher polynomial terms
821	 * Stage 1 (with unlimited parallelism)
822	 * P3 = C1_lo + C2 * Z
823	 */
824	mulps	%xmm8, %xmm2
825	mulps	%xmm15, %xmm14
826	addps	%xmm2, %xmm3
827
828	/*
829	 * Multiply by sRecip_ok to make sR_lo relative to sR_hi
830	 * Since sR_lo is shifted off by about 12 bits, this is accurate enough.
831	 */
832	mulps	%xmm14, %xmm1
833
834	/*
835	 * Now create a low reciprocal using
836	 * (Recip_hi + Er * Recip_ok) * (1 + sR_lo^2 - sR_lo)
837	 * =~= Recip_hi + Recip_ok * (Er + sR_lo^2 - sR_lo)
838	 */
839	movaps	%xmm1, %xmm15
840	mulps	%xmm1, %xmm1
841	subps	(%rsp), %xmm15
842
843	/* P4 = C3 + C4 * Z */
844	movups	32(%rsp), %xmm2
845	subps	%xmm15, %xmm1
846	mulps	%xmm8, %xmm2
847	mulps	%xmm1, %xmm14
848	addps	%xmm2, %xmm10
849	mulps	%xmm14, %xmm7
850
851	/* H2 = high(C0_hi + C1_hi * Z) */
852	movdqa	%xmm6, %xmm14
853	addps	%xmm4, %xmm14
854
855	/* H4 = high(H2 + Recip_hi) */
856	movaps	%xmm14, %xmm1
857
858	/* intermediate in compensated sum */
859	subps	%xmm14, %xmm6
860	addps	%xmm9, %xmm1
861
862	/* H5 = low(C0_hi + C1_hi * Z) */
863	addps	%xmm6, %xmm4
864
865	/* intermediate in compensated sum */
866	subps	%xmm1, %xmm9
867
868	/* H7 = low(C0_hi + C1_hi * Z) + Recip_lo */
869	addps	%xmm4, %xmm7
870
871	/* H8 = low(H2 + Recip_hi) */
872	addps	%xmm9, %xmm14
873
874	/* Z2 = Z^2 */
875	movaps	%xmm8, %xmm4
876
877	/* Now H4 + H9 should be that part */
878	addps	%xmm14, %xmm7
879	mulps	%xmm8, %xmm4
880
881	/* P9 = trail(dominant part) + C0_lo */
882	addps	%xmm7, %xmm5
883
884	/*
885	 * Stage 2 (with unlimited parallelism)
886	 * P6 = C1_lo + C2 * Z + C3 * Z^2 + C4 * Z^3
887	 */
888	mulps	%xmm4, %xmm10
889	addps	%xmm10, %xmm3
890
891	/* Final accumulation of low part */
892	mulps	%xmm3, %xmm8
893
894	/* Merge results from main and large paths: */
895	movaps	%xmm11, %xmm3
896	andnps	%xmm0, %xmm3
897	addps	%xmm8, %xmm5
898	movaps	%xmm3, %xmm0
899
900	/* And now the very final summation */
901	addps	%xmm5, %xmm1
902
903	/*
904	 *  The end of implementation (LA with huge args reduction)
905	 * End of large arguments path (_HA_, _LA_ and _EP_)
906	 */
907
908	pxor	%xmm12, %xmm1
909	andps	%xmm11, %xmm1
910	orps	%xmm1, %xmm0
911
912	/* Return to main vector processing path */
913	jmp	L(AUX_BRANCH_RETURN)
914	# LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm13
915END(_ZGVbN4v_tanf_sse4)
916
917	.section .rodata, "a"
918	.align	16
919
920#ifdef __svml_stan_data_internal_typedef
921typedef unsigned int VUINT32;
922typedef struct {
923	__declspec(align(16)) VUINT32 _sInvPI_uisa[4][1];
924	__declspec(align(16)) VUINT32 _sPI1_uisa[4][1];
925	__declspec(align(16)) VUINT32 _sPI2_uisa[4][1];
926	__declspec(align(16)) VUINT32 _sPI3_uisa[4][1];
927	__declspec(align(16)) VUINT32 _sPI2_ha_uisa[4][1];
928	__declspec(align(16)) VUINT32 _sPI3_ha_uisa[4][1];
929	__declspec(align(16)) VUINT32 Th_tbl_uisa[32][1];
930	__declspec(align(16)) VUINT32 Tl_tbl_uisa[32][1];
931	__declspec(align(16)) VUINT32 _sPC3_uisa[4][1];
932	__declspec(align(16)) VUINT32 _sPC5_uisa[4][1];
933	__declspec(align(16)) VUINT32 _sRangeReductionVal_uisa[4][1];
934	__declspec(align(16)) VUINT32 _sInvPi[4][1];
935	__declspec(align(16)) VUINT32 _sSignMask[4][1];
936	__declspec(align(16)) VUINT32 _sAbsMask[4][1];
937	__declspec(align(16)) VUINT32 _sRangeVal[4][1];
938	__declspec(align(16)) VUINT32 _sRShifter[4][1];
939	__declspec(align(16)) VUINT32 _sOne[4][1];
940	__declspec(align(16)) VUINT32 _sRangeReductionVal[4][1];
941	__declspec(align(16)) VUINT32 _sPI1[4][1];
942	__declspec(align(16)) VUINT32 _sPI2[4][1];
943	__declspec(align(16)) VUINT32 _sPI3[4][1];
944	__declspec(align(16)) VUINT32 _sPI4[4][1];
945	__declspec(align(16)) VUINT32 _sPI1_FMA[4][1];
946	__declspec(align(16)) VUINT32 _sPI2_FMA[4][1];
947	__declspec(align(16)) VUINT32 _sPI3_FMA[4][1];
948	__declspec(align(16)) VUINT32 _sP0[4][1];
949	__declspec(align(16)) VUINT32 _sP1[4][1];
950	__declspec(align(16)) VUINT32 _sQ0[4][1];
951	__declspec(align(16)) VUINT32 _sQ1[4][1];
952	__declspec(align(16)) VUINT32 _sQ2[4][1];
953	__declspec(align(16)) VUINT32 _sTwo[4][1];
954	__declspec(align(16)) VUINT32 _sCoeffs[128][10][1];
955} __svml_stan_data_internal;
956#endif
957__svml_stan_data_internal:
958	/* UISA */
959	.long	0x4122f983, 0x4122f983, 0x4122f983, 0x4122f983 /* _sInvPI_uisa */
960	.align	16
961	.long	0x3dc90fda, 0x3dc90fda, 0x3dc90fda, 0x3dc90fda /* _sPI1_uisa */
962	.align	16
963	.long	0x31a22168, 0x31a22168, 0x31a22168, 0x31a22168 /* _sPI2_uisa */
964	.align	16
965	.long	0x25c234c5, 0x25c234c5, 0x25c234c5, 0x25c234c5 /* _sPI3_uisa */
966	.align	16
967	.long	0x31a22000, 0x31a22000, 0x31a22000, 0x31a22000 /* _sPI2_ha_uisa */
968	.align	16
969	.long	0x2a34611a, 0x2a34611a, 0x2a34611a, 0x2a34611a /* _sPI3_ha_uisa */
970	/* Th_tbl_uisa for i from 0 to 31 do printsingle(tan(i*Pi/32)); */
971	.align	16
972	.long	0x80000000, 0x3dc9b5dc, 0x3e4bafaf, 0x3e9b5042
973	.long	0x3ed413cd, 0x3f08d5b9, 0x3f2b0dc1, 0x3f521801
974	.long	0x3f800000, 0x3f9bf7ec, 0x3fbf90c7, 0x3fef789e
975	.long	0x401a827a, 0x4052facf, 0x40a0dff7, 0x41227363
976	.long	0xff7fffff, 0xc1227363, 0xc0a0dff7, 0xc052facf
977	.long	0xc01a827a, 0xbfef789e, 0xbfbf90c7, 0xbf9bf7ec
978	.long	0xbf800000, 0xbf521801, 0xbf2b0dc1, 0xbf08d5b9
979	.long	0xbed413cd, 0xbe9b5042, 0xbe4bafaf, 0xbdc9b5dc
980	/* Tl_tbl_uisa for i from 0 to 31 do printsingle(tan(i*Pi/32)-round(tan(i*Pi/32), SG, RN)); */
981	.align	16
982	.long	0x80000000, 0x3145b2da, 0x2f2a62b0, 0xb22a39c2
983	.long	0xb1c0621a, 0xb25ef963, 0x32ab7f99, 0x32ae4285
984	.long	0x00000000, 0x33587608, 0x32169d18, 0xb30c3ec0
985	.long	0xb3cc0622, 0x3390600e, 0x331091dc, 0xb454a046
986	.long	0xf3800000, 0x3454a046, 0xb31091dc, 0xb390600e
987	.long	0x33cc0622, 0x330c3ec0, 0xb2169d18, 0xb3587608
988	.long	0x00000000, 0xb2ae4285, 0xb2ab7f99, 0x325ef963
989	.long	0x31c0621a, 0x322a39c2, 0xaf2a62b0, 0xb145b2da
990	.align	16
991	.long	0x3eaaaaa6, 0x3eaaaaa6, 0x3eaaaaa6, 0x3eaaaaa6 /* _sPC3_uisa */
992	.align	16
993	.long	0x3e08b888, 0x3e08b888, 0x3e08b888, 0x3e08b888 /* _sPC5_uisa */
994	.align	16
995	.long	0x46010000, 0x46010000, 0x46010000, 0x46010000 /* _sRangeReductionVal_uisa */
996	.align	16
997	.long	0x3F22F983, 0x3F22F983, 0x3F22F983, 0x3F22F983 /* _sInvPi */
998	.align	16
999	.long	0x80000000, 0x80000000, 0x80000000, 0x80000000 /* _sSignMask */
1000	.align	16
1001	.long	0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF /* _sAbsMask */
1002	.align	16
1003	.long	0x7f800000, 0x7f800000, 0x7f800000, 0x7f800000 /* _sRangeVal */
1004	.align	16
1005	.long	0x4B400000, 0x4B400000, 0x4B400000, 0x4B400000 /* _sRShifter */
1006	.align	16
1007	.long	0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 /* _sOne */
1008	.align	16
1009	.long	0x46010000, 0x46010000, 0x46010000, 0x46010000 /* _sRangeVal */
1010	.align	16
1011	.long	0x3FC90000, 0x3FC90000, 0x3FC90000, 0x3FC90000 /* _sPI1 */
1012	.align	16
1013	.long	0x39FDA000, 0x39FDA000, 0x39FDA000, 0x39FDA000 /* _sPI2 */
1014	.align	16
1015	.long	0x33A22000, 0x33A22000, 0x33A22000, 0x33A22000 /* _sPI3 */
1016	.align	16
1017	.long	0x2C34611A, 0x2C34611A, 0x2C34611A, 0x2C34611A /* _sPI4 */
1018	// PI1, PI2, and PI3 when FMA is available
1019	.align	16
1020	.long	0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB, 0x3FC90FDB /* _sPI1_FMA */
1021	.align	16
1022	.long	0xB33BBD2E, 0xB33BBD2E, 0xB33BBD2E, 0xB33BBD2E /* _sPI2_FMA */
1023	.align	16
1024	.long	0xA6F72CED, 0xA6F72CED, 0xA6F72CED, 0xA6F72CED /* _sPI3_FMA */
1025	.align	16
1026	.long	0x3F7FFFFC, 0x3F7FFFFC, 0x3F7FFFFC, 0x3F7FFFFC /* _sP0 */
1027	.align	16
1028	.long	0xBDC433B4, 0xBDC433B4, 0xBDC433B4, 0xBDC433B4 /* _sP1 */
1029	.align	16
1030	.long	0x3F7FFFFC, 0x3F7FFFFC, 0x3F7FFFFC, 0x3F7FFFFC /* _sQ0 */
1031	.align	16
1032	.long	0xBEDBB7AB, 0xBEDBB7AB, 0xBEDBB7AB, 0xBEDBB7AB /* _sQ1 */
1033	.align	16
1034	.long	0x3C1F336B, 0x3C1F336B, 0x3C1F336B, 0x3C1F336B /* _sQ2 */
1035	.align	16
1036	.long	0x40000000, 0x40000000, 0x40000000, 0x40000000 /* _sTwo */
1037	// _sCoeffs Breakpoint B = 0 * pi/128, function tan(B + x)
1038	.align	16
1039	.long	0x3FC90FDB // B' = pi/2 - B (high single)
1040	.long	0xB33BBD2E // B' = pi/2 - B (low single)
1041	.long	0x00000000 // tau (1 for cot path)
1042	.long	0x00000000 // c0 (high single)
1043	.long	0x00000000 // c0 (low single)
1044	.long	0x3F800000 // c1 (high 1 bit)
1045	.long	0x00000000 // c1 (low single)
1046	.long	0x00000000 // c2
1047	.long	0x3EAAACDD // c3
1048	.long	0x00000000 // c4
1049	.long	0x3FC5EB9B // B' = pi/2 - B (high single)
1050	.long	0x32DE638C // B' = pi/2 - B (low single)
1051	.long	0x00000000 // tau (1 for cot path)
1052	.long	0x3CC91A31 // c0 (high single)
1053	.long	0x2F8E8D1A // c0 (low single)
1054	.long	0x3F800000 // c1 (high 1 bit)
1055	.long	0x3A1DFA00 // c1 (low single)
1056	.long	0x3CC9392D // c2
1057	.long	0x3EAB1889 // c3
1058	.long	0x3C885D3B // c4
1059	.long	0x3FC2C75C // B' = pi/2 - B (high single)
1060	.long	0xB2CBBE8A // B' = pi/2 - B (low single)
1061	.long	0x00000000 // tau (1 for cot path)
1062	.long	0x3D49393C // c0 (high single)
1063	.long	0x30A39F5B // c0 (low single)
1064	.long	0x3F800000 // c1 (high 1 bit)
1065	.long	0x3B1E2B00 // c1 (low single)
1066	.long	0x3D49B5D4 // c2
1067	.long	0x3EAC4F10 // c3
1068	.long	0x3CFD9425 // c4
1069	.long	0x3FBFA31C // B' = pi/2 - B (high single)
1070	.long	0x33450FB0 // B' = pi/2 - B (low single)
1071	.long	0x00000000 // tau (1 for cot path)
1072	.long	0x3D9711CE // c0 (high single)
1073	.long	0x314FEB28 // c0 (low single)
1074	.long	0x3F800000 // c1 (high 1 bit)
1075	.long	0x3BB24C00 // c1 (low single)
1076	.long	0x3D97E43A // c2
1077	.long	0x3EAE6A89 // c3
1078	.long	0x3D4D07E0 // c4
1079	.long	0x3FBC7EDD // B' = pi/2 - B (high single)
1080	.long	0xB1800ADD // B' = pi/2 - B (low single)
1081	.long	0x00000000 // tau (1 for cot path)
1082	.long	0x3DC9B5DC // c0 (high single)
1083	.long	0x3145AD86 // c0 (low single)
1084	.long	0x3F800000 // c1 (high 1 bit)
1085	.long	0x3C1EEF20 // c1 (low single)
1086	.long	0x3DCBAAEA // c2
1087	.long	0x3EB14E5E // c3
1088	.long	0x3D858BB2 // c4
1089	.long	0x3FB95A9E // B' = pi/2 - B (high single)
1090	.long	0xB3651267 // B' = pi/2 - B (low single)
1091	.long	0x00000000 // tau (1 for cot path)
1092	.long	0x3DFC98C2 // c0 (high single)
1093	.long	0xB0AE525C // c0 (low single)
1094	.long	0x3F800000 // c1 (high 1 bit)
1095	.long	0x3C793D20 // c1 (low single)
1096	.long	0x3E003845 // c2
1097	.long	0x3EB5271F // c3
1098	.long	0x3DAC669E // c4
1099	.long	0x3FB6365E // B' = pi/2 - B (high single)
1100	.long	0x328BB91C // B' = pi/2 - B (low single)
1101	.long	0x00000000 // tau (1 for cot path)
1102	.long	0x3E17E564 // c0 (high single)
1103	.long	0xB1C5A2E4 // c0 (low single)
1104	.long	0x3F800000 // c1 (high 1 bit)
1105	.long	0x3CB440D0 // c1 (low single)
1106	.long	0x3E1B3D00 // c2
1107	.long	0x3EB9F664 // c3
1108	.long	0x3DD647C0 // c4
1109	.long	0x3FB3121F // B' = pi/2 - B (high single)
1110	.long	0xB30F347D // B' = pi/2 - B (low single)
1111	.long	0x00000000 // tau (1 for cot path)
1112	.long	0x3E31AE4D // c0 (high single)
1113	.long	0xB1F32251 // c0 (low single)
1114	.long	0x3F800000 // c1 (high 1 bit)
1115	.long	0x3CF6A500 // c1 (low single)
1116	.long	0x3E3707DA // c2
1117	.long	0x3EBFA489 // c3
1118	.long	0x3DFBD9C7 // c4
1119	.long	0x3FAFEDDF // B' = pi/2 - B (high single)
1120	.long	0x331BBA77 // B' = pi/2 - B (low single)
1121	.long	0x00000000 // tau (1 for cot path)
1122	.long	0x3E4BAFAF // c0 (high single)
1123	.long	0x2F2A29E0 // c0 (low single)
1124	.long	0x3F800000 // c1 (high 1 bit)
1125	.long	0x3D221018 // c1 (low single)
1126	.long	0x3E53BED0 // c2
1127	.long	0x3EC67E26 // c3
1128	.long	0x3E1568E2 // c4
1129	.long	0x3FACC9A0 // B' = pi/2 - B (high single)
1130	.long	0xB2655A50 // B' = pi/2 - B (low single)
1131	.long	0x00000000 // tau (1 for cot path)
1132	.long	0x3E65F267 // c0 (high single)
1133	.long	0x31B4B1DF // c0 (low single)
1134	.long	0x3F800000 // c1 (high 1 bit)
1135	.long	0x3D4E8B90 // c1 (low single)
1136	.long	0x3E718ACA // c2
1137	.long	0x3ECE7164 // c3
1138	.long	0x3E2DC161 // c4
1139	.long	0x3FA9A560 // B' = pi/2 - B (high single)
1140	.long	0x33719861 // B' = pi/2 - B (low single)
1141	.long	0x00000000 // tau (1 for cot path)
1142	.long	0x3E803FD4 // c0 (high single)
1143	.long	0xB2279E66 // c0 (low single)
1144	.long	0x3F800000 // c1 (high 1 bit)
1145	.long	0x3D807FC8 // c1 (low single)
1146	.long	0x3E884BD4 // c2
1147	.long	0x3ED7812D // c3
1148	.long	0x3E4636EB // c4
1149	.long	0x3FA68121 // B' = pi/2 - B (high single)
1150	.long	0x31E43AAC // B' = pi/2 - B (low single)
1151	.long	0x00000000 // tau (1 for cot path)
1152	.long	0x3E8DB082 // c0 (high single)
1153	.long	0xB132A234 // c0 (low single)
1154	.long	0x3F800000 // c1 (high 1 bit)
1155	.long	0x3D9CD7D0 // c1 (low single)
1156	.long	0x3E988A60 // c2
1157	.long	0x3EE203E3 // c3
1158	.long	0x3E63582C // c4
1159	.long	0x3FA35CE2 // B' = pi/2 - B (high single)
1160	.long	0xB33889B6 // B' = pi/2 - B (low single)
1161	.long	0x00000000 // tau (1 for cot path)
1162	.long	0x3E9B5042 // c0 (high single)
1163	.long	0xB22A3AEE // c0 (low single)
1164	.long	0x3F800000 // c1 (high 1 bit)
1165	.long	0x3DBC7490 // c1 (low single)
1166	.long	0x3EA99AF5 // c2
1167	.long	0x3EEDE107 // c3
1168	.long	0x3E80E9AA // c4
1169	.long	0x3FA038A2 // B' = pi/2 - B (high single)
1170	.long	0x32E4CA7E // B' = pi/2 - B (low single)
1171	.long	0x00000000 // tau (1 for cot path)
1172	.long	0x3EA92457 // c0 (high single)
1173	.long	0x30B80830 // c0 (low single)
1174	.long	0x3F800000 // c1 (high 1 bit)
1175	.long	0x3DDF8200 // c1 (low single)
1176	.long	0x3EBB99E9 // c2
1177	.long	0x3EFB4AA8 // c3
1178	.long	0x3E9182BE // c4
1179	.long	0x3F9D1463 // B' = pi/2 - B (high single)
1180	.long	0xB2C55799 // B' = pi/2 - B (low single)
1181	.long	0x00000000 // tau (1 for cot path)
1182	.long	0x3EB73250 // c0 (high single)
1183	.long	0xB2028823 // c0 (low single)
1184	.long	0x3F800000 // c1 (high 1 bit)
1185	.long	0x3E0318F8 // c1 (low single)
1186	.long	0x3ECEA678 // c2
1187	.long	0x3F053C67 // c3
1188	.long	0x3EA41E53 // c4
1189	.long	0x3F99F023 // B' = pi/2 - B (high single)
1190	.long	0x33484328 // B' = pi/2 - B (low single)
1191	.long	0x00000000 // tau (1 for cot path)
1192	.long	0x3EC5800D // c0 (high single)
1193	.long	0xB214C3C1 // c0 (low single)
1194	.long	0x3F800000 // c1 (high 1 bit)
1195	.long	0x3E185E54 // c1 (low single)
1196	.long	0x3EE2E342 // c2
1197	.long	0x3F0DCA73 // c3
1198	.long	0x3EB8CC21 // c4
1199	.long	0x3F96CBE4 // B' = pi/2 - B (high single)
1200	.long	0xB14CDE2E // B' = pi/2 - B (low single)
1201	.long	0x00000000 // tau (1 for cot path)
1202	.long	0x3ED413CD // c0 (high single)
1203	.long	0xB1C06152 // c0 (low single)
1204	.long	0x3F800000 // c1 (high 1 bit)
1205	.long	0x3E2FB0CC // c1 (low single)
1206	.long	0x3EF876CB // c2
1207	.long	0x3F177807 // c3
1208	.long	0x3ED08437 // c4
1209	.long	0x3F93A7A5 // B' = pi/2 - B (high single)
1210	.long	0xB361DEEE // B' = pi/2 - B (low single)
1211	.long	0x00000000 // tau (1 for cot path)
1212	.long	0x3EE2F439 // c0 (high single)
1213	.long	0xB1F4399E // c0 (low single)
1214	.long	0x3F800000 // c1 (high 1 bit)
1215	.long	0x3E49341C // c1 (low single)
1216	.long	0x3F07C61A // c2
1217	.long	0x3F22560F // c3
1218	.long	0x3EEAA81E // c4
1219	.long	0x3F908365 // B' = pi/2 - B (high single)
1220	.long	0x3292200D // B' = pi/2 - B (low single)
1221	.long	0x00000000 // tau (1 for cot path)
1222	.long	0x3EF22870 // c0 (high single)
1223	.long	0x325271F4 // c0 (low single)
1224	.long	0x3F800000 // c1 (high 1 bit)
1225	.long	0x3E65107A // c1 (low single)
1226	.long	0x3F1429F0 // c2
1227	.long	0x3F2E8AFC // c3
1228	.long	0x3F040498 // c4
1229	.long	0x3F8D5F26 // B' = pi/2 - B (high single)
1230	.long	0xB30C0105 // B' = pi/2 - B (low single)
1231	.long	0x00000000 // tau (1 for cot path)
1232	.long	0x3F00DC0D // c0 (high single)
1233	.long	0xB214AF72 // c0 (low single)
1234	.long	0x3F800000 // c1 (high 1 bit)
1235	.long	0x3E81B994 // c1 (low single)
1236	.long	0x3F218233 // c2
1237	.long	0x3F3C4531 // c3
1238	.long	0x3F149688 // c4
1239	.long	0x3F8A3AE6 // B' = pi/2 - B (high single)
1240	.long	0x331EEDF0 // B' = pi/2 - B (low single)
1241	.long	0x00000000 // tau (1 for cot path)
1242	.long	0x3F08D5B9 // c0 (high single)
1243	.long	0xB25EF98E // c0 (low single)
1244	.long	0x3F800000 // c1 (high 1 bit)
1245	.long	0x3E92478D // c1 (low single)
1246	.long	0x3F2FEDC9 // c2
1247	.long	0x3F4BCD58 // c3
1248	.long	0x3F27AE9E // c4
1249	.long	0x3F8716A7 // B' = pi/2 - B (high single)
1250	.long	0xB2588C6D // B' = pi/2 - B (low single)
1251	.long	0x00000000 // tau (1 for cot path)
1252	.long	0x3F1105AF // c0 (high single)
1253	.long	0x32F045B0 // c0 (low single)
1254	.long	0x3F800000 // c1 (high 1 bit)
1255	.long	0x3EA44EE2 // c1 (low single)
1256	.long	0x3F3F8FDB // c2
1257	.long	0x3F5D3FD0 // c3
1258	.long	0x3F3D0A23 // c4
1259	.long	0x3F83F267 // B' = pi/2 - B (high single)
1260	.long	0x3374CBD9 // B' = pi/2 - B (low single)
1261	.long	0x00000000 // tau (1 for cot path)
1262	.long	0x3F1970C4 // c0 (high single)
1263	.long	0x32904848 // c0 (low single)
1264	.long	0x3F800000 // c1 (high 1 bit)
1265	.long	0x3EB7EFF8 // c1 (low single)
1266	.long	0x3F50907C // c2
1267	.long	0x3F710FEA // c3
1268	.long	0x3F561FED // c4
1269	.long	0x3F80CE28 // B' = pi/2 - B (high single)
1270	.long	0x31FDD672 // B' = pi/2 - B (low single)
1271	.long	0x00000000 // tau (1 for cot path)
1272	.long	0x3F221C37 // c0 (high single)
1273	.long	0xB20C61DC // c0 (low single)
1274	.long	0x3F800000 // c1 (high 1 bit)
1275	.long	0x3ECD4F71 // c1 (low single)
1276	.long	0x3F631DAA // c2
1277	.long	0x3F83B471 // c3
1278	.long	0x3F7281EA // c4
1279	.long	0x3F7B53D1 // B' = pi/2 - B (high single)
1280	.long	0x32955386 // B' = pi/2 - B (low single)
1281	.long	0x00000000 // tau (1 for cot path)
1282	.long	0x3F2B0DC1 // c0 (high single)
1283	.long	0x32AB7EBA // c0 (low single)
1284	.long	0x3F800000 // c1 (high 1 bit)
1285	.long	0x3EE496C2 // c1 (low single)
1286	.long	0x3F776C40 // c2
1287	.long	0x3F9065C1 // c3
1288	.long	0x3F89AFB6 // c4
1289	.long	0x3F750B52 // B' = pi/2 - B (high single)
1290	.long	0x32EB316F // B' = pi/2 - B (low single)
1291	.long	0x00000000 // tau (1 for cot path)
1292	.long	0x3F344BA9 // c0 (high single)
1293	.long	0xB2B8B0EA // c0 (low single)
1294	.long	0x3F800000 // c1 (high 1 bit)
1295	.long	0x3EFDF4F7 // c1 (low single)
1296	.long	0x3F86DCA8 // c2
1297	.long	0x3F9ED53B // c3
1298	.long	0x3F9CBEDE // c4
1299	.long	0x3F6EC2D4 // B' = pi/2 - B (high single)
1300	.long	0xB2BEF0A7 // B' = pi/2 - B (low single)
1301	.long	0x00000000 // tau (1 for cot path)
1302	.long	0x3F3DDCCF // c0 (high single)
1303	.long	0x32D29606 // c0 (low single)
1304	.long	0x40000000 // c1 (high 1 bit)
1305	.long	0xBEE6606F // c1 (low single)
1306	.long	0x3F9325D6 // c2
1307	.long	0x3FAF4E69 // c3
1308	.long	0x3FB3080C // c4
1309	.long	0x3F687A55 // B' = pi/2 - B (high single)
1310	.long	0xB252257B // B' = pi/2 - B (low single)
1311	.long	0x00000000 // tau (1 for cot path)
1312	.long	0x3F47C8CC // c0 (high single)
1313	.long	0xB200F51A // c0 (low single)
1314	.long	0x40000000 // c1 (high 1 bit)
1315	.long	0xBEC82C6C // c1 (low single)
1316	.long	0x3FA0BAE9 // c2
1317	.long	0x3FC2252F // c3
1318	.long	0x3FCD24C7 // c4
1319	.long	0x3F6231D6 // B' = pi/2 - B (high single)
1320	.long	0xB119A6A2 // B' = pi/2 - B (low single)
1321	.long	0x00000000 // tau (1 for cot path)
1322	.long	0x3F521801 // c0 (high single)
1323	.long	0x32AE4178 // c0 (low single)
1324	.long	0x40000000 // c1 (high 1 bit)
1325	.long	0xBEA72938 // c1 (low single)
1326	.long	0x3FAFCC22 // c2
1327	.long	0x3FD7BD4A // c3
1328	.long	0x3FEBB01B // c4
1329	.long	0x3F5BE957 // B' = pi/2 - B (high single)
1330	.long	0x3205522A // B' = pi/2 - B (low single)
1331	.long	0x00000000 // tau (1 for cot path)
1332	.long	0x3F5CD3BE // c0 (high single)
1333	.long	0x31460308 // c0 (low single)
1334	.long	0x40000000 // c1 (high 1 bit)
1335	.long	0xBE8306C5 // c1 (low single)
1336	.long	0x3FC09232 // c2
1337	.long	0x3FF09632 // c3
1338	.long	0x4007DB00 // c4
1339	.long	0x3F55A0D8 // B' = pi/2 - B (high single)
1340	.long	0x329886FF // B' = pi/2 - B (low single)
1341	.long	0x00000000 // tau (1 for cot path)
1342	.long	0x3F68065E // c0 (high single)
1343	.long	0x32670D1A // c0 (low single)
1344	.long	0x40000000 // c1 (high 1 bit)
1345	.long	0xBE36D1D6 // c1 (low single)
1346	.long	0x3FD35007 // c2
1347	.long	0x4006A861 // c3
1348	.long	0x401D4BDA // c4
1349	.long	0x3F4F5859 // B' = pi/2 - B (high single)
1350	.long	0x32EE64E8 // B' = pi/2 - B (low single)
1351	.long	0x00000000 // tau (1 for cot path)
1352	.long	0x3F73BB75 // c0 (high single)
1353	.long	0x32FC908D // c0 (low single)
1354	.long	0x40000000 // c1 (high 1 bit)
1355	.long	0xBDBF94B0 // c1 (low single)
1356	.long	0x3FE8550F // c2
1357	.long	0x40174F67 // c3
1358	.long	0x4036C608 // c4
1359	.long	0x3F490FDB // B' = pi/2 - B (high single)
1360	.long	0xB2BBBD2E // B' = pi/2 - B (low single)
1361	.long	0x3F800000 // tau (1 for cot path)
1362	.long	0xBE8BE60E // c0 (high single)
1363	.long	0x320D8D84 // c0 (low single)
1364	.long	0x3F000000 // c1 (high 1 bit)
1365	.long	0xBDF817B1 // c1 (low single)
1366	.long	0xBD8345EB // c2
1367	.long	0x3D1DFDAC // c3
1368	.long	0xBC52CF6F // c4
1369	.long	0x3F42C75C // B' = pi/2 - B (high single)
1370	.long	0xB24BBE8A // B' = pi/2 - B (low single)
1371	.long	0x3F800000 // tau (1 for cot path)
1372	.long	0xBE87283F // c0 (high single)
1373	.long	0xB268B966 // c0 (low single)
1374	.long	0x3F000000 // c1 (high 1 bit)
1375	.long	0xBDFE6529 // c1 (low single)
1376	.long	0xBD7B1953 // c2
1377	.long	0x3D18E109 // c3
1378	.long	0xBC4570B0 // c4
1379	.long	0x3F3C7EDD // B' = pi/2 - B (high single)
1380	.long	0xB1000ADD // B' = pi/2 - B (low single)
1381	.long	0x3F800000 // tau (1 for cot path)
1382	.long	0xBE827420 // c0 (high single)
1383	.long	0x320B8B4D // c0 (low single)
1384	.long	0x3E800000 // c1 (high 1 bit)
1385	.long	0x3DFB9428 // c1 (low single)
1386	.long	0xBD7002B4 // c2
1387	.long	0x3D142A6C // c3
1388	.long	0xBC3A47FF // c4
1389	.long	0x3F36365E // B' = pi/2 - B (high single)
1390	.long	0x320BB91C // B' = pi/2 - B (low single)
1391	.long	0x3F800000 // tau (1 for cot path)
1392	.long	0xBE7B9282 // c0 (high single)
1393	.long	0xB13383D2 // c0 (low single)
1394	.long	0x3E800000 // c1 (high 1 bit)
1395	.long	0x3DF5D211 // c1 (low single)
1396	.long	0xBD6542B3 // c2
1397	.long	0x3D0FE5E5 // c3
1398	.long	0xBC31FB14 // c4
1399	.long	0x3F2FEDDF // B' = pi/2 - B (high single)
1400	.long	0x329BBA77 // B' = pi/2 - B (low single)
1401	.long	0x3F800000 // tau (1 for cot path)
1402	.long	0xBE724E73 // c0 (high single)
1403	.long	0x3120C3E2 // c0 (low single)
1404	.long	0x3E800000 // c1 (high 1 bit)
1405	.long	0x3DF05283 // c1 (low single)
1406	.long	0xBD5AD45E // c2
1407	.long	0x3D0BAFBF // c3
1408	.long	0xBC27B8BB // c4
1409	.long	0x3F29A560 // B' = pi/2 - B (high single)
1410	.long	0x32F19861 // B' = pi/2 - B (low single)
1411	.long	0x3F800000 // tau (1 for cot path)
1412	.long	0xBE691B44 // c0 (high single)
1413	.long	0x31F18936 // c0 (low single)
1414	.long	0x3E800000 // c1 (high 1 bit)
1415	.long	0x3DEB138B // c1 (low single)
1416	.long	0xBD50B2F7 // c2
1417	.long	0x3D07BE3A // c3
1418	.long	0xBC1E46A7 // c4
1419	.long	0x3F235CE2 // B' = pi/2 - B (high single)
1420	.long	0xB2B889B6 // B' = pi/2 - B (low single)
1421	.long	0x3F800000 // tau (1 for cot path)
1422	.long	0xBE5FF82C // c0 (high single)
1423	.long	0xB170723A // c0 (low single)
1424	.long	0x3E800000 // c1 (high 1 bit)
1425	.long	0x3DE61354 // c1 (low single)
1426	.long	0xBD46DA06 // c2
1427	.long	0x3D0401F8 // c3
1428	.long	0xBC14E013 // c4
1429	.long	0x3F1D1463 // B' = pi/2 - B (high single)
1430	.long	0xB2455799 // B' = pi/2 - B (low single)
1431	.long	0x3F800000 // tau (1 for cot path)
1432	.long	0xBE56E46B // c0 (high single)
1433	.long	0x31E3F001 // c0 (low single)
1434	.long	0x3E800000 // c1 (high 1 bit)
1435	.long	0x3DE15025 // c1 (low single)
1436	.long	0xBD3D4550 // c2
1437	.long	0x3D00462D // c3
1438	.long	0xBC092C98 // c4
1439	.long	0x3F16CBE4 // B' = pi/2 - B (high single)
1440	.long	0xB0CCDE2E // B' = pi/2 - B (low single)
1441	.long	0x3F800000 // tau (1 for cot path)
1442	.long	0xBE4DDF41 // c0 (high single)
1443	.long	0xB1AEA094 // c0 (low single)
1444	.long	0x3E800000 // c1 (high 1 bit)
1445	.long	0x3DDCC85C // c1 (low single)
1446	.long	0xBD33F0BE // c2
1447	.long	0x3CFA23B0 // c3
1448	.long	0xBC01FCF7 // c4
1449	.long	0x3F108365 // B' = pi/2 - B (high single)
1450	.long	0x3212200D // B' = pi/2 - B (low single)
1451	.long	0x3F800000 // tau (1 for cot path)
1452	.long	0xBE44E7F8 // c0 (high single)
1453	.long	0xB1CAA3CB // c0 (low single)
1454	.long	0x3E800000 // c1 (high 1 bit)
1455	.long	0x3DD87A74 // c1 (low single)
1456	.long	0xBD2AD885 // c2
1457	.long	0x3CF3C785 // c3
1458	.long	0xBBF1E348 // c4
1459	.long	0x3F0A3AE6 // B' = pi/2 - B (high single)
1460	.long	0x329EEDF0 // B' = pi/2 - B (low single)
1461	.long	0x3F800000 // tau (1 for cot path)
1462	.long	0xBE3BFDDC // c0 (high single)
1463	.long	0xB132521A // c0 (low single)
1464	.long	0x3E800000 // c1 (high 1 bit)
1465	.long	0x3DD464FC // c1 (low single)
1466	.long	0xBD21F8F1 // c2
1467	.long	0x3CEE3076 // c3
1468	.long	0xBBE6D263 // c4
1469	.long	0x3F03F267 // B' = pi/2 - B (high single)
1470	.long	0x32F4CBD9 // B' = pi/2 - B (low single)
1471	.long	0x3F800000 // tau (1 for cot path)
1472	.long	0xBE33203E // c0 (high single)
1473	.long	0x31FEF5BE // c0 (low single)
1474	.long	0x3E800000 // c1 (high 1 bit)
1475	.long	0x3DD0869C // c1 (low single)
1476	.long	0xBD194E8C // c2
1477	.long	0x3CE8DCA9 // c3
1478	.long	0xBBDADA55 // c4
1479	.long	0x3EFB53D1 // B' = pi/2 - B (high single)
1480	.long	0x32155386 // B' = pi/2 - B (low single)
1481	.long	0x3F800000 // tau (1 for cot path)
1482	.long	0xBE2A4E71 // c0 (high single)
1483	.long	0xB19CFCEC // c0 (low single)
1484	.long	0x3E800000 // c1 (high 1 bit)
1485	.long	0x3DCCDE11 // c1 (low single)
1486	.long	0xBD10D605 // c2
1487	.long	0x3CE382A7 // c3
1488	.long	0xBBC8BD97 // c4
1489	.long	0x3EEEC2D4 // B' = pi/2 - B (high single)
1490	.long	0xB23EF0A7 // B' = pi/2 - B (low single)
1491	.long	0x3F800000 // tau (1 for cot path)
1492	.long	0xBE2187D0 // c0 (high single)
1493	.long	0xB1B7C7F7 // c0 (low single)
1494	.long	0x3E800000 // c1 (high 1 bit)
1495	.long	0x3DC96A2B // c1 (low single)
1496	.long	0xBD088C22 // c2
1497	.long	0x3CDE950E // c3
1498	.long	0xBBB89AD1 // c4
1499	.long	0x3EE231D6 // B' = pi/2 - B (high single)
1500	.long	0xB099A6A2 // B' = pi/2 - B (low single)
1501	.long	0x3F800000 // tau (1 for cot path)
1502	.long	0xBE18CBB7 // c0 (high single)
1503	.long	0xAFE28430 // c0 (low single)
1504	.long	0x3E800000 // c1 (high 1 bit)
1505	.long	0x3DC629CE // c1 (low single)
1506	.long	0xBD006DCD // c2
1507	.long	0x3CDA5A2C // c3
1508	.long	0xBBB0B3D2 // c4
1509	.long	0x3ED5A0D8 // B' = pi/2 - B (high single)
1510	.long	0x321886FF // B' = pi/2 - B (low single)
1511	.long	0x3F800000 // tau (1 for cot path)
1512	.long	0xBE101985 // c0 (high single)
1513	.long	0xB02FB2B8 // c0 (low single)
1514	.long	0x3E800000 // c1 (high 1 bit)
1515	.long	0x3DC31BF3 // c1 (low single)
1516	.long	0xBCF0F04D // c2
1517	.long	0x3CD60BC7 // c3
1518	.long	0xBBA138BA // c4
1519	.long	0x3EC90FDB // B' = pi/2 - B (high single)
1520	.long	0xB23BBD2E // B' = pi/2 - B (low single)
1521	.long	0x3F800000 // tau (1 for cot path)
1522	.long	0xBE07709D // c0 (high single)
1523	.long	0xB18A2A83 // c0 (low single)
1524	.long	0x3E800000 // c1 (high 1 bit)
1525	.long	0x3DC03FA2 // c1 (low single)
1526	.long	0xBCE15096 // c2
1527	.long	0x3CD26472 // c3
1528	.long	0xBB9A1270 // c4
1529	.long	0x3EBC7EDD // B' = pi/2 - B (high single)
1530	.long	0xB0800ADD // B' = pi/2 - B (low single)
1531	.long	0x3F800000 // tau (1 for cot path)
1532	.long	0xBDFDA0CB // c0 (high single)
1533	.long	0x2F14FCA0 // c0 (low single)
1534	.long	0x3E800000 // c1 (high 1 bit)
1535	.long	0x3DBD93F7 // c1 (low single)
1536	.long	0xBCD1F71B // c2
1537	.long	0x3CCEDD2B // c3
1538	.long	0xBB905946 // c4
1539	.long	0x3EAFEDDF // B' = pi/2 - B (high single)
1540	.long	0x321BBA77 // B' = pi/2 - B (low single)
1541	.long	0x3F800000 // tau (1 for cot path)
1542	.long	0xBDEC708C // c0 (high single)
1543	.long	0xB14895C4 // c0 (low single)
1544	.long	0x3E800000 // c1 (high 1 bit)
1545	.long	0x3DBB181E // c1 (low single)
1546	.long	0xBCC2DEA6 // c2
1547	.long	0x3CCB5027 // c3
1548	.long	0xBB7F3969 // c4
1549	.long	0x3EA35CE2 // B' = pi/2 - B (high single)
1550	.long	0xB23889B6 // B' = pi/2 - B (low single)
1551	.long	0x3F800000 // tau (1 for cot path)
1552	.long	0xBDDB4F55 // c0 (high single)
1553	.long	0x30F6437E // c0 (low single)
1554	.long	0x3E800000 // c1 (high 1 bit)
1555	.long	0x3DB8CB52 // c1 (low single)
1556	.long	0xBCB40210 // c2
1557	.long	0x3CC82D45 // c3
1558	.long	0xBB643075 // c4
1559	.long	0x3E96CBE4 // B' = pi/2 - B (high single)
1560	.long	0xB04CDE2E // B' = pi/2 - B (low single)
1561	.long	0x3F800000 // tau (1 for cot path)
1562	.long	0xBDCA3BFF // c0 (high single)
1563	.long	0x311C95EA // c0 (low single)
1564	.long	0x3E800000 // c1 (high 1 bit)
1565	.long	0x3DB6ACDE // c1 (low single)
1566	.long	0xBCA55C5B // c2
1567	.long	0x3CC5BC04 // c3
1568	.long	0xBB63A969 // c4
1569	.long	0x3E8A3AE6 // B' = pi/2 - B (high single)
1570	.long	0x321EEDF0 // B' = pi/2 - B (low single)
1571	.long	0x3F800000 // tau (1 for cot path)
1572	.long	0xBDB93569 // c0 (high single)
1573	.long	0xAFB9ED00 // c0 (low single)
1574	.long	0x3E800000 // c1 (high 1 bit)
1575	.long	0x3DB4BC1F // c1 (low single)
1576	.long	0xBC96E905 // c2
1577	.long	0x3CC2E6F5 // c3
1578	.long	0xBB3E10A6 // c4
1579	.long	0x3E7B53D1 // B' = pi/2 - B (high single)
1580	.long	0x31955386 // B' = pi/2 - B (low single)
1581	.long	0x3F800000 // tau (1 for cot path)
1582	.long	0xBDA83A77 // c0 (high single)
1583	.long	0x316D967A // c0 (low single)
1584	.long	0x3E800000 // c1 (high 1 bit)
1585	.long	0x3DB2F87C // c1 (low single)
1586	.long	0xBC88A31F // c2
1587	.long	0x3CC0E763 // c3
1588	.long	0xBB3F1666 // c4
1589	.long	0x3E6231D6 // B' = pi/2 - B (high single)
1590	.long	0xB019A6A2 // B' = pi/2 - B (low single)
1591	.long	0x3F800000 // tau (1 for cot path)
1592	.long	0xBD974A0D // c0 (high single)
1593	.long	0xB14F365B // c0 (low single)
1594	.long	0x3E800000 // c1 (high 1 bit)
1595	.long	0x3DB1616F // c1 (low single)
1596	.long	0xBC750CD8 // c2
1597	.long	0x3CBEB595 // c3
1598	.long	0xBB22B883 // c4
1599	.long	0x3E490FDB // B' = pi/2 - B (high single)
1600	.long	0xB1BBBD2E // B' = pi/2 - B (low single)
1601	.long	0x3F800000 // tau (1 for cot path)
1602	.long	0xBD866317 // c0 (high single)
1603	.long	0xAFF02140 // c0 (low single)
1604	.long	0x3E800000 // c1 (high 1 bit)
1605	.long	0x3DAFF67D // c1 (low single)
1606	.long	0xBC591CD0 // c2
1607	.long	0x3CBCBEAD // c3
1608	.long	0xBB04BBEC // c4
1609	.long	0x3E2FEDDF // B' = pi/2 - B (high single)
1610	.long	0x319BBA77 // B' = pi/2 - B (low single)
1611	.long	0x3F800000 // tau (1 for cot path)
1612	.long	0xBD6B08FF // c0 (high single)
1613	.long	0xB0EED236 // c0 (low single)
1614	.long	0x3E800000 // c1 (high 1 bit)
1615	.long	0x3DAEB739 // c1 (low single)
1616	.long	0xBC3D6D51 // c2
1617	.long	0x3CBB485D // c3
1618	.long	0xBAFFF5BA // c4
1619	.long	0x3E16CBE4 // B' = pi/2 - B (high single)
1620	.long	0xAFCCDE2E // B' = pi/2 - B (low single)
1621	.long	0x3F800000 // tau (1 for cot path)
1622	.long	0xBD495A6C // c0 (high single)
1623	.long	0xB0A427BD // c0 (low single)
1624	.long	0x3E800000 // c1 (high 1 bit)
1625	.long	0x3DADA345 // c1 (low single)
1626	.long	0xBC21F648 // c2
1627	.long	0x3CB9D1B4 // c3
1628	.long	0xBACB5567 // c4
1629	.long	0x3DFB53D1 // B' = pi/2 - B (high single)
1630	.long	0x31155386 // B' = pi/2 - B (low single)
1631	.long	0x3F800000 // tau (1 for cot path)
1632	.long	0xBD27B856 // c0 (high single)
1633	.long	0xB0F7EE91 // c0 (low single)
1634	.long	0x3E800000 // c1 (high 1 bit)
1635	.long	0x3DACBA4E // c1 (low single)
1636	.long	0xBC06AEE3 // c2
1637	.long	0x3CB8E5DC // c3
1638	.long	0xBAEC00EE // c4
1639	.long	0x3DC90FDB // B' = pi/2 - B (high single)
1640	.long	0xB13BBD2E // B' = pi/2 - B (low single)
1641	.long	0x3F800000 // tau (1 for cot path)
1642	.long	0xBD0620A3 // c0 (high single)
1643	.long	0xB0ECAB40 // c0 (low single)
1644	.long	0x3E800000 // c1 (high 1 bit)
1645	.long	0x3DABFC11 // c1 (low single)
1646	.long	0xBBD7200F // c2
1647	.long	0x3CB79475 // c3
1648	.long	0xBA2B0ADC // c4
1649	.long	0x3D96CBE4 // B' = pi/2 - B (high single)
1650	.long	0xAF4CDE2E // B' = pi/2 - B (low single)
1651	.long	0x3F800000 // tau (1 for cot path)
1652	.long	0xBCC92278 // c0 (high single)
1653	.long	0x302F2E68 // c0 (low single)
1654	.long	0x3E800000 // c1 (high 1 bit)
1655	.long	0x3DAB6854 // c1 (low single)
1656	.long	0xBBA1214F // c2
1657	.long	0x3CB6C1E9 // c3
1658	.long	0x3843C2F3 // c4
1659	.long	0x3D490FDB // B' = pi/2 - B (high single)
1660	.long	0xB0BBBD2E // B' = pi/2 - B (low single)
1661	.long	0x3F800000 // tau (1 for cot path)
1662	.long	0xBC861015 // c0 (high single)
1663	.long	0xAFD68E2E // c0 (low single)
1664	.long	0x3E800000 // c1 (high 1 bit)
1665	.long	0x3DAAFEEB // c1 (low single)
1666	.long	0xBB569F3F // c2
1667	.long	0x3CB6A84E // c3
1668	.long	0xBAC64194 // c4
1669	.long	0x3CC90FDB // B' = pi/2 - B (high single)
1670	.long	0xB03BBD2E // B' = pi/2 - B (low single)
1671	.long	0x3F800000 // tau (1 for cot path)
1672	.long	0xBC060BF3 // c0 (high single)
1673	.long	0x2FE251AE // c0 (low single)
1674	.long	0x3E800000 // c1 (high 1 bit)
1675	.long	0x3DAABFB9 // c1 (low single)
1676	.long	0xBAD67C60 // c2
1677	.long	0x3CB64CA5 // c3
1678	.long	0xBACDE881 // c4
1679	.long	0x00000000 // B' = pi/2 - B (high single)
1680	.long	0x00000000 // B' = pi/2 - B (low single)
1681	.long	0x3F800000 // tau (1 for cot path)
1682	.long	0x00000000 // c0 (high single)
1683	.long	0x00000000 // c0 (low single)
1684	.long	0x3E800000 // c1 (high 1 bit)
1685	.long	0x3DAAAAAB // c1 (low single)
1686	.long	0x00000000 // c2
1687	.long	0x3CB5E28B // c3
1688	.long	0x00000000 // c4
1689	.long	0xBCC90FDB // B' = pi/2 - B (high single)
1690	.long	0x303BBD2E // B' = pi/2 - B (low single)
1691	.long	0x3F800000 // tau (1 for cot path)
1692	.long	0x3C060BF3 // c0 (high single)
1693	.long	0xAFE251AE // c0 (low single)
1694	.long	0x3E800000 // c1 (high 1 bit)
1695	.long	0x3DAABFB9 // c1 (low single)
1696	.long	0x3AD67C60 // c2
1697	.long	0x3CB64CA5 // c3
1698	.long	0x3ACDE881 // c4
1699	.long	0xBD490FDB // B' = pi/2 - B (high single)
1700	.long	0x30BBBD2E // B' = pi/2 - B (low single)
1701	.long	0x3F800000 // tau (1 for cot path)
1702	.long	0x3C861015 // c0 (high single)
1703	.long	0x2FD68E2E // c0 (low single)
1704	.long	0x3E800000 // c1 (high 1 bit)
1705	.long	0x3DAAFEEB // c1 (low single)
1706	.long	0x3B569F3F // c2
1707	.long	0x3CB6A84E // c3
1708	.long	0x3AC64194 // c4
1709	.long	0xBD96CBE4 // B' = pi/2 - B (high single)
1710	.long	0x2F4CDE2E // B' = pi/2 - B (low single)
1711	.long	0x3F800000 // tau (1 for cot path)
1712	.long	0x3CC92278 // c0 (high single)
1713	.long	0xB02F2E68 // c0 (low single)
1714	.long	0x3E800000 // c1 (high 1 bit)
1715	.long	0x3DAB6854 // c1 (low single)
1716	.long	0x3BA1214F // c2
1717	.long	0x3CB6C1E9 // c3
1718	.long	0xB843C2F2 // c4
1719	.long	0xBDC90FDB // B' = pi/2 - B (high single)
1720	.long	0x313BBD2E // B' = pi/2 - B (low single)
1721	.long	0x3F800000 // tau (1 for cot path)
1722	.long	0x3D0620A3 // c0 (high single)
1723	.long	0x30ECAB40 // c0 (low single)
1724	.long	0x3E800000 // c1 (high 1 bit)
1725	.long	0x3DABFC11 // c1 (low single)
1726	.long	0x3BD7200F // c2
1727	.long	0x3CB79475 // c3
1728	.long	0x3A2B0ADC // c4
1729	.long	0xBDFB53D1 // B' = pi/2 - B (high single)
1730	.long	0xB1155386 // B' = pi/2 - B (low single)
1731	.long	0x3F800000 // tau (1 for cot path)
1732	.long	0x3D27B856 // c0 (high single)
1733	.long	0x30F7EE91 // c0 (low single)
1734	.long	0x3E800000 // c1 (high 1 bit)
1735	.long	0x3DACBA4E // c1 (low single)
1736	.long	0x3C06AEE3 // c2
1737	.long	0x3CB8E5DC // c3
1738	.long	0x3AEC00EE // c4
1739	.long	0xBE16CBE4 // B' = pi/2 - B (high single)
1740	.long	0x2FCCDE2E // B' = pi/2 - B (low single)
1741	.long	0x3F800000 // tau (1 for cot path)
1742	.long	0x3D495A6C // c0 (high single)
1743	.long	0x30A427BD // c0 (low single)
1744	.long	0x3E800000 // c1 (high 1 bit)
1745	.long	0x3DADA345 // c1 (low single)
1746	.long	0x3C21F648 // c2
1747	.long	0x3CB9D1B4 // c3
1748	.long	0x3ACB5567 // c4
1749	.long	0xBE2FEDDF // B' = pi/2 - B (high single)
1750	.long	0xB19BBA77 // B' = pi/2 - B (low single)
1751	.long	0x3F800000 // tau (1 for cot path)
1752	.long	0x3D6B08FF // c0 (high single)
1753	.long	0x30EED236 // c0 (low single)
1754	.long	0x3E800000 // c1 (high 1 bit)
1755	.long	0x3DAEB739 // c1 (low single)
1756	.long	0x3C3D6D51 // c2
1757	.long	0x3CBB485D // c3
1758	.long	0x3AFFF5BA // c4
1759	.long	0xBE490FDB // B' = pi/2 - B (high single)
1760	.long	0x31BBBD2E // B' = pi/2 - B (low single)
1761	.long	0x3F800000 // tau (1 for cot path)
1762	.long	0x3D866317 // c0 (high single)
1763	.long	0x2FF02140 // c0 (low single)
1764	.long	0x3E800000 // c1 (high 1 bit)
1765	.long	0x3DAFF67D // c1 (low single)
1766	.long	0x3C591CD0 // c2
1767	.long	0x3CBCBEAD // c3
1768	.long	0x3B04BBEC // c4
1769	.long	0xBE6231D6 // B' = pi/2 - B (high single)
1770	.long	0x3019A6A2 // B' = pi/2 - B (low single)
1771	.long	0x3F800000 // tau (1 for cot path)
1772	.long	0x3D974A0D // c0 (high single)
1773	.long	0x314F365B // c0 (low single)
1774	.long	0x3E800000 // c1 (high 1 bit)
1775	.long	0x3DB1616F // c1 (low single)
1776	.long	0x3C750CD8 // c2
1777	.long	0x3CBEB595 // c3
1778	.long	0x3B22B883 // c4
1779	.long	0xBE7B53D1 // B' = pi/2 - B (high single)
1780	.long	0xB1955386 // B' = pi/2 - B (low single)
1781	.long	0x3F800000 // tau (1 for cot path)
1782	.long	0x3DA83A77 // c0 (high single)
1783	.long	0xB16D967A // c0 (low single)
1784	.long	0x3E800000 // c1 (high 1 bit)
1785	.long	0x3DB2F87C // c1 (low single)
1786	.long	0x3C88A31F // c2
1787	.long	0x3CC0E763 // c3
1788	.long	0x3B3F1666 // c4
1789	.long	0xBE8A3AE6 // B' = pi/2 - B (high single)
1790	.long	0xB21EEDF0 // B' = pi/2 - B (low single)
1791	.long	0x3F800000 // tau (1 for cot path)
1792	.long	0x3DB93569 // c0 (high single)
1793	.long	0x2FB9ED00 // c0 (low single)
1794	.long	0x3E800000 // c1 (high 1 bit)
1795	.long	0x3DB4BC1F // c1 (low single)
1796	.long	0x3C96E905 // c2
1797	.long	0x3CC2E6F5 // c3
1798	.long	0x3B3E10A6 // c4
1799	.long	0xBE96CBE4 // B' = pi/2 - B (high single)
1800	.long	0x304CDE2E // B' = pi/2 - B (low single)
1801	.long	0x3F800000 // tau (1 for cot path)
1802	.long	0x3DCA3BFF // c0 (high single)
1803	.long	0xB11C95EA // c0 (low single)
1804	.long	0x3E800000 // c1 (high 1 bit)
1805	.long	0x3DB6ACDE // c1 (low single)
1806	.long	0x3CA55C5B // c2
1807	.long	0x3CC5BC04 // c3
1808	.long	0x3B63A969 // c4
1809	.long	0xBEA35CE2 // B' = pi/2 - B (high single)
1810	.long	0x323889B6 // B' = pi/2 - B (low single)
1811	.long	0x3F800000 // tau (1 for cot path)
1812	.long	0x3DDB4F55 // c0 (high single)
1813	.long	0xB0F6437E // c0 (low single)
1814	.long	0x3E800000 // c1 (high 1 bit)
1815	.long	0x3DB8CB52 // c1 (low single)
1816	.long	0x3CB40210 // c2
1817	.long	0x3CC82D45 // c3
1818	.long	0x3B643075 // c4
1819	.long	0xBEAFEDDF // B' = pi/2 - B (high single)
1820	.long	0xB21BBA77 // B' = pi/2 - B (low single)
1821	.long	0x3F800000 // tau (1 for cot path)
1822	.long	0x3DEC708C // c0 (high single)
1823	.long	0x314895C4 // c0 (low single)
1824	.long	0x3E800000 // c1 (high 1 bit)
1825	.long	0x3DBB181E // c1 (low single)
1826	.long	0x3CC2DEA6 // c2
1827	.long	0x3CCB5027 // c3
1828	.long	0x3B7F3969 // c4
1829	.long	0xBEBC7EDD // B' = pi/2 - B (high single)
1830	.long	0x30800ADD // B' = pi/2 - B (low single)
1831	.long	0x3F800000 // tau (1 for cot path)
1832	.long	0x3DFDA0CB // c0 (high single)
1833	.long	0xAF14FCA0 // c0 (low single)
1834	.long	0x3E800000 // c1 (high 1 bit)
1835	.long	0x3DBD93F7 // c1 (low single)
1836	.long	0x3CD1F71B // c2
1837	.long	0x3CCEDD2B // c3
1838	.long	0x3B905946 // c4
1839	.long	0xBEC90FDB // B' = pi/2 - B (high single)
1840	.long	0x323BBD2E // B' = pi/2 - B (low single)
1841	.long	0x3F800000 // tau (1 for cot path)
1842	.long	0x3E07709D // c0 (high single)
1843	.long	0x318A2A83 // c0 (low single)
1844	.long	0x3E800000 // c1 (high 1 bit)
1845	.long	0x3DC03FA2 // c1 (low single)
1846	.long	0x3CE15096 // c2
1847	.long	0x3CD26472 // c3
1848	.long	0x3B9A1270 // c4
1849	.long	0xBED5A0D8 // B' = pi/2 - B (high single)
1850	.long	0xB21886FF // B' = pi/2 - B (low single)
1851	.long	0x3F800000 // tau (1 for cot path)
1852	.long	0x3E101985 // c0 (high single)
1853	.long	0x302FB2B8 // c0 (low single)
1854	.long	0x3E800000 // c1 (high 1 bit)
1855	.long	0x3DC31BF3 // c1 (low single)
1856	.long	0x3CF0F04D // c2
1857	.long	0x3CD60BC7 // c3
1858	.long	0x3BA138BA // c4
1859	.long	0xBEE231D6 // B' = pi/2 - B (high single)
1860	.long	0x3099A6A2 // B' = pi/2 - B (low single)
1861	.long	0x3F800000 // tau (1 for cot path)
1862	.long	0x3E18CBB7 // c0 (high single)
1863	.long	0x2FE28430 // c0 (low single)
1864	.long	0x3E800000 // c1 (high 1 bit)
1865	.long	0x3DC629CE // c1 (low single)
1866	.long	0x3D006DCD // c2
1867	.long	0x3CDA5A2C // c3
1868	.long	0x3BB0B3D2 // c4
1869	.long	0xBEEEC2D4 // B' = pi/2 - B (high single)
1870	.long	0x323EF0A7 // B' = pi/2 - B (low single)
1871	.long	0x3F800000 // tau (1 for cot path)
1872	.long	0x3E2187D0 // c0 (high single)
1873	.long	0x31B7C7F7 // c0 (low single)
1874	.long	0x3E800000 // c1 (high 1 bit)
1875	.long	0x3DC96A2B // c1 (low single)
1876	.long	0x3D088C22 // c2
1877	.long	0x3CDE950E // c3
1878	.long	0x3BB89AD1 // c4
1879	.long	0xBEFB53D1 // B' = pi/2 - B (high single)
1880	.long	0xB2155386 // B' = pi/2 - B (low single)
1881	.long	0x3F800000 // tau (1 for cot path)
1882	.long	0x3E2A4E71 // c0 (high single)
1883	.long	0x319CFCEC // c0 (low single)
1884	.long	0x3E800000 // c1 (high 1 bit)
1885	.long	0x3DCCDE11 // c1 (low single)
1886	.long	0x3D10D605 // c2
1887	.long	0x3CE382A7 // c3
1888	.long	0x3BC8BD97 // c4
1889	.long	0xBF03F267 // B' = pi/2 - B (high single)
1890	.long	0xB2F4CBD9 // B' = pi/2 - B (low single)
1891	.long	0x3F800000 // tau (1 for cot path)
1892	.long	0x3E33203E // c0 (high single)
1893	.long	0xB1FEF5BE // c0 (low single)
1894	.long	0x3E800000 // c1 (high 1 bit)
1895	.long	0x3DD0869C // c1 (low single)
1896	.long	0x3D194E8C // c2
1897	.long	0x3CE8DCA9 // c3
1898	.long	0x3BDADA55 // c4
1899	.long	0xBF0A3AE6 // B' = pi/2 - B (high single)
1900	.long	0xB29EEDF0 // B' = pi/2 - B (low single)
1901	.long	0x3F800000 // tau (1 for cot path)
1902	.long	0x3E3BFDDC // c0 (high single)
1903	.long	0x3132521A // c0 (low single)
1904	.long	0x3E800000 // c1 (high 1 bit)
1905	.long	0x3DD464FC // c1 (low single)
1906	.long	0x3D21F8F1 // c2
1907	.long	0x3CEE3076 // c3
1908	.long	0x3BE6D263 // c4
1909	.long	0xBF108365 // B' = pi/2 - B (high single)
1910	.long	0xB212200D // B' = pi/2 - B (low single)
1911	.long	0x3F800000 // tau (1 for cot path)
1912	.long	0x3E44E7F8 // c0 (high single)
1913	.long	0x31CAA3CB // c0 (low single)
1914	.long	0x3E800000 // c1 (high 1 bit)
1915	.long	0x3DD87A74 // c1 (low single)
1916	.long	0x3D2AD885 // c2
1917	.long	0x3CF3C785 // c3
1918	.long	0x3BF1E348 // c4
1919	.long	0xBF16CBE4 // B' = pi/2 - B (high single)
1920	.long	0x30CCDE2E // B' = pi/2 - B (low single)
1921	.long	0x3F800000 // tau (1 for cot path)
1922	.long	0x3E4DDF41 // c0 (high single)
1923	.long	0x31AEA094 // c0 (low single)
1924	.long	0x3E800000 // c1 (high 1 bit)
1925	.long	0x3DDCC85C // c1 (low single)
1926	.long	0x3D33F0BE // c2
1927	.long	0x3CFA23B0 // c3
1928	.long	0x3C01FCF7 // c4
1929	.long	0xBF1D1463 // B' = pi/2 - B (high single)
1930	.long	0x32455799 // B' = pi/2 - B (low single)
1931	.long	0x3F800000 // tau (1 for cot path)
1932	.long	0x3E56E46B // c0 (high single)
1933	.long	0xB1E3F001 // c0 (low single)
1934	.long	0x3E800000 // c1 (high 1 bit)
1935	.long	0x3DE15025 // c1 (low single)
1936	.long	0x3D3D4550 // c2
1937	.long	0x3D00462D // c3
1938	.long	0x3C092C98 // c4
1939	.long	0xBF235CE2 // B' = pi/2 - B (high single)
1940	.long	0x32B889B6 // B' = pi/2 - B (low single)
1941	.long	0x3F800000 // tau (1 for cot path)
1942	.long	0x3E5FF82C // c0 (high single)
1943	.long	0x3170723A // c0 (low single)
1944	.long	0x3E800000 // c1 (high 1 bit)
1945	.long	0x3DE61354 // c1 (low single)
1946	.long	0x3D46DA06 // c2
1947	.long	0x3D0401F8 // c3
1948	.long	0x3C14E013 // c4
1949	.long	0xBF29A560 // B' = pi/2 - B (high single)
1950	.long	0xB2F19861 // B' = pi/2 - B (low single)
1951	.long	0x3F800000 // tau (1 for cot path)
1952	.long	0x3E691B44 // c0 (high single)
1953	.long	0xB1F18936 // c0 (low single)
1954	.long	0x3E800000 // c1 (high 1 bit)
1955	.long	0x3DEB138B // c1 (low single)
1956	.long	0x3D50B2F7 // c2
1957	.long	0x3D07BE3A // c3
1958	.long	0x3C1E46A7 // c4
1959	.long	0xBF2FEDDF // B' = pi/2 - B (high single)
1960	.long	0xB29BBA77 // B' = pi/2 - B (low single)
1961	.long	0x3F800000 // tau (1 for cot path)
1962	.long	0x3E724E73 // c0 (high single)
1963	.long	0xB120C3E2 // c0 (low single)
1964	.long	0x3E800000 // c1 (high 1 bit)
1965	.long	0x3DF05283 // c1 (low single)
1966	.long	0x3D5AD45E // c2
1967	.long	0x3D0BAFBF // c3
1968	.long	0x3C27B8BB // c4
1969	.long	0xBF36365E // B' = pi/2 - B (high single)
1970	.long	0xB20BB91C // B' = pi/2 - B (low single)
1971	.long	0x3F800000 // tau (1 for cot path)
1972	.long	0x3E7B9282 // c0 (high single)
1973	.long	0x313383D2 // c0 (low single)
1974	.long	0x3E800000 // c1 (high 1 bit)
1975	.long	0x3DF5D211 // c1 (low single)
1976	.long	0x3D6542B3 // c2
1977	.long	0x3D0FE5E5 // c3
1978	.long	0x3C31FB14 // c4
1979	.long	0xBF3C7EDD // B' = pi/2 - B (high single)
1980	.long	0x31000ADD // B' = pi/2 - B (low single)
1981	.long	0x3F800000 // tau (1 for cot path)
1982	.long	0x3E827420 // c0 (high single)
1983	.long	0xB20B8B4D // c0 (low single)
1984	.long	0x3E800000 // c1 (high 1 bit)
1985	.long	0x3DFB9428 // c1 (low single)
1986	.long	0x3D7002B4 // c2
1987	.long	0x3D142A6C // c3
1988	.long	0x3C3A47FF // c4
1989	.long	0xBF42C75C // B' = pi/2 - B (high single)
1990	.long	0x324BBE8A // B' = pi/2 - B (low single)
1991	.long	0x3F800000 // tau (1 for cot path)
1992	.long	0x3E87283F // c0 (high single)
1993	.long	0x3268B966 // c0 (low single)
1994	.long	0x3F000000 // c1 (high 1 bit)
1995	.long	0xBDFE6529 // c1 (low single)
1996	.long	0x3D7B1953 // c2
1997	.long	0x3D18E109 // c3
1998	.long	0x3C4570B0 // c4
1999	.long	0xBF490FDB // B' = pi/2 - B (high single)
2000	.long	0x32BBBD2E // B' = pi/2 - B (low single)
2001	.long	0x00000000 // tau (1 for cot path)
2002	.long	0xBF800000 // c0 (high single)
2003	.long	0x2B410000 // c0 (low single)
2004	.long	0x40000000 // c1 (high 1 bit)
2005	.long	0xB3000000 // c1 (low single)
2006	.long	0xC0000000 // c2
2007	.long	0x402AB7C8 // c3
2008	.long	0xC05561DB // c4
2009	.long	0xBF4F5859 // B' = pi/2 - B (high single)
2010	.long	0xB2EE64E8 // B' = pi/2 - B (low single)
2011	.long	0x00000000 // tau (1 for cot path)
2012	.long	0xBF73BB75 // c0 (high single)
2013	.long	0xB2FC908D // c0 (low single)
2014	.long	0x40000000 // c1 (high 1 bit)
2015	.long	0xBDBF94B0 // c1 (low single)
2016	.long	0xBFE8550F // c2
2017	.long	0x40174F67 // c3
2018	.long	0xC036C608 // c4
2019	.long	0xBF55A0D8 // B' = pi/2 - B (high single)
2020	.long	0xB29886FF // B' = pi/2 - B (low single)
2021	.long	0x00000000 // tau (1 for cot path)
2022	.long	0xBF68065E // c0 (high single)
2023	.long	0xB2670D1A // c0 (low single)
2024	.long	0x40000000 // c1 (high 1 bit)
2025	.long	0xBE36D1D6 // c1 (low single)
2026	.long	0xBFD35007 // c2
2027	.long	0x4006A861 // c3
2028	.long	0xC01D4BDA // c4
2029	.long	0xBF5BE957 // B' = pi/2 - B (high single)
2030	.long	0xB205522A // B' = pi/2 - B (low single)
2031	.long	0x00000000 // tau (1 for cot path)
2032	.long	0xBF5CD3BE // c0 (high single)
2033	.long	0xB1460308 // c0 (low single)
2034	.long	0x40000000 // c1 (high 1 bit)
2035	.long	0xBE8306C5 // c1 (low single)
2036	.long	0xBFC09232 // c2
2037	.long	0x3FF09632 // c3
2038	.long	0xC007DB00 // c4
2039	.long	0xBF6231D6 // B' = pi/2 - B (high single)
2040	.long	0x3119A6A2 // B' = pi/2 - B (low single)
2041	.long	0x00000000 // tau (1 for cot path)
2042	.long	0xBF521801 // c0 (high single)
2043	.long	0xB2AE4178 // c0 (low single)
2044	.long	0x40000000 // c1 (high 1 bit)
2045	.long	0xBEA72938 // c1 (low single)
2046	.long	0xBFAFCC22 // c2
2047	.long	0x3FD7BD4A // c3
2048	.long	0xBFEBB01B // c4
2049	.long	0xBF687A55 // B' = pi/2 - B (high single)
2050	.long	0x3252257B // B' = pi/2 - B (low single)
2051	.long	0x00000000 // tau (1 for cot path)
2052	.long	0xBF47C8CC // c0 (high single)
2053	.long	0x3200F51A // c0 (low single)
2054	.long	0x40000000 // c1 (high 1 bit)
2055	.long	0xBEC82C6C // c1 (low single)
2056	.long	0xBFA0BAE9 // c2
2057	.long	0x3FC2252F // c3
2058	.long	0xBFCD24C7 // c4
2059	.long	0xBF6EC2D4 // B' = pi/2 - B (high single)
2060	.long	0x32BEF0A7 // B' = pi/2 - B (low single)
2061	.long	0x00000000 // tau (1 for cot path)
2062	.long	0xBF3DDCCF // c0 (high single)
2063	.long	0xB2D29606 // c0 (low single)
2064	.long	0x40000000 // c1 (high 1 bit)
2065	.long	0xBEE6606F // c1 (low single)
2066	.long	0xBF9325D6 // c2
2067	.long	0x3FAF4E69 // c3
2068	.long	0xBFB3080C // c4
2069	.long	0xBF750B52 // B' = pi/2 - B (high single)
2070	.long	0xB2EB316F // B' = pi/2 - B (low single)
2071	.long	0x00000000 // tau (1 for cot path)
2072	.long	0xBF344BA9 // c0 (high single)
2073	.long	0x32B8B0EA // c0 (low single)
2074	.long	0x3F800000 // c1 (high 1 bit)
2075	.long	0x3EFDF4F7 // c1 (low single)
2076	.long	0xBF86DCA8 // c2
2077	.long	0x3F9ED53B // c3
2078	.long	0xBF9CBEDE // c4
2079	.long	0xBF7B53D1 // B' = pi/2 - B (high single)
2080	.long	0xB2955386 // B' = pi/2 - B (low single)
2081	.long	0x00000000 // tau (1 for cot path)
2082	.long	0xBF2B0DC1 // c0 (high single)
2083	.long	0xB2AB7EBA // c0 (low single)
2084	.long	0x3F800000 // c1 (high 1 bit)
2085	.long	0x3EE496C2 // c1 (low single)
2086	.long	0xBF776C40 // c2
2087	.long	0x3F9065C1 // c3
2088	.long	0xBF89AFB6 // c4
2089	.long	0xBF80CE28 // B' = pi/2 - B (high single)
2090	.long	0xB1FDD672 // B' = pi/2 - B (low single)
2091	.long	0x00000000 // tau (1 for cot path)
2092	.long	0xBF221C37 // c0 (high single)
2093	.long	0x320C61DC // c0 (low single)
2094	.long	0x3F800000 // c1 (high 1 bit)
2095	.long	0x3ECD4F71 // c1 (low single)
2096	.long	0xBF631DAA // c2
2097	.long	0x3F83B471 // c3
2098	.long	0xBF7281EA // c4
2099	.long	0xBF83F267 // B' = pi/2 - B (high single)
2100	.long	0xB374CBD9 // B' = pi/2 - B (low single)
2101	.long	0x00000000 // tau (1 for cot path)
2102	.long	0xBF1970C4 // c0 (high single)
2103	.long	0xB2904848 // c0 (low single)
2104	.long	0x3F800000 // c1 (high 1 bit)
2105	.long	0x3EB7EFF8 // c1 (low single)
2106	.long	0xBF50907C // c2
2107	.long	0x3F710FEA // c3
2108	.long	0xBF561FED // c4
2109	.long	0xBF8716A7 // B' = pi/2 - B (high single)
2110	.long	0x32588C6D // B' = pi/2 - B (low single)
2111	.long	0x00000000 // tau (1 for cot path)
2112	.long	0xBF1105AF // c0 (high single)
2113	.long	0xB2F045B0 // c0 (low single)
2114	.long	0x3F800000 // c1 (high 1 bit)
2115	.long	0x3EA44EE2 // c1 (low single)
2116	.long	0xBF3F8FDB // c2
2117	.long	0x3F5D3FD0 // c3
2118	.long	0xBF3D0A23 // c4
2119	.long	0xBF8A3AE6 // B' = pi/2 - B (high single)
2120	.long	0xB31EEDF0 // B' = pi/2 - B (low single)
2121	.long	0x00000000 // tau (1 for cot path)
2122	.long	0xBF08D5B9 // c0 (high single)
2123	.long	0x325EF98E // c0 (low single)
2124	.long	0x3F800000 // c1 (high 1 bit)
2125	.long	0x3E92478D // c1 (low single)
2126	.long	0xBF2FEDC9 // c2
2127	.long	0x3F4BCD58 // c3
2128	.long	0xBF27AE9E // c4
2129	.long	0xBF8D5F26 // B' = pi/2 - B (high single)
2130	.long	0x330C0105 // B' = pi/2 - B (low single)
2131	.long	0x00000000 // tau (1 for cot path)
2132	.long	0xBF00DC0D // c0 (high single)
2133	.long	0x3214AF72 // c0 (low single)
2134	.long	0x3F800000 // c1 (high 1 bit)
2135	.long	0x3E81B994 // c1 (low single)
2136	.long	0xBF218233 // c2
2137	.long	0x3F3C4531 // c3
2138	.long	0xBF149688 // c4
2139	.long	0xBF908365 // B' = pi/2 - B (high single)
2140	.long	0xB292200D // B' = pi/2 - B (low single)
2141	.long	0x00000000 // tau (1 for cot path)
2142	.long	0xBEF22870 // c0 (high single)
2143	.long	0xB25271F4 // c0 (low single)
2144	.long	0x3F800000 // c1 (high 1 bit)
2145	.long	0x3E65107A // c1 (low single)
2146	.long	0xBF1429F0 // c2
2147	.long	0x3F2E8AFC // c3
2148	.long	0xBF040498 // c4
2149	.long	0xBF93A7A5 // B' = pi/2 - B (high single)
2150	.long	0x3361DEEE // B' = pi/2 - B (low single)
2151	.long	0x00000000 // tau (1 for cot path)
2152	.long	0xBEE2F439 // c0 (high single)
2153	.long	0x31F4399E // c0 (low single)
2154	.long	0x3F800000 // c1 (high 1 bit)
2155	.long	0x3E49341C // c1 (low single)
2156	.long	0xBF07C61A // c2
2157	.long	0x3F22560F // c3
2158	.long	0xBEEAA81E // c4
2159	.long	0xBF96CBE4 // B' = pi/2 - B (high single)
2160	.long	0x314CDE2E // B' = pi/2 - B (low single)
2161	.long	0x00000000 // tau (1 for cot path)
2162	.long	0xBED413CD // c0 (high single)
2163	.long	0x31C06152 // c0 (low single)
2164	.long	0x3F800000 // c1 (high 1 bit)
2165	.long	0x3E2FB0CC // c1 (low single)
2166	.long	0xBEF876CB // c2
2167	.long	0x3F177807 // c3
2168	.long	0xBED08437 // c4
2169	.long	0xBF99F023 // B' = pi/2 - B (high single)
2170	.long	0xB3484328 // B' = pi/2 - B (low single)
2171	.long	0x00000000 // tau (1 for cot path)
2172	.long	0xBEC5800D // c0 (high single)
2173	.long	0x3214C3C1 // c0 (low single)
2174	.long	0x3F800000 // c1 (high 1 bit)
2175	.long	0x3E185E54 // c1 (low single)
2176	.long	0xBEE2E342 // c2
2177	.long	0x3F0DCA73 // c3
2178	.long	0xBEB8CC21 // c4
2179	.long	0xBF9D1463 // B' = pi/2 - B (high single)
2180	.long	0x32C55799 // B' = pi/2 - B (low single)
2181	.long	0x00000000 // tau (1 for cot path)
2182	.long	0xBEB73250 // c0 (high single)
2183	.long	0x32028823 // c0 (low single)
2184	.long	0x3F800000 // c1 (high 1 bit)
2185	.long	0x3E0318F8 // c1 (low single)
2186	.long	0xBECEA678 // c2
2187	.long	0x3F053C67 // c3
2188	.long	0xBEA41E53 // c4
2189	.long	0xBFA038A2 // B' = pi/2 - B (high single)
2190	.long	0xB2E4CA7E // B' = pi/2 - B (low single)
2191	.long	0x00000000 // tau (1 for cot path)
2192	.long	0xBEA92457 // c0 (high single)
2193	.long	0xB0B80830 // c0 (low single)
2194	.long	0x3F800000 // c1 (high 1 bit)
2195	.long	0x3DDF8200 // c1 (low single)
2196	.long	0xBEBB99E9 // c2
2197	.long	0x3EFB4AA8 // c3
2198	.long	0xBE9182BE // c4
2199	.long	0xBFA35CE2 // B' = pi/2 - B (high single)
2200	.long	0x333889B6 // B' = pi/2 - B (low single)
2201	.long	0x00000000 // tau (1 for cot path)
2202	.long	0xBE9B5042 // c0 (high single)
2203	.long	0x322A3AEE // c0 (low single)
2204	.long	0x3F800000 // c1 (high 1 bit)
2205	.long	0x3DBC7490 // c1 (low single)
2206	.long	0xBEA99AF5 // c2
2207	.long	0x3EEDE107 // c3
2208	.long	0xBE80E9AA // c4
2209	.long	0xBFA68121 // B' = pi/2 - B (high single)
2210	.long	0xB1E43AAC // B' = pi/2 - B (low single)
2211	.long	0x00000000 // tau (1 for cot path)
2212	.long	0xBE8DB082 // c0 (high single)
2213	.long	0x3132A234 // c0 (low single)
2214	.long	0x3F800000 // c1 (high 1 bit)
2215	.long	0x3D9CD7D0 // c1 (low single)
2216	.long	0xBE988A60 // c2
2217	.long	0x3EE203E3 // c3
2218	.long	0xBE63582C // c4
2219	.long	0xBFA9A560 // B' = pi/2 - B (high single)
2220	.long	0xB3719861 // B' = pi/2 - B (low single)
2221	.long	0x00000000 // tau (1 for cot path)
2222	.long	0xBE803FD4 // c0 (high single)
2223	.long	0x32279E66 // c0 (low single)
2224	.long	0x3F800000 // c1 (high 1 bit)
2225	.long	0x3D807FC8 // c1 (low single)
2226	.long	0xBE884BD4 // c2
2227	.long	0x3ED7812D // c3
2228	.long	0xBE4636EB // c4
2229	.long	0xBFACC9A0 // B' = pi/2 - B (high single)
2230	.long	0x32655A50 // B' = pi/2 - B (low single)
2231	.long	0x00000000 // tau (1 for cot path)
2232	.long	0xBE65F267 // c0 (high single)
2233	.long	0xB1B4B1DF // c0 (low single)
2234	.long	0x3F800000 // c1 (high 1 bit)
2235	.long	0x3D4E8B90 // c1 (low single)
2236	.long	0xBE718ACA // c2
2237	.long	0x3ECE7164 // c3
2238	.long	0xBE2DC161 // c4
2239	.long	0xBFAFEDDF // B' = pi/2 - B (high single)
2240	.long	0xB31BBA77 // B' = pi/2 - B (low single)
2241	.long	0x00000000 // tau (1 for cot path)
2242	.long	0xBE4BAFAF // c0 (high single)
2243	.long	0xAF2A29E0 // c0 (low single)
2244	.long	0x3F800000 // c1 (high 1 bit)
2245	.long	0x3D221018 // c1 (low single)
2246	.long	0xBE53BED0 // c2
2247	.long	0x3EC67E26 // c3
2248	.long	0xBE1568E2 // c4
2249	.long	0xBFB3121F // B' = pi/2 - B (high single)
2250	.long	0x330F347D // B' = pi/2 - B (low single)
2251	.long	0x00000000 // tau (1 for cot path)
2252	.long	0xBE31AE4D // c0 (high single)
2253	.long	0x31F32251 // c0 (low single)
2254	.long	0x3F800000 // c1 (high 1 bit)
2255	.long	0x3CF6A500 // c1 (low single)
2256	.long	0xBE3707DA // c2
2257	.long	0x3EBFA489 // c3
2258	.long	0xBDFBD9C7 // c4
2259	.long	0xBFB6365E // B' = pi/2 - B (high single)
2260	.long	0xB28BB91C // B' = pi/2 - B (low single)
2261	.long	0x00000000 // tau (1 for cot path)
2262	.long	0xBE17E564 // c0 (high single)
2263	.long	0x31C5A2E4 // c0 (low single)
2264	.long	0x3F800000 // c1 (high 1 bit)
2265	.long	0x3CB440D0 // c1 (low single)
2266	.long	0xBE1B3D00 // c2
2267	.long	0x3EB9F664 // c3
2268	.long	0xBDD647C0 // c4
2269	.long	0xBFB95A9E // B' = pi/2 - B (high single)
2270	.long	0x33651267 // B' = pi/2 - B (low single)
2271	.long	0x00000000 // tau (1 for cot path)
2272	.long	0xBDFC98C2 // c0 (high single)
2273	.long	0x30AE525C // c0 (low single)
2274	.long	0x3F800000 // c1 (high 1 bit)
2275	.long	0x3C793D20 // c1 (low single)
2276	.long	0xBE003845 // c2
2277	.long	0x3EB5271F // c3
2278	.long	0xBDAC669E // c4
2279	.long	0xBFBC7EDD // B' = pi/2 - B (high single)
2280	.long	0x31800ADD // B' = pi/2 - B (low single)
2281	.long	0x00000000 // tau (1 for cot path)
2282	.long	0xBDC9B5DC // c0 (high single)
2283	.long	0xB145AD86 // c0 (low single)
2284	.long	0x3F800000 // c1 (high 1 bit)
2285	.long	0x3C1EEF20 // c1 (low single)
2286	.long	0xBDCBAAEA // c2
2287	.long	0x3EB14E5E // c3
2288	.long	0xBD858BB2 // c4
2289	.long	0xBFBFA31C // B' = pi/2 - B (high single)
2290	.long	0xB3450FB0 // B' = pi/2 - B (low single)
2291	.long	0x00000000 // tau (1 for cot path)
2292	.long	0xBD9711CE // c0 (high single)
2293	.long	0xB14FEB28 // c0 (low single)
2294	.long	0x3F800000 // c1 (high 1 bit)
2295	.long	0x3BB24C00 // c1 (low single)
2296	.long	0xBD97E43A // c2
2297	.long	0x3EAE6A89 // c3
2298	.long	0xBD4D07E0 // c4
2299	.long	0xBFC2C75C // B' = pi/2 - B (high single)
2300	.long	0x32CBBE8A // B' = pi/2 - B (low single)
2301	.long	0x00000000 // tau (1 for cot path)
2302	.long	0xBD49393C // c0 (high single)
2303	.long	0xB0A39F5B // c0 (low single)
2304	.long	0x3F800000 // c1 (high 1 bit)
2305	.long	0x3B1E2B00 // c1 (low single)
2306	.long	0xBD49B5D4 // c2
2307	.long	0x3EAC4F10 // c3
2308	.long	0xBCFD9425 // c4
2309	.long	0xBFC5EB9B // B' = pi/2 - B (high single)
2310	.long	0xB2DE638C // B' = pi/2 - B (low single)
2311	.long	0x00000000 // tau (1 for cot path)
2312	.long	0xBCC91A31 // c0 (high single)
2313	.long	0xAF8E8D1A // c0 (low single)
2314	.long	0x3F800000 // c1 (high 1 bit)
2315	.long	0x3A1DFA00 // c1 (low single)
2316	.long	0xBCC9392D // c2
2317	.long	0x3EAB1889 // c3
2318	.long	0xBC885D3B // c4
2319	.align	16
2320	.type	__svml_stan_data_internal, @object
2321	.size	__svml_stan_data_internal, .-__svml_stan_data_internal
2322	.space	16, 0x00
2323	.align	16
2324
2325#ifdef __svml_stan_reduction_data_internal_typedef
2326typedef unsigned int VUINT32;
2327typedef struct {
2328	__declspec(align(16)) VUINT32 _sPtable[256][3][1];
2329} __svml_stan_reduction_data_internal;
2330#endif
2331__svml_stan_reduction_data_internal:
2332	/*     P_hi                  P_med               P_lo                */
2333	.long	0x00000000, 0x00000000, 0x00000000 /* 0 */
2334	.long	0x00000000, 0x00000000, 0x00000000 /* 1 */
2335	.long	0x00000000, 0x00000000, 0x00000000 /* 2 */
2336	.long	0x00000000, 0x00000000, 0x00000000 /* 3 */
2337	.long	0x00000000, 0x00000000, 0x00000000 /* 4 */
2338	.long	0x00000000, 0x00000000, 0x00000000 /* 5 */
2339	.long	0x00000000, 0x00000000, 0x00000000 /* 6 */
2340	.long	0x00000000, 0x00000000, 0x00000000 /* 7 */
2341	.long	0x00000000, 0x00000000, 0x00000000 /* 8 */
2342	.long	0x00000000, 0x00000000, 0x00000000 /* 9 */
2343	.long	0x00000000, 0x00000000, 0x00000000 /* 10 */
2344	.long	0x00000000, 0x00000000, 0x00000000 /* 11 */
2345	.long	0x00000000, 0x00000000, 0x00000000 /* 12 */
2346	.long	0x00000000, 0x00000000, 0x00000000 /* 13 */
2347	.long	0x00000000, 0x00000000, 0x00000000 /* 14 */
2348	.long	0x00000000, 0x00000000, 0x00000000 /* 15 */
2349	.long	0x00000000, 0x00000000, 0x00000000 /* 16 */
2350	.long	0x00000000, 0x00000000, 0x00000000 /* 17 */
2351	.long	0x00000000, 0x00000000, 0x00000000 /* 18 */
2352	.long	0x00000000, 0x00000000, 0x00000000 /* 19 */
2353	.long	0x00000000, 0x00000000, 0x00000000 /* 20 */
2354	.long	0x00000000, 0x00000000, 0x00000000 /* 21 */
2355	.long	0x00000000, 0x00000000, 0x00000000 /* 22 */
2356	.long	0x00000000, 0x00000000, 0x00000000 /* 23 */
2357	.long	0x00000000, 0x00000000, 0x00000000 /* 24 */
2358	.long	0x00000000, 0x00000000, 0x00000000 /* 25 */
2359	.long	0x00000000, 0x00000000, 0x00000000 /* 26 */
2360	.long	0x00000000, 0x00000000, 0x00000000 /* 27 */
2361	.long	0x00000000, 0x00000000, 0x00000000 /* 28 */
2362	.long	0x00000000, 0x00000000, 0x00000000 /* 29 */
2363	.long	0x00000000, 0x00000000, 0x00000000 /* 30 */
2364	.long	0x00000000, 0x00000000, 0x00000000 /* 31 */
2365	.long	0x00000000, 0x00000000, 0x00000000 /* 32 */
2366	.long	0x00000000, 0x00000000, 0x00000000 /* 33 */
2367	.long	0x00000000, 0x00000000, 0x00000000 /* 34 */
2368	.long	0x00000000, 0x00000000, 0x00000000 /* 35 */
2369	.long	0x00000000, 0x00000000, 0x00000000 /* 36 */
2370	.long	0x00000000, 0x00000000, 0x00000000 /* 37 */
2371	.long	0x00000000, 0x00000000, 0x00000000 /* 38 */
2372	.long	0x00000000, 0x00000000, 0x00000000 /* 39 */
2373	.long	0x00000000, 0x00000000, 0x00000000 /* 40 */
2374	.long	0x00000000, 0x00000000, 0x00000000 /* 41 */
2375	.long	0x00000000, 0x00000000, 0x00000000 /* 42 */
2376	.long	0x00000000, 0x00000000, 0x00000000 /* 43 */
2377	.long	0x00000000, 0x00000000, 0x00000000 /* 44 */
2378	.long	0x00000000, 0x00000000, 0x00000000 /* 45 */
2379	.long	0x00000000, 0x00000000, 0x00000000 /* 46 */
2380	.long	0x00000000, 0x00000000, 0x00000000 /* 47 */
2381	.long	0x00000000, 0x00000000, 0x00000000 /* 48 */
2382	.long	0x00000000, 0x00000000, 0x00000000 /* 49 */
2383	.long	0x00000000, 0x00000000, 0x00000000 /* 50 */
2384	.long	0x00000000, 0x00000000, 0x00000000 /* 51 */
2385	.long	0x00000000, 0x00000000, 0x00000000 /* 52 */
2386	.long	0x00000000, 0x00000000, 0x00000000 /* 53 */
2387	.long	0x00000000, 0x00000000, 0x00000000 /* 54 */
2388	.long	0x00000000, 0x00000000, 0x00000000 /* 55 */
2389	.long	0x00000000, 0x00000000, 0x00000000 /* 56 */
2390	.long	0x00000000, 0x00000000, 0x00000001 /* 57 */
2391	.long	0x00000000, 0x00000000, 0x00000002 /* 58 */
2392	.long	0x00000000, 0x00000000, 0x00000005 /* 59 */
2393	.long	0x00000000, 0x00000000, 0x0000000A /* 60 */
2394	.long	0x00000000, 0x00000000, 0x00000014 /* 61 */
2395	.long	0x00000000, 0x00000000, 0x00000028 /* 62 */
2396	.long	0x00000000, 0x00000000, 0x00000051 /* 63 */
2397	.long	0x00000000, 0x00000000, 0x000000A2 /* 64 */
2398	.long	0x00000000, 0x00000000, 0x00000145 /* 65 */
2399	.long	0x00000000, 0x00000000, 0x0000028B /* 66 */
2400	.long	0x00000000, 0x00000000, 0x00000517 /* 67 */
2401	.long	0x00000000, 0x00000000, 0x00000A2F /* 68 */
2402	.long	0x00000000, 0x00000000, 0x0000145F /* 69 */
2403	.long	0x00000000, 0x00000000, 0x000028BE /* 70 */
2404	.long	0x00000000, 0x00000000, 0x0000517C /* 71 */
2405	.long	0x00000000, 0x00000000, 0x0000A2F9 /* 72 */
2406	.long	0x00000000, 0x00000000, 0x000145F3 /* 73 */
2407	.long	0x00000000, 0x00000000, 0x00028BE6 /* 74 */
2408	.long	0x00000000, 0x00000000, 0x000517CC /* 75 */
2409	.long	0x00000000, 0x00000000, 0x000A2F98 /* 76 */
2410	.long	0x00000000, 0x00000000, 0x00145F30 /* 77 */
2411	.long	0x00000000, 0x00000000, 0x0028BE60 /* 78 */
2412	.long	0x00000000, 0x00000000, 0x00517CC1 /* 79 */
2413	.long	0x00000000, 0x00000000, 0x00A2F983 /* 80 */
2414	.long	0x00000000, 0x00000000, 0x0145F306 /* 81 */
2415	.long	0x00000000, 0x00000000, 0x028BE60D /* 82 */
2416	.long	0x00000000, 0x00000000, 0x0517CC1B /* 83 */
2417	.long	0x00000000, 0x00000000, 0x0A2F9836 /* 84 */
2418	.long	0x00000000, 0x00000000, 0x145F306D /* 85 */
2419	.long	0x00000000, 0x00000000, 0x28BE60DB /* 86 */
2420	.long	0x00000000, 0x00000000, 0x517CC1B7 /* 87 */
2421	.long	0x00000000, 0x00000000, 0xA2F9836E /* 88 */
2422	.long	0x00000000, 0x00000001, 0x45F306DC /* 89 */
2423	.long	0x00000000, 0x00000002, 0x8BE60DB9 /* 90 */
2424	.long	0x00000000, 0x00000005, 0x17CC1B72 /* 91 */
2425	.long	0x00000000, 0x0000000A, 0x2F9836E4 /* 92 */
2426	.long	0x00000000, 0x00000014, 0x5F306DC9 /* 93 */
2427	.long	0x00000000, 0x00000028, 0xBE60DB93 /* 94 */
2428	.long	0x00000000, 0x00000051, 0x7CC1B727 /* 95 */
2429	.long	0x00000000, 0x000000A2, 0xF9836E4E /* 96 */
2430	.long	0x00000000, 0x00000145, 0xF306DC9C /* 97 */
2431	.long	0x00000000, 0x0000028B, 0xE60DB939 /* 98 */
2432	.long	0x00000000, 0x00000517, 0xCC1B7272 /* 99 */
2433	.long	0x00000000, 0x00000A2F, 0x9836E4E4 /* 100 */
2434	.long	0x00000000, 0x0000145F, 0x306DC9C8 /* 101 */
2435	.long	0x00000000, 0x000028BE, 0x60DB9391 /* 102 */
2436	.long	0x00000000, 0x0000517C, 0xC1B72722 /* 103 */
2437	.long	0x00000000, 0x0000A2F9, 0x836E4E44 /* 104 */
2438	.long	0x00000000, 0x000145F3, 0x06DC9C88 /* 105 */
2439	.long	0x00000000, 0x00028BE6, 0x0DB93910 /* 106 */
2440	.long	0x00000000, 0x000517CC, 0x1B727220 /* 107 */
2441	.long	0x00000000, 0x000A2F98, 0x36E4E441 /* 108 */
2442	.long	0x00000000, 0x00145F30, 0x6DC9C882 /* 109 */
2443	.long	0x00000000, 0x0028BE60, 0xDB939105 /* 110 */
2444	.long	0x00000000, 0x00517CC1, 0xB727220A /* 111 */
2445	.long	0x00000000, 0x00A2F983, 0x6E4E4415 /* 112 */
2446	.long	0x00000000, 0x0145F306, 0xDC9C882A /* 113 */
2447	.long	0x00000000, 0x028BE60D, 0xB9391054 /* 114 */
2448	.long	0x00000000, 0x0517CC1B, 0x727220A9 /* 115 */
2449	.long	0x00000000, 0x0A2F9836, 0xE4E44152 /* 116 */
2450	.long	0x00000000, 0x145F306D, 0xC9C882A5 /* 117 */
2451	.long	0x00000000, 0x28BE60DB, 0x9391054A /* 118 */
2452	.long	0x00000000, 0x517CC1B7, 0x27220A94 /* 119 */
2453	.long	0x00000000, 0xA2F9836E, 0x4E441529 /* 120 */
2454	.long	0x00000001, 0x45F306DC, 0x9C882A53 /* 121 */
2455	.long	0x00000002, 0x8BE60DB9, 0x391054A7 /* 122 */
2456	.long	0x00000005, 0x17CC1B72, 0x7220A94F /* 123 */
2457	.long	0x0000000A, 0x2F9836E4, 0xE441529F /* 124 */
2458	.long	0x00000014, 0x5F306DC9, 0xC882A53F /* 125 */
2459	.long	0x00000028, 0xBE60DB93, 0x91054A7F /* 126 */
2460	.long	0x00000051, 0x7CC1B727, 0x220A94FE /* 127 */
2461	.long	0x000000A2, 0xF9836E4E, 0x441529FC /* 128 */
2462	.long	0x00000145, 0xF306DC9C, 0x882A53F8 /* 129 */
2463	.long	0x0000028B, 0xE60DB939, 0x1054A7F0 /* 130 */
2464	.long	0x00000517, 0xCC1B7272, 0x20A94FE1 /* 131 */
2465	.long	0x00000A2F, 0x9836E4E4, 0x41529FC2 /* 132 */
2466	.long	0x0000145F, 0x306DC9C8, 0x82A53F84 /* 133 */
2467	.long	0x000028BE, 0x60DB9391, 0x054A7F09 /* 134 */
2468	.long	0x0000517C, 0xC1B72722, 0x0A94FE13 /* 135 */
2469	.long	0x0000A2F9, 0x836E4E44, 0x1529FC27 /* 136 */
2470	.long	0x000145F3, 0x06DC9C88, 0x2A53F84E /* 137 */
2471	.long	0x00028BE6, 0x0DB93910, 0x54A7F09D /* 138 */
2472	.long	0x000517CC, 0x1B727220, 0xA94FE13A /* 139 */
2473	.long	0x000A2F98, 0x36E4E441, 0x529FC275 /* 140 */
2474	.long	0x00145F30, 0x6DC9C882, 0xA53F84EA /* 141 */
2475	.long	0x0028BE60, 0xDB939105, 0x4A7F09D5 /* 142 */
2476	.long	0x00517CC1, 0xB727220A, 0x94FE13AB /* 143 */
2477	.long	0x00A2F983, 0x6E4E4415, 0x29FC2757 /* 144 */
2478	.long	0x0145F306, 0xDC9C882A, 0x53F84EAF /* 145 */
2479	.long	0x028BE60D, 0xB9391054, 0xA7F09D5F /* 146 */
2480	.long	0x0517CC1B, 0x727220A9, 0x4FE13ABE /* 147 */
2481	.long	0x0A2F9836, 0xE4E44152, 0x9FC2757D /* 148 */
2482	.long	0x145F306D, 0xC9C882A5, 0x3F84EAFA /* 149 */
2483	.long	0x28BE60DB, 0x9391054A, 0x7F09D5F4 /* 150 */
2484	.long	0x517CC1B7, 0x27220A94, 0xFE13ABE8 /* 151 */
2485	.long	0xA2F9836E, 0x4E441529, 0xFC2757D1 /* 152 */
2486	.long	0x45F306DC, 0x9C882A53, 0xF84EAFA3 /* 153 */
2487	.long	0x8BE60DB9, 0x391054A7, 0xF09D5F47 /* 154 */
2488	.long	0x17CC1B72, 0x7220A94F, 0xE13ABE8F /* 155 */
2489	.long	0x2F9836E4, 0xE441529F, 0xC2757D1F /* 156 */
2490	.long	0x5F306DC9, 0xC882A53F, 0x84EAFA3E /* 157 */
2491	.long	0xBE60DB93, 0x91054A7F, 0x09D5F47D /* 158 */
2492	.long	0x7CC1B727, 0x220A94FE, 0x13ABE8FA /* 159 */
2493	.long	0xF9836E4E, 0x441529FC, 0x2757D1F5 /* 160 */
2494	.long	0xF306DC9C, 0x882A53F8, 0x4EAFA3EA /* 161 */
2495	.long	0xE60DB939, 0x1054A7F0, 0x9D5F47D4 /* 162 */
2496	.long	0xCC1B7272, 0x20A94FE1, 0x3ABE8FA9 /* 163 */
2497	.long	0x9836E4E4, 0x41529FC2, 0x757D1F53 /* 164 */
2498	.long	0x306DC9C8, 0x82A53F84, 0xEAFA3EA6 /* 165 */
2499	.long	0x60DB9391, 0x054A7F09, 0xD5F47D4D /* 166 */
2500	.long	0xC1B72722, 0x0A94FE13, 0xABE8FA9A /* 167 */
2501	.long	0x836E4E44, 0x1529FC27, 0x57D1F534 /* 168 */
2502	.long	0x06DC9C88, 0x2A53F84E, 0xAFA3EA69 /* 169 */
2503	.long	0x0DB93910, 0x54A7F09D, 0x5F47D4D3 /* 170 */
2504	.long	0x1B727220, 0xA94FE13A, 0xBE8FA9A6 /* 171 */
2505	.long	0x36E4E441, 0x529FC275, 0x7D1F534D /* 172 */
2506	.long	0x6DC9C882, 0xA53F84EA, 0xFA3EA69B /* 173 */
2507	.long	0xDB939105, 0x4A7F09D5, 0xF47D4D37 /* 174 */
2508	.long	0xB727220A, 0x94FE13AB, 0xE8FA9A6E /* 175 */
2509	.long	0x6E4E4415, 0x29FC2757, 0xD1F534DD /* 176 */
2510	.long	0xDC9C882A, 0x53F84EAF, 0xA3EA69BB /* 177 */
2511	.long	0xB9391054, 0xA7F09D5F, 0x47D4D377 /* 178 */
2512	.long	0x727220A9, 0x4FE13ABE, 0x8FA9A6EE /* 179 */
2513	.long	0xE4E44152, 0x9FC2757D, 0x1F534DDC /* 180 */
2514	.long	0xC9C882A5, 0x3F84EAFA, 0x3EA69BB8 /* 181 */
2515	.long	0x9391054A, 0x7F09D5F4, 0x7D4D3770 /* 182 */
2516	.long	0x27220A94, 0xFE13ABE8, 0xFA9A6EE0 /* 183 */
2517	.long	0x4E441529, 0xFC2757D1, 0xF534DDC0 /* 184 */
2518	.long	0x9C882A53, 0xF84EAFA3, 0xEA69BB81 /* 185 */
2519	.long	0x391054A7, 0xF09D5F47, 0xD4D37703 /* 186 */
2520	.long	0x7220A94F, 0xE13ABE8F, 0xA9A6EE06 /* 187 */
2521	.long	0xE441529F, 0xC2757D1F, 0x534DDC0D /* 188 */
2522	.long	0xC882A53F, 0x84EAFA3E, 0xA69BB81B /* 189 */
2523	.long	0x91054A7F, 0x09D5F47D, 0x4D377036 /* 190 */
2524	.long	0x220A94FE, 0x13ABE8FA, 0x9A6EE06D /* 191 */
2525	.long	0x441529FC, 0x2757D1F5, 0x34DDC0DB /* 192 */
2526	.long	0x882A53F8, 0x4EAFA3EA, 0x69BB81B6 /* 193 */
2527	.long	0x1054A7F0, 0x9D5F47D4, 0xD377036D /* 194 */
2528	.long	0x20A94FE1, 0x3ABE8FA9, 0xA6EE06DB /* 195 */
2529	.long	0x41529FC2, 0x757D1F53, 0x4DDC0DB6 /* 196 */
2530	.long	0x82A53F84, 0xEAFA3EA6, 0x9BB81B6C /* 197 */
2531	.long	0x054A7F09, 0xD5F47D4D, 0x377036D8 /* 198 */
2532	.long	0x0A94FE13, 0xABE8FA9A, 0x6EE06DB1 /* 199 */
2533	.long	0x1529FC27, 0x57D1F534, 0xDDC0DB62 /* 200 */
2534	.long	0x2A53F84E, 0xAFA3EA69, 0xBB81B6C5 /* 201 */
2535	.long	0x54A7F09D, 0x5F47D4D3, 0x77036D8A /* 202 */
2536	.long	0xA94FE13A, 0xBE8FA9A6, 0xEE06DB14 /* 203 */
2537	.long	0x529FC275, 0x7D1F534D, 0xDC0DB629 /* 204 */
2538	.long	0xA53F84EA, 0xFA3EA69B, 0xB81B6C52 /* 205 */
2539	.long	0x4A7F09D5, 0xF47D4D37, 0x7036D8A5 /* 206 */
2540	.long	0x94FE13AB, 0xE8FA9A6E, 0xE06DB14A /* 207 */
2541	.long	0x29FC2757, 0xD1F534DD, 0xC0DB6295 /* 208 */
2542	.long	0x53F84EAF, 0xA3EA69BB, 0x81B6C52B /* 209 */
2543	.long	0xA7F09D5F, 0x47D4D377, 0x036D8A56 /* 210 */
2544	.long	0x4FE13ABE, 0x8FA9A6EE, 0x06DB14AC /* 211 */
2545	.long	0x9FC2757D, 0x1F534DDC, 0x0DB62959 /* 212 */
2546	.long	0x3F84EAFA, 0x3EA69BB8, 0x1B6C52B3 /* 213 */
2547	.long	0x7F09D5F4, 0x7D4D3770, 0x36D8A566 /* 214 */
2548	.long	0xFE13ABE8, 0xFA9A6EE0, 0x6DB14ACC /* 215 */
2549	.long	0xFC2757D1, 0xF534DDC0, 0xDB629599 /* 216 */
2550	.long	0xF84EAFA3, 0xEA69BB81, 0xB6C52B32 /* 217 */
2551	.long	0xF09D5F47, 0xD4D37703, 0x6D8A5664 /* 218 */
2552	.long	0xE13ABE8F, 0xA9A6EE06, 0xDB14ACC9 /* 219 */
2553	.long	0xC2757D1F, 0x534DDC0D, 0xB6295993 /* 220 */
2554	.long	0x84EAFA3E, 0xA69BB81B, 0x6C52B327 /* 221 */
2555	.long	0x09D5F47D, 0x4D377036, 0xD8A5664F /* 222 */
2556	.long	0x13ABE8FA, 0x9A6EE06D, 0xB14ACC9E /* 223 */
2557	.long	0x2757D1F5, 0x34DDC0DB, 0x6295993C /* 224 */
2558	.long	0x4EAFA3EA, 0x69BB81B6, 0xC52B3278 /* 225 */
2559	.long	0x9D5F47D4, 0xD377036D, 0x8A5664F1 /* 226 */
2560	.long	0x3ABE8FA9, 0xA6EE06DB, 0x14ACC9E2 /* 227 */
2561	.long	0x757D1F53, 0x4DDC0DB6, 0x295993C4 /* 228 */
2562	.long	0xEAFA3EA6, 0x9BB81B6C, 0x52B32788 /* 229 */
2563	.long	0xD5F47D4D, 0x377036D8, 0xA5664F10 /* 230 */
2564	.long	0xABE8FA9A, 0x6EE06DB1, 0x4ACC9E21 /* 231 */
2565	.long	0x57D1F534, 0xDDC0DB62, 0x95993C43 /* 232 */
2566	.long	0xAFA3EA69, 0xBB81B6C5, 0x2B327887 /* 233 */
2567	.long	0x5F47D4D3, 0x77036D8A, 0x5664F10E /* 234 */
2568	.long	0xBE8FA9A6, 0xEE06DB14, 0xACC9E21C /* 235 */
2569	.long	0x7D1F534D, 0xDC0DB629, 0x5993C439 /* 236 */
2570	.long	0xFA3EA69B, 0xB81B6C52, 0xB3278872 /* 237 */
2571	.long	0xF47D4D37, 0x7036D8A5, 0x664F10E4 /* 238 */
2572	.long	0xE8FA9A6E, 0xE06DB14A, 0xCC9E21C8 /* 239 */
2573	.long	0xD1F534DD, 0xC0DB6295, 0x993C4390 /* 240 */
2574	.long	0xA3EA69BB, 0x81B6C52B, 0x32788720 /* 241 */
2575	.long	0x47D4D377, 0x036D8A56, 0x64F10E41 /* 242 */
2576	.long	0x8FA9A6EE, 0x06DB14AC, 0xC9E21C82 /* 243 */
2577	.long	0x1F534DDC, 0x0DB62959, 0x93C43904 /* 244 */
2578	.long	0x3EA69BB8, 0x1B6C52B3, 0x27887208 /* 245 */
2579	.long	0x7D4D3770, 0x36D8A566, 0x4F10E410 /* 246 */
2580	.long	0xFA9A6EE0, 0x6DB14ACC, 0x9E21C820 /* 247 */
2581	.long	0xF534DDC0, 0xDB629599, 0x3C439041 /* 248 */
2582	.long	0xEA69BB81, 0xB6C52B32, 0x78872083 /* 249 */
2583	.long	0xD4D37703, 0x6D8A5664, 0xF10E4107 /* 250 */
2584	.long	0xA9A6EE06, 0xDB14ACC9, 0xE21C820F /* 251 */
2585	.long	0x534DDC0D, 0xB6295993, 0xC439041F /* 252 */
2586	.long	0xA69BB81B, 0x6C52B327, 0x8872083F /* 253 */
2587	.long	0x4D377036, 0xD8A5664F, 0x10E4107F /* 254 */
2588	.long	0x9A6EE06D, 0xB14ACC9E, 0x21C820FF /* 255 */
2589	.align	16
2590	.type	__svml_stan_reduction_data_internal, @object
2591	.size	__svml_stan_reduction_data_internal, .-__svml_stan_reduction_data_internal
2592	.align	16
2593
2594.FLT_16:
2595	.long	0xffffffff, 0x00000000, 0xffffffff, 0x00000000
2596	.type	.FLT_16, @object
2597	.size	.FLT_16, 16
2598