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 5.5 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226
  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 r3
  41. #define X r4
  42. #define INCX r5
  43. #define Y r6
  44. #define INCY r7
  45. #define PREA r8
  46. #define STACKSIZE 16
  47. PROLOGUE
  48. PROFCODE
  49. addi SP, SP, -STACKSIZE
  50. li r0, 0
  51. stfd f14, 0(SP)
  52. stfd f15, 8(SP)
  53. slwi INCX, INCX, BASE_SHIFT
  54. slwi INCY, INCY, BASE_SHIFT
  55. #ifdef L1_DUALFETCH
  56. li PREA, (L1_PREFETCHSIZE) / 2
  57. #else
  58. li PREA, (L1_PREFETCHSIZE)
  59. #endif
  60. cmpwi cr0, N, 0
  61. ble- LL(999)
  62. cmpwi cr0, INCX, SIZE
  63. bne- cr0, LL(100)
  64. cmpwi cr0, INCY, SIZE
  65. bne- cr0, LL(100)
  66. srawi. r0, N, 4
  67. mtspr CTR, r0
  68. beq- cr0, LL(50)
  69. .align 4
  70. LL(10):
  71. LFD f0, 0 * SIZE(X)
  72. LFD f1, 1 * SIZE(X)
  73. LFD f2, 2 * SIZE(X)
  74. LFD f3, 3 * SIZE(X)
  75. STFD f0, 0 * SIZE(Y)
  76. STFD f1, 1 * SIZE(Y)
  77. STFD f2, 2 * SIZE(Y)
  78. STFD f3, 3 * SIZE(Y)
  79. LFD f4, 4 * SIZE(X)
  80. LFD f5, 5 * SIZE(X)
  81. LFD f6, 6 * SIZE(X)
  82. LFD f7, 7 * SIZE(X)
  83. STFD f4, 4 * SIZE(Y)
  84. STFD f5, 5 * SIZE(Y)
  85. STFD f6, 6 * SIZE(Y)
  86. STFD f7, 7 * SIZE(Y)
  87. LFD f8, 8 * SIZE(X)
  88. LFD f9, 9 * SIZE(X)
  89. LFD f10, 10 * SIZE(X)
  90. LFD f11, 11 * SIZE(X)
  91. STFD f8, 8 * SIZE(Y)
  92. STFD f9, 9 * SIZE(Y)
  93. STFD f10, 10 * SIZE(Y)
  94. STFD f11, 11 * SIZE(Y)
  95. LFD f12, 12 * SIZE(X)
  96. LFD f13, 13 * SIZE(X)
  97. LFD f14, 14 * SIZE(X)
  98. LFD f15, 15 * SIZE(X)
  99. STFD f12, 12 * SIZE(Y)
  100. STFD f13, 13 * SIZE(Y)
  101. STFD f14, 14 * SIZE(Y)
  102. STFD f15, 15 * SIZE(Y)
  103. #ifndef POWER6
  104. dcbtst Y, PREA
  105. #ifdef L1_DUALFETCH
  106. dcbt X, PREA
  107. #endif
  108. #endif
  109. addi X, X, 16 * SIZE
  110. addi Y, Y, 16 * SIZE
  111. #ifdef POWER6
  112. dcbtst Y, PREA
  113. L1_PREFETCH X, PREA
  114. #endif
  115. bdnz LL(10)
  116. .align 4
  117. LL(50):
  118. andi. r0, N, 15
  119. mtspr CTR, r0
  120. beq LL(999)
  121. .align 4
  122. LL(60):
  123. LFD f8, 0 * SIZE(X)
  124. addi X, X, 1 * SIZE
  125. STFD f8, 0 * SIZE(Y)
  126. addi Y, Y, 1 * SIZE
  127. bdnz LL(60)
  128. b LL(999)
  129. .align 4
  130. LL(100):
  131. sub X, X, INCX
  132. sub Y, Y, INCY
  133. srawi. r0, N, 4
  134. mtspr CTR, r0
  135. beq- LL(150)
  136. .align 4
  137. LL(110):
  138. LFDUX f0, X, INCX
  139. LFDUX f1, X, INCX
  140. LFDUX f2, X, INCX
  141. LFDUX f3, X, INCX
  142. LFDUX f4, X, INCX
  143. LFDUX f5, X, INCX
  144. LFDUX f6, X, INCX
  145. LFDUX f7, X, INCX
  146. LFDUX f8, X, INCX
  147. LFDUX f9, X, INCX
  148. LFDUX f10, X, INCX
  149. LFDUX f11, X, INCX
  150. LFDUX f12, X, INCX
  151. LFDUX f13, X, INCX
  152. LFDUX f14, X, INCX
  153. LFDUX f15, X, INCX
  154. STFDUX f0, Y, INCY
  155. STFDUX f1, Y, INCY
  156. STFDUX f2, Y, INCY
  157. STFDUX f3, Y, INCY
  158. STFDUX f4, Y, INCY
  159. STFDUX f5, Y, INCY
  160. STFDUX f6, Y, INCY
  161. STFDUX f7, Y, INCY
  162. STFDUX f8, Y, INCY
  163. STFDUX f9, Y, INCY
  164. STFDUX f10, Y, INCY
  165. STFDUX f11, Y, INCY
  166. STFDUX f12, Y, INCY
  167. STFDUX f13, Y, INCY
  168. STFDUX f14, Y, INCY
  169. STFDUX f15, Y, INCY
  170. bdnz LL(110)
  171. .align 4
  172. LL(150):
  173. andi. r0, N, 15
  174. mtspr CTR, r0
  175. beq LL(999)
  176. .align 4
  177. LL(160):
  178. LFDUX f8, X, INCX
  179. STFDUX f8, Y, INCY
  180. bdnz LL(160)
  181. .align 4
  182. LL(999):
  183. lfd f14, 0(SP)
  184. lfd f15, 8(SP)
  185. addi SP, SP, STACKSIZE
  186. blr
  187. EPILOGUE