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_kernel_power9.S 6.4 kB

7 years ago
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286
  1. /***************************************************************************
  2. Copyright (c) 2013-2019, 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. #define ASSEMBLER
  28. #include "common.h"
  29. #include "def_vsx.h"
  30. #define LOAD ld
  31. #define STACKSIZE (512 )
  32. #define M r3
  33. #define N r4
  34. #define K r5
  35. #define A r7
  36. #define B r8
  37. #define C r9
  38. #define LDC r10
  39. #define OFFSET r6
  40. #define alpha_r vs20
  41. #define save_permute_1 vs21
  42. #define save_permute_2 vs22
  43. #define permute_mask vs23
  44. #define o0 0
  45. #define T1 r11
  46. #define T2 r12
  47. #define T3 r14
  48. #define T4 r15
  49. #define T5 r16
  50. #define T6 r17
  51. #define L r18
  52. #define T7 r19
  53. #define T8 r20
  54. #define TEMP_REG r21
  55. #define I r22
  56. #define J r23
  57. #define AO r24
  58. #define BO r25
  59. #define CO r26
  60. #define T9 r27
  61. #define T10 r28
  62. #define T11 r29
  63. #define T12 r30
  64. #define T13 r31
  65. #include "sgemm_macros_power9.S"
  66. .equ perm_const1, 0x0405060700010203
  67. .equ perm_const2, 0x0c0d0e0f08090a0b
  68. .equ save_permute_11, 0x1415161718191a1b
  69. .equ save_permute_12, 0x0405060708090a0b
  70. .equ save_permute_21, 0x101112131c1d1e1f
  71. .equ save_permute_22, 0x000102030c0d0e0f
  72. #ifndef NEEDPARAM
  73. PROLOGUE
  74. PROFCODE
  75. addi SP, SP, -STACKSIZE
  76. li r0, 0
  77. stfd f14, 0(SP)
  78. stfd f15, 8(SP)
  79. stfd f16, 16(SP)
  80. stfd f17, 24(SP)
  81. stfd f18, 32(SP)
  82. stfd f19, 40(SP)
  83. stfd f20, 48(SP)
  84. stfd f21, 56(SP)
  85. stfd f22, 64(SP)
  86. stfd f23, 72(SP)
  87. stfd f24, 80(SP)
  88. stfd f25, 88(SP)
  89. stfd f26, 96(SP)
  90. stfd f27, 104(SP)
  91. stfd f28, 112(SP)
  92. stfd f29, 120(SP)
  93. stfd f30, 128(SP)
  94. stfd f31, 136(SP)
  95. std r31, 144(SP)
  96. std r30, 152(SP)
  97. std r29, 160(SP)
  98. std r28, 168(SP)
  99. std r27, 176(SP)
  100. std r26, 184(SP)
  101. std r25, 192(SP)
  102. std r24, 200(SP)
  103. std r23, 208(SP)
  104. std r22, 216(SP)
  105. std r21, 224(SP)
  106. std r20, 232(SP)
  107. std r19, 240(SP)
  108. std r18, 248(SP)
  109. std r17, 256(SP)
  110. std r16, 264(SP)
  111. std r15, 272(SP)
  112. std r14, 280(SP)
  113. stxv v20, 288(SP)
  114. stxv v21, 304(SP)
  115. stxv v22, 320(SP)
  116. stxv v23, 336(SP)
  117. stxv v24, 352(SP)
  118. stxv v25, 368(SP)
  119. stxv v26, 384(SP)
  120. stxv v27, 400(SP)
  121. stxv v28, 416(SP)
  122. stxv v29, 432(SP)
  123. stxv v30, 448(SP)
  124. stxv v31, 464(SP)
  125. #if defined(TRMMKERNEL)
  126. ld OFFSET, FRAMESLOT(0) + STACKSIZE(SP)
  127. #endif
  128. slwi LDC, LDC, 2
  129. /* cmpwi cr0, M, 0
  130. ble .L999_H1
  131. cmpwi cr0, N, 0
  132. ble .L999_H1
  133. cmpwi cr0, K, 0
  134. ble .L999_H1
  135. */
  136. /*alpha is stored in f1. convert to single and splat*/
  137. xscvdpspn alpha_r,vs1
  138. xxspltw alpha_r,alpha_r,0
  139. /*load reverse permute mask for big endian
  140. uint128 = 0xc0d0e0f08090a0b0405060700010203
  141. */
  142. lis T2, perm_const2@highest
  143. ori T2, T2, perm_const2@higher
  144. rldicr T2, T2, 32, 31
  145. oris T2, T2, perm_const2@h
  146. ori T2, T2, perm_const2@l
  147. lis T1, perm_const1@highest
  148. ori T1, T1, perm_const1@higher
  149. rldicr T1, T1, 32, 31
  150. oris T1, T1, perm_const1@h
  151. ori T1, T1, perm_const1@l
  152. mtvsrdd permute_mask,T2,T1
  153. lis T2, save_permute_12@highest
  154. ori T2, T2, save_permute_12@higher
  155. rldicr T2, T2, 32, 31
  156. oris T2, T2, save_permute_12@h
  157. ori T2, T2, save_permute_12@l
  158. lis T1, save_permute_11@highest
  159. ori T1, T1, save_permute_11@higher
  160. rldicr T1, T1, 32, 31
  161. oris T1, T1, save_permute_11@h
  162. ori T1, T1, save_permute_11@l
  163. mtvsrdd save_permute_1,T2,T1
  164. lis T2, save_permute_22@highest
  165. ori T2, T2, save_permute_22@higher
  166. rldicr T2, T2, 32, 31
  167. oris T2, T2, save_permute_22@h
  168. ori T2, T2, save_permute_22@l
  169. lis T1, save_permute_21@highest
  170. ori T1, T1, save_permute_21@higher
  171. rldicr T1, T1, 32, 31
  172. oris T1, T1, save_permute_21@h
  173. ori T1, T1, save_permute_21@l
  174. mtvsrdd save_permute_2,T2,T1
  175. #include "sgemm_logic_power9.S"
  176. .L999:
  177. addi r3, 0, 0
  178. lfd f14, 0(SP)
  179. lfd f15, 8(SP)
  180. lfd f16, 16(SP)
  181. lfd f17, 24(SP)
  182. lfd f18, 32(SP)
  183. lfd f19, 40(SP)
  184. lfd f20, 48(SP)
  185. lfd f21, 56(SP)
  186. lfd f22, 64(SP)
  187. lfd f23, 72(SP)
  188. lfd f24, 80(SP)
  189. lfd f25, 88(SP)
  190. lfd f26, 96(SP)
  191. lfd f27, 104(SP)
  192. lfd f28, 112(SP)
  193. lfd f29, 120(SP)
  194. lfd f30, 128(SP)
  195. lfd f31, 136(SP)
  196. ld r31, 144(SP)
  197. ld r30, 152(SP)
  198. ld r29, 160(SP)
  199. ld r28, 168(SP)
  200. ld r27, 176(SP)
  201. ld r26, 184(SP)
  202. ld r25, 192(SP)
  203. ld r24, 200(SP)
  204. ld r23, 208(SP)
  205. ld r22, 216(SP)
  206. ld r21, 224(SP)
  207. ld r20, 232(SP)
  208. ld r19, 240(SP)
  209. ld r18, 248(SP)
  210. ld r17, 256(SP)
  211. ld r16, 264(SP)
  212. ld r15, 272(SP)
  213. ld r14, 280(SP)
  214. lxv v20, 288(SP)
  215. lxv v21, 304(SP)
  216. lxv v22, 320(SP)
  217. lxv v23, 336(SP)
  218. lxv v24, 352(SP)
  219. lxv v25, 368(SP)
  220. lxv v26, 384(SP)
  221. lxv v27, 400(SP)
  222. lxv v28, 416(SP)
  223. lxv v29, 432(SP)
  224. lxv v30, 448(SP)
  225. lxv v31, 464(SP)
  226. addi SP, SP, STACKSIZE
  227. blr
  228. EPILOGUE
  229. #endif