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.

copy.S 5.9 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277
  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. #define ASSEMBLER
  39. #include "common.h"
  40. #define N $4
  41. #define X $5
  42. #define INCX $6
  43. #define Y $7
  44. #define INCY $8
  45. #define I $2
  46. #define TEMP $3
  47. #define a1 $f0
  48. #define a2 $f1
  49. #define a3 $f2
  50. #define a4 $f3
  51. #define a5 $f4
  52. #define a6 $f5
  53. #define a7 $f6
  54. #define a8 $f7
  55. PROLOGUE
  56. #ifdef F_INTERFACE
  57. LDINT N, 0(N)
  58. LDINT INCX, 0(INCX)
  59. LDINT INCY, 0(INCY)
  60. #endif
  61. li TEMP, SIZE
  62. NOP
  63. blez N, .L999
  64. dsll INCX, INCX, BASE_SHIFT
  65. bne INCX, TEMP, .L20
  66. dsll INCY, INCY, BASE_SHIFT
  67. bne INCY, TEMP, .L20
  68. dsra I, N, 3
  69. blez I, .L15
  70. daddiu I, I, -1
  71. LD a1, 0 * SIZE(X)
  72. LD a2, 1 * SIZE(X)
  73. LD a3, 2 * SIZE(X)
  74. LD a4, 3 * SIZE(X)
  75. LD a5, 4 * SIZE(X)
  76. LD a6, 5 * SIZE(X)
  77. LD a7, 6 * SIZE(X)
  78. LD a8, 7 * SIZE(X)
  79. blez I, .L13
  80. NOP
  81. .align 3
  82. .L12:
  83. ST a1, 0 * SIZE(Y)
  84. LD a1, 8 * SIZE(X)
  85. ST a2, 1 * SIZE(Y)
  86. LD a2, 9 * SIZE(X)
  87. ST a3, 2 * SIZE(Y)
  88. LD a3, 10 * SIZE(X)
  89. ST a4, 3 * SIZE(Y)
  90. LD a4, 11 * SIZE(X)
  91. ST a5, 4 * SIZE(Y)
  92. LD a5, 12 * SIZE(X)
  93. ST a6, 5 * SIZE(Y)
  94. LD a6, 13 * SIZE(X)
  95. ST a7, 6 * SIZE(Y)
  96. LD a7, 14 * SIZE(X)
  97. ST a8, 7 * SIZE(Y)
  98. LD a8, 15 * SIZE(X)
  99. daddiu I, I, -1
  100. daddiu X, X, 8 * SIZE
  101. bgtz I, .L12
  102. daddiu Y, Y, 8 * SIZE
  103. .align 3
  104. .L13:
  105. ST a1, 0 * SIZE(Y)
  106. ST a2, 1 * SIZE(Y)
  107. ST a3, 2 * SIZE(Y)
  108. ST a4, 3 * SIZE(Y)
  109. ST a5, 4 * SIZE(Y)
  110. ST a6, 5 * SIZE(Y)
  111. ST a7, 6 * SIZE(Y)
  112. ST a8, 7 * SIZE(Y)
  113. daddiu X, X, 8 * SIZE
  114. daddiu Y, Y, 8 * SIZE
  115. .align 3
  116. .L15:
  117. andi I, N, 7
  118. blez I, .L999
  119. NOP
  120. .align 3
  121. .L16:
  122. LD a1, 0 * SIZE(X)
  123. daddiu X, X, SIZE
  124. daddiu I, I, -1
  125. daddiu Y, Y, SIZE
  126. bgtz I, .L16
  127. ST a1, -1 * SIZE(Y)
  128. j .L999
  129. NOP
  130. .align 3
  131. .L20:
  132. dsra I, N, 3
  133. blez I, .L25
  134. daddiu I, I, -1
  135. LD a1, 0 * SIZE(X)
  136. daddu X, X, INCX
  137. LD a2, 0 * SIZE(X)
  138. daddu X, X, INCX
  139. LD a3, 0 * SIZE(X)
  140. daddu X, X, INCX
  141. LD a4, 0 * SIZE(X)
  142. daddu X, X, INCX
  143. LD a5, 0 * SIZE(X)
  144. daddu X, X, INCX
  145. LD a6, 0 * SIZE(X)
  146. daddu X, X, INCX
  147. LD a7, 0 * SIZE(X)
  148. daddu X, X, INCX
  149. LD a8, 0 * SIZE(X)
  150. daddu X, X, INCX
  151. blez I, .L23
  152. NOP
  153. .align 3
  154. .L22:
  155. ST a1, 0 * SIZE(Y)
  156. daddu Y, Y, INCY
  157. LD a1, 0 * SIZE(X)
  158. daddu X, X, INCX
  159. ST a2, 0 * SIZE(Y)
  160. daddu Y, Y, INCY
  161. LD a2, 0 * SIZE(X)
  162. daddu X, X, INCX
  163. ST a3, 0 * SIZE(Y)
  164. daddu Y, Y, INCY
  165. LD a3, 0 * SIZE(X)
  166. daddu X, X, INCX
  167. ST a4, 0 * SIZE(Y)
  168. daddu Y, Y, INCY
  169. LD a4, 0 * SIZE(X)
  170. daddu X, X, INCX
  171. ST a5, 0 * SIZE(Y)
  172. daddu Y, Y, INCY
  173. LD a5, 0 * SIZE(X)
  174. daddu X, X, INCX
  175. ST a6, 0 * SIZE(Y)
  176. daddu Y, Y, INCY
  177. LD a6, 0 * SIZE(X)
  178. daddu X, X, INCX
  179. ST a7, 0 * SIZE(Y)
  180. daddu Y, Y, INCY
  181. LD a7, 0 * SIZE(X)
  182. daddu X, X, INCX
  183. ST a8, 0 * SIZE(Y)
  184. daddu Y, Y, INCY
  185. LD a8, 0 * SIZE(X)
  186. daddiu I, I, -1
  187. bgtz I, .L22
  188. daddu X, X, INCX
  189. .align 3
  190. .L23:
  191. ST a1, 0 * SIZE(Y)
  192. daddu Y, Y, INCY
  193. ST a2, 0 * SIZE(Y)
  194. daddu Y, Y, INCY
  195. ST a3, 0 * SIZE(Y)
  196. daddu Y, Y, INCY
  197. ST a4, 0 * SIZE(Y)
  198. daddu Y, Y, INCY
  199. ST a5, 0 * SIZE(Y)
  200. daddu Y, Y, INCY
  201. ST a6, 0 * SIZE(Y)
  202. daddu Y, Y, INCY
  203. ST a7, 0 * SIZE(Y)
  204. daddu Y, Y, INCY
  205. ST a8, 0 * SIZE(Y)
  206. daddu Y, Y, INCY
  207. .align 3
  208. .L25:
  209. andi I, N, 7
  210. blez I, .L999
  211. NOP
  212. .align 3
  213. .L26:
  214. LD a1, 0 * SIZE(X)
  215. daddu X, X, INCX
  216. daddiu I, I, -1
  217. ST a1, 0 * SIZE(Y)
  218. bgtz I, .L26
  219. daddu Y, Y, INCY
  220. .align 3
  221. .L999:
  222. j $31
  223. NOP
  224. EPILOGUE