]>
Commit | Line | Data |
---|---|---|
3352b62b N |
1 | /* $NetBSD: timesoftfloat.c,v 1.1 2000/06/06 08:15:11 bjh21 Exp $ */\r |
2 | \r | |
3 | /*\r | |
4 | ===============================================================================\r | |
5 | \r | |
6 | This C source file is part of the SoftFloat IEC/IEEE Floating-point\r | |
7 | Arithmetic Package, Release 2a.\r | |
8 | \r | |
9 | Written by John R. Hauser. This work was made possible in part by the\r | |
10 | International Computer Science Institute, located at Suite 600, 1947 Center\r | |
11 | Street, Berkeley, California 94704. Funding was partially provided by the\r | |
12 | National Science Foundation under grant MIP-9311980. The original version\r | |
13 | of this code was written as part of a project to build a fixed-point vector\r | |
14 | processor in collaboration with the University of California at Berkeley,\r | |
15 | overseen by Profs. Nelson Morgan and John Wawrzynek. More information\r | |
16 | is available through the Web page `http://HTTP.CS.Berkeley.EDU/~jhauser/\r | |
17 | arithmetic/SoftFloat.html'.\r | |
18 | \r | |
19 | THIS SOFTWARE IS DISTRIBUTED AS IS, FOR FREE. Although reasonable effort\r | |
20 | has been made to avoid it, THIS SOFTWARE MAY CONTAIN FAULTS THAT WILL AT\r | |
21 | TIMES RESULT IN INCORRECT BEHAVIOR. USE OF THIS SOFTWARE IS RESTRICTED TO\r | |
22 | PERSONS AND ORGANIZATIONS WHO CAN AND WILL TAKE FULL RESPONSIBILITY FOR ANY\r | |
23 | AND ALL LOSSES, COSTS, OR OTHER PROBLEMS ARISING FROM ITS USE.\r | |
24 | \r | |
25 | Derivative works are acceptable, even for commercial purposes, so long as\r | |
26 | (1) they include prominent notice that the work is derivative, and (2) they\r | |
27 | include prominent notice akin to these four paragraphs for those parts of\r | |
28 | this code that are retained.\r | |
29 | \r | |
30 | ===============================================================================\r | |
31 | */\r | |
32 | \r | |
33 | #include <sys/cdefs.h>\r | |
34 | #if defined(LIBC_SCCS) && !defined(lint)\r | |
35 | __RCSID("$NetBSD: timesoftfloat.c,v 1.1 2000/06/06 08:15:11 bjh21 Exp $");\r | |
36 | #endif /* LIBC_SCCS and not lint */\r | |
37 | \r | |
38 | #include <stdlib.h>\r | |
39 | #include <stdarg.h>\r | |
40 | #include <string.h>\r | |
41 | #include <stdio.h>\r | |
42 | #include <time.h>\r | |
43 | #include "milieu.h"\r | |
44 | #include "softfloat.h"\r | |
45 | \r | |
46 | enum {\r | |
47 | minIterations = 1000\r | |
48 | };\r | |
49 | \r | |
50 | static void fail( const char *message, ... )\r | |
51 | {\r | |
52 | va_list varArgs;\r | |
53 | \r | |
54 | fputs( "timesoftfloat: ", stderr );\r | |
55 | va_start( varArgs, message );\r | |
56 | vfprintf( stderr, message, varArgs );\r | |
57 | va_end( varArgs );\r | |
58 | fputs( ".\n", stderr );\r | |
59 | exit( EXIT_FAILURE );\r | |
60 | \r | |
61 | }\r | |
62 | \r | |
63 | static char *functionName;\r | |
64 | static char *roundingPrecisionName, *roundingModeName, *tininessModeName;\r | |
65 | \r | |
66 | static void reportTime( int32 count, long clocks )\r | |
67 | {\r | |
68 | \r | |
69 | printf(\r | |
70 | "%8.1f kops/s: %s",\r | |
71 | ( count / ( ( (float) clocks ) / CLOCKS_PER_SEC ) ) / 1000,\r | |
72 | functionName\r | |
73 | );\r | |
74 | if ( roundingModeName ) {\r | |
75 | if ( roundingPrecisionName ) {\r | |
76 | fputs( ", precision ", stdout );\r | |
77 | fputs( roundingPrecisionName, stdout );\r | |
78 | }\r | |
79 | fputs( ", rounding ", stdout );\r | |
80 | fputs( roundingModeName, stdout );\r | |
81 | if ( tininessModeName ) {\r | |
82 | fputs( ", tininess ", stdout );\r | |
83 | fputs( tininessModeName, stdout );\r | |
84 | fputs( " rounding", stdout );\r | |
85 | }\r | |
86 | }\r | |
87 | fputc( '\n', stdout );\r | |
88 | \r | |
89 | }\r | |
90 | \r | |
91 | enum {\r | |
92 | numInputs_int32 = 32\r | |
93 | };\r | |
94 | \r | |
95 | static const int32 inputs_int32[ numInputs_int32 ] = {\r | |
96 | 0xFFFFBB79, 0x405CF80F, 0x00000000, 0xFFFFFD04,\r | |
97 | 0xFFF20002, 0x0C8EF795, 0xF00011FF, 0x000006CA,\r | |
98 | 0x00009BFE, 0xFF4862E3, 0x9FFFEFFE, 0xFFFFFFB7,\r | |
99 | 0x0BFF7FFF, 0x0000F37A, 0x0011DFFE, 0x00000006,\r | |
100 | 0xFFF02006, 0xFFFFF7D1, 0x10200003, 0xDE8DF765,\r | |
101 | 0x00003E02, 0x000019E8, 0x0008FFFE, 0xFFFFFB5C,\r | |
102 | 0xFFDF7FFE, 0x07C42FBF, 0x0FFFE3FF, 0x040B9F13,\r | |
103 | 0xBFFFFFF8, 0x0001BF56, 0x000017F6, 0x000A908A\r | |
104 | };\r | |
105 | \r | |
106 | static void time_a_int32_z_float32( float32 function( int32 ) )\r | |
107 | {\r | |
108 | clock_t startClock, endClock;\r | |
109 | int32 count, i;\r | |
110 | int8 inputNum;\r | |
111 | \r | |
112 | count = 0;\r | |
113 | inputNum = 0;\r | |
114 | startClock = clock();\r | |
115 | do {\r | |
116 | for ( i = minIterations; i; --i ) {\r | |
117 | function( inputs_int32[ inputNum ] );\r | |
118 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
119 | }\r | |
120 | count += minIterations;\r | |
121 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
122 | inputNum = 0;\r | |
123 | startClock = clock();\r | |
124 | for ( i = count; i; --i ) {\r | |
125 | function( inputs_int32[ inputNum ] );\r | |
126 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
127 | }\r | |
128 | endClock = clock();\r | |
129 | reportTime( count, endClock - startClock );\r | |
130 | \r | |
131 | }\r | |
132 | \r | |
133 | static void time_a_int32_z_float64( float64 function( int32 ) )\r | |
134 | {\r | |
135 | clock_t startClock, endClock;\r | |
136 | int32 count, i;\r | |
137 | int8 inputNum;\r | |
138 | \r | |
139 | count = 0;\r | |
140 | inputNum = 0;\r | |
141 | startClock = clock();\r | |
142 | do {\r | |
143 | for ( i = minIterations; i; --i ) {\r | |
144 | function( inputs_int32[ inputNum ] );\r | |
145 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
146 | }\r | |
147 | count += minIterations;\r | |
148 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
149 | inputNum = 0;\r | |
150 | startClock = clock();\r | |
151 | for ( i = count; i; --i ) {\r | |
152 | function( inputs_int32[ inputNum ] );\r | |
153 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
154 | }\r | |
155 | endClock = clock();\r | |
156 | reportTime( count, endClock - startClock );\r | |
157 | \r | |
158 | }\r | |
159 | \r | |
160 | #ifdef FLOATX80\r | |
161 | \r | |
162 | static void time_a_int32_z_floatx80( floatx80 function( int32 ) )\r | |
163 | {\r | |
164 | clock_t startClock, endClock;\r | |
165 | int32 count, i;\r | |
166 | int8 inputNum;\r | |
167 | \r | |
168 | count = 0;\r | |
169 | inputNum = 0;\r | |
170 | startClock = clock();\r | |
171 | do {\r | |
172 | for ( i = minIterations; i; --i ) {\r | |
173 | function( inputs_int32[ inputNum ] );\r | |
174 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
175 | }\r | |
176 | count += minIterations;\r | |
177 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
178 | inputNum = 0;\r | |
179 | startClock = clock();\r | |
180 | for ( i = count; i; --i ) {\r | |
181 | function( inputs_int32[ inputNum ] );\r | |
182 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
183 | }\r | |
184 | endClock = clock();\r | |
185 | reportTime( count, endClock - startClock );\r | |
186 | \r | |
187 | }\r | |
188 | \r | |
189 | #endif\r | |
190 | \r | |
191 | #ifdef FLOAT128\r | |
192 | \r | |
193 | static void time_a_int32_z_float128( float128 function( int32 ) )\r | |
194 | {\r | |
195 | clock_t startClock, endClock;\r | |
196 | int32 count, i;\r | |
197 | int8 inputNum;\r | |
198 | \r | |
199 | count = 0;\r | |
200 | inputNum = 0;\r | |
201 | startClock = clock();\r | |
202 | do {\r | |
203 | for ( i = minIterations; i; --i ) {\r | |
204 | function( inputs_int32[ inputNum ] );\r | |
205 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
206 | }\r | |
207 | count += minIterations;\r | |
208 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
209 | inputNum = 0;\r | |
210 | startClock = clock();\r | |
211 | for ( i = count; i; --i ) {\r | |
212 | function( inputs_int32[ inputNum ] );\r | |
213 | inputNum = ( inputNum + 1 ) & ( numInputs_int32 - 1 );\r | |
214 | }\r | |
215 | endClock = clock();\r | |
216 | reportTime( count, endClock - startClock );\r | |
217 | \r | |
218 | }\r | |
219 | \r | |
220 | #endif\r | |
221 | \r | |
222 | enum {\r | |
223 | numInputs_int64 = 32\r | |
224 | };\r | |
225 | \r | |
226 | static const int64 inputs_int64[ numInputs_int64 ] = {\r | |
227 | LIT64( 0xFBFFC3FFFFFFFFFF ),\r | |
228 | LIT64( 0x0000000003C589BC ),\r | |
229 | LIT64( 0x00000000400013FE ),\r | |
230 | LIT64( 0x0000000000186171 ),\r | |
231 | LIT64( 0xFFFFFFFFFFFEFBFA ),\r | |
232 | LIT64( 0xFFFFFD79E6DFFC73 ),\r | |
233 | LIT64( 0x0000000010001DFF ),\r | |
234 | LIT64( 0xDD1A0F0C78513710 ),\r | |
235 | LIT64( 0xFFFF83FFFFFEFFFE ),\r | |
236 | LIT64( 0x00756EBD1AD0C1C7 ),\r | |
237 | LIT64( 0x0003FDFFFFFFFFBE ),\r | |
238 | LIT64( 0x0007D0FB2C2CA951 ),\r | |
239 | LIT64( 0x0007FC0007FFFFFE ),\r | |
240 | LIT64( 0x0000001F942B18BB ),\r | |
241 | LIT64( 0x0000080101FFFFFE ),\r | |
242 | LIT64( 0xFFFFFFFFFFFF0978 ),\r | |
243 | LIT64( 0x000000000008BFFF ),\r | |
244 | LIT64( 0x0000000006F5AF08 ),\r | |
245 | LIT64( 0xFFDEFF7FFFFFFFFE ),\r | |
246 | LIT64( 0x0000000000000003 ),\r | |
247 | LIT64( 0x3FFFFFFFFF80007D ),\r | |
248 | LIT64( 0x0000000000000078 ),\r | |
249 | LIT64( 0xFFF80000007FDFFD ),\r | |
250 | LIT64( 0x1BBC775B78016AB0 ),\r | |
251 | LIT64( 0xFFF9001FFFFFFFFE ),\r | |
252 | LIT64( 0xFFFD4767AB98E43F ),\r | |
253 | LIT64( 0xFFFFFEFFFE00001E ),\r | |
254 | LIT64( 0xFFFFFFFFFFF04EFD ),\r | |
255 | LIT64( 0x07FFFFFFFFFFF7FF ),\r | |
256 | LIT64( 0xFFFC9EAA38F89050 ),\r | |
257 | LIT64( 0x00000020FBFFFFFE ),\r | |
258 | LIT64( 0x0000099AE6455357 )\r | |
259 | };\r | |
260 | \r | |
261 | static void time_a_int64_z_float32( float32 function( int64 ) )\r | |
262 | {\r | |
263 | clock_t startClock, endClock;\r | |
264 | int32 count, i;\r | |
265 | int8 inputNum;\r | |
266 | \r | |
267 | count = 0;\r | |
268 | inputNum = 0;\r | |
269 | startClock = clock();\r | |
270 | do {\r | |
271 | for ( i = minIterations; i; --i ) {\r | |
272 | function( inputs_int64[ inputNum ] );\r | |
273 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
274 | }\r | |
275 | count += minIterations;\r | |
276 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
277 | inputNum = 0;\r | |
278 | startClock = clock();\r | |
279 | for ( i = count; i; --i ) {\r | |
280 | function( inputs_int64[ inputNum ] );\r | |
281 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
282 | }\r | |
283 | endClock = clock();\r | |
284 | reportTime( count, endClock - startClock );\r | |
285 | \r | |
286 | }\r | |
287 | \r | |
288 | static void time_a_int64_z_float64( float64 function( int64 ) )\r | |
289 | {\r | |
290 | clock_t startClock, endClock;\r | |
291 | int32 count, i;\r | |
292 | int8 inputNum;\r | |
293 | \r | |
294 | count = 0;\r | |
295 | inputNum = 0;\r | |
296 | startClock = clock();\r | |
297 | do {\r | |
298 | for ( i = minIterations; i; --i ) {\r | |
299 | function( inputs_int64[ inputNum ] );\r | |
300 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
301 | }\r | |
302 | count += minIterations;\r | |
303 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
304 | inputNum = 0;\r | |
305 | startClock = clock();\r | |
306 | for ( i = count; i; --i ) {\r | |
307 | function( inputs_int64[ inputNum ] );\r | |
308 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
309 | }\r | |
310 | endClock = clock();\r | |
311 | reportTime( count, endClock - startClock );\r | |
312 | \r | |
313 | }\r | |
314 | \r | |
315 | #ifdef FLOATX80\r | |
316 | \r | |
317 | static void time_a_int64_z_floatx80( floatx80 function( int64 ) )\r | |
318 | {\r | |
319 | clock_t startClock, endClock;\r | |
320 | int32 count, i;\r | |
321 | int8 inputNum;\r | |
322 | \r | |
323 | count = 0;\r | |
324 | inputNum = 0;\r | |
325 | startClock = clock();\r | |
326 | do {\r | |
327 | for ( i = minIterations; i; --i ) {\r | |
328 | function( inputs_int64[ inputNum ] );\r | |
329 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
330 | }\r | |
331 | count += minIterations;\r | |
332 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
333 | inputNum = 0;\r | |
334 | startClock = clock();\r | |
335 | for ( i = count; i; --i ) {\r | |
336 | function( inputs_int64[ inputNum ] );\r | |
337 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
338 | }\r | |
339 | endClock = clock();\r | |
340 | reportTime( count, endClock - startClock );\r | |
341 | \r | |
342 | }\r | |
343 | \r | |
344 | #endif\r | |
345 | \r | |
346 | #ifdef FLOAT128\r | |
347 | \r | |
348 | static void time_a_int64_z_float128( float128 function( int64 ) )\r | |
349 | {\r | |
350 | clock_t startClock, endClock;\r | |
351 | int32 count, i;\r | |
352 | int8 inputNum;\r | |
353 | \r | |
354 | count = 0;\r | |
355 | inputNum = 0;\r | |
356 | startClock = clock();\r | |
357 | do {\r | |
358 | for ( i = minIterations; i; --i ) {\r | |
359 | function( inputs_int64[ inputNum ] );\r | |
360 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
361 | }\r | |
362 | count += minIterations;\r | |
363 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
364 | inputNum = 0;\r | |
365 | startClock = clock();\r | |
366 | for ( i = count; i; --i ) {\r | |
367 | function( inputs_int64[ inputNum ] );\r | |
368 | inputNum = ( inputNum + 1 ) & ( numInputs_int64 - 1 );\r | |
369 | }\r | |
370 | endClock = clock();\r | |
371 | reportTime( count, endClock - startClock );\r | |
372 | \r | |
373 | }\r | |
374 | \r | |
375 | #endif\r | |
376 | \r | |
377 | enum {\r | |
378 | numInputs_float32 = 32\r | |
379 | };\r | |
380 | \r | |
381 | static const float32 inputs_float32[ numInputs_float32 ] = {\r | |
382 | 0x4EFA0000, 0xC1D0B328, 0x80000000, 0x3E69A31E,\r | |
383 | 0xAF803EFF, 0x3F800000, 0x17BF8000, 0xE74A301A,\r | |
384 | 0x4E010003, 0x7EE3C75D, 0xBD803FE0, 0xBFFEFF00,\r | |
385 | 0x7981F800, 0x431FFFFC, 0xC100C000, 0x3D87EFFF,\r | |
386 | 0x4103FEFE, 0xBC000007, 0xBF01F7FF, 0x4E6C6B5C,\r | |
387 | 0xC187FFFE, 0xC58B9F13, 0x4F88007F, 0xDF004007,\r | |
388 | 0xB7FFD7FE, 0x7E8001FB, 0x46EFFBFF, 0x31C10000,\r | |
389 | 0xDB428661, 0x33F89B1F, 0xA3BFEFFF, 0x537BFFBE\r | |
390 | };\r | |
391 | \r | |
392 | static void time_a_float32_z_int32( int32 function( float32 ) )\r | |
393 | {\r | |
394 | clock_t startClock, endClock;\r | |
395 | int32 count, i;\r | |
396 | int8 inputNum;\r | |
397 | \r | |
398 | count = 0;\r | |
399 | inputNum = 0;\r | |
400 | startClock = clock();\r | |
401 | do {\r | |
402 | for ( i = minIterations; i; --i ) {\r | |
403 | function( inputs_float32[ inputNum ] );\r | |
404 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
405 | }\r | |
406 | count += minIterations;\r | |
407 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
408 | inputNum = 0;\r | |
409 | startClock = clock();\r | |
410 | for ( i = count; i; --i ) {\r | |
411 | function( inputs_float32[ inputNum ] );\r | |
412 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
413 | }\r | |
414 | endClock = clock();\r | |
415 | reportTime( count, endClock - startClock );\r | |
416 | \r | |
417 | }\r | |
418 | \r | |
419 | static void time_a_float32_z_int64( int64 function( float32 ) )\r | |
420 | {\r | |
421 | clock_t startClock, endClock;\r | |
422 | int32 count, i;\r | |
423 | int8 inputNum;\r | |
424 | \r | |
425 | count = 0;\r | |
426 | inputNum = 0;\r | |
427 | startClock = clock();\r | |
428 | do {\r | |
429 | for ( i = minIterations; i; --i ) {\r | |
430 | function( inputs_float32[ inputNum ] );\r | |
431 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
432 | }\r | |
433 | count += minIterations;\r | |
434 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
435 | inputNum = 0;\r | |
436 | startClock = clock();\r | |
437 | for ( i = count; i; --i ) {\r | |
438 | function( inputs_float32[ inputNum ] );\r | |
439 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
440 | }\r | |
441 | endClock = clock();\r | |
442 | reportTime( count, endClock - startClock );\r | |
443 | \r | |
444 | }\r | |
445 | \r | |
446 | static void time_a_float32_z_float64( float64 function( float32 ) )\r | |
447 | {\r | |
448 | clock_t startClock, endClock;\r | |
449 | int32 count, i;\r | |
450 | int8 inputNum;\r | |
451 | \r | |
452 | count = 0;\r | |
453 | inputNum = 0;\r | |
454 | startClock = clock();\r | |
455 | do {\r | |
456 | for ( i = minIterations; i; --i ) {\r | |
457 | function( inputs_float32[ inputNum ] );\r | |
458 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
459 | }\r | |
460 | count += minIterations;\r | |
461 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
462 | inputNum = 0;\r | |
463 | startClock = clock();\r | |
464 | for ( i = count; i; --i ) {\r | |
465 | function( inputs_float32[ inputNum ] );\r | |
466 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
467 | }\r | |
468 | endClock = clock();\r | |
469 | reportTime( count, endClock - startClock );\r | |
470 | \r | |
471 | }\r | |
472 | \r | |
473 | #ifdef FLOATX80\r | |
474 | \r | |
475 | static void time_a_float32_z_floatx80( floatx80 function( float32 ) )\r | |
476 | {\r | |
477 | clock_t startClock, endClock;\r | |
478 | int32 count, i;\r | |
479 | int8 inputNum;\r | |
480 | \r | |
481 | count = 0;\r | |
482 | inputNum = 0;\r | |
483 | startClock = clock();\r | |
484 | do {\r | |
485 | for ( i = minIterations; i; --i ) {\r | |
486 | function( inputs_float32[ inputNum ] );\r | |
487 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
488 | }\r | |
489 | count += minIterations;\r | |
490 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
491 | inputNum = 0;\r | |
492 | startClock = clock();\r | |
493 | for ( i = count; i; --i ) {\r | |
494 | function( inputs_float32[ inputNum ] );\r | |
495 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
496 | }\r | |
497 | endClock = clock();\r | |
498 | reportTime( count, endClock - startClock );\r | |
499 | \r | |
500 | }\r | |
501 | \r | |
502 | #endif\r | |
503 | \r | |
504 | #ifdef FLOAT128\r | |
505 | \r | |
506 | static void time_a_float32_z_float128( float128 function( float32 ) )\r | |
507 | {\r | |
508 | clock_t startClock, endClock;\r | |
509 | int32 count, i;\r | |
510 | int8 inputNum;\r | |
511 | \r | |
512 | count = 0;\r | |
513 | inputNum = 0;\r | |
514 | startClock = clock();\r | |
515 | do {\r | |
516 | for ( i = minIterations; i; --i ) {\r | |
517 | function( inputs_float32[ inputNum ] );\r | |
518 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
519 | }\r | |
520 | count += minIterations;\r | |
521 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
522 | inputNum = 0;\r | |
523 | startClock = clock();\r | |
524 | for ( i = count; i; --i ) {\r | |
525 | function( inputs_float32[ inputNum ] );\r | |
526 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
527 | }\r | |
528 | endClock = clock();\r | |
529 | reportTime( count, endClock - startClock );\r | |
530 | \r | |
531 | }\r | |
532 | \r | |
533 | #endif\r | |
534 | \r | |
535 | static void time_az_float32( float32 function( float32 ) )\r | |
536 | {\r | |
537 | clock_t startClock, endClock;\r | |
538 | int32 count, i;\r | |
539 | int8 inputNum;\r | |
540 | \r | |
541 | count = 0;\r | |
542 | inputNum = 0;\r | |
543 | startClock = clock();\r | |
544 | do {\r | |
545 | for ( i = minIterations; i; --i ) {\r | |
546 | function( inputs_float32[ inputNum ] );\r | |
547 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
548 | }\r | |
549 | count += minIterations;\r | |
550 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
551 | inputNum = 0;\r | |
552 | startClock = clock();\r | |
553 | for ( i = count; i; --i ) {\r | |
554 | function( inputs_float32[ inputNum ] );\r | |
555 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
556 | }\r | |
557 | endClock = clock();\r | |
558 | reportTime( count, endClock - startClock );\r | |
559 | \r | |
560 | }\r | |
561 | \r | |
562 | static void time_ab_float32_z_flag( flag function( float32, float32 ) )\r | |
563 | {\r | |
564 | clock_t startClock, endClock;\r | |
565 | int32 count, i;\r | |
566 | int8 inputNumA, inputNumB;\r | |
567 | \r | |
568 | count = 0;\r | |
569 | inputNumA = 0;\r | |
570 | inputNumB = 0;\r | |
571 | startClock = clock();\r | |
572 | do {\r | |
573 | for ( i = minIterations; i; --i ) {\r | |
574 | function(\r | |
575 | inputs_float32[ inputNumA ], inputs_float32[ inputNumB ] );\r | |
576 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float32 - 1 );\r | |
577 | if ( inputNumA == 0 ) ++inputNumB;\r | |
578 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float32 - 1 );\r | |
579 | }\r | |
580 | count += minIterations;\r | |
581 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
582 | inputNumA = 0;\r | |
583 | inputNumB = 0;\r | |
584 | startClock = clock();\r | |
585 | for ( i = count; i; --i ) {\r | |
586 | function(\r | |
587 | inputs_float32[ inputNumA ], inputs_float32[ inputNumB ] );\r | |
588 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float32 - 1 );\r | |
589 | if ( inputNumA == 0 ) ++inputNumB;\r | |
590 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float32 - 1 );\r | |
591 | }\r | |
592 | endClock = clock();\r | |
593 | reportTime( count, endClock - startClock );\r | |
594 | \r | |
595 | }\r | |
596 | \r | |
597 | static void time_abz_float32( float32 function( float32, float32 ) )\r | |
598 | {\r | |
599 | clock_t startClock, endClock;\r | |
600 | int32 count, i;\r | |
601 | int8 inputNumA, inputNumB;\r | |
602 | \r | |
603 | count = 0;\r | |
604 | inputNumA = 0;\r | |
605 | inputNumB = 0;\r | |
606 | startClock = clock();\r | |
607 | do {\r | |
608 | for ( i = minIterations; i; --i ) {\r | |
609 | function(\r | |
610 | inputs_float32[ inputNumA ], inputs_float32[ inputNumB ] );\r | |
611 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float32 - 1 );\r | |
612 | if ( inputNumA == 0 ) ++inputNumB;\r | |
613 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float32 - 1 );\r | |
614 | }\r | |
615 | count += minIterations;\r | |
616 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
617 | inputNumA = 0;\r | |
618 | inputNumB = 0;\r | |
619 | startClock = clock();\r | |
620 | for ( i = count; i; --i ) {\r | |
621 | function(\r | |
622 | inputs_float32[ inputNumA ], inputs_float32[ inputNumB ] );\r | |
623 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float32 - 1 );\r | |
624 | if ( inputNumA == 0 ) ++inputNumB;\r | |
625 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float32 - 1 );\r | |
626 | }\r | |
627 | endClock = clock();\r | |
628 | reportTime( count, endClock - startClock );\r | |
629 | \r | |
630 | }\r | |
631 | \r | |
632 | static const float32 inputs_float32_pos[ numInputs_float32 ] = {\r | |
633 | 0x4EFA0000, 0x41D0B328, 0x00000000, 0x3E69A31E,\r | |
634 | 0x2F803EFF, 0x3F800000, 0x17BF8000, 0x674A301A,\r | |
635 | 0x4E010003, 0x7EE3C75D, 0x3D803FE0, 0x3FFEFF00,\r | |
636 | 0x7981F800, 0x431FFFFC, 0x4100C000, 0x3D87EFFF,\r | |
637 | 0x4103FEFE, 0x3C000007, 0x3F01F7FF, 0x4E6C6B5C,\r | |
638 | 0x4187FFFE, 0x458B9F13, 0x4F88007F, 0x5F004007,\r | |
639 | 0x37FFD7FE, 0x7E8001FB, 0x46EFFBFF, 0x31C10000,\r | |
640 | 0x5B428661, 0x33F89B1F, 0x23BFEFFF, 0x537BFFBE\r | |
641 | };\r | |
642 | \r | |
643 | static void time_az_float32_pos( float32 function( float32 ) )\r | |
644 | {\r | |
645 | clock_t startClock, endClock;\r | |
646 | int32 count, i;\r | |
647 | int8 inputNum;\r | |
648 | \r | |
649 | count = 0;\r | |
650 | inputNum = 0;\r | |
651 | startClock = clock();\r | |
652 | do {\r | |
653 | for ( i = minIterations; i; --i ) {\r | |
654 | function( inputs_float32_pos[ inputNum ] );\r | |
655 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
656 | }\r | |
657 | count += minIterations;\r | |
658 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
659 | inputNum = 0;\r | |
660 | startClock = clock();\r | |
661 | for ( i = count; i; --i ) {\r | |
662 | function( inputs_float32_pos[ inputNum ] );\r | |
663 | inputNum = ( inputNum + 1 ) & ( numInputs_float32 - 1 );\r | |
664 | }\r | |
665 | endClock = clock();\r | |
666 | reportTime( count, endClock - startClock );\r | |
667 | \r | |
668 | }\r | |
669 | \r | |
670 | enum {\r | |
671 | numInputs_float64 = 32\r | |
672 | };\r | |
673 | \r | |
674 | static const float64 inputs_float64[ numInputs_float64 ] = {\r | |
675 | LIT64( 0x422FFFC008000000 ),\r | |
676 | LIT64( 0xB7E0000480000000 ),\r | |
677 | LIT64( 0xF3FD2546120B7935 ),\r | |
678 | LIT64( 0x3FF0000000000000 ),\r | |
679 | LIT64( 0xCE07F766F09588D6 ),\r | |
680 | LIT64( 0x8000000000000000 ),\r | |
681 | LIT64( 0x3FCE000400000000 ),\r | |
682 | LIT64( 0x8313B60F0032BED8 ),\r | |
683 | LIT64( 0xC1EFFFFFC0002000 ),\r | |
684 | LIT64( 0x3FB3C75D224F2B0F ),\r | |
685 | LIT64( 0x7FD00000004000FF ),\r | |
686 | LIT64( 0xA12FFF8000001FFF ),\r | |
687 | LIT64( 0x3EE0000000FE0000 ),\r | |
688 | LIT64( 0x0010000080000004 ),\r | |
689 | LIT64( 0x41CFFFFE00000020 ),\r | |
690 | LIT64( 0x40303FFFFFFFFFFD ),\r | |
691 | LIT64( 0x3FD000003FEFFFFF ),\r | |
692 | LIT64( 0xBFD0000010000000 ),\r | |
693 | LIT64( 0xB7FC6B5C16CA55CF ),\r | |
694 | LIT64( 0x413EEB940B9D1301 ),\r | |
695 | LIT64( 0xC7E00200001FFFFF ),\r | |
696 | LIT64( 0x47F00021FFFFFFFE ),\r | |
697 | LIT64( 0xBFFFFFFFF80000FF ),\r | |
698 | LIT64( 0xC07FFFFFE00FFFFF ),\r | |
699 | LIT64( 0x001497A63740C5E8 ),\r | |
700 | LIT64( 0xC4BFFFE0001FFFFF ),\r | |
701 | LIT64( 0x96FFDFFEFFFFFFFF ),\r | |
702 | LIT64( 0x403FC000000001FE ),\r | |
703 | LIT64( 0xFFD00000000001F6 ),\r | |
704 | LIT64( 0x0640400002000000 ),\r | |
705 | LIT64( 0x479CEE1E4F789FE0 ),\r | |
706 | LIT64( 0xC237FFFFFFFFFDFE )\r | |
707 | };\r | |
708 | \r | |
709 | static void time_a_float64_z_int32( int32 function( float64 ) )\r | |
710 | {\r | |
711 | clock_t startClock, endClock;\r | |
712 | int32 count, i;\r | |
713 | int8 inputNum;\r | |
714 | \r | |
715 | count = 0;\r | |
716 | inputNum = 0;\r | |
717 | startClock = clock();\r | |
718 | do {\r | |
719 | for ( i = minIterations; i; --i ) {\r | |
720 | function( inputs_float64[ inputNum ] );\r | |
721 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
722 | }\r | |
723 | count += minIterations;\r | |
724 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
725 | inputNum = 0;\r | |
726 | startClock = clock();\r | |
727 | for ( i = count; i; --i ) {\r | |
728 | function( inputs_float64[ inputNum ] );\r | |
729 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
730 | }\r | |
731 | endClock = clock();\r | |
732 | reportTime( count, endClock - startClock );\r | |
733 | \r | |
734 | }\r | |
735 | \r | |
736 | static void time_a_float64_z_int64( int64 function( float64 ) )\r | |
737 | {\r | |
738 | clock_t startClock, endClock;\r | |
739 | int32 count, i;\r | |
740 | int8 inputNum;\r | |
741 | \r | |
742 | count = 0;\r | |
743 | inputNum = 0;\r | |
744 | startClock = clock();\r | |
745 | do {\r | |
746 | for ( i = minIterations; i; --i ) {\r | |
747 | function( inputs_float64[ inputNum ] );\r | |
748 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
749 | }\r | |
750 | count += minIterations;\r | |
751 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
752 | inputNum = 0;\r | |
753 | startClock = clock();\r | |
754 | for ( i = count; i; --i ) {\r | |
755 | function( inputs_float64[ inputNum ] );\r | |
756 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
757 | }\r | |
758 | endClock = clock();\r | |
759 | reportTime( count, endClock - startClock );\r | |
760 | \r | |
761 | }\r | |
762 | \r | |
763 | static void time_a_float64_z_float32( float32 function( float64 ) )\r | |
764 | {\r | |
765 | clock_t startClock, endClock;\r | |
766 | int32 count, i;\r | |
767 | int8 inputNum;\r | |
768 | \r | |
769 | count = 0;\r | |
770 | inputNum = 0;\r | |
771 | startClock = clock();\r | |
772 | do {\r | |
773 | for ( i = minIterations; i; --i ) {\r | |
774 | function( inputs_float64[ inputNum ] );\r | |
775 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
776 | }\r | |
777 | count += minIterations;\r | |
778 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
779 | inputNum = 0;\r | |
780 | startClock = clock();\r | |
781 | for ( i = count; i; --i ) {\r | |
782 | function( inputs_float64[ inputNum ] );\r | |
783 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
784 | }\r | |
785 | endClock = clock();\r | |
786 | reportTime( count, endClock - startClock );\r | |
787 | \r | |
788 | }\r | |
789 | \r | |
790 | #ifdef FLOATX80\r | |
791 | \r | |
792 | static void time_a_float64_z_floatx80( floatx80 function( float64 ) )\r | |
793 | {\r | |
794 | clock_t startClock, endClock;\r | |
795 | int32 count, i;\r | |
796 | int8 inputNum;\r | |
797 | \r | |
798 | count = 0;\r | |
799 | inputNum = 0;\r | |
800 | startClock = clock();\r | |
801 | do {\r | |
802 | for ( i = minIterations; i; --i ) {\r | |
803 | function( inputs_float64[ inputNum ] );\r | |
804 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
805 | }\r | |
806 | count += minIterations;\r | |
807 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
808 | inputNum = 0;\r | |
809 | startClock = clock();\r | |
810 | for ( i = count; i; --i ) {\r | |
811 | function( inputs_float64[ inputNum ] );\r | |
812 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
813 | }\r | |
814 | endClock = clock();\r | |
815 | reportTime( count, endClock - startClock );\r | |
816 | \r | |
817 | }\r | |
818 | \r | |
819 | #endif\r | |
820 | \r | |
821 | #ifdef FLOAT128\r | |
822 | \r | |
823 | static void time_a_float64_z_float128( float128 function( float64 ) )\r | |
824 | {\r | |
825 | clock_t startClock, endClock;\r | |
826 | int32 count, i;\r | |
827 | int8 inputNum;\r | |
828 | \r | |
829 | count = 0;\r | |
830 | inputNum = 0;\r | |
831 | startClock = clock();\r | |
832 | do {\r | |
833 | for ( i = minIterations; i; --i ) {\r | |
834 | function( inputs_float64[ inputNum ] );\r | |
835 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
836 | }\r | |
837 | count += minIterations;\r | |
838 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
839 | inputNum = 0;\r | |
840 | startClock = clock();\r | |
841 | for ( i = count; i; --i ) {\r | |
842 | function( inputs_float64[ inputNum ] );\r | |
843 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
844 | }\r | |
845 | endClock = clock();\r | |
846 | reportTime( count, endClock - startClock );\r | |
847 | \r | |
848 | }\r | |
849 | \r | |
850 | #endif\r | |
851 | \r | |
852 | static void time_az_float64( float64 function( float64 ) )\r | |
853 | {\r | |
854 | clock_t startClock, endClock;\r | |
855 | int32 count, i;\r | |
856 | int8 inputNum;\r | |
857 | \r | |
858 | count = 0;\r | |
859 | inputNum = 0;\r | |
860 | startClock = clock();\r | |
861 | do {\r | |
862 | for ( i = minIterations; i; --i ) {\r | |
863 | function( inputs_float64[ inputNum ] );\r | |
864 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
865 | }\r | |
866 | count += minIterations;\r | |
867 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
868 | inputNum = 0;\r | |
869 | startClock = clock();\r | |
870 | for ( i = count; i; --i ) {\r | |
871 | function( inputs_float64[ inputNum ] );\r | |
872 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
873 | }\r | |
874 | endClock = clock();\r | |
875 | reportTime( count, endClock - startClock );\r | |
876 | \r | |
877 | }\r | |
878 | \r | |
879 | static void time_ab_float64_z_flag( flag function( float64, float64 ) )\r | |
880 | {\r | |
881 | clock_t startClock, endClock;\r | |
882 | int32 count, i;\r | |
883 | int8 inputNumA, inputNumB;\r | |
884 | \r | |
885 | count = 0;\r | |
886 | inputNumA = 0;\r | |
887 | inputNumB = 0;\r | |
888 | startClock = clock();\r | |
889 | do {\r | |
890 | for ( i = minIterations; i; --i ) {\r | |
891 | function(\r | |
892 | inputs_float64[ inputNumA ], inputs_float64[ inputNumB ] );\r | |
893 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float64 - 1 );\r | |
894 | if ( inputNumA == 0 ) ++inputNumB;\r | |
895 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float64 - 1 );\r | |
896 | }\r | |
897 | count += minIterations;\r | |
898 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
899 | inputNumA = 0;\r | |
900 | inputNumB = 0;\r | |
901 | startClock = clock();\r | |
902 | for ( i = count; i; --i ) {\r | |
903 | function(\r | |
904 | inputs_float64[ inputNumA ], inputs_float64[ inputNumB ] );\r | |
905 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float64 - 1 );\r | |
906 | if ( inputNumA == 0 ) ++inputNumB;\r | |
907 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float64 - 1 );\r | |
908 | }\r | |
909 | endClock = clock();\r | |
910 | reportTime( count, endClock - startClock );\r | |
911 | \r | |
912 | }\r | |
913 | \r | |
914 | static void time_abz_float64( float64 function( float64, float64 ) )\r | |
915 | {\r | |
916 | clock_t startClock, endClock;\r | |
917 | int32 count, i;\r | |
918 | int8 inputNumA, inputNumB;\r | |
919 | \r | |
920 | count = 0;\r | |
921 | inputNumA = 0;\r | |
922 | inputNumB = 0;\r | |
923 | startClock = clock();\r | |
924 | do {\r | |
925 | for ( i = minIterations; i; --i ) {\r | |
926 | function(\r | |
927 | inputs_float64[ inputNumA ], inputs_float64[ inputNumB ] );\r | |
928 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float64 - 1 );\r | |
929 | if ( inputNumA == 0 ) ++inputNumB;\r | |
930 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float64 - 1 );\r | |
931 | }\r | |
932 | count += minIterations;\r | |
933 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
934 | inputNumA = 0;\r | |
935 | inputNumB = 0;\r | |
936 | startClock = clock();\r | |
937 | for ( i = count; i; --i ) {\r | |
938 | function(\r | |
939 | inputs_float64[ inputNumA ], inputs_float64[ inputNumB ] );\r | |
940 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float64 - 1 );\r | |
941 | if ( inputNumA == 0 ) ++inputNumB;\r | |
942 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float64 - 1 );\r | |
943 | }\r | |
944 | endClock = clock();\r | |
945 | reportTime( count, endClock - startClock );\r | |
946 | \r | |
947 | }\r | |
948 | \r | |
949 | static const float64 inputs_float64_pos[ numInputs_float64 ] = {\r | |
950 | LIT64( 0x422FFFC008000000 ),\r | |
951 | LIT64( 0x37E0000480000000 ),\r | |
952 | LIT64( 0x73FD2546120B7935 ),\r | |
953 | LIT64( 0x3FF0000000000000 ),\r | |
954 | LIT64( 0x4E07F766F09588D6 ),\r | |
955 | LIT64( 0x0000000000000000 ),\r | |
956 | LIT64( 0x3FCE000400000000 ),\r | |
957 | LIT64( 0x0313B60F0032BED8 ),\r | |
958 | LIT64( 0x41EFFFFFC0002000 ),\r | |
959 | LIT64( 0x3FB3C75D224F2B0F ),\r | |
960 | LIT64( 0x7FD00000004000FF ),\r | |
961 | LIT64( 0x212FFF8000001FFF ),\r | |
962 | LIT64( 0x3EE0000000FE0000 ),\r | |
963 | LIT64( 0x0010000080000004 ),\r | |
964 | LIT64( 0x41CFFFFE00000020 ),\r | |
965 | LIT64( 0x40303FFFFFFFFFFD ),\r | |
966 | LIT64( 0x3FD000003FEFFFFF ),\r | |
967 | LIT64( 0x3FD0000010000000 ),\r | |
968 | LIT64( 0x37FC6B5C16CA55CF ),\r | |
969 | LIT64( 0x413EEB940B9D1301 ),\r | |
970 | LIT64( 0x47E00200001FFFFF ),\r | |
971 | LIT64( 0x47F00021FFFFFFFE ),\r | |
972 | LIT64( 0x3FFFFFFFF80000FF ),\r | |
973 | LIT64( 0x407FFFFFE00FFFFF ),\r | |
974 | LIT64( 0x001497A63740C5E8 ),\r | |
975 | LIT64( 0x44BFFFE0001FFFFF ),\r | |
976 | LIT64( 0x16FFDFFEFFFFFFFF ),\r | |
977 | LIT64( 0x403FC000000001FE ),\r | |
978 | LIT64( 0x7FD00000000001F6 ),\r | |
979 | LIT64( 0x0640400002000000 ),\r | |
980 | LIT64( 0x479CEE1E4F789FE0 ),\r | |
981 | LIT64( 0x4237FFFFFFFFFDFE )\r | |
982 | };\r | |
983 | \r | |
984 | static void time_az_float64_pos( float64 function( float64 ) )\r | |
985 | {\r | |
986 | clock_t startClock, endClock;\r | |
987 | int32 count, i;\r | |
988 | int8 inputNum;\r | |
989 | \r | |
990 | count = 0;\r | |
991 | inputNum = 0;\r | |
992 | startClock = clock();\r | |
993 | do {\r | |
994 | for ( i = minIterations; i; --i ) {\r | |
995 | function( inputs_float64_pos[ inputNum ] );\r | |
996 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
997 | }\r | |
998 | count += minIterations;\r | |
999 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1000 | inputNum = 0;\r | |
1001 | startClock = clock();\r | |
1002 | for ( i = count; i; --i ) {\r | |
1003 | function( inputs_float64_pos[ inputNum ] );\r | |
1004 | inputNum = ( inputNum + 1 ) & ( numInputs_float64 - 1 );\r | |
1005 | }\r | |
1006 | endClock = clock();\r | |
1007 | reportTime( count, endClock - startClock );\r | |
1008 | \r | |
1009 | }\r | |
1010 | \r | |
1011 | #ifdef FLOATX80\r | |
1012 | \r | |
1013 | enum {\r | |
1014 | numInputs_floatx80 = 32\r | |
1015 | };\r | |
1016 | \r | |
1017 | static const struct {\r | |
1018 | bits16 high;\r | |
1019 | bits64 low;\r | |
1020 | } inputs_floatx80[ numInputs_floatx80 ] = {\r | |
1021 | { 0xC03F, LIT64( 0xA9BE15A19C1E8B62 ) },\r | |
1022 | { 0x8000, LIT64( 0x0000000000000000 ) },\r | |
1023 | { 0x75A8, LIT64( 0xE59591E4788957A5 ) },\r | |
1024 | { 0xBFFF, LIT64( 0xFFF0000000000040 ) },\r | |
1025 | { 0x0CD8, LIT64( 0xFC000000000007FE ) },\r | |
1026 | { 0x43BA, LIT64( 0x99A4000000000000 ) },\r | |
1027 | { 0x3FFF, LIT64( 0x8000000000000000 ) },\r | |
1028 | { 0x4081, LIT64( 0x94FBF1BCEB5545F0 ) },\r | |
1029 | { 0x403E, LIT64( 0xFFF0000000002000 ) },\r | |
1030 | { 0x3FFE, LIT64( 0xC860E3C75D224F28 ) },\r | |
1031 | { 0x407E, LIT64( 0xFC00000FFFFFFFFE ) },\r | |
1032 | { 0x737A, LIT64( 0x800000007FFDFFFE ) },\r | |
1033 | { 0x4044, LIT64( 0xFFFFFF80000FFFFF ) },\r | |
1034 | { 0xBBFE, LIT64( 0x8000040000001FFE ) },\r | |
1035 | { 0xC002, LIT64( 0xFF80000000000020 ) },\r | |
1036 | { 0xDE8D, LIT64( 0xFFFFFFFFFFE00004 ) },\r | |
1037 | { 0xC004, LIT64( 0x8000000000003FFB ) },\r | |
1038 | { 0x407F, LIT64( 0x800000000003FFFE ) },\r | |
1039 | { 0xC000, LIT64( 0xA459EE6A5C16CA55 ) },\r | |
1040 | { 0x8003, LIT64( 0xC42CBF7399AEEB94 ) },\r | |
1041 | { 0xBF7F, LIT64( 0xF800000000000006 ) },\r | |
1042 | { 0xC07F, LIT64( 0xBF56BE8871F28FEA ) },\r | |
1043 | { 0xC07E, LIT64( 0xFFFF77FFFFFFFFFE ) },\r | |
1044 | { 0xADC9, LIT64( 0x8000000FFFFFFFDE ) },\r | |
1045 | { 0xC001, LIT64( 0xEFF7FFFFFFFFFFFF ) },\r | |
1046 | { 0x4001, LIT64( 0xBE84F30125C497A6 ) },\r | |
1047 | { 0xC06B, LIT64( 0xEFFFFFFFFFFFFFFF ) },\r | |
1048 | { 0x4080, LIT64( 0xFFFFFFFFBFFFFFFF ) },\r | |
1049 | { 0x87E9, LIT64( 0x81FFFFFFFFFFFBFF ) },\r | |
1050 | { 0xA63F, LIT64( 0x801FFFFFFEFFFFFE ) },\r | |
1051 | { 0x403C, LIT64( 0x801FFFFFFFF7FFFF ) },\r | |
1052 | { 0x4018, LIT64( 0x8000000000080003 ) }\r | |
1053 | };\r | |
1054 | \r | |
1055 | static void time_a_floatx80_z_int32( int32 function( floatx80 ) )\r | |
1056 | {\r | |
1057 | clock_t startClock, endClock;\r | |
1058 | int32 count, i;\r | |
1059 | int8 inputNum;\r | |
1060 | floatx80 a;\r | |
1061 | \r | |
1062 | count = 0;\r | |
1063 | inputNum = 0;\r | |
1064 | startClock = clock();\r | |
1065 | do {\r | |
1066 | for ( i = minIterations; i; --i ) {\r | |
1067 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1068 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1069 | function( a );\r | |
1070 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1071 | }\r | |
1072 | count += minIterations;\r | |
1073 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1074 | inputNum = 0;\r | |
1075 | startClock = clock();\r | |
1076 | for ( i = count; i; --i ) {\r | |
1077 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1078 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1079 | function( a );\r | |
1080 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1081 | }\r | |
1082 | endClock = clock();\r | |
1083 | reportTime( count, endClock - startClock );\r | |
1084 | \r | |
1085 | }\r | |
1086 | \r | |
1087 | static void time_a_floatx80_z_int64( int64 function( floatx80 ) )\r | |
1088 | {\r | |
1089 | clock_t startClock, endClock;\r | |
1090 | int32 count, i;\r | |
1091 | int8 inputNum;\r | |
1092 | floatx80 a;\r | |
1093 | \r | |
1094 | count = 0;\r | |
1095 | inputNum = 0;\r | |
1096 | startClock = clock();\r | |
1097 | do {\r | |
1098 | for ( i = minIterations; i; --i ) {\r | |
1099 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1100 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1101 | function( a );\r | |
1102 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1103 | }\r | |
1104 | count += minIterations;\r | |
1105 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1106 | inputNum = 0;\r | |
1107 | startClock = clock();\r | |
1108 | for ( i = count; i; --i ) {\r | |
1109 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1110 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1111 | function( a );\r | |
1112 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1113 | }\r | |
1114 | endClock = clock();\r | |
1115 | reportTime( count, endClock - startClock );\r | |
1116 | \r | |
1117 | }\r | |
1118 | \r | |
1119 | static void time_a_floatx80_z_float32( float32 function( floatx80 ) )\r | |
1120 | {\r | |
1121 | clock_t startClock, endClock;\r | |
1122 | int32 count, i;\r | |
1123 | int8 inputNum;\r | |
1124 | floatx80 a;\r | |
1125 | \r | |
1126 | count = 0;\r | |
1127 | inputNum = 0;\r | |
1128 | startClock = clock();\r | |
1129 | do {\r | |
1130 | for ( i = minIterations; i; --i ) {\r | |
1131 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1132 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1133 | function( a );\r | |
1134 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1135 | }\r | |
1136 | count += minIterations;\r | |
1137 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1138 | inputNum = 0;\r | |
1139 | startClock = clock();\r | |
1140 | for ( i = count; i; --i ) {\r | |
1141 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1142 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1143 | function( a );\r | |
1144 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1145 | }\r | |
1146 | endClock = clock();\r | |
1147 | reportTime( count, endClock - startClock );\r | |
1148 | \r | |
1149 | }\r | |
1150 | \r | |
1151 | static void time_a_floatx80_z_float64( float64 function( floatx80 ) )\r | |
1152 | {\r | |
1153 | clock_t startClock, endClock;\r | |
1154 | int32 count, i;\r | |
1155 | int8 inputNum;\r | |
1156 | floatx80 a;\r | |
1157 | \r | |
1158 | count = 0;\r | |
1159 | inputNum = 0;\r | |
1160 | startClock = clock();\r | |
1161 | do {\r | |
1162 | for ( i = minIterations; i; --i ) {\r | |
1163 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1164 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1165 | function( a );\r | |
1166 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1167 | }\r | |
1168 | count += minIterations;\r | |
1169 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1170 | inputNum = 0;\r | |
1171 | startClock = clock();\r | |
1172 | for ( i = count; i; --i ) {\r | |
1173 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1174 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1175 | function( a );\r | |
1176 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1177 | }\r | |
1178 | endClock = clock();\r | |
1179 | reportTime( count, endClock - startClock );\r | |
1180 | \r | |
1181 | }\r | |
1182 | \r | |
1183 | #ifdef FLOAT128\r | |
1184 | \r | |
1185 | static void time_a_floatx80_z_float128( float128 function( floatx80 ) )\r | |
1186 | {\r | |
1187 | clock_t startClock, endClock;\r | |
1188 | int32 count, i;\r | |
1189 | int8 inputNum;\r | |
1190 | floatx80 a;\r | |
1191 | \r | |
1192 | count = 0;\r | |
1193 | inputNum = 0;\r | |
1194 | startClock = clock();\r | |
1195 | do {\r | |
1196 | for ( i = minIterations; i; --i ) {\r | |
1197 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1198 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1199 | function( a );\r | |
1200 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1201 | }\r | |
1202 | count += minIterations;\r | |
1203 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1204 | inputNum = 0;\r | |
1205 | startClock = clock();\r | |
1206 | for ( i = count; i; --i ) {\r | |
1207 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1208 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1209 | function( a );\r | |
1210 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1211 | }\r | |
1212 | endClock = clock();\r | |
1213 | reportTime( count, endClock - startClock );\r | |
1214 | \r | |
1215 | }\r | |
1216 | \r | |
1217 | #endif\r | |
1218 | \r | |
1219 | static void time_az_floatx80( floatx80 function( floatx80 ) )\r | |
1220 | {\r | |
1221 | clock_t startClock, endClock;\r | |
1222 | int32 count, i;\r | |
1223 | int8 inputNum;\r | |
1224 | floatx80 a;\r | |
1225 | \r | |
1226 | count = 0;\r | |
1227 | inputNum = 0;\r | |
1228 | startClock = clock();\r | |
1229 | do {\r | |
1230 | for ( i = minIterations; i; --i ) {\r | |
1231 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1232 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1233 | function( a );\r | |
1234 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1235 | }\r | |
1236 | count += minIterations;\r | |
1237 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1238 | inputNum = 0;\r | |
1239 | startClock = clock();\r | |
1240 | for ( i = count; i; --i ) {\r | |
1241 | a.low = inputs_floatx80[ inputNum ].low;\r | |
1242 | a.high = inputs_floatx80[ inputNum ].high;\r | |
1243 | function( a );\r | |
1244 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1245 | }\r | |
1246 | endClock = clock();\r | |
1247 | reportTime( count, endClock - startClock );\r | |
1248 | \r | |
1249 | }\r | |
1250 | \r | |
1251 | static void time_ab_floatx80_z_flag( flag function( floatx80, floatx80 ) )\r | |
1252 | {\r | |
1253 | clock_t startClock, endClock;\r | |
1254 | int32 count, i;\r | |
1255 | int8 inputNumA, inputNumB;\r | |
1256 | floatx80 a, b;\r | |
1257 | \r | |
1258 | count = 0;\r | |
1259 | inputNumA = 0;\r | |
1260 | inputNumB = 0;\r | |
1261 | startClock = clock();\r | |
1262 | do {\r | |
1263 | for ( i = minIterations; i; --i ) {\r | |
1264 | a.low = inputs_floatx80[ inputNumA ].low;\r | |
1265 | a.high = inputs_floatx80[ inputNumA ].high;\r | |
1266 | b.low = inputs_floatx80[ inputNumB ].low;\r | |
1267 | b.high = inputs_floatx80[ inputNumB ].high;\r | |
1268 | function( a, b );\r | |
1269 | inputNumA = ( inputNumA + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1270 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1271 | inputNumB = ( inputNumB + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1272 | }\r | |
1273 | count += minIterations;\r | |
1274 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1275 | inputNumA = 0;\r | |
1276 | inputNumB = 0;\r | |
1277 | startClock = clock();\r | |
1278 | for ( i = count; i; --i ) {\r | |
1279 | a.low = inputs_floatx80[ inputNumA ].low;\r | |
1280 | a.high = inputs_floatx80[ inputNumA ].high;\r | |
1281 | b.low = inputs_floatx80[ inputNumB ].low;\r | |
1282 | b.high = inputs_floatx80[ inputNumB ].high;\r | |
1283 | function( a, b );\r | |
1284 | inputNumA = ( inputNumA + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1285 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1286 | inputNumB = ( inputNumB + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1287 | }\r | |
1288 | endClock = clock();\r | |
1289 | reportTime( count, endClock - startClock );\r | |
1290 | \r | |
1291 | }\r | |
1292 | \r | |
1293 | static void time_abz_floatx80( floatx80 function( floatx80, floatx80 ) )\r | |
1294 | {\r | |
1295 | clock_t startClock, endClock;\r | |
1296 | int32 count, i;\r | |
1297 | int8 inputNumA, inputNumB;\r | |
1298 | floatx80 a, b;\r | |
1299 | \r | |
1300 | count = 0;\r | |
1301 | inputNumA = 0;\r | |
1302 | inputNumB = 0;\r | |
1303 | startClock = clock();\r | |
1304 | do {\r | |
1305 | for ( i = minIterations; i; --i ) {\r | |
1306 | a.low = inputs_floatx80[ inputNumA ].low;\r | |
1307 | a.high = inputs_floatx80[ inputNumA ].high;\r | |
1308 | b.low = inputs_floatx80[ inputNumB ].low;\r | |
1309 | b.high = inputs_floatx80[ inputNumB ].high;\r | |
1310 | function( a, b );\r | |
1311 | inputNumA = ( inputNumA + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1312 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1313 | inputNumB = ( inputNumB + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1314 | }\r | |
1315 | count += minIterations;\r | |
1316 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1317 | inputNumA = 0;\r | |
1318 | inputNumB = 0;\r | |
1319 | startClock = clock();\r | |
1320 | for ( i = count; i; --i ) {\r | |
1321 | a.low = inputs_floatx80[ inputNumA ].low;\r | |
1322 | a.high = inputs_floatx80[ inputNumA ].high;\r | |
1323 | b.low = inputs_floatx80[ inputNumB ].low;\r | |
1324 | b.high = inputs_floatx80[ inputNumB ].high;\r | |
1325 | function( a, b );\r | |
1326 | inputNumA = ( inputNumA + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1327 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1328 | inputNumB = ( inputNumB + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1329 | }\r | |
1330 | endClock = clock();\r | |
1331 | reportTime( count, endClock - startClock );\r | |
1332 | \r | |
1333 | }\r | |
1334 | \r | |
1335 | static const struct {\r | |
1336 | bits16 high;\r | |
1337 | bits64 low;\r | |
1338 | } inputs_floatx80_pos[ numInputs_floatx80 ] = {\r | |
1339 | { 0x403F, LIT64( 0xA9BE15A19C1E8B62 ) },\r | |
1340 | { 0x0000, LIT64( 0x0000000000000000 ) },\r | |
1341 | { 0x75A8, LIT64( 0xE59591E4788957A5 ) },\r | |
1342 | { 0x3FFF, LIT64( 0xFFF0000000000040 ) },\r | |
1343 | { 0x0CD8, LIT64( 0xFC000000000007FE ) },\r | |
1344 | { 0x43BA, LIT64( 0x99A4000000000000 ) },\r | |
1345 | { 0x3FFF, LIT64( 0x8000000000000000 ) },\r | |
1346 | { 0x4081, LIT64( 0x94FBF1BCEB5545F0 ) },\r | |
1347 | { 0x403E, LIT64( 0xFFF0000000002000 ) },\r | |
1348 | { 0x3FFE, LIT64( 0xC860E3C75D224F28 ) },\r | |
1349 | { 0x407E, LIT64( 0xFC00000FFFFFFFFE ) },\r | |
1350 | { 0x737A, LIT64( 0x800000007FFDFFFE ) },\r | |
1351 | { 0x4044, LIT64( 0xFFFFFF80000FFFFF ) },\r | |
1352 | { 0x3BFE, LIT64( 0x8000040000001FFE ) },\r | |
1353 | { 0x4002, LIT64( 0xFF80000000000020 ) },\r | |
1354 | { 0x5E8D, LIT64( 0xFFFFFFFFFFE00004 ) },\r | |
1355 | { 0x4004, LIT64( 0x8000000000003FFB ) },\r | |
1356 | { 0x407F, LIT64( 0x800000000003FFFE ) },\r | |
1357 | { 0x4000, LIT64( 0xA459EE6A5C16CA55 ) },\r | |
1358 | { 0x0003, LIT64( 0xC42CBF7399AEEB94 ) },\r | |
1359 | { 0x3F7F, LIT64( 0xF800000000000006 ) },\r | |
1360 | { 0x407F, LIT64( 0xBF56BE8871F28FEA ) },\r | |
1361 | { 0x407E, LIT64( 0xFFFF77FFFFFFFFFE ) },\r | |
1362 | { 0x2DC9, LIT64( 0x8000000FFFFFFFDE ) },\r | |
1363 | { 0x4001, LIT64( 0xEFF7FFFFFFFFFFFF ) },\r | |
1364 | { 0x4001, LIT64( 0xBE84F30125C497A6 ) },\r | |
1365 | { 0x406B, LIT64( 0xEFFFFFFFFFFFFFFF ) },\r | |
1366 | { 0x4080, LIT64( 0xFFFFFFFFBFFFFFFF ) },\r | |
1367 | { 0x07E9, LIT64( 0x81FFFFFFFFFFFBFF ) },\r | |
1368 | { 0x263F, LIT64( 0x801FFFFFFEFFFFFE ) },\r | |
1369 | { 0x403C, LIT64( 0x801FFFFFFFF7FFFF ) },\r | |
1370 | { 0x4018, LIT64( 0x8000000000080003 ) }\r | |
1371 | };\r | |
1372 | \r | |
1373 | static void time_az_floatx80_pos( floatx80 function( floatx80 ) )\r | |
1374 | {\r | |
1375 | clock_t startClock, endClock;\r | |
1376 | int32 count, i;\r | |
1377 | int8 inputNum;\r | |
1378 | floatx80 a;\r | |
1379 | \r | |
1380 | count = 0;\r | |
1381 | inputNum = 0;\r | |
1382 | startClock = clock();\r | |
1383 | do {\r | |
1384 | for ( i = minIterations; i; --i ) {\r | |
1385 | a.low = inputs_floatx80_pos[ inputNum ].low;\r | |
1386 | a.high = inputs_floatx80_pos[ inputNum ].high;\r | |
1387 | function( a );\r | |
1388 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1389 | }\r | |
1390 | count += minIterations;\r | |
1391 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1392 | inputNum = 0;\r | |
1393 | startClock = clock();\r | |
1394 | for ( i = count; i; --i ) {\r | |
1395 | a.low = inputs_floatx80_pos[ inputNum ].low;\r | |
1396 | a.high = inputs_floatx80_pos[ inputNum ].high;\r | |
1397 | function( a );\r | |
1398 | inputNum = ( inputNum + 1 ) & ( numInputs_floatx80 - 1 );\r | |
1399 | }\r | |
1400 | endClock = clock();\r | |
1401 | reportTime( count, endClock - startClock );\r | |
1402 | \r | |
1403 | }\r | |
1404 | \r | |
1405 | #endif\r | |
1406 | \r | |
1407 | #ifdef FLOAT128\r | |
1408 | \r | |
1409 | enum {\r | |
1410 | numInputs_float128 = 32\r | |
1411 | };\r | |
1412 | \r | |
1413 | static const struct {\r | |
1414 | bits64 high, low;\r | |
1415 | } inputs_float128[ numInputs_float128 ] = {\r | |
1416 | { LIT64( 0x3FDA200000100000 ), LIT64( 0x0000000000000000 ) },\r | |
1417 | { LIT64( 0x3FFF000000000000 ), LIT64( 0x0000000000000000 ) },\r | |
1418 | { LIT64( 0x85F14776190C8306 ), LIT64( 0xD8715F4E3D54BB92 ) },\r | |
1419 | { LIT64( 0xF2B00000007FFFFF ), LIT64( 0xFFFFFFFFFFF7FFFF ) },\r | |
1420 | { LIT64( 0x8000000000000000 ), LIT64( 0x0000000000000000 ) },\r | |
1421 | { LIT64( 0xBFFFFFFFFFE00000 ), LIT64( 0x0000008000000000 ) },\r | |
1422 | { LIT64( 0x407F1719CE722F3E ), LIT64( 0xDA6B3FE5FF29425B ) },\r | |
1423 | { LIT64( 0x43FFFF8000000000 ), LIT64( 0x0000000000400000 ) },\r | |
1424 | { LIT64( 0x401E000000000100 ), LIT64( 0x0000000000002000 ) },\r | |
1425 | { LIT64( 0x3FFED71DACDA8E47 ), LIT64( 0x4860E3C75D224F28 ) },\r | |
1426 | { LIT64( 0xBF7ECFC1E90647D1 ), LIT64( 0x7A124FE55623EE44 ) },\r | |
1427 | { LIT64( 0x0DF7007FFFFFFFFF ), LIT64( 0xFFFFFFFFEFFFFFFF ) },\r | |
1428 | { LIT64( 0x3FE5FFEFFFFFFFFF ), LIT64( 0xFFFFFFFFFFFFEFFF ) },\r | |
1429 | { LIT64( 0x403FFFFFFFFFFFFF ), LIT64( 0xFFFFFFFFFFFFFBFE ) },\r | |
1430 | { LIT64( 0xBFFB2FBF7399AFEB ), LIT64( 0xA459EE6A5C16CA55 ) },\r | |
1431 | { LIT64( 0xBDB8FFFFFFFFFFFC ), LIT64( 0x0000000000000400 ) },\r | |
1432 | { LIT64( 0x3FC8FFDFFFFFFFFF ), LIT64( 0xFFFFFFFFF0000000 ) },\r | |
1433 | { LIT64( 0x3FFBFFFFFFDFFFFF ), LIT64( 0xFFF8000000000000 ) },\r | |
1434 | { LIT64( 0x407043C11737BE84 ), LIT64( 0xDDD58212ADC937F4 ) },\r | |
1435 | { LIT64( 0x8001000000000000 ), LIT64( 0x0000001000000001 ) },\r | |
1436 | { LIT64( 0xC036FFFFFFFFFFFF ), LIT64( 0xFE40000000000000 ) },\r | |
1437 | { LIT64( 0x4002FFFFFE000002 ), LIT64( 0x0000000000000000 ) },\r | |
1438 | { LIT64( 0x4000C3FEDE897773 ), LIT64( 0x326AC4FD8EFBE6DC ) },\r | |
1439 | { LIT64( 0xBFFF0000000FFFFF ), LIT64( 0xFFFFFE0000000000 ) },\r | |
1440 | { LIT64( 0x62C3E502146E426D ), LIT64( 0x43F3CAA0DC7DF1A0 ) },\r | |
1441 | { LIT64( 0xB5CBD32E52BB570E ), LIT64( 0xBCC477CB11C6236C ) },\r | |
1442 | { LIT64( 0xE228FFFFFFC00000 ), LIT64( 0x0000000000000000 ) },\r | |
1443 | { LIT64( 0x3F80000000000000 ), LIT64( 0x0000000080000008 ) },\r | |
1444 | { LIT64( 0xC1AFFFDFFFFFFFFF ), LIT64( 0xFFFC000000000000 ) },\r | |
1445 | { LIT64( 0xC96F000000000000 ), LIT64( 0x00000001FFFBFFFF ) },\r | |
1446 | { LIT64( 0x3DE09BFE7923A338 ), LIT64( 0xBCC8FBBD7CEC1F4F ) },\r | |
1447 | { LIT64( 0x401CFFFFFFFFFFFF ), LIT64( 0xFFFFFFFEFFFFFF80 ) }\r | |
1448 | };\r | |
1449 | \r | |
1450 | static void time_a_float128_z_int32( int32 function( float128 ) )\r | |
1451 | {\r | |
1452 | clock_t startClock, endClock;\r | |
1453 | int32 count, i;\r | |
1454 | int8 inputNum;\r | |
1455 | float128 a;\r | |
1456 | \r | |
1457 | count = 0;\r | |
1458 | inputNum = 0;\r | |
1459 | startClock = clock();\r | |
1460 | do {\r | |
1461 | for ( i = minIterations; i; --i ) {\r | |
1462 | a.low = inputs_float128[ inputNum ].low;\r | |
1463 | a.high = inputs_float128[ inputNum ].high;\r | |
1464 | function( a );\r | |
1465 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1466 | }\r | |
1467 | count += minIterations;\r | |
1468 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1469 | inputNum = 0;\r | |
1470 | startClock = clock();\r | |
1471 | for ( i = count; i; --i ) {\r | |
1472 | a.low = inputs_float128[ inputNum ].low;\r | |
1473 | a.high = inputs_float128[ inputNum ].high;\r | |
1474 | function( a );\r | |
1475 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1476 | }\r | |
1477 | endClock = clock();\r | |
1478 | reportTime( count, endClock - startClock );\r | |
1479 | \r | |
1480 | }\r | |
1481 | \r | |
1482 | static void time_a_float128_z_int64( int64 function( float128 ) )\r | |
1483 | {\r | |
1484 | clock_t startClock, endClock;\r | |
1485 | int32 count, i;\r | |
1486 | int8 inputNum;\r | |
1487 | float128 a;\r | |
1488 | \r | |
1489 | count = 0;\r | |
1490 | inputNum = 0;\r | |
1491 | startClock = clock();\r | |
1492 | do {\r | |
1493 | for ( i = minIterations; i; --i ) {\r | |
1494 | a.low = inputs_float128[ inputNum ].low;\r | |
1495 | a.high = inputs_float128[ inputNum ].high;\r | |
1496 | function( a );\r | |
1497 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1498 | }\r | |
1499 | count += minIterations;\r | |
1500 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1501 | inputNum = 0;\r | |
1502 | startClock = clock();\r | |
1503 | for ( i = count; i; --i ) {\r | |
1504 | a.low = inputs_float128[ inputNum ].low;\r | |
1505 | a.high = inputs_float128[ inputNum ].high;\r | |
1506 | function( a );\r | |
1507 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1508 | }\r | |
1509 | endClock = clock();\r | |
1510 | reportTime( count, endClock - startClock );\r | |
1511 | \r | |
1512 | }\r | |
1513 | \r | |
1514 | static void time_a_float128_z_float32( float32 function( float128 ) )\r | |
1515 | {\r | |
1516 | clock_t startClock, endClock;\r | |
1517 | int32 count, i;\r | |
1518 | int8 inputNum;\r | |
1519 | float128 a;\r | |
1520 | \r | |
1521 | count = 0;\r | |
1522 | inputNum = 0;\r | |
1523 | startClock = clock();\r | |
1524 | do {\r | |
1525 | for ( i = minIterations; i; --i ) {\r | |
1526 | a.low = inputs_float128[ inputNum ].low;\r | |
1527 | a.high = inputs_float128[ inputNum ].high;\r | |
1528 | function( a );\r | |
1529 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1530 | }\r | |
1531 | count += minIterations;\r | |
1532 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1533 | inputNum = 0;\r | |
1534 | startClock = clock();\r | |
1535 | for ( i = count; i; --i ) {\r | |
1536 | a.low = inputs_float128[ inputNum ].low;\r | |
1537 | a.high = inputs_float128[ inputNum ].high;\r | |
1538 | function( a );\r | |
1539 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1540 | }\r | |
1541 | endClock = clock();\r | |
1542 | reportTime( count, endClock - startClock );\r | |
1543 | \r | |
1544 | }\r | |
1545 | \r | |
1546 | static void time_a_float128_z_float64( float64 function( float128 ) )\r | |
1547 | {\r | |
1548 | clock_t startClock, endClock;\r | |
1549 | int32 count, i;\r | |
1550 | int8 inputNum;\r | |
1551 | float128 a;\r | |
1552 | \r | |
1553 | count = 0;\r | |
1554 | inputNum = 0;\r | |
1555 | startClock = clock();\r | |
1556 | do {\r | |
1557 | for ( i = minIterations; i; --i ) {\r | |
1558 | a.low = inputs_float128[ inputNum ].low;\r | |
1559 | a.high = inputs_float128[ inputNum ].high;\r | |
1560 | function( a );\r | |
1561 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1562 | }\r | |
1563 | count += minIterations;\r | |
1564 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1565 | inputNum = 0;\r | |
1566 | startClock = clock();\r | |
1567 | for ( i = count; i; --i ) {\r | |
1568 | a.low = inputs_float128[ inputNum ].low;\r | |
1569 | a.high = inputs_float128[ inputNum ].high;\r | |
1570 | function( a );\r | |
1571 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1572 | }\r | |
1573 | endClock = clock();\r | |
1574 | reportTime( count, endClock - startClock );\r | |
1575 | \r | |
1576 | }\r | |
1577 | \r | |
1578 | #ifdef FLOATX80\r | |
1579 | \r | |
1580 | static void time_a_float128_z_floatx80( floatx80 function( float128 ) )\r | |
1581 | {\r | |
1582 | clock_t startClock, endClock;\r | |
1583 | int32 count, i;\r | |
1584 | int8 inputNum;\r | |
1585 | float128 a;\r | |
1586 | \r | |
1587 | count = 0;\r | |
1588 | inputNum = 0;\r | |
1589 | startClock = clock();\r | |
1590 | do {\r | |
1591 | for ( i = minIterations; i; --i ) {\r | |
1592 | a.low = inputs_float128[ inputNum ].low;\r | |
1593 | a.high = inputs_float128[ inputNum ].high;\r | |
1594 | function( a );\r | |
1595 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1596 | }\r | |
1597 | count += minIterations;\r | |
1598 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1599 | inputNum = 0;\r | |
1600 | startClock = clock();\r | |
1601 | for ( i = count; i; --i ) {\r | |
1602 | a.low = inputs_float128[ inputNum ].low;\r | |
1603 | a.high = inputs_float128[ inputNum ].high;\r | |
1604 | function( a );\r | |
1605 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1606 | }\r | |
1607 | endClock = clock();\r | |
1608 | reportTime( count, endClock - startClock );\r | |
1609 | \r | |
1610 | }\r | |
1611 | \r | |
1612 | #endif\r | |
1613 | \r | |
1614 | static void time_az_float128( float128 function( float128 ) )\r | |
1615 | {\r | |
1616 | clock_t startClock, endClock;\r | |
1617 | int32 count, i;\r | |
1618 | int8 inputNum;\r | |
1619 | float128 a;\r | |
1620 | \r | |
1621 | count = 0;\r | |
1622 | inputNum = 0;\r | |
1623 | startClock = clock();\r | |
1624 | do {\r | |
1625 | for ( i = minIterations; i; --i ) {\r | |
1626 | a.low = inputs_float128[ inputNum ].low;\r | |
1627 | a.high = inputs_float128[ inputNum ].high;\r | |
1628 | function( a );\r | |
1629 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1630 | }\r | |
1631 | count += minIterations;\r | |
1632 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1633 | inputNum = 0;\r | |
1634 | startClock = clock();\r | |
1635 | for ( i = count; i; --i ) {\r | |
1636 | a.low = inputs_float128[ inputNum ].low;\r | |
1637 | a.high = inputs_float128[ inputNum ].high;\r | |
1638 | function( a );\r | |
1639 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1640 | }\r | |
1641 | endClock = clock();\r | |
1642 | reportTime( count, endClock - startClock );\r | |
1643 | \r | |
1644 | }\r | |
1645 | \r | |
1646 | static void time_ab_float128_z_flag( flag function( float128, float128 ) )\r | |
1647 | {\r | |
1648 | clock_t startClock, endClock;\r | |
1649 | int32 count, i;\r | |
1650 | int8 inputNumA, inputNumB;\r | |
1651 | float128 a, b;\r | |
1652 | \r | |
1653 | count = 0;\r | |
1654 | inputNumA = 0;\r | |
1655 | inputNumB = 0;\r | |
1656 | startClock = clock();\r | |
1657 | do {\r | |
1658 | for ( i = minIterations; i; --i ) {\r | |
1659 | a.low = inputs_float128[ inputNumA ].low;\r | |
1660 | a.high = inputs_float128[ inputNumA ].high;\r | |
1661 | b.low = inputs_float128[ inputNumB ].low;\r | |
1662 | b.high = inputs_float128[ inputNumB ].high;\r | |
1663 | function( a, b );\r | |
1664 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float128 - 1 );\r | |
1665 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1666 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float128 - 1 );\r | |
1667 | }\r | |
1668 | count += minIterations;\r | |
1669 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1670 | inputNumA = 0;\r | |
1671 | inputNumB = 0;\r | |
1672 | startClock = clock();\r | |
1673 | for ( i = count; i; --i ) {\r | |
1674 | a.low = inputs_float128[ inputNumA ].low;\r | |
1675 | a.high = inputs_float128[ inputNumA ].high;\r | |
1676 | b.low = inputs_float128[ inputNumB ].low;\r | |
1677 | b.high = inputs_float128[ inputNumB ].high;\r | |
1678 | function( a, b );\r | |
1679 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float128 - 1 );\r | |
1680 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1681 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float128 - 1 );\r | |
1682 | }\r | |
1683 | endClock = clock();\r | |
1684 | reportTime( count, endClock - startClock );\r | |
1685 | \r | |
1686 | }\r | |
1687 | \r | |
1688 | static void time_abz_float128( float128 function( float128, float128 ) )\r | |
1689 | {\r | |
1690 | clock_t startClock, endClock;\r | |
1691 | int32 count, i;\r | |
1692 | int8 inputNumA, inputNumB;\r | |
1693 | float128 a, b;\r | |
1694 | \r | |
1695 | count = 0;\r | |
1696 | inputNumA = 0;\r | |
1697 | inputNumB = 0;\r | |
1698 | startClock = clock();\r | |
1699 | do {\r | |
1700 | for ( i = minIterations; i; --i ) {\r | |
1701 | a.low = inputs_float128[ inputNumA ].low;\r | |
1702 | a.high = inputs_float128[ inputNumA ].high;\r | |
1703 | b.low = inputs_float128[ inputNumB ].low;\r | |
1704 | b.high = inputs_float128[ inputNumB ].high;\r | |
1705 | function( a, b );\r | |
1706 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float128 - 1 );\r | |
1707 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1708 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float128 - 1 );\r | |
1709 | }\r | |
1710 | count += minIterations;\r | |
1711 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1712 | inputNumA = 0;\r | |
1713 | inputNumB = 0;\r | |
1714 | startClock = clock();\r | |
1715 | for ( i = count; i; --i ) {\r | |
1716 | a.low = inputs_float128[ inputNumA ].low;\r | |
1717 | a.high = inputs_float128[ inputNumA ].high;\r | |
1718 | b.low = inputs_float128[ inputNumB ].low;\r | |
1719 | b.high = inputs_float128[ inputNumB ].high;\r | |
1720 | function( a, b );\r | |
1721 | inputNumA = ( inputNumA + 1 ) & ( numInputs_float128 - 1 );\r | |
1722 | if ( inputNumA == 0 ) ++inputNumB;\r | |
1723 | inputNumB = ( inputNumB + 1 ) & ( numInputs_float128 - 1 );\r | |
1724 | }\r | |
1725 | endClock = clock();\r | |
1726 | reportTime( count, endClock - startClock );\r | |
1727 | \r | |
1728 | }\r | |
1729 | \r | |
1730 | static const struct {\r | |
1731 | bits64 high, low;\r | |
1732 | } inputs_float128_pos[ numInputs_float128 ] = {\r | |
1733 | { LIT64( 0x3FDA200000100000 ), LIT64( 0x0000000000000000 ) },\r | |
1734 | { LIT64( 0x3FFF000000000000 ), LIT64( 0x0000000000000000 ) },\r | |
1735 | { LIT64( 0x05F14776190C8306 ), LIT64( 0xD8715F4E3D54BB92 ) },\r | |
1736 | { LIT64( 0x72B00000007FFFFF ), LIT64( 0xFFFFFFFFFFF7FFFF ) },\r | |
1737 | { LIT64( 0x0000000000000000 ), LIT64( 0x0000000000000000 ) },\r | |
1738 | { LIT64( 0x3FFFFFFFFFE00000 ), LIT64( 0x0000008000000000 ) },\r | |
1739 | { LIT64( 0x407F1719CE722F3E ), LIT64( 0xDA6B3FE5FF29425B ) },\r | |
1740 | { LIT64( 0x43FFFF8000000000 ), LIT64( 0x0000000000400000 ) },\r | |
1741 | { LIT64( 0x401E000000000100 ), LIT64( 0x0000000000002000 ) },\r | |
1742 | { LIT64( 0x3FFED71DACDA8E47 ), LIT64( 0x4860E3C75D224F28 ) },\r | |
1743 | { LIT64( 0x3F7ECFC1E90647D1 ), LIT64( 0x7A124FE55623EE44 ) },\r | |
1744 | { LIT64( 0x0DF7007FFFFFFFFF ), LIT64( 0xFFFFFFFFEFFFFFFF ) },\r | |
1745 | { LIT64( 0x3FE5FFEFFFFFFFFF ), LIT64( 0xFFFFFFFFFFFFEFFF ) },\r | |
1746 | { LIT64( 0x403FFFFFFFFFFFFF ), LIT64( 0xFFFFFFFFFFFFFBFE ) },\r | |
1747 | { LIT64( 0x3FFB2FBF7399AFEB ), LIT64( 0xA459EE6A5C16CA55 ) },\r | |
1748 | { LIT64( 0x3DB8FFFFFFFFFFFC ), LIT64( 0x0000000000000400 ) },\r | |
1749 | { LIT64( 0x3FC8FFDFFFFFFFFF ), LIT64( 0xFFFFFFFFF0000000 ) },\r | |
1750 | { LIT64( 0x3FFBFFFFFFDFFFFF ), LIT64( 0xFFF8000000000000 ) },\r | |
1751 | { LIT64( 0x407043C11737BE84 ), LIT64( 0xDDD58212ADC937F4 ) },\r | |
1752 | { LIT64( 0x0001000000000000 ), LIT64( 0x0000001000000001 ) },\r | |
1753 | { LIT64( 0x4036FFFFFFFFFFFF ), LIT64( 0xFE40000000000000 ) },\r | |
1754 | { LIT64( 0x4002FFFFFE000002 ), LIT64( 0x0000000000000000 ) },\r | |
1755 | { LIT64( 0x4000C3FEDE897773 ), LIT64( 0x326AC4FD8EFBE6DC ) },\r | |
1756 | { LIT64( 0x3FFF0000000FFFFF ), LIT64( 0xFFFFFE0000000000 ) },\r | |
1757 | { LIT64( 0x62C3E502146E426D ), LIT64( 0x43F3CAA0DC7DF1A0 ) },\r | |
1758 | { LIT64( 0x35CBD32E52BB570E ), LIT64( 0xBCC477CB11C6236C ) },\r | |
1759 | { LIT64( 0x6228FFFFFFC00000 ), LIT64( 0x0000000000000000 ) },\r | |
1760 | { LIT64( 0x3F80000000000000 ), LIT64( 0x0000000080000008 ) },\r | |
1761 | { LIT64( 0x41AFFFDFFFFFFFFF ), LIT64( 0xFFFC000000000000 ) },\r | |
1762 | { LIT64( 0x496F000000000000 ), LIT64( 0x00000001FFFBFFFF ) },\r | |
1763 | { LIT64( 0x3DE09BFE7923A338 ), LIT64( 0xBCC8FBBD7CEC1F4F ) },\r | |
1764 | { LIT64( 0x401CFFFFFFFFFFFF ), LIT64( 0xFFFFFFFEFFFFFF80 ) }\r | |
1765 | };\r | |
1766 | \r | |
1767 | static void time_az_float128_pos( float128 function( float128 ) )\r | |
1768 | {\r | |
1769 | clock_t startClock, endClock;\r | |
1770 | int32 count, i;\r | |
1771 | int8 inputNum;\r | |
1772 | float128 a;\r | |
1773 | \r | |
1774 | count = 0;\r | |
1775 | inputNum = 0;\r | |
1776 | startClock = clock();\r | |
1777 | do {\r | |
1778 | for ( i = minIterations; i; --i ) {\r | |
1779 | a.low = inputs_float128_pos[ inputNum ].low;\r | |
1780 | a.high = inputs_float128_pos[ inputNum ].high;\r | |
1781 | function( a );\r | |
1782 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1783 | }\r | |
1784 | count += minIterations;\r | |
1785 | } while ( clock() - startClock < CLOCKS_PER_SEC );\r | |
1786 | inputNum = 0;\r | |
1787 | startClock = clock();\r | |
1788 | for ( i = count; i; --i ) {\r | |
1789 | a.low = inputs_float128_pos[ inputNum ].low;\r | |
1790 | a.high = inputs_float128_pos[ inputNum ].high;\r | |
1791 | function( a );\r | |
1792 | inputNum = ( inputNum + 1 ) & ( numInputs_float128 - 1 );\r | |
1793 | }\r | |
1794 | endClock = clock();\r | |
1795 | reportTime( count, endClock - startClock );\r | |
1796 | \r | |
1797 | }\r | |
1798 | \r | |
1799 | #endif\r | |
1800 | \r | |
1801 | enum {\r | |
1802 | INT32_TO_FLOAT32 = 1,\r | |
1803 | INT32_TO_FLOAT64,\r | |
1804 | #ifdef FLOATX80\r | |
1805 | INT32_TO_FLOATX80,\r | |
1806 | #endif\r | |
1807 | #ifdef FLOAT128\r | |
1808 | INT32_TO_FLOAT128,\r | |
1809 | #endif\r | |
1810 | INT64_TO_FLOAT32,\r | |
1811 | INT64_TO_FLOAT64,\r | |
1812 | #ifdef FLOATX80\r | |
1813 | INT64_TO_FLOATX80,\r | |
1814 | #endif\r | |
1815 | #ifdef FLOAT128\r | |
1816 | INT64_TO_FLOAT128,\r | |
1817 | #endif\r | |
1818 | FLOAT32_TO_INT32,\r | |
1819 | FLOAT32_TO_INT32_ROUND_TO_ZERO,\r | |
1820 | FLOAT32_TO_INT64,\r | |
1821 | FLOAT32_TO_INT64_ROUND_TO_ZERO,\r | |
1822 | FLOAT32_TO_FLOAT64,\r | |
1823 | #ifdef FLOATX80\r | |
1824 | FLOAT32_TO_FLOATX80,\r | |
1825 | #endif\r | |
1826 | #ifdef FLOAT128\r | |
1827 | FLOAT32_TO_FLOAT128,\r | |
1828 | #endif\r | |
1829 | FLOAT32_ROUND_TO_INT,\r | |
1830 | FLOAT32_ADD,\r | |
1831 | FLOAT32_SUB,\r | |
1832 | FLOAT32_MUL,\r | |
1833 | FLOAT32_DIV,\r | |
1834 | FLOAT32_REM,\r | |
1835 | FLOAT32_SQRT,\r | |
1836 | FLOAT32_EQ,\r | |
1837 | FLOAT32_LE,\r | |
1838 | FLOAT32_LT,\r | |
1839 | FLOAT32_EQ_SIGNALING,\r | |
1840 | FLOAT32_LE_QUIET,\r | |
1841 | FLOAT32_LT_QUIET,\r | |
1842 | FLOAT64_TO_INT32,\r | |
1843 | FLOAT64_TO_INT32_ROUND_TO_ZERO,\r | |
1844 | FLOAT64_TO_INT64,\r | |
1845 | FLOAT64_TO_INT64_ROUND_TO_ZERO,\r | |
1846 | FLOAT64_TO_FLOAT32,\r | |
1847 | #ifdef FLOATX80\r | |
1848 | FLOAT64_TO_FLOATX80,\r | |
1849 | #endif\r | |
1850 | #ifdef FLOAT128\r | |
1851 | FLOAT64_TO_FLOAT128,\r | |
1852 | #endif\r | |
1853 | FLOAT64_ROUND_TO_INT,\r | |
1854 | FLOAT64_ADD,\r | |
1855 | FLOAT64_SUB,\r | |
1856 | FLOAT64_MUL,\r | |
1857 | FLOAT64_DIV,\r | |
1858 | FLOAT64_REM,\r | |
1859 | FLOAT64_SQRT,\r | |
1860 | FLOAT64_EQ,\r | |
1861 | FLOAT64_LE,\r | |
1862 | FLOAT64_LT,\r | |
1863 | FLOAT64_EQ_SIGNALING,\r | |
1864 | FLOAT64_LE_QUIET,\r | |
1865 | FLOAT64_LT_QUIET,\r | |
1866 | #ifdef FLOATX80\r | |
1867 | FLOATX80_TO_INT32,\r | |
1868 | FLOATX80_TO_INT32_ROUND_TO_ZERO,\r | |
1869 | FLOATX80_TO_INT64,\r | |
1870 | FLOATX80_TO_INT64_ROUND_TO_ZERO,\r | |
1871 | FLOATX80_TO_FLOAT32,\r | |
1872 | FLOATX80_TO_FLOAT64,\r | |
1873 | #ifdef FLOAT128\r | |
1874 | FLOATX80_TO_FLOAT128,\r | |
1875 | #endif\r | |
1876 | FLOATX80_ROUND_TO_INT,\r | |
1877 | FLOATX80_ADD,\r | |
1878 | FLOATX80_SUB,\r | |
1879 | FLOATX80_MUL,\r | |
1880 | FLOATX80_DIV,\r | |
1881 | FLOATX80_REM,\r | |
1882 | FLOATX80_SQRT,\r | |
1883 | FLOATX80_EQ,\r | |
1884 | FLOATX80_LE,\r | |
1885 | FLOATX80_LT,\r | |
1886 | FLOATX80_EQ_SIGNALING,\r | |
1887 | FLOATX80_LE_QUIET,\r | |
1888 | FLOATX80_LT_QUIET,\r | |
1889 | #endif\r | |
1890 | #ifdef FLOAT128\r | |
1891 | FLOAT128_TO_INT32,\r | |
1892 | FLOAT128_TO_INT32_ROUND_TO_ZERO,\r | |
1893 | FLOAT128_TO_INT64,\r | |
1894 | FLOAT128_TO_INT64_ROUND_TO_ZERO,\r | |
1895 | FLOAT128_TO_FLOAT32,\r | |
1896 | FLOAT128_TO_FLOAT64,\r | |
1897 | #ifdef FLOATX80\r | |
1898 | FLOAT128_TO_FLOATX80,\r | |
1899 | #endif\r | |
1900 | FLOAT128_ROUND_TO_INT,\r | |
1901 | FLOAT128_ADD,\r | |
1902 | FLOAT128_SUB,\r | |
1903 | FLOAT128_MUL,\r | |
1904 | FLOAT128_DIV,\r | |
1905 | FLOAT128_REM,\r | |
1906 | FLOAT128_SQRT,\r | |
1907 | FLOAT128_EQ,\r | |
1908 | FLOAT128_LE,\r | |
1909 | FLOAT128_LT,\r | |
1910 | FLOAT128_EQ_SIGNALING,\r | |
1911 | FLOAT128_LE_QUIET,\r | |
1912 | FLOAT128_LT_QUIET,\r | |
1913 | #endif\r | |
1914 | NUM_FUNCTIONS\r | |
1915 | };\r | |
1916 | \r | |
1917 | static struct {\r | |
1918 | char *name;\r | |
1919 | int8 numInputs;\r | |
1920 | flag roundingPrecision, roundingMode;\r | |
1921 | flag tininessMode, tininessModeAtReducedPrecision;\r | |
1922 | } functions[ NUM_FUNCTIONS ] = {\r | |
1923 | { 0, 0, 0, 0, 0, 0 },\r | |
1924 | { "int32_to_float32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1925 | { "int32_to_float64", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1926 | #ifdef FLOATX80\r | |
1927 | { "int32_to_floatx80", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1928 | #endif\r | |
1929 | #ifdef FLOAT128\r | |
1930 | { "int32_to_float128", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1931 | #endif\r | |
1932 | { "int64_to_float32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1933 | { "int64_to_float64", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1934 | #ifdef FLOATX80\r | |
1935 | { "int64_to_floatx80", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1936 | #endif\r | |
1937 | #ifdef FLOAT128\r | |
1938 | { "int64_to_float128", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1939 | #endif\r | |
1940 | { "float32_to_int32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1941 | { "float32_to_int32_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1942 | { "float32_to_int64", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1943 | { "float32_to_int64_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1944 | { "float32_to_float64", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1945 | #ifdef FLOATX80\r | |
1946 | { "float32_to_floatx80", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1947 | #endif\r | |
1948 | #ifdef FLOAT128\r | |
1949 | { "float32_to_float128", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1950 | #endif\r | |
1951 | { "float32_round_to_int", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1952 | { "float32_add", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1953 | { "float32_sub", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1954 | { "float32_mul", 2, FALSE, TRUE, TRUE, FALSE },\r | |
1955 | { "float32_div", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1956 | { "float32_rem", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1957 | { "float32_sqrt", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1958 | { "float32_eq", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1959 | { "float32_le", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1960 | { "float32_lt", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1961 | { "float32_eq_signaling", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1962 | { "float32_le_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1963 | { "float32_lt_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1964 | { "float64_to_int32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1965 | { "float64_to_int32_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1966 | { "float64_to_int64", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1967 | { "float64_to_int64_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1968 | { "float64_to_float32", 1, FALSE, TRUE, TRUE, FALSE },\r | |
1969 | #ifdef FLOATX80\r | |
1970 | { "float64_to_floatx80", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1971 | #endif\r | |
1972 | #ifdef FLOAT128\r | |
1973 | { "float64_to_float128", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1974 | #endif\r | |
1975 | { "float64_round_to_int", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1976 | { "float64_add", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1977 | { "float64_sub", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1978 | { "float64_mul", 2, FALSE, TRUE, TRUE, FALSE },\r | |
1979 | { "float64_div", 2, FALSE, TRUE, FALSE, FALSE },\r | |
1980 | { "float64_rem", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1981 | { "float64_sqrt", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1982 | { "float64_eq", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1983 | { "float64_le", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1984 | { "float64_lt", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1985 | { "float64_eq_signaling", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1986 | { "float64_le_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1987 | { "float64_lt_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
1988 | #ifdef FLOATX80\r | |
1989 | { "floatx80_to_int32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1990 | { "floatx80_to_int32_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1991 | { "floatx80_to_int64", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1992 | { "floatx80_to_int64_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1993 | { "floatx80_to_float32", 1, FALSE, TRUE, TRUE, FALSE },\r | |
1994 | { "floatx80_to_float64", 1, FALSE, TRUE, TRUE, FALSE },\r | |
1995 | #ifdef FLOAT128\r | |
1996 | { "floatx80_to_float128", 1, FALSE, FALSE, FALSE, FALSE },\r | |
1997 | #endif\r | |
1998 | { "floatx80_round_to_int", 1, FALSE, TRUE, FALSE, FALSE },\r | |
1999 | { "floatx80_add", 2, TRUE, TRUE, FALSE, TRUE },\r | |
2000 | { "floatx80_sub", 2, TRUE, TRUE, FALSE, TRUE },\r | |
2001 | { "floatx80_mul", 2, TRUE, TRUE, TRUE, TRUE },\r | |
2002 | { "floatx80_div", 2, TRUE, TRUE, FALSE, TRUE },\r | |
2003 | { "floatx80_rem", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2004 | { "floatx80_sqrt", 1, TRUE, TRUE, FALSE, FALSE },\r | |
2005 | { "floatx80_eq", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2006 | { "floatx80_le", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2007 | { "floatx80_lt", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2008 | { "floatx80_eq_signaling", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2009 | { "floatx80_le_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2010 | { "floatx80_lt_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2011 | #endif\r | |
2012 | #ifdef FLOAT128\r | |
2013 | { "float128_to_int32", 1, FALSE, TRUE, FALSE, FALSE },\r | |
2014 | { "float128_to_int32_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
2015 | { "float128_to_int64", 1, FALSE, TRUE, FALSE, FALSE },\r | |
2016 | { "float128_to_int64_round_to_zero", 1, FALSE, FALSE, FALSE, FALSE },\r | |
2017 | { "float128_to_float32", 1, FALSE, TRUE, TRUE, FALSE },\r | |
2018 | { "float128_to_float64", 1, FALSE, TRUE, TRUE, FALSE },\r | |
2019 | #ifdef FLOATX80\r | |
2020 | { "float128_to_floatx80", 1, FALSE, TRUE, TRUE, FALSE },\r | |
2021 | #endif\r | |
2022 | { "float128_round_to_int", 1, FALSE, TRUE, FALSE, FALSE },\r | |
2023 | { "float128_add", 2, FALSE, TRUE, FALSE, FALSE },\r | |
2024 | { "float128_sub", 2, FALSE, TRUE, FALSE, FALSE },\r | |
2025 | { "float128_mul", 2, FALSE, TRUE, TRUE, FALSE },\r | |
2026 | { "float128_div", 2, FALSE, TRUE, FALSE, FALSE },\r | |
2027 | { "float128_rem", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2028 | { "float128_sqrt", 1, FALSE, TRUE, FALSE, FALSE },\r | |
2029 | { "float128_eq", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2030 | { "float128_le", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2031 | { "float128_lt", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2032 | { "float128_eq_signaling", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2033 | { "float128_le_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2034 | { "float128_lt_quiet", 2, FALSE, FALSE, FALSE, FALSE },\r | |
2035 | #endif\r | |
2036 | };\r | |
2037 | \r | |
2038 | enum {\r | |
2039 | ROUND_NEAREST_EVEN = 1,\r | |
2040 | ROUND_TO_ZERO,\r | |
2041 | ROUND_DOWN,\r | |
2042 | ROUND_UP,\r | |
2043 | NUM_ROUNDINGMODES\r | |
2044 | };\r | |
2045 | enum {\r | |
2046 | TININESS_BEFORE_ROUNDING = 1,\r | |
2047 | TININESS_AFTER_ROUNDING,\r | |
2048 | NUM_TININESSMODES\r | |
2049 | };\r | |
2050 | \r | |
2051 | static void\r | |
2052 | timeFunctionVariety(\r | |
2053 | uint8 functionCode,\r | |
2054 | int8 roundingPrecision,\r | |
2055 | int8 roundingMode,\r | |
2056 | int8 tininessMode\r | |
2057 | )\r | |
2058 | {\r | |
2059 | uint8 roundingCode;\r | |
2060 | int8 tininessCode;\r | |
2061 | \r | |
2062 | functionName = functions[ functionCode ].name;\r | |
2063 | if ( roundingPrecision == 32 ) {\r | |
2064 | roundingPrecisionName = "32";\r | |
2065 | }\r | |
2066 | else if ( roundingPrecision == 64 ) {\r | |
2067 | roundingPrecisionName = "64";\r | |
2068 | }\r | |
2069 | else if ( roundingPrecision == 80 ) {\r | |
2070 | roundingPrecisionName = "80";\r | |
2071 | }\r | |
2072 | else {\r | |
2073 | roundingPrecisionName = 0;\r | |
2074 | }\r | |
2075 | #ifdef FLOATX80\r | |
2076 | floatx80_rounding_precision = roundingPrecision;\r | |
2077 | #endif\r | |
2078 | switch ( roundingMode ) {\r | |
2079 | case 0:\r | |
2080 | roundingModeName = 0;\r | |
2081 | roundingCode = float_round_nearest_even;\r | |
2082 | break;\r | |
2083 | case ROUND_NEAREST_EVEN:\r | |
2084 | roundingModeName = "nearest_even";\r | |
2085 | roundingCode = float_round_nearest_even;\r | |
2086 | break;\r | |
2087 | case ROUND_TO_ZERO:\r | |
2088 | roundingModeName = "to_zero";\r | |
2089 | roundingCode = float_round_to_zero;\r | |
2090 | break;\r | |
2091 | case ROUND_DOWN:\r | |
2092 | roundingModeName = "down";\r | |
2093 | roundingCode = float_round_down;\r | |
2094 | break;\r | |
2095 | case ROUND_UP:\r | |
2096 | roundingModeName = "up";\r | |
2097 | roundingCode = float_round_up;\r | |
2098 | break;\r | |
2099 | }\r | |
2100 | float_rounding_mode = roundingCode;\r | |
2101 | switch ( tininessMode ) {\r | |
2102 | case 0:\r | |
2103 | tininessModeName = 0;\r | |
2104 | tininessCode = float_tininess_after_rounding;\r | |
2105 | break;\r | |
2106 | case TININESS_BEFORE_ROUNDING:\r | |
2107 | tininessModeName = "before";\r | |
2108 | tininessCode = float_tininess_before_rounding;\r | |
2109 | break;\r | |
2110 | case TININESS_AFTER_ROUNDING:\r | |
2111 | tininessModeName = "after";\r | |
2112 | tininessCode = float_tininess_after_rounding;\r | |
2113 | break;\r | |
2114 | }\r | |
2115 | float_detect_tininess = tininessCode;\r | |
2116 | switch ( functionCode ) {\r | |
2117 | case INT32_TO_FLOAT32:\r | |
2118 | time_a_int32_z_float32( int32_to_float32 );\r | |
2119 | break;\r | |
2120 | case INT32_TO_FLOAT64:\r | |
2121 | time_a_int32_z_float64( int32_to_float64 );\r | |
2122 | break;\r | |
2123 | #ifdef FLOATX80\r | |
2124 | case INT32_TO_FLOATX80:\r | |
2125 | time_a_int32_z_floatx80( int32_to_floatx80 );\r | |
2126 | break;\r | |
2127 | #endif\r | |
2128 | #ifdef FLOAT128\r | |
2129 | case INT32_TO_FLOAT128:\r | |
2130 | time_a_int32_z_float128( int32_to_float128 );\r | |
2131 | break;\r | |
2132 | #endif\r | |
2133 | case INT64_TO_FLOAT32:\r | |
2134 | time_a_int64_z_float32( int64_to_float32 );\r | |
2135 | break;\r | |
2136 | case INT64_TO_FLOAT64:\r | |
2137 | time_a_int64_z_float64( int64_to_float64 );\r | |
2138 | break;\r | |
2139 | #ifdef FLOATX80\r | |
2140 | case INT64_TO_FLOATX80:\r | |
2141 | time_a_int64_z_floatx80( int64_to_floatx80 );\r | |
2142 | break;\r | |
2143 | #endif\r | |
2144 | #ifdef FLOAT128\r | |
2145 | case INT64_TO_FLOAT128:\r | |
2146 | time_a_int64_z_float128( int64_to_float128 );\r | |
2147 | break;\r | |
2148 | #endif\r | |
2149 | case FLOAT32_TO_INT32:\r | |
2150 | time_a_float32_z_int32( float32_to_int32 );\r | |
2151 | break;\r | |
2152 | case FLOAT32_TO_INT32_ROUND_TO_ZERO:\r | |
2153 | time_a_float32_z_int32( float32_to_int32_round_to_zero );\r | |
2154 | break;\r | |
2155 | case FLOAT32_TO_INT64:\r | |
2156 | time_a_float32_z_int64( float32_to_int64 );\r | |
2157 | break;\r | |
2158 | case FLOAT32_TO_INT64_ROUND_TO_ZERO:\r | |
2159 | time_a_float32_z_int64( float32_to_int64_round_to_zero );\r | |
2160 | break;\r | |
2161 | case FLOAT32_TO_FLOAT64:\r | |
2162 | time_a_float32_z_float64( float32_to_float64 );\r | |
2163 | break;\r | |
2164 | #ifdef FLOATX80\r | |
2165 | case FLOAT32_TO_FLOATX80:\r | |
2166 | time_a_float32_z_floatx80( float32_to_floatx80 );\r | |
2167 | break;\r | |
2168 | #endif\r | |
2169 | #ifdef FLOAT128\r | |
2170 | case FLOAT32_TO_FLOAT128:\r | |
2171 | time_a_float32_z_float128( float32_to_float128 );\r | |
2172 | break;\r | |
2173 | #endif\r | |
2174 | case FLOAT32_ROUND_TO_INT:\r | |
2175 | time_az_float32( float32_round_to_int );\r | |
2176 | break;\r | |
2177 | case FLOAT32_ADD:\r | |
2178 | time_abz_float32( float32_add );\r | |
2179 | break;\r | |
2180 | case FLOAT32_SUB:\r | |
2181 | time_abz_float32( float32_sub );\r | |
2182 | break;\r | |
2183 | case FLOAT32_MUL:\r | |
2184 | time_abz_float32( float32_mul );\r | |
2185 | break;\r | |
2186 | case FLOAT32_DIV:\r | |
2187 | time_abz_float32( float32_div );\r | |
2188 | break;\r | |
2189 | case FLOAT32_REM:\r | |
2190 | time_abz_float32( float32_rem );\r | |
2191 | break;\r | |
2192 | case FLOAT32_SQRT:\r | |
2193 | time_az_float32_pos( float32_sqrt );\r | |
2194 | break;\r | |
2195 | case FLOAT32_EQ:\r | |
2196 | time_ab_float32_z_flag( float32_eq );\r | |
2197 | break;\r | |
2198 | case FLOAT32_LE:\r | |
2199 | time_ab_float32_z_flag( float32_le );\r | |
2200 | break;\r | |
2201 | case FLOAT32_LT:\r | |
2202 | time_ab_float32_z_flag( float32_lt );\r | |
2203 | break;\r | |
2204 | case FLOAT32_EQ_SIGNALING:\r | |
2205 | time_ab_float32_z_flag( float32_eq_signaling );\r | |
2206 | break;\r | |
2207 | case FLOAT32_LE_QUIET:\r | |
2208 | time_ab_float32_z_flag( float32_le_quiet );\r | |
2209 | break;\r | |
2210 | case FLOAT32_LT_QUIET:\r | |
2211 | time_ab_float32_z_flag( float32_lt_quiet );\r | |
2212 | break;\r | |
2213 | case FLOAT64_TO_INT32:\r | |
2214 | time_a_float64_z_int32( float64_to_int32 );\r | |
2215 | break;\r | |
2216 | case FLOAT64_TO_INT32_ROUND_TO_ZERO:\r | |
2217 | time_a_float64_z_int32( float64_to_int32_round_to_zero );\r | |
2218 | break;\r | |
2219 | case FLOAT64_TO_INT64:\r | |
2220 | time_a_float64_z_int64( float64_to_int64 );\r | |
2221 | break;\r | |
2222 | case FLOAT64_TO_INT64_ROUND_TO_ZERO:\r | |
2223 | time_a_float64_z_int64( float64_to_int64_round_to_zero );\r | |
2224 | break;\r | |
2225 | case FLOAT64_TO_FLOAT32:\r | |
2226 | time_a_float64_z_float32( float64_to_float32 );\r | |
2227 | break;\r | |
2228 | #ifdef FLOATX80\r | |
2229 | case FLOAT64_TO_FLOATX80:\r | |
2230 | time_a_float64_z_floatx80( float64_to_floatx80 );\r | |
2231 | break;\r | |
2232 | #endif\r | |
2233 | #ifdef FLOAT128\r | |
2234 | case FLOAT64_TO_FLOAT128:\r | |
2235 | time_a_float64_z_float128( float64_to_float128 );\r | |
2236 | break;\r | |
2237 | #endif\r | |
2238 | case FLOAT64_ROUND_TO_INT:\r | |
2239 | time_az_float64( float64_round_to_int );\r | |
2240 | break;\r | |
2241 | case FLOAT64_ADD:\r | |
2242 | time_abz_float64( float64_add );\r | |
2243 | break;\r | |
2244 | case FLOAT64_SUB:\r | |
2245 | time_abz_float64( float64_sub );\r | |
2246 | break;\r | |
2247 | case FLOAT64_MUL:\r | |
2248 | time_abz_float64( float64_mul );\r | |
2249 | break;\r | |
2250 | case FLOAT64_DIV:\r | |
2251 | time_abz_float64( float64_div );\r | |
2252 | break;\r | |
2253 | case FLOAT64_REM:\r | |
2254 | time_abz_float64( float64_rem );\r | |
2255 | break;\r | |
2256 | case FLOAT64_SQRT:\r | |
2257 | time_az_float64_pos( float64_sqrt );\r | |
2258 | break;\r | |
2259 | case FLOAT64_EQ:\r | |
2260 | time_ab_float64_z_flag( float64_eq );\r | |
2261 | break;\r | |
2262 | case FLOAT64_LE:\r | |
2263 | time_ab_float64_z_flag( float64_le );\r | |
2264 | break;\r | |
2265 | case FLOAT64_LT:\r | |
2266 | time_ab_float64_z_flag( float64_lt );\r | |
2267 | break;\r | |
2268 | case FLOAT64_EQ_SIGNALING:\r | |
2269 | time_ab_float64_z_flag( float64_eq_signaling );\r | |
2270 | break;\r | |
2271 | case FLOAT64_LE_QUIET:\r | |
2272 | time_ab_float64_z_flag( float64_le_quiet );\r | |
2273 | break;\r | |
2274 | case FLOAT64_LT_QUIET:\r | |
2275 | time_ab_float64_z_flag( float64_lt_quiet );\r | |
2276 | break;\r | |
2277 | #ifdef FLOATX80\r | |
2278 | case FLOATX80_TO_INT32:\r | |
2279 | time_a_floatx80_z_int32( floatx80_to_int32 );\r | |
2280 | break;\r | |
2281 | case FLOATX80_TO_INT32_ROUND_TO_ZERO:\r | |
2282 | time_a_floatx80_z_int32( floatx80_to_int32_round_to_zero );\r | |
2283 | break;\r | |
2284 | case FLOATX80_TO_INT64:\r | |
2285 | time_a_floatx80_z_int64( floatx80_to_int64 );\r | |
2286 | break;\r | |
2287 | case FLOATX80_TO_INT64_ROUND_TO_ZERO:\r | |
2288 | time_a_floatx80_z_int64( floatx80_to_int64_round_to_zero );\r | |
2289 | break;\r | |
2290 | case FLOATX80_TO_FLOAT32:\r | |
2291 | time_a_floatx80_z_float32( floatx80_to_float32 );\r | |
2292 | break;\r | |
2293 | case FLOATX80_TO_FLOAT64:\r | |
2294 | time_a_floatx80_z_float64( floatx80_to_float64 );\r | |
2295 | break;\r | |
2296 | #ifdef FLOAT128\r | |
2297 | case FLOATX80_TO_FLOAT128:\r | |
2298 | time_a_floatx80_z_float128( floatx80_to_float128 );\r | |
2299 | break;\r | |
2300 | #endif\r | |
2301 | case FLOATX80_ROUND_TO_INT:\r | |
2302 | time_az_floatx80( floatx80_round_to_int );\r | |
2303 | break;\r | |
2304 | case FLOATX80_ADD:\r | |
2305 | time_abz_floatx80( floatx80_add );\r | |
2306 | break;\r | |
2307 | case FLOATX80_SUB:\r | |
2308 | time_abz_floatx80( floatx80_sub );\r | |
2309 | break;\r | |
2310 | case FLOATX80_MUL:\r | |
2311 | time_abz_floatx80( floatx80_mul );\r | |
2312 | break;\r | |
2313 | case FLOATX80_DIV:\r | |
2314 | time_abz_floatx80( floatx80_div );\r | |
2315 | break;\r | |
2316 | case FLOATX80_REM:\r | |
2317 | time_abz_floatx80( floatx80_rem );\r | |
2318 | break;\r | |
2319 | case FLOATX80_SQRT:\r | |
2320 | time_az_floatx80_pos( floatx80_sqrt );\r | |
2321 | break;\r | |
2322 | case FLOATX80_EQ:\r | |
2323 | time_ab_floatx80_z_flag( floatx80_eq );\r | |
2324 | break;\r | |
2325 | case FLOATX80_LE:\r | |
2326 | time_ab_floatx80_z_flag( floatx80_le );\r | |
2327 | break;\r | |
2328 | case FLOATX80_LT:\r | |
2329 | time_ab_floatx80_z_flag( floatx80_lt );\r | |
2330 | break;\r | |
2331 | case FLOATX80_EQ_SIGNALING:\r | |
2332 | time_ab_floatx80_z_flag( floatx80_eq_signaling );\r | |
2333 | break;\r | |
2334 | case FLOATX80_LE_QUIET:\r | |
2335 | time_ab_floatx80_z_flag( floatx80_le_quiet );\r | |
2336 | break;\r | |
2337 | case FLOATX80_LT_QUIET:\r | |
2338 | time_ab_floatx80_z_flag( floatx80_lt_quiet );\r | |
2339 | break;\r | |
2340 | #endif\r | |
2341 | #ifdef FLOAT128\r | |
2342 | case FLOAT128_TO_INT32:\r | |
2343 | time_a_float128_z_int32( float128_to_int32 );\r | |
2344 | break;\r | |
2345 | case FLOAT128_TO_INT32_ROUND_TO_ZERO:\r | |
2346 | time_a_float128_z_int32( float128_to_int32_round_to_zero );\r | |
2347 | break;\r | |
2348 | case FLOAT128_TO_INT64:\r | |
2349 | time_a_float128_z_int64( float128_to_int64 );\r | |
2350 | break;\r | |
2351 | case FLOAT128_TO_INT64_ROUND_TO_ZERO:\r | |
2352 | time_a_float128_z_int64( float128_to_int64_round_to_zero );\r | |
2353 | break;\r | |
2354 | case FLOAT128_TO_FLOAT32:\r | |
2355 | time_a_float128_z_float32( float128_to_float32 );\r | |
2356 | break;\r | |
2357 | case FLOAT128_TO_FLOAT64:\r | |
2358 | time_a_float128_z_float64( float128_to_float64 );\r | |
2359 | break;\r | |
2360 | #ifdef FLOATX80\r | |
2361 | case FLOAT128_TO_FLOATX80:\r | |
2362 | time_a_float128_z_floatx80( float128_to_floatx80 );\r | |
2363 | break;\r | |
2364 | #endif\r | |
2365 | case FLOAT128_ROUND_TO_INT:\r | |
2366 | time_az_float128( float128_round_to_int );\r | |
2367 | break;\r | |
2368 | case FLOAT128_ADD:\r | |
2369 | time_abz_float128( float128_add );\r | |
2370 | break;\r | |
2371 | case FLOAT128_SUB:\r | |
2372 | time_abz_float128( float128_sub );\r | |
2373 | break;\r | |
2374 | case FLOAT128_MUL:\r | |
2375 | time_abz_float128( float128_mul );\r | |
2376 | break;\r | |
2377 | case FLOAT128_DIV:\r | |
2378 | time_abz_float128( float128_div );\r | |
2379 | break;\r | |
2380 | case FLOAT128_REM:\r | |
2381 | time_abz_float128( float128_rem );\r | |
2382 | break;\r | |
2383 | case FLOAT128_SQRT:\r | |
2384 | time_az_float128_pos( float128_sqrt );\r | |
2385 | break;\r | |
2386 | case FLOAT128_EQ:\r | |
2387 | time_ab_float128_z_flag( float128_eq );\r | |
2388 | break;\r | |
2389 | case FLOAT128_LE:\r | |
2390 | time_ab_float128_z_flag( float128_le );\r | |
2391 | break;\r | |
2392 | case FLOAT128_LT:\r | |
2393 | time_ab_float128_z_flag( float128_lt );\r | |
2394 | break;\r | |
2395 | case FLOAT128_EQ_SIGNALING:\r | |
2396 | time_ab_float128_z_flag( float128_eq_signaling );\r | |
2397 | break;\r | |
2398 | case FLOAT128_LE_QUIET:\r | |
2399 | time_ab_float128_z_flag( float128_le_quiet );\r | |
2400 | break;\r | |
2401 | case FLOAT128_LT_QUIET:\r | |
2402 | time_ab_float128_z_flag( float128_lt_quiet );\r | |
2403 | break;\r | |
2404 | #endif\r | |
2405 | }\r | |
2406 | \r | |
2407 | }\r | |
2408 | \r | |
2409 | static void\r | |
2410 | timeFunction(\r | |
2411 | uint8 functionCode,\r | |
2412 | int8 roundingPrecisionIn,\r | |
2413 | int8 roundingModeIn,\r | |
2414 | int8 tininessModeIn\r | |
2415 | )\r | |
2416 | {\r | |
2417 | int8 roundingPrecision, roundingMode, tininessMode;\r | |
2418 | \r | |
2419 | roundingPrecision = 32;\r | |
2420 | for (;;) {\r | |
2421 | if ( ! functions[ functionCode ].roundingPrecision ) {\r | |
2422 | roundingPrecision = 0;\r | |
2423 | }\r | |
2424 | else if ( roundingPrecisionIn ) {\r | |
2425 | roundingPrecision = roundingPrecisionIn;\r | |
2426 | }\r | |
2427 | for ( roundingMode = 1;\r | |
2428 | roundingMode < NUM_ROUNDINGMODES;\r | |
2429 | ++roundingMode\r | |
2430 | ) {\r | |
2431 | if ( ! functions[ functionCode ].roundingMode ) {\r | |
2432 | roundingMode = 0;\r | |
2433 | }\r | |
2434 | else if ( roundingModeIn ) {\r | |
2435 | roundingMode = roundingModeIn;\r | |
2436 | }\r | |
2437 | for ( tininessMode = 1;\r | |
2438 | tininessMode < NUM_TININESSMODES;\r | |
2439 | ++tininessMode\r | |
2440 | ) {\r | |
2441 | if ( ( roundingPrecision == 32 )\r | |
2442 | || ( roundingPrecision == 64 ) ) {\r | |
2443 | if ( ! functions[ functionCode ]\r | |
2444 | .tininessModeAtReducedPrecision\r | |
2445 | ) {\r | |
2446 | tininessMode = 0;\r | |
2447 | }\r | |
2448 | else if ( tininessModeIn ) {\r | |
2449 | tininessMode = tininessModeIn;\r | |
2450 | }\r | |
2451 | }\r | |
2452 | else {\r | |
2453 | if ( ! functions[ functionCode ].tininessMode ) {\r | |
2454 | tininessMode = 0;\r | |
2455 | }\r | |
2456 | else if ( tininessModeIn ) {\r | |
2457 | tininessMode = tininessModeIn;\r | |
2458 | }\r | |
2459 | }\r | |
2460 | timeFunctionVariety(\r | |
2461 | functionCode, roundingPrecision, roundingMode, tininessMode\r | |
2462 | );\r | |
2463 | if ( tininessModeIn || ! tininessMode ) break;\r | |
2464 | }\r | |
2465 | if ( roundingModeIn || ! roundingMode ) break;\r | |
2466 | }\r | |
2467 | if ( roundingPrecisionIn || ! roundingPrecision ) break;\r | |
2468 | if ( roundingPrecision == 80 ) {\r | |
2469 | break;\r | |
2470 | }\r | |
2471 | else if ( roundingPrecision == 64 ) {\r | |
2472 | roundingPrecision = 80;\r | |
2473 | }\r | |
2474 | else if ( roundingPrecision == 32 ) {\r | |
2475 | roundingPrecision = 64;\r | |
2476 | }\r | |
2477 | }\r | |
2478 | \r | |
2479 | }\r | |
2480 | \r | |
2481 | main( int argc, char **argv )\r | |
2482 | {\r | |
2483 | char *argPtr;\r | |
2484 | flag functionArgument;\r | |
2485 | uint8 functionCode;\r | |
2486 | int8 operands, roundingPrecision, roundingMode, tininessMode;\r | |
2487 | \r | |
2488 | if ( argc <= 1 ) goto writeHelpMessage;\r | |
2489 | functionArgument = FALSE;\r | |
2490 | functionCode = 0;\r | |
2491 | operands = 0;\r | |
2492 | roundingPrecision = 0;\r | |
2493 | roundingMode = 0;\r | |
2494 | tininessMode = 0;\r | |
2495 | --argc;\r | |
2496 | ++argv;\r | |
2497 | while ( argc && ( argPtr = argv[ 0 ] ) ) {\r | |
2498 | if ( argPtr[ 0 ] == '-' ) ++argPtr;\r | |
2499 | if ( strcmp( argPtr, "help" ) == 0 ) {\r | |
2500 | writeHelpMessage:\r | |
2501 | fputs(\r | |
2502 | "timesoftfloat [<option>...] <function>\n"\r | |
2503 | " <option>: (* is default)\n"\r | |
2504 | " -help --Write this message and exit.\n"\r | |
2505 | #ifdef FLOATX80\r | |
2506 | " -precision32 --Only time rounding precision equivalent to float32.\n"\r | |
2507 | " -precision64 --Only time rounding precision equivalent to float64.\n"\r | |
2508 | " -precision80 --Only time maximum rounding precision.\n"\r | |
2509 | #endif\r | |
2510 | " -nearesteven --Only time rounding to nearest/even.\n"\r | |
2511 | " -tozero --Only time rounding to zero.\n"\r | |
2512 | " -down --Only time rounding down.\n"\r | |
2513 | " -up --Only time rounding up.\n"\r | |
2514 | " -tininessbefore --Only time underflow tininess before rounding.\n"\r | |
2515 | " -tininessafter --Only time underflow tininess after rounding.\n"\r | |
2516 | " <function>:\n"\r | |
2517 | " int32_to_<float> <float>_add <float>_eq\n"\r | |
2518 | " <float>_to_int32 <float>_sub <float>_le\n"\r | |
2519 | " <float>_to_int32_round_to_zero <float>_mul <float>_lt\n"\r | |
2520 | " int64_to_<float> <float>_div <float>_eq_signaling\n"\r | |
2521 | " <float>_to_int64 <float>_rem <float>_le_quiet\n"\r | |
2522 | " <float>_to_int64_round_to_zero <float>_lt_quiet\n"\r | |
2523 | " <float>_to_<float>\n"\r | |
2524 | " <float>_round_to_int\n"\r | |
2525 | " <float>_sqrt\n"\r | |
2526 | " -all1 --All 1-operand functions.\n"\r | |
2527 | " -all2 --All 2-operand functions.\n"\r | |
2528 | " -all --All functions.\n"\r | |
2529 | " <float>:\n"\r | |
2530 | " float32 --Single precision.\n"\r | |
2531 | " float64 --Double precision.\n"\r | |
2532 | #ifdef FLOATX80\r | |
2533 | " floatx80 --Extended double precision.\n"\r | |
2534 | #endif\r | |
2535 | #ifdef FLOAT128\r | |
2536 | " float128 --Quadruple precision.\n"\r | |
2537 | #endif\r | |
2538 | ,\r | |
2539 | stdout\r | |
2540 | );\r | |
2541 | return EXIT_SUCCESS;\r | |
2542 | }\r | |
2543 | #ifdef FLOATX80\r | |
2544 | else if ( strcmp( argPtr, "precision32" ) == 0 ) {\r | |
2545 | roundingPrecision = 32;\r | |
2546 | }\r | |
2547 | else if ( strcmp( argPtr, "precision64" ) == 0 ) {\r | |
2548 | roundingPrecision = 64;\r | |
2549 | }\r | |
2550 | else if ( strcmp( argPtr, "precision80" ) == 0 ) {\r | |
2551 | roundingPrecision = 80;\r | |
2552 | }\r | |
2553 | #endif\r | |
2554 | else if ( ( strcmp( argPtr, "nearesteven" ) == 0 )\r | |
2555 | || ( strcmp( argPtr, "nearest_even" ) == 0 ) ) {\r | |
2556 | roundingMode = ROUND_NEAREST_EVEN;\r | |
2557 | }\r | |
2558 | else if ( ( strcmp( argPtr, "tozero" ) == 0 )\r | |
2559 | || ( strcmp( argPtr, "to_zero" ) == 0 ) ) {\r | |
2560 | roundingMode = ROUND_TO_ZERO;\r | |
2561 | }\r | |
2562 | else if ( strcmp( argPtr, "down" ) == 0 ) {\r | |
2563 | roundingMode = ROUND_DOWN;\r | |
2564 | }\r | |
2565 | else if ( strcmp( argPtr, "up" ) == 0 ) {\r | |
2566 | roundingMode = ROUND_UP;\r | |
2567 | }\r | |
2568 | else if ( strcmp( argPtr, "tininessbefore" ) == 0 ) {\r | |
2569 | tininessMode = TININESS_BEFORE_ROUNDING;\r | |
2570 | }\r | |
2571 | else if ( strcmp( argPtr, "tininessafter" ) == 0 ) {\r | |
2572 | tininessMode = TININESS_AFTER_ROUNDING;\r | |
2573 | }\r | |
2574 | else if ( strcmp( argPtr, "all1" ) == 0 ) {\r | |
2575 | functionArgument = TRUE;\r | |
2576 | functionCode = 0;\r | |
2577 | operands = 1;\r | |
2578 | }\r | |
2579 | else if ( strcmp( argPtr, "all2" ) == 0 ) {\r | |
2580 | functionArgument = TRUE;\r | |
2581 | functionCode = 0;\r | |
2582 | operands = 2;\r | |
2583 | }\r | |
2584 | else if ( strcmp( argPtr, "all" ) == 0 ) {\r | |
2585 | functionArgument = TRUE;\r | |
2586 | functionCode = 0;\r | |
2587 | operands = 0;\r | |
2588 | }\r | |
2589 | else {\r | |
2590 | for ( functionCode = 1;\r | |
2591 | functionCode < NUM_FUNCTIONS;\r | |
2592 | ++functionCode\r | |
2593 | ) {\r | |
2594 | if ( strcmp( argPtr, functions[ functionCode ].name ) == 0 ) {\r | |
2595 | break;\r | |
2596 | }\r | |
2597 | }\r | |
2598 | if ( functionCode == NUM_FUNCTIONS ) {\r | |
2599 | fail( "Invalid option or function `%s'", argv[ 0 ] );\r | |
2600 | }\r | |
2601 | functionArgument = TRUE;\r | |
2602 | }\r | |
2603 | --argc;\r | |
2604 | ++argv;\r | |
2605 | }\r | |
2606 | if ( ! functionArgument ) fail( "Function argument required" );\r | |
2607 | if ( functionCode ) {\r | |
2608 | timeFunction(\r | |
2609 | functionCode, roundingPrecision, roundingMode, tininessMode );\r | |
2610 | }\r | |
2611 | else if ( operands == 1 ) {\r | |
2612 | for ( functionCode = 1; functionCode < NUM_FUNCTIONS; ++functionCode\r | |
2613 | ) {\r | |
2614 | if ( functions[ functionCode ].numInputs == 1 ) {\r | |
2615 | timeFunction(\r | |
2616 | functionCode, roundingPrecision, roundingMode, tininessMode\r | |
2617 | );\r | |
2618 | }\r | |
2619 | }\r | |
2620 | }\r | |
2621 | else if ( operands == 2 ) {\r | |
2622 | for ( functionCode = 1; functionCode < NUM_FUNCTIONS; ++functionCode\r | |
2623 | ) {\r | |
2624 | if ( functions[ functionCode ].numInputs == 2 ) {\r | |
2625 | timeFunction(\r | |
2626 | functionCode, roundingPrecision, roundingMode, tininessMode\r | |
2627 | );\r | |
2628 | }\r | |
2629 | }\r | |
2630 | }\r | |
2631 | else {\r | |
2632 | for ( functionCode = 1; functionCode < NUM_FUNCTIONS; ++functionCode\r | |
2633 | ) {\r | |
2634 | timeFunction(\r | |
2635 | functionCode, roundingPrecision, roundingMode, tininessMode );\r | |
2636 | }\r | |
2637 | }\r | |
2638 | return EXIT_SUCCESS;\r | |
2639 | \r | |
2640 | }\r | |
2641 | \r |