OSchip
/
OpenBLAS
Not watched
1
0
Fork 0
Code
Releases 66 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
Tree: de139337b8
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'de139337b8'
${ noResults }
OpenBLAS/kernel/x86/iamax.S

365 lines
6.5 kB
Raw Blame History 

  1. /*********************************************************************/

  2. /* Copyright 2009, 2010 The University of Texas at Austin.           */

  3. /* All rights reserved.                                              */

  4. /*                                                                   */

  5. /* Redistribution and use in source and binary forms, with or        */

  6. /* without modification, are permitted provided that the following   */

  7. /* conditions are met:                                               */

  8. /*                                                                   */

  9. /*   1. Redistributions of source code must retain the above         */

  10. /*      copyright notice, this list of conditions and the following  */

  11. /*      disclaimer.                                                  */

  12. /*                                                                   */

  13. /*   2. Redistributions in binary form must reproduce the above      */

  14. /*      copyright notice, this list of conditions and the following  */

  15. /*      disclaimer in the documentation and/or other materials       */

  16. /*      provided with the distribution.                              */

  17. /*                                                                   */

  18. /*    THIS  SOFTWARE IS PROVIDED  BY THE  UNIVERSITY OF  TEXAS AT    */

  19. /*    AUSTIN  ``AS IS''  AND ANY  EXPRESS OR  IMPLIED WARRANTIES,    */

  20. /*    INCLUDING, BUT  NOT LIMITED  TO, THE IMPLIED  WARRANTIES OF    */

  21. /*    MERCHANTABILITY  AND FITNESS FOR  A PARTICULAR  PURPOSE ARE    */

  22. /*    DISCLAIMED.  IN  NO EVENT SHALL THE UNIVERSITY  OF TEXAS AT    */

  23. /*    AUSTIN OR CONTRIBUTORS BE  LIABLE FOR ANY DIRECT, INDIRECT,    */

  24. /*    INCIDENTAL,  SPECIAL, EXEMPLARY,  OR  CONSEQUENTIAL DAMAGES    */

  25. /*    (INCLUDING, BUT  NOT LIMITED TO,  PROCUREMENT OF SUBSTITUTE    */

  26. /*    GOODS  OR  SERVICES; LOSS  OF  USE,  DATA,  OR PROFITS;  OR    */

  27. /*    BUSINESS INTERRUPTION) HOWEVER CAUSED  AND ON ANY THEORY OF    */

  28. /*    LIABILITY, WHETHER  IN CONTRACT, STRICT  LIABILITY, OR TORT    */

  29. /*    (INCLUDING NEGLIGENCE OR OTHERWISE)  ARISING IN ANY WAY OUT    */

  30. /*    OF  THE  USE OF  THIS  SOFTWARE,  EVEN  IF ADVISED  OF  THE    */

  31. /*    POSSIBILITY OF SUCH DAMAGE.                                    */

  32. /*                                                                   */

  33. /* The views and conclusions contained in the software and           */

  34. /* documentation are those of the authors and should not be          */

  35. /* interpreted as representing official policies, either expressed   */

  36. /* or implied, of The University of Texas at Austin.                 */

  37. /*********************************************************************/

  38. 

  39. #define ASSEMBLER

  40. #include "common.h"

  41. 

  42. #define STACK	16

  43. #define ARGS	 0

  44. 

  45. #define STACK_M		 4 + STACK + ARGS(%esp)

  46. #define STACK_X	 	 8 + STACK + ARGS(%esp)

  47. #define STACK_INCX	12 + STACK + ARGS(%esp)

  48. 

  49. #define M	%ebx

  50. #define INCX	%esi

  51. #define X	%ecx

  52. #define I	%edx

  53. #define NUM	%edi

  54. #define RET	%eax

  55. 

  56. #ifndef USE_MIN

  57. #define FMOV	fcmovbe

  58. #define IMOV	cmovnbe

  59. #else

  60. #define FMOV	fcmovnbe

  61. #define IMOV	cmovb

  62. #endif

  63. 

  64. #include "l1param.h"

  65. 

  66. 	PROLOGUE

  67. 

  68. 	pushl	%ebp

  69. 	pushl	%edi

  70. 	pushl	%esi

  71. 	pushl	%ebx

  72. 

  73. 	PROFCODE

  74. 

  75. 	movl	STACK_M,    M

  76. 	movl	STACK_INCX, INCX

  77. 	movl	STACK_X,    X

  78. 

  79. #ifdef F_INTERFACE

  80. 	movl	(M), M

  81. 	movl	(INCX), INCX

  82. #endif

  83. 

  84. #if defined(F_INTERFACE_GFORT) || defined(F_INTERFACE_G95)

  85. 	EMMS

  86. #endif

  87. 

  88. 	sall	$BASE_SHIFT, INCX

  89. 

  90. 	fldz

  91. 	xorl	RET, RET

  92. 

  93. 	testl	M, M

  94. 	jle	.L999

  95. 	testl	INCX, INCX

  96. 	jle	.L999

  97. 

  98. 	fstp	%st(0)

  99. 	movl	$2, NUM

  100. 	movl	$1, RET

  101. 

  102. 	FLD	(X)

  103. #ifdef USE_ABS

  104. 	fabs

  105. #endif

  106. 	addl	INCX, X

  107. 	decl	M

  108. 	jle	.L999

  109. 

  110. 	cmpl	$SIZE, INCX

  111. 	jne	.L40

  112. 

  113. 	movl	M, I

  114. 	sarl	$3,   I

  115. 	jle	.L20

  116. 	ALIGN_4

  117. 

  118. .L10:

  119. #ifdef PREFETCH

  120. 	PREFETCH (PREFETCHSIZE +  0) - PREOFFSET(X)

  121. #endif

  122. 

  123. 	FLD	0 * SIZE(X)

  124. #ifdef USE_ABS

  125. 	fabs

  126. #endif

  127. 	fcomi	%st(1), %st

  128. 	FMOV	%st(1), %st(0)

  129. 	IMOV	NUM, RET

  130. 	fstp	%st(1)

  131. 	incl	NUM

  132. 

  133. 	FLD	1 * SIZE(X)

  134. #ifdef USE_ABS

  135. 	fabs

  136. #endif

  137. 	fcomi	%st(1), %st

  138. 	FMOV	%st(1), %st(0)

  139. 	IMOV	NUM, RET

  140. 	fstp	%st(1)

  141. 	incl	NUM

  142. 

  143. 	FLD	2 * SIZE(X)

  144. #ifdef USE_ABS

  145. 	fabs

  146. #endif

  147. 	fcomi	%st(1), %st

  148. 	FMOV	%st(1), %st(0)

  149. 	IMOV	NUM, RET

  150. 	fstp	%st(1)

  151. 	incl	NUM

  152. 

  153. 	FLD	3 * SIZE(X)

  154. #ifdef USE_ABS

  155. 	fabs

  156. #endif

  157. 	fcomi	%st(1), %st

  158. 	FMOV	%st(1), %st(0)

  159. 	IMOV	NUM, RET

  160. 	fstp	%st(1)

  161. 	incl	NUM

  162. 

  163. 	FLD	4 * SIZE(X)

  164. #ifdef USE_ABS

  165. 	fabs

  166. #endif

  167. 	fcomi	%st(1), %st

  168. 	FMOV	%st(1), %st(0)

  169. 	IMOV	NUM, RET

  170. 	fstp	%st(1)

  171. 	incl	NUM

  172. 

  173. 	FLD	5 * SIZE(X)

  174. #ifdef USE_ABS

  175. 	fabs

  176. #endif

  177. 	fcomi	%st(1), %st

  178. 	FMOV	%st(1), %st(0)

  179. 	IMOV	NUM, RET

  180. 	fstp	%st(1)

  181. 	incl	NUM

  182. 

  183. 	FLD	6 * SIZE(X)

  184. #ifdef USE_ABS

  185. 	fabs

  186. #endif

  187. 	fcomi	%st(1), %st

  188. 	FMOV	%st(1), %st(0)

  189. 	IMOV	NUM, RET

  190. 	fstp	%st(1)

  191. 	incl	NUM

  192. 

  193. 	FLD	7 * SIZE(X)

  194. #ifdef USE_ABS

  195. 	fabs

  196. #endif

  197. 	fcomi	%st(1), %st

  198. 	FMOV	%st(1), %st(0)

  199. 	IMOV	NUM, RET

  200. 	fstp	%st(1)

  201. 	incl	NUM

  202. 

  203. 	addl	$8 * SIZE, X

  204. 

  205. 	decl	I

  206. 	jg	.L10

  207. 	ALIGN_4

  208. 

  209. .L20:

  210. 	movl	M,  I

  211. 	andl	$7, I

  212. 	jle	.L999

  213. 	ALIGN_4

  214. 

  215. 

  216. .L21:

  217. 	FLD	0 * SIZE(X)

  218. #ifdef USE_ABS

  219. 	fabs

  220. #endif

  221. 	fcomi	%st(1), %st

  222. 	FMOV	%st(1), %st(0)

  223. 	IMOV	NUM, RET

  224. 	fstp	%st(1)

  225. 

  226. 	addl	$1 * SIZE, X

  227. 	incl	NUM

  228. 	decl	I

  229. 	jg	.L21

  230. 	jmp	.L999

  231. 	ALIGN_4

  232. 

  233. .L40:

  234. 	movl	M, I

  235. 	sarl	$3,   I

  236. 	jle	.L60

  237. 	ALIGN_4

  238. 

  239. .L50:

  240. 	FLD	0 * SIZE(X)

  241. 	addl	INCX, X

  242. #ifdef USE_ABS

  243. 	fabs

  244. #endif

  245. 	fcomi	%st(1), %st

  246. 	FMOV	%st(1), %st(0)

  247. 	IMOV	NUM, RET

  248. 	fstp	%st(1)

  249. 	incl	NUM

  250. 

  251. 	FLD	0 * SIZE(X)

  252. 	addl	INCX, X

  253. #ifdef USE_ABS

  254. 	fabs

  255. #endif

  256. 	fcomi	%st(1), %st

  257. 	FMOV	%st(1), %st(0)

  258. 	IMOV	NUM, RET

  259. 	fstp	%st(1)

  260. 	incl	NUM

  261. 

  262. 	FLD	0 * SIZE(X)

  263. 	addl	INCX, X

  264. #ifdef USE_ABS

  265. 	fabs

  266. #endif

  267. 	fcomi	%st(1), %st

  268. 	FMOV	%st(1), %st(0)

  269. 	IMOV	NUM, RET

  270. 	fstp	%st(1)

  271. 	incl	NUM

  272. 

  273. 	FLD	0 * SIZE(X)

  274. 	addl	INCX, X

  275. #ifdef USE_ABS

  276. 	fabs

  277. #endif

  278. 	fcomi	%st(1), %st

  279. 	FMOV	%st(1), %st(0)

  280. 	IMOV	NUM, RET

  281. 	fstp	%st(1)

  282. 	incl	NUM

  283. 

  284. 	FLD	0 * SIZE(X)

  285. 	addl	INCX, X

  286. #ifdef USE_ABS

  287. 	fabs

  288. #endif

  289. 	fcomi	%st(1), %st

  290. 	FMOV	%st(1), %st(0)

  291. 	IMOV	NUM, RET

  292. 	fstp	%st(1)

  293. 	incl	NUM

  294. 

  295. 	FLD	0 * SIZE(X)

  296. 	addl	INCX, X

  297. #ifdef USE_ABS

  298. 	fabs

  299. #endif

  300. 	fcomi	%st(1), %st

  301. 	FMOV	%st(1), %st(0)

  302. 	IMOV	NUM, RET

  303. 	fstp	%st(1)

  304. 	incl	NUM

  305. 

  306. 	FLD	0 * SIZE(X)

  307. 	addl	INCX, X

  308. #ifdef USE_ABS

  309. 	fabs

  310. #endif

  311. 	fcomi	%st(1), %st

  312. 	FMOV	%st(1), %st(0)

  313. 	IMOV	NUM, RET

  314. 	fstp	%st(1)

  315. 	incl	NUM

  316. 

  317. 	FLD	0 * SIZE(X)

  318. 	addl	INCX, X

  319. #ifdef USE_ABS

  320. 	fabs

  321. #endif

  322. 	fcomi	%st(1), %st

  323. 	FMOV	%st(1), %st(0)

  324. 	IMOV	NUM, RET

  325. 	fstp	%st(1)

  326. 	incl	NUM

  327. 

  328. 	decl	I

  329. 	jg	.L50

  330. 	ALIGN_4

  331. 

  332. .L60:

  333. 	movl	M,  I

  334. 	andl	$7, I

  335. 	jle	.L999

  336. 	ALIGN_4

  337. 

  338. 

  339. .L61:

  340. 	FLD	0 * SIZE(X)

  341. #ifdef USE_ABS

  342. 	fabs

  343. #endif

  344. 	fcomi	%st(1), %st

  345. 	FMOV	%st(1), %st(0)

  346. 	IMOV	NUM, RET

  347. 	fstp	%st(1)

  348. 	incl	NUM

  349. 

  350. 	addl	INCX, X

  351. 	decl	I

  352. 	jg	.L61

  353. 	ALIGN_4

  354. 

  355. .L999:

  356. 	fstp	%st(0)

  357. 

  358. 	popl	%ebx

  359. 	popl	%esi

  360. 	popl	%edi

  361. 	popl	%ebp

  362. 	ret

  363. 

  364. 	EPILOGUE