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.

sgemm_tcopy_macros_8_power8.S 7.1 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308
  1. /***************************************************************************
  2. Copyright (c) 2013-2016, The OpenBLAS Project
  3. All rights reserved.
  4. Redistribution and use in source and binary forms, with or without
  5. modification, are permitted provided that the following conditions are
  6. met:
  7. 1. Redistributions of source code must retain the above copyright
  8. notice, this list of conditions and the following disclaimer.
  9. 2. Redistributions in binary form must reproduce the above copyright
  10. notice, this list of conditions and the following disclaimer in
  11. the documentation and/or other materials provided with the
  12. distribution.
  13. 3. Neither the name of the OpenBLAS project nor the names of
  14. its contributors may be used to endorse or promote products
  15. derived from this software without specific prior written permission.
  16. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  17. AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  18. IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  19. ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
  20. LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  21. DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  22. SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
  23. CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  24. OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
  25. USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  26. *****************************************************************************/
  27. /**************************************************************************************
  28. * 2016/04/23 Werner Saar (wernsaar@googlemail.com)
  29. * BLASTEST : OK
  30. * CTEST : OK
  31. * TEST : OK
  32. * LAPACK-TEST : OK
  33. **************************************************************************************/
  34. /**********************************************************************************************
  35. * Macros for N=4 and M=8
  36. **********************************************************************************************/
  37. .macro COPY_4x8
  38. lxvw4x vs32, o0, A0
  39. lxvw4x vs33, o16, A0
  40. lxvw4x vs34, o0, A1
  41. lxvw4x vs35, o16, A1
  42. lxvw4x vs36, o0, A2
  43. lxvw4x vs37, o16, A2
  44. lxvw4x vs38, o0, A3
  45. lxvw4x vs39, o16, A3
  46. mr T1, BO
  47. stxvw4x vs32, o0, T1
  48. stxvw4x vs33, o16, T1
  49. stxvw4x vs34, o32, T1
  50. stxvw4x vs35, o48, T1
  51. addi T1, T1, 64
  52. stxvw4x vs36, o0, T1
  53. stxvw4x vs37, o16, T1
  54. stxvw4x vs38, o32, T1
  55. stxvw4x vs39, o48, T1
  56. .endm
  57. /**********************************************************************************************
  58. * Macros for N=4 and M=4
  59. **********************************************************************************************/
  60. .macro COPY_4x4
  61. lxvw4x vs32, o0, A0
  62. lxvw4x vs33, o0, A1
  63. lxvw4x vs34, o0, A2
  64. lxvw4x vs35, o0, A3
  65. mr T1, BO
  66. stxvw4x vs32, o0, T1
  67. stxvw4x vs33, o16, T1
  68. stxvw4x vs34, o32, T1
  69. stxvw4x vs35, o48, T1
  70. .endm
  71. /**********************************************************************************************
  72. * Macros for N=4 and M=2
  73. **********************************************************************************************/
  74. .macro COPY_4x2
  75. lxsspx vs32, o0, A0
  76. lxsspx vs33, o4, A0
  77. lxsspx vs34, o0, A1
  78. lxsspx vs35, o4, A1
  79. lxsspx vs36, o0, A2
  80. lxsspx vs37, o4, A2
  81. lxsspx vs38, o0, A3
  82. lxsspx vs39, o4, A3
  83. mr T1, BO
  84. stxsspx vs32, o0, T1
  85. stxsspx vs33, o4, T1
  86. addi T1, T1, 8
  87. stxsspx vs34, o0, T1
  88. stxsspx vs35, o4, T1
  89. addi T1, T1, 8
  90. stxsspx vs36, o0, T1
  91. stxsspx vs37, o4, T1
  92. addi T1, T1, 8
  93. stxsspx vs38, o0, T1
  94. stxsspx vs39, o4, T1
  95. .endm
  96. /**********************************************************************************************
  97. * Macros for N=4 and M=1
  98. **********************************************************************************************/
  99. .macro COPY_4x1
  100. lxsspx vs32, o0, A0
  101. lxsspx vs33, o0, A1
  102. lxsspx vs34, o0, A2
  103. lxsspx vs35, o0, A3
  104. mr T1, BO
  105. stxsspx vs32, o0, T1
  106. stxsspx vs33, o4, T1
  107. addi T1, T1, 8
  108. stxsspx vs34, o0, T1
  109. stxsspx vs35, o4, T1
  110. .endm
  111. /**********************************************************************************************
  112. * Macros for N=2 and M=8
  113. **********************************************************************************************/
  114. .macro COPY_2x8
  115. lxvw4x vs32, o0, A0
  116. lxvw4x vs33, o16, A0
  117. lxvw4x vs34, o0, A1
  118. lxvw4x vs35, o16, A1
  119. mr T1, BO
  120. stxvw4x vs32, o0, T1
  121. stxvw4x vs33, o16, T1
  122. stxvw4x vs34, o32, T1
  123. stxvw4x vs35, o48, T1
  124. .endm
  125. /**********************************************************************************************
  126. * Macros for N=2 and M=4
  127. **********************************************************************************************/
  128. .macro COPY_2x4
  129. lxvw4x vs32, o0, A0
  130. lxvw4x vs33, o0, A1
  131. mr T1, BO
  132. stxvw4x vs32, o0, T1
  133. stxvw4x vs33, o16, T1
  134. .endm
  135. /**********************************************************************************************
  136. * Macros for N=2 and M=2
  137. **********************************************************************************************/
  138. .macro COPY_2x2
  139. lxsspx vs32, o0, A0
  140. lxsspx vs33, o4, A0
  141. lxsspx vs34, o0, A1
  142. lxsspx vs35, o4, A1
  143. mr T1, BO
  144. stxsspx vs32, o0, T1
  145. stxsspx vs33, o4, T1
  146. addi T1, T1, 8
  147. stxsspx vs34, o0, T1
  148. stxsspx vs35, o4, T1
  149. .endm
  150. /**********************************************************************************************
  151. * Macros for N=2 and M=1
  152. **********************************************************************************************/
  153. .macro COPY_2x1
  154. lxsspx vs32, o0, A0
  155. lxsspx vs33, o0, A1
  156. mr T1, BO
  157. stxsspx vs32, o0, T1
  158. stxsspx vs33, o4, T1
  159. .endm
  160. /**********************************************************************************************
  161. * Macros for N=1 and M=8
  162. **********************************************************************************************/
  163. .macro COPY_1x8
  164. lxvw4x vs32, o0, A0
  165. lxvw4x vs33, o16, A0
  166. mr T1, BO
  167. stxvw4x vs32, o0, T1
  168. stxvw4x vs33, o16, T1
  169. .endm
  170. /**********************************************************************************************
  171. * Macros for N=1 and M=4
  172. **********************************************************************************************/
  173. .macro COPY_1x4
  174. lxvw4x vs32, o0, A0
  175. mr T1, BO
  176. stxvw4x vs32, o0, T1
  177. .endm
  178. /**********************************************************************************************
  179. * Macros for N=1 and M=2
  180. **********************************************************************************************/
  181. .macro COPY_1x2
  182. lxsspx vs32, o0, A0
  183. lxsspx vs33, o4, A0
  184. mr T1, BO
  185. stxsspx vs32, o0, T1
  186. stxsspx vs33, o4, T1
  187. .endm
  188. /**********************************************************************************************
  189. * Macros for N=1 and M=1
  190. **********************************************************************************************/
  191. .macro COPY_1x1
  192. lxsspx vs32, o0, A0
  193. mr T1, BO
  194. stxsspx vs32, o0, T1
  195. .endm