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.

zcopy.S 5.8 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237
  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 INCXM1 r9
  47. #define INCYM1 r10
  48. #define STACKSIZE 16
  49. PROLOGUE
  50. PROFCODE
  51. addi SP, SP, -STACKSIZE
  52. li r0, 0
  53. stfd f14, 0(SP)
  54. stfd f15, 8(SP)
  55. slwi INCX, INCX, ZBASE_SHIFT
  56. slwi INCY, INCY, ZBASE_SHIFT
  57. subi INCXM1, INCX, SIZE
  58. subi INCYM1, INCY, SIZE
  59. #ifdef L1_DUALFETCH
  60. li PREA, (L1_PREFETCHSIZE) / 2
  61. #else
  62. li PREA, (L1_PREFETCHSIZE)
  63. #endif
  64. cmpwi cr0, N, 0
  65. ble- LL(999)
  66. cmpwi cr0, INCX, 2 * SIZE
  67. bne- cr0, LL(100)
  68. cmpwi cr0, INCY, 2 * SIZE
  69. bne- cr0, LL(100)
  70. srawi. r0, N, 3
  71. mtspr CTR, r0
  72. beq- cr0, LL(50)
  73. .align 4
  74. LL(10):
  75. LFD f0, 0 * SIZE(X)
  76. LFD f1, 1 * SIZE(X)
  77. LFD f2, 2 * SIZE(X)
  78. LFD f3, 3 * SIZE(X)
  79. STFD f0, 0 * SIZE(Y)
  80. STFD f1, 1 * SIZE(Y)
  81. STFD f2, 2 * SIZE(Y)
  82. STFD f3, 3 * SIZE(Y)
  83. LFD f4, 4 * SIZE(X)
  84. LFD f5, 5 * SIZE(X)
  85. LFD f6, 6 * SIZE(X)
  86. LFD f7, 7 * SIZE(X)
  87. STFD f4, 4 * SIZE(Y)
  88. STFD f5, 5 * SIZE(Y)
  89. STFD f6, 6 * SIZE(Y)
  90. STFD f7, 7 * SIZE(Y)
  91. LFD f8, 8 * SIZE(X)
  92. LFD f9, 9 * SIZE(X)
  93. LFD f10, 10 * SIZE(X)
  94. LFD f11, 11 * SIZE(X)
  95. STFD f8, 8 * SIZE(Y)
  96. STFD f9, 9 * SIZE(Y)
  97. STFD f10, 10 * SIZE(Y)
  98. STFD f11, 11 * SIZE(Y)
  99. LFD f12, 12 * SIZE(X)
  100. LFD f13, 13 * SIZE(X)
  101. LFD f14, 14 * SIZE(X)
  102. LFD f15, 15 * SIZE(X)
  103. STFD f12, 12 * SIZE(Y)
  104. STFD f13, 13 * SIZE(Y)
  105. STFD f14, 14 * SIZE(Y)
  106. STFD f15, 15 * SIZE(Y)
  107. #ifndef POWER6
  108. dcbtst Y, PREA
  109. #ifdef L1_DUALFETCH
  110. dcbt X, PREA
  111. #endif
  112. #endif
  113. addi X, X, 16 * SIZE
  114. addi Y, Y, 16 * SIZE
  115. #ifdef POWER6
  116. dcbtst Y, PREA
  117. L1_PREFETCH X, PREA
  118. #endif
  119. bdnz LL(10)
  120. .align 4
  121. LL(50):
  122. andi. r0, N, 7
  123. mtspr CTR, r0
  124. beq LL(999)
  125. .align 4
  126. LL(60):
  127. LFD f8, 0 * SIZE(X)
  128. LFD f9, 1 * SIZE(X)
  129. addi X, X, 2 * SIZE
  130. STFD f8, 0 * SIZE(Y)
  131. STFD f9, 1 * SIZE(Y)
  132. addi Y, Y, 2 * SIZE
  133. bdnz LL(60)
  134. b LL(999)
  135. .align 4
  136. LL(100):
  137. sub X, X, INCXM1
  138. sub Y, Y, INCYM1
  139. srawi. r0, N, 3
  140. mtspr CTR, r0
  141. beq- LL(150)
  142. .align 4
  143. LL(110):
  144. LFDX f0, X, INCXM1
  145. LFDUX f1, X, INCX
  146. LFDX f2, X, INCXM1
  147. LFDUX f3, X, INCX
  148. LFDX f4, X, INCXM1
  149. LFDUX f5, X, INCX
  150. LFDX f6, X, INCXM1
  151. LFDUX f7, X, INCX
  152. LFDX f8, X, INCXM1
  153. LFDUX f9, X, INCX
  154. LFDX f10, X, INCXM1
  155. LFDUX f11, X, INCX
  156. LFDX f12, X, INCXM1
  157. LFDUX f13, X, INCX
  158. LFDX f14, X, INCXM1
  159. LFDUX f15, X, INCX
  160. STFDX f0, Y, INCYM1
  161. STFDUX f1, Y, INCY
  162. STFDX f2, Y, INCYM1
  163. STFDUX f3, Y, INCY
  164. STFDX f4, Y, INCYM1
  165. STFDUX f5, Y, INCY
  166. STFDX f6, Y, INCYM1
  167. STFDUX f7, Y, INCY
  168. STFDX f8, Y, INCYM1
  169. STFDUX f9, Y, INCY
  170. STFDX f10, Y, INCYM1
  171. STFDUX f11, Y, INCY
  172. STFDX f12, Y, INCYM1
  173. STFDUX f13, Y, INCY
  174. STFDX f14, Y, INCYM1
  175. STFDUX f15, Y, INCY
  176. bdnz LL(110)
  177. .align 4
  178. LL(150):
  179. andi. r0, N, 7
  180. mtspr CTR, r0
  181. beq LL(999)
  182. .align 4
  183. LL(160):
  184. LFDX f8, X, INCXM1
  185. LFDUX f9, X, INCX
  186. STFDX f8, Y, INCYM1
  187. STFDUX f9, Y, INCY
  188. bdnz LL(160)
  189. .align 4
  190. LL(999):
  191. lfd f14, 0(SP)
  192. lfd f15, 8(SP)
  193. addi SP, SP, STACKSIZE
  194. blr
  195. EPILOGUE