xref: /freebsd/contrib/arm-optimized-routines/math/test/ulp.c (revision 2e3507c25e42292b45a5482e116d278f5515d04d)
1 /*
2  * ULP error checking tool for math functions.
3  *
4  * Copyright (c) 2019-2022, Arm Limited.
5  * SPDX-License-Identifier: MIT OR Apache-2.0 WITH LLVM-exception
6  */
7 
8 #include <ctype.h>
9 #include <fenv.h>
10 #include <float.h>
11 #include <math.h>
12 #include <stdint.h>
13 #include <stdio.h>
14 #include <stdlib.h>
15 #include <string.h>
16 #include "mathlib.h"
17 
18 /* Don't depend on mpfr by default.  */
19 #ifndef USE_MPFR
20 # define USE_MPFR 0
21 #endif
22 #if USE_MPFR
23 # include <mpfr.h>
24 #endif
25 
26 #ifndef WANT_VMATH
27 /* Enable the build of vector math code.  */
28 # define WANT_VMATH 1
29 #endif
30 
31 static inline uint64_t
32 asuint64 (double f)
33 {
34   union
35   {
36     double f;
37     uint64_t i;
38   } u = {f};
39   return u.i;
40 }
41 
42 static inline double
43 asdouble (uint64_t i)
44 {
45   union
46   {
47     uint64_t i;
48     double f;
49   } u = {i};
50   return u.f;
51 }
52 
53 static inline uint32_t
54 asuint (float f)
55 {
56   union
57   {
58     float f;
59     uint32_t i;
60   } u = {f};
61   return u.i;
62 }
63 
64 static inline float
65 asfloat (uint32_t i)
66 {
67   union
68   {
69     uint32_t i;
70     float f;
71   } u = {i};
72   return u.f;
73 }
74 
75 static uint64_t seed = 0x0123456789abcdef;
76 static uint64_t
77 rand64 (void)
78 {
79   seed = 6364136223846793005ull * seed + 1;
80   return seed ^ (seed >> 32);
81 }
82 
83 /* Uniform random in [0,n].  */
84 static uint64_t
85 randn (uint64_t n)
86 {
87   uint64_t r, m;
88 
89   if (n == 0)
90     return 0;
91   n++;
92   if (n == 0)
93     return rand64 ();
94   for (;;)
95     {
96       r = rand64 ();
97       m = r % n;
98       if (r - m <= -n)
99 	return m;
100     }
101 }
102 
103 struct gen
104 {
105   uint64_t start;
106   uint64_t len;
107   uint64_t start2;
108   uint64_t len2;
109   uint64_t off;
110   uint64_t step;
111   uint64_t cnt;
112 };
113 
114 struct args_f1
115 {
116   float x;
117 };
118 
119 struct args_f2
120 {
121   float x;
122   float x2;
123 };
124 
125 struct args_d1
126 {
127   double x;
128 };
129 
130 struct args_d2
131 {
132   double x;
133   double x2;
134 };
135 
136 /* result = y + tail*2^ulpexp.  */
137 struct ret_f
138 {
139   float y;
140   double tail;
141   int ulpexp;
142   int ex;
143   int ex_may;
144 };
145 
146 struct ret_d
147 {
148   double y;
149   double tail;
150   int ulpexp;
151   int ex;
152   int ex_may;
153 };
154 
155 static inline uint64_t
156 next1 (struct gen *g)
157 {
158   /* For single argument use randomized incremental steps,
159      that produce dense sampling without collisions and allow
160      testing all inputs in a range.  */
161   uint64_t r = g->start + g->off;
162   g->off += g->step + randn (g->step / 2);
163   if (g->off > g->len)
164     g->off -= g->len; /* hack.  */
165   return r;
166 }
167 
168 static inline uint64_t
169 next2 (uint64_t *x2, struct gen *g)
170 {
171   /* For two arguments use uniform random sampling.  */
172   uint64_t r = g->start + randn (g->len);
173   *x2 = g->start2 + randn (g->len2);
174   return r;
175 }
176 
177 static struct args_f1
178 next_f1 (void *g)
179 {
180   return (struct args_f1){asfloat (next1 (g))};
181 }
182 
183 static struct args_f2
184 next_f2 (void *g)
185 {
186   uint64_t x2;
187   uint64_t x = next2 (&x2, g);
188   return (struct args_f2){asfloat (x), asfloat (x2)};
189 }
190 
191 static struct args_d1
192 next_d1 (void *g)
193 {
194   return (struct args_d1){asdouble (next1 (g))};
195 }
196 
197 static struct args_d2
198 next_d2 (void *g)
199 {
200   uint64_t x2;
201   uint64_t x = next2 (&x2, g);
202   return (struct args_d2){asdouble (x), asdouble (x2)};
203 }
204 
205 struct conf
206 {
207   int r;
208   int rc;
209   int quiet;
210   int mpfr;
211   int fenv;
212   unsigned long long n;
213   double softlim;
214   double errlim;
215 };
216 
217 /* A bit of a hack: call vector functions twice with the same
218    input in lane 0 but a different value in other lanes: once
219    with an in-range value and then with a special case value.  */
220 static int secondcall;
221 
222 /* Wrappers for vector functions.  */
223 #if __aarch64__ && WANT_VMATH
224 typedef __f32x4_t v_float;
225 typedef __f64x2_t v_double;
226 /* First element of fv and dv may be changed by -c argument.  */
227 static float fv[2] = {1.0f, -INFINITY};
228 static double dv[2] = {1.0, -INFINITY};
229 static inline v_float argf(float x) { return (v_float){x,x,x,fv[secondcall]}; }
230 static inline v_double argd(double x) { return (v_double){x,dv[secondcall]}; }
231 #if WANT_SVE_MATH
232 #include <arm_sve.h>
233 typedef __SVFloat32_t sv_float;
234 typedef __SVFloat64_t sv_double;
235 
236 static inline sv_float svargf(float x)  {
237 	int n = svcntw();
238 	float base[n];
239 	for (int i=0; i<n; i++)
240 		base[i] = (float)x;
241 	base[n-1] = (float) fv[secondcall];
242 	return svld1(svptrue_b32(), base);
243 }
244 static inline sv_double svargd(double x) {
245 	int n = svcntd();
246 	double base[n];
247 	for (int i=0; i<n; i++)
248 		base[i] = x;
249 	base[n-1] = dv[secondcall];
250 	return svld1(svptrue_b64(), base);
251 }
252 static inline float svretf(sv_float vec)  {
253 	int n = svcntw();
254 	float res[n];
255 	svst1(svptrue_b32(), res, vec);
256 	return res[0];
257 }
258 static inline double svretd(sv_double vec) {
259 	int n = svcntd();
260 	double res[n];
261 	svst1(svptrue_b64(), res, vec);
262 	return res[0];
263 }
264 #endif
265 #endif
266 
267 #if WANT_SVE_MATH
268 long double
269 dummyl (long double x)
270 {
271   return x;
272 }
273 
274 double
275 dummy (double x)
276 {
277   return x;
278 }
279 
280 static sv_double
281 __sv_dummy (sv_double x)
282 {
283   return x;
284 }
285 
286 static sv_float
287 __sv_dummyf (sv_float x)
288 {
289   return x;
290 }
291 #endif
292 
293 #include "test/ulp_wrappers.h"
294 
295 /* Wrappers for SVE functions.  */
296 #if WANT_SVE_MATH
297 static double sv_dummy (double x) { return svretd (__sv_dummy (svargd (x))); }
298 static float sv_dummyf (float x) { return svretf (__sv_dummyf (svargf (x))); }
299 #endif
300 
301 struct fun
302 {
303   const char *name;
304   int arity;
305   int singleprec;
306   int twice;
307   union
308   {
309     float (*f1) (float);
310     float (*f2) (float, float);
311     double (*d1) (double);
312     double (*d2) (double, double);
313   } fun;
314   union
315   {
316     double (*f1) (double);
317     double (*f2) (double, double);
318     long double (*d1) (long double);
319     long double (*d2) (long double, long double);
320   } fun_long;
321 #if USE_MPFR
322   union
323   {
324     int (*f1) (mpfr_t, const mpfr_t, mpfr_rnd_t);
325     int (*f2) (mpfr_t, const mpfr_t, const mpfr_t, mpfr_rnd_t);
326     int (*d1) (mpfr_t, const mpfr_t, mpfr_rnd_t);
327     int (*d2) (mpfr_t, const mpfr_t, const mpfr_t, mpfr_rnd_t);
328   } fun_mpfr;
329 #endif
330 };
331 
332 static const struct fun fun[] = {
333 #if USE_MPFR
334 # define F(x, x_wrap, x_long, x_mpfr, a, s, t, twice) \
335   {#x, a, s, twice, {.t = x_wrap}, {.t = x_long}, {.t = x_mpfr}},
336 #else
337 # define F(x, x_wrap, x_long, x_mpfr, a, s, t, twice) \
338   {#x, a, s, twice, {.t = x_wrap}, {.t = x_long}},
339 #endif
340 #define F1(x) F (x##f, x##f, x, mpfr_##x, 1, 1, f1, 0)
341 #define F2(x) F (x##f, x##f, x, mpfr_##x, 2, 1, f2, 0)
342 #define D1(x) F (x, x, x##l, mpfr_##x, 1, 0, d1, 0)
343 #define D2(x) F (x, x, x##l, mpfr_##x, 2, 0, d2, 0)
344 /* Neon routines.  */
345 #define VF1(x) F (__v_##x##f, v_##x##f, x, mpfr_##x, 1, 1, f1, 0)
346 #define VF2(x) F (__v_##x##f, v_##x##f, x, mpfr_##x, 2, 1, f2, 0)
347 #define VD1(x) F (__v_##x, v_##x, x##l, mpfr_##x, 1, 0, d1, 0)
348 #define VD2(x) F (__v_##x, v_##x, x##l, mpfr_##x, 2, 0, d2, 0)
349 #define VNF1(x) F (__vn_##x##f, vn_##x##f, x, mpfr_##x, 1, 1, f1, 0)
350 #define VNF2(x) F (__vn_##x##f, vn_##x##f, x, mpfr_##x, 2, 1, f2, 0)
351 #define VND1(x) F (__vn_##x, vn_##x, x##l, mpfr_##x, 1, 0, d1, 0)
352 #define VND2(x) F (__vn_##x, vn_##x, x##l, mpfr_##x, 2, 0, d2, 0)
353 #define ZVF1(x) F (_ZGVnN4v_##x##f, Z_##x##f, x, mpfr_##x, 1, 1, f1, 0)
354 #define ZVF2(x) F (_ZGVnN4vv_##x##f, Z_##x##f, x, mpfr_##x, 2, 1, f2, 0)
355 #define ZVD1(x) F (_ZGVnN2v_##x, Z_##x, x##l, mpfr_##x, 1, 0, d1, 0)
356 #define ZVD2(x) F (_ZGVnN2vv_##x, Z_##x, x##l, mpfr_##x, 2, 0, d2, 0)
357 #define ZVNF1(x) VNF1 (x) ZVF1 (x)
358 #define ZVNF2(x) VNF2 (x) ZVF2 (x)
359 #define ZVND1(x) VND1 (x) ZVD1 (x)
360 #define ZVND2(x) VND2 (x) ZVD2 (x)
361 #define SF1(x) F (__s_##x##f, __s_##x##f, x, mpfr_##x, 1, 1, f1, 0)
362 #define SF2(x) F (__s_##x##f, __s_##x##f, x, mpfr_##x, 2, 1, f2, 0)
363 #define SD1(x) F (__s_##x, __s_##x, x##l, mpfr_##x, 1, 0, d1, 0)
364 #define SD2(x) F (__s_##x, __s_##x, x##l, mpfr_##x, 2, 0, d2, 0)
365 /* SVE routines.  */
366 #define SVF1(x) F (__sv_##x##f, sv_##x##f, x, mpfr_##x, 1, 1, f1, 0)
367 #define SVF2(x) F (__sv_##x##f, sv_##x##f, x, mpfr_##x, 2, 1, f2, 0)
368 #define SVD1(x) F (__sv_##x, sv_##x, x##l, mpfr_##x, 1, 0, d1, 0)
369 #define SVD2(x) F (__sv_##x, sv_##x, x##l, mpfr_##x, 2, 0, d2, 0)
370 #define ZSVF1(x) F (_ZGVsMxv_##x##f, Z_sv_##x##f, x, mpfr_##x, 1, 1, f1, 0)
371 #define ZSVF2(x) F (_ZGVsMxvv_##x##f, Z_sv_##x##f, x, mpfr_##x, 2, 1, f2, 0)
372 #define ZSVD1(x) F (_ZGVsMxv_##x, Z_sv_##x, x##l, mpfr_##x, 1, 0, d1, 0)
373 #define ZSVD2(x) F (_ZGVsMxvv_##x, Z_sv_##x, x##l, mpfr_##x, 2, 0, d2, 0)
374 
375 #include "test/ulp_funcs.h"
376 
377 #if WANT_SVE_MATH
378  SVD1 (dummy)
379  SVF1 (dummy)
380 #endif
381 
382 #undef F
383 #undef F1
384 #undef F2
385 #undef D1
386 #undef D2
387 #undef SVF1
388 #undef SVF2
389 #undef SVD1
390 #undef SVD2
391  {0}};
392 
393 /* Boilerplate for generic calls.  */
394 
395 static inline int
396 ulpscale_f (float x)
397 {
398   int e = asuint (x) >> 23 & 0xff;
399   if (!e)
400     e++;
401   return e - 0x7f - 23;
402 }
403 static inline int
404 ulpscale_d (double x)
405 {
406   int e = asuint64 (x) >> 52 & 0x7ff;
407   if (!e)
408     e++;
409   return e - 0x3ff - 52;
410 }
411 static inline float
412 call_f1 (const struct fun *f, struct args_f1 a)
413 {
414   return f->fun.f1 (a.x);
415 }
416 static inline float
417 call_f2 (const struct fun *f, struct args_f2 a)
418 {
419   return f->fun.f2 (a.x, a.x2);
420 }
421 
422 static inline double
423 call_d1 (const struct fun *f, struct args_d1 a)
424 {
425   return f->fun.d1 (a.x);
426 }
427 static inline double
428 call_d2 (const struct fun *f, struct args_d2 a)
429 {
430   return f->fun.d2 (a.x, a.x2);
431 }
432 static inline double
433 call_long_f1 (const struct fun *f, struct args_f1 a)
434 {
435   return f->fun_long.f1 (a.x);
436 }
437 static inline double
438 call_long_f2 (const struct fun *f, struct args_f2 a)
439 {
440   return f->fun_long.f2 (a.x, a.x2);
441 }
442 static inline long double
443 call_long_d1 (const struct fun *f, struct args_d1 a)
444 {
445   return f->fun_long.d1 (a.x);
446 }
447 static inline long double
448 call_long_d2 (const struct fun *f, struct args_d2 a)
449 {
450   return f->fun_long.d2 (a.x, a.x2);
451 }
452 static inline void
453 printcall_f1 (const struct fun *f, struct args_f1 a)
454 {
455   printf ("%s(%a)", f->name, a.x);
456 }
457 static inline void
458 printcall_f2 (const struct fun *f, struct args_f2 a)
459 {
460   printf ("%s(%a, %a)", f->name, a.x, a.x2);
461 }
462 static inline void
463 printcall_d1 (const struct fun *f, struct args_d1 a)
464 {
465   printf ("%s(%a)", f->name, a.x);
466 }
467 static inline void
468 printcall_d2 (const struct fun *f, struct args_d2 a)
469 {
470   printf ("%s(%a, %a)", f->name, a.x, a.x2);
471 }
472 static inline void
473 printgen_f1 (const struct fun *f, struct gen *gen)
474 {
475   printf ("%s in [%a;%a]", f->name, asfloat (gen->start),
476 	  asfloat (gen->start + gen->len));
477 }
478 static inline void
479 printgen_f2 (const struct fun *f, struct gen *gen)
480 {
481   printf ("%s in [%a;%a] x [%a;%a]", f->name, asfloat (gen->start),
482 	  asfloat (gen->start + gen->len), asfloat (gen->start2),
483 	  asfloat (gen->start2 + gen->len2));
484 }
485 static inline void
486 printgen_d1 (const struct fun *f, struct gen *gen)
487 {
488   printf ("%s in [%a;%a]", f->name, asdouble (gen->start),
489 	  asdouble (gen->start + gen->len));
490 }
491 static inline void
492 printgen_d2 (const struct fun *f, struct gen *gen)
493 {
494   printf ("%s in [%a;%a] x [%a;%a]", f->name, asdouble (gen->start),
495 	  asdouble (gen->start + gen->len), asdouble (gen->start2),
496 	  asdouble (gen->start2 + gen->len2));
497 }
498 
499 #define reduce_f1(a, f, op) (f (a.x))
500 #define reduce_f2(a, f, op) (f (a.x) op f (a.x2))
501 #define reduce_d1(a, f, op) (f (a.x))
502 #define reduce_d2(a, f, op) (f (a.x) op f (a.x2))
503 
504 #ifndef IEEE_754_2008_SNAN
505 # define IEEE_754_2008_SNAN 1
506 #endif
507 static inline int
508 issignaling_f (float x)
509 {
510   uint32_t ix = asuint (x);
511   if (!IEEE_754_2008_SNAN)
512     return (ix & 0x7fc00000) == 0x7fc00000;
513   return 2 * (ix ^ 0x00400000) > 2u * 0x7fc00000;
514 }
515 static inline int
516 issignaling_d (double x)
517 {
518   uint64_t ix = asuint64 (x);
519   if (!IEEE_754_2008_SNAN)
520     return (ix & 0x7ff8000000000000) == 0x7ff8000000000000;
521   return 2 * (ix ^ 0x0008000000000000) > 2 * 0x7ff8000000000000ULL;
522 }
523 
524 #if USE_MPFR
525 static mpfr_rnd_t
526 rmap (int r)
527 {
528   switch (r)
529     {
530     case FE_TONEAREST:
531       return MPFR_RNDN;
532     case FE_TOWARDZERO:
533       return MPFR_RNDZ;
534     case FE_UPWARD:
535       return MPFR_RNDU;
536     case FE_DOWNWARD:
537       return MPFR_RNDD;
538     }
539   return -1;
540 }
541 
542 #define prec_mpfr_f 50
543 #define prec_mpfr_d 80
544 #define prec_f 24
545 #define prec_d 53
546 #define emin_f -148
547 #define emin_d -1073
548 #define emax_f 128
549 #define emax_d 1024
550 static inline int
551 call_mpfr_f1 (mpfr_t y, const struct fun *f, struct args_f1 a, mpfr_rnd_t r)
552 {
553   MPFR_DECL_INIT (x, prec_f);
554   mpfr_set_flt (x, a.x, MPFR_RNDN);
555   return f->fun_mpfr.f1 (y, x, r);
556 }
557 static inline int
558 call_mpfr_f2 (mpfr_t y, const struct fun *f, struct args_f2 a, mpfr_rnd_t r)
559 {
560   MPFR_DECL_INIT (x, prec_f);
561   MPFR_DECL_INIT (x2, prec_f);
562   mpfr_set_flt (x, a.x, MPFR_RNDN);
563   mpfr_set_flt (x2, a.x2, MPFR_RNDN);
564   return f->fun_mpfr.f2 (y, x, x2, r);
565 }
566 static inline int
567 call_mpfr_d1 (mpfr_t y, const struct fun *f, struct args_d1 a, mpfr_rnd_t r)
568 {
569   MPFR_DECL_INIT (x, prec_d);
570   mpfr_set_d (x, a.x, MPFR_RNDN);
571   return f->fun_mpfr.d1 (y, x, r);
572 }
573 static inline int
574 call_mpfr_d2 (mpfr_t y, const struct fun *f, struct args_d2 a, mpfr_rnd_t r)
575 {
576   MPFR_DECL_INIT (x, prec_d);
577   MPFR_DECL_INIT (x2, prec_d);
578   mpfr_set_d (x, a.x, MPFR_RNDN);
579   mpfr_set_d (x2, a.x2, MPFR_RNDN);
580   return f->fun_mpfr.d2 (y, x, x2, r);
581 }
582 #endif
583 
584 #define float_f float
585 #define double_f double
586 #define copysign_f copysignf
587 #define nextafter_f nextafterf
588 #define fabs_f fabsf
589 #define asuint_f asuint
590 #define asfloat_f asfloat
591 #define scalbn_f scalbnf
592 #define lscalbn_f scalbn
593 #define halfinf_f 0x1p127f
594 #define min_normal_f 0x1p-126f
595 
596 #define float_d double
597 #define double_d long double
598 #define copysign_d copysign
599 #define nextafter_d nextafter
600 #define fabs_d fabs
601 #define asuint_d asuint64
602 #define asfloat_d asdouble
603 #define scalbn_d scalbn
604 #define lscalbn_d scalbnl
605 #define halfinf_d 0x1p1023
606 #define min_normal_d 0x1p-1022
607 
608 #define NEW_RT
609 #define RT(x) x##_f
610 #define T(x) x##_f1
611 #include "ulp.h"
612 #undef T
613 #define T(x) x##_f2
614 #include "ulp.h"
615 #undef T
616 #undef RT
617 
618 #define NEW_RT
619 #define RT(x) x##_d
620 #define T(x) x##_d1
621 #include "ulp.h"
622 #undef T
623 #define T(x) x##_d2
624 #include "ulp.h"
625 #undef T
626 #undef RT
627 
628 static void
629 usage (void)
630 {
631   puts ("./ulp [-q] [-m] [-f] [-r nudz] [-l soft-ulplimit] [-e ulplimit] func "
632 	"lo [hi [x lo2 hi2] [count]]");
633   puts ("Compares func against a higher precision implementation in [lo; hi].");
634   puts ("-q: quiet.");
635   puts ("-m: use mpfr even if faster method is available.");
636   puts ("-f: disable fenv testing (rounding modes and exceptions).");
637 #if __aarch64__ && WANT_VMATH
638   puts ("-c: neutral 'control value' to test behaviour when one lane can affect another. \n"
639 	"    This should be different from tested input in other lanes, and non-special \n"
640 	"    (i.e. should not trigger fenv exceptions). Default is 1.");
641 #endif
642   puts ("Supported func:");
643   for (const struct fun *f = fun; f->name; f++)
644     printf ("\t%s\n", f->name);
645   exit (1);
646 }
647 
648 static int
649 cmp (const struct fun *f, struct gen *gen, const struct conf *conf)
650 {
651   int r = 1;
652   if (f->arity == 1 && f->singleprec)
653     r = cmp_f1 (f, gen, conf);
654   else if (f->arity == 2 && f->singleprec)
655     r = cmp_f2 (f, gen, conf);
656   else if (f->arity == 1 && !f->singleprec)
657     r = cmp_d1 (f, gen, conf);
658   else if (f->arity == 2 && !f->singleprec)
659     r = cmp_d2 (f, gen, conf);
660   else
661     usage ();
662   return r;
663 }
664 
665 static uint64_t
666 getnum (const char *s, int singleprec)
667 {
668   //	int i;
669   uint64_t sign = 0;
670   //	char buf[12];
671 
672   if (s[0] == '+')
673     s++;
674   else if (s[0] == '-')
675     {
676       sign = singleprec ? 1ULL << 31 : 1ULL << 63;
677       s++;
678     }
679   /* 0xXXXX is treated as bit representation, '-' flips the sign bit.  */
680   if (s[0] == '0' && tolower (s[1]) == 'x' && strchr (s, 'p') == 0)
681     return sign ^ strtoull (s, 0, 0);
682   //	/* SNaN, QNaN, NaN, Inf.  */
683   //	for (i=0; s[i] && i < sizeof buf; i++)
684   //		buf[i] = tolower(s[i]);
685   //	buf[i] = 0;
686   //	if (strcmp(buf, "snan") == 0)
687   //		return sign | (singleprec ? 0x7fa00000 : 0x7ff4000000000000);
688   //	if (strcmp(buf, "qnan") == 0 || strcmp(buf, "nan") == 0)
689   //		return sign | (singleprec ? 0x7fc00000 : 0x7ff8000000000000);
690   //	if (strcmp(buf, "inf") == 0 || strcmp(buf, "infinity") == 0)
691   //		return sign | (singleprec ? 0x7f800000 : 0x7ff0000000000000);
692   /* Otherwise assume it's a floating-point literal.  */
693   return sign
694 	 | (singleprec ? asuint (strtof (s, 0)) : asuint64 (strtod (s, 0)));
695 }
696 
697 static void
698 parsegen (struct gen *g, int argc, char *argv[], const struct fun *f)
699 {
700   int singleprec = f->singleprec;
701   int arity = f->arity;
702   uint64_t a, b, a2, b2, n;
703   if (argc < 1)
704     usage ();
705   b = a = getnum (argv[0], singleprec);
706   n = 0;
707   if (argc > 1 && strcmp (argv[1], "x") == 0)
708     {
709       argc -= 2;
710       argv += 2;
711     }
712   else if (argc > 1)
713     {
714       b = getnum (argv[1], singleprec);
715       if (argc > 2 && strcmp (argv[2], "x") == 0)
716 	{
717 	  argc -= 3;
718 	  argv += 3;
719 	}
720     }
721   b2 = a2 = getnum (argv[0], singleprec);
722   if (argc > 1)
723     b2 = getnum (argv[1], singleprec);
724   if (argc > 2)
725     n = strtoull (argv[2], 0, 0);
726   if (argc > 3)
727     usage ();
728   //printf("ab %lx %lx ab2 %lx %lx n %lu\n", a, b, a2, b2, n);
729   if (arity == 1)
730     {
731       g->start = a;
732       g->len = b - a;
733       if (n - 1 > b - a)
734 	n = b - a + 1;
735       g->off = 0;
736       g->step = n ? (g->len + 1) / n : 1;
737       g->start2 = g->len2 = 0;
738       g->cnt = n;
739     }
740   else if (arity == 2)
741     {
742       g->start = a;
743       g->len = b - a;
744       g->off = g->step = 0;
745       g->start2 = a2;
746       g->len2 = b2 - a2;
747       g->cnt = n;
748     }
749   else
750     usage ();
751 }
752 
753 int
754 main (int argc, char *argv[])
755 {
756   const struct fun *f;
757   struct gen gen;
758   struct conf conf;
759   conf.rc = 'n';
760   conf.quiet = 0;
761   conf.mpfr = 0;
762   conf.fenv = 1;
763   conf.softlim = 0;
764   conf.errlim = INFINITY;
765   for (;;)
766     {
767       argc--;
768       argv++;
769       if (argc < 1)
770 	usage ();
771       if (argv[0][0] != '-')
772 	break;
773       switch (argv[0][1])
774 	{
775 	case 'e':
776 	  argc--;
777 	  argv++;
778 	  if (argc < 1)
779 	    usage ();
780 	  conf.errlim = strtod (argv[0], 0);
781 	  break;
782 	case 'f':
783 	  conf.fenv = 0;
784 	  break;
785 	case 'l':
786 	  argc--;
787 	  argv++;
788 	  if (argc < 1)
789 	    usage ();
790 	  conf.softlim = strtod (argv[0], 0);
791 	  break;
792 	case 'm':
793 	  conf.mpfr = 1;
794 	  break;
795 	case 'q':
796 	  conf.quiet = 1;
797 	  break;
798 	case 'r':
799 	  conf.rc = argv[0][2];
800 	  if (!conf.rc)
801 	    {
802 	      argc--;
803 	      argv++;
804 	      if (argc < 1)
805 		usage ();
806 	      conf.rc = argv[0][0];
807 	    }
808 	  break;
809 #if __aarch64__ && WANT_VMATH
810 	case 'c':
811 	  argc--;
812 	  argv++;
813 	  fv[0] = strtof(argv[0], 0);
814 	  dv[0] = strtod(argv[0], 0);
815 	  break;
816 #endif
817 	default:
818 	  usage ();
819 	}
820     }
821   switch (conf.rc)
822     {
823     case 'n':
824       conf.r = FE_TONEAREST;
825       break;
826     case 'u':
827       conf.r = FE_UPWARD;
828       break;
829     case 'd':
830       conf.r = FE_DOWNWARD;
831       break;
832     case 'z':
833       conf.r = FE_TOWARDZERO;
834       break;
835     default:
836       usage ();
837     }
838   for (f = fun; f->name; f++)
839     if (strcmp (argv[0], f->name) == 0)
840       break;
841   if (!f->name)
842     usage ();
843   if (!f->singleprec && LDBL_MANT_DIG == DBL_MANT_DIG)
844     conf.mpfr = 1; /* Use mpfr if long double has no extra precision.  */
845   if (!USE_MPFR && conf.mpfr)
846     {
847       puts ("mpfr is not available.");
848       return 0;
849     }
850   argc--;
851   argv++;
852   parsegen (&gen, argc, argv, f);
853   conf.n = gen.cnt;
854   return cmp (f, &gen, &conf);
855 }
856