memcpy_64.S 4.1 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225
  1. /*
  2. * Copyright (C) 2002 Paul Mackerras, IBM Corp.
  3. *
  4. * This program is free software; you can redistribute it and/or
  5. * modify it under the terms of the GNU General Public License
  6. * as published by the Free Software Foundation; either version
  7. * 2 of the License, or (at your option) any later version.
  8. */
  9. #include <asm/processor.h>
  10. #include <asm/ppc_asm.h>
  11. #include <asm/export.h>
  12. .align 7
  13. _GLOBAL_TOC(memcpy)
  14. BEGIN_FTR_SECTION
  15. #ifdef __LITTLE_ENDIAN__
  16. cmpdi cr7,r5,0
  17. #else
  18. std r3,-STACKFRAMESIZE+STK_REG(R31)(r1) /* save destination pointer for return value */
  19. #endif
  20. FTR_SECTION_ELSE
  21. #ifdef CONFIG_PPC_BOOK3S_64
  22. #ifndef SELFTEST
  23. b memcpy_power7
  24. #endif
  25. #endif
  26. ALT_FTR_SECTION_END_IFCLR(CPU_FTR_VMX_COPY)
  27. #ifdef __LITTLE_ENDIAN__
  28. /* dumb little-endian memcpy that will get replaced at runtime */
  29. addi r9,r3,-1
  30. addi r4,r4,-1
  31. beqlr cr7
  32. mtctr r5
  33. 1: lbzu r10,1(r4)
  34. stbu r10,1(r9)
  35. bdnz 1b
  36. blr
  37. #else
  38. PPC_MTOCRF(0x01,r5)
  39. cmpldi cr1,r5,16
  40. neg r6,r3 # LS 3 bits = # bytes to 8-byte dest bdry
  41. andi. r6,r6,7
  42. dcbt 0,r4
  43. blt cr1,.Lshort_copy
  44. /* Below we want to nop out the bne if we're on a CPU that has the
  45. CPU_FTR_UNALIGNED_LD_STD bit set and the CPU_FTR_CP_USE_DCBTZ bit
  46. cleared.
  47. At the time of writing the only CPU that has this combination of bits
  48. set is Power6. */
  49. BEGIN_FTR_SECTION
  50. nop
  51. FTR_SECTION_ELSE
  52. bne .Ldst_unaligned
  53. ALT_FTR_SECTION_END(CPU_FTR_UNALIGNED_LD_STD | CPU_FTR_CP_USE_DCBTZ, \
  54. CPU_FTR_UNALIGNED_LD_STD)
  55. .Ldst_aligned:
  56. addi r3,r3,-16
  57. BEGIN_FTR_SECTION
  58. andi. r0,r4,7
  59. bne .Lsrc_unaligned
  60. END_FTR_SECTION_IFCLR(CPU_FTR_UNALIGNED_LD_STD)
  61. srdi r7,r5,4
  62. ld r9,0(r4)
  63. addi r4,r4,-8
  64. mtctr r7
  65. andi. r5,r5,7
  66. bf cr7*4+0,2f
  67. addi r3,r3,8
  68. addi r4,r4,8
  69. mr r8,r9
  70. blt cr1,3f
  71. 1: ld r9,8(r4)
  72. std r8,8(r3)
  73. 2: ldu r8,16(r4)
  74. stdu r9,16(r3)
  75. bdnz 1b
  76. 3: std r8,8(r3)
  77. beq 3f
  78. addi r3,r3,16
  79. .Ldo_tail:
  80. bf cr7*4+1,1f
  81. lwz r9,8(r4)
  82. addi r4,r4,4
  83. stw r9,0(r3)
  84. addi r3,r3,4
  85. 1: bf cr7*4+2,2f
  86. lhz r9,8(r4)
  87. addi r4,r4,2
  88. sth r9,0(r3)
  89. addi r3,r3,2
  90. 2: bf cr7*4+3,3f
  91. lbz r9,8(r4)
  92. stb r9,0(r3)
  93. 3: ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1) /* return dest pointer */
  94. blr
  95. .Lsrc_unaligned:
  96. srdi r6,r5,3
  97. addi r5,r5,-16
  98. subf r4,r0,r4
  99. srdi r7,r5,4
  100. sldi r10,r0,3
  101. cmpdi cr6,r6,3
  102. andi. r5,r5,7
  103. mtctr r7
  104. subfic r11,r10,64
  105. add r5,r5,r0
  106. bt cr7*4+0,0f
  107. ld r9,0(r4) # 3+2n loads, 2+2n stores
  108. ld r0,8(r4)
  109. sld r6,r9,r10
  110. ldu r9,16(r4)
  111. srd r7,r0,r11
  112. sld r8,r0,r10
  113. or r7,r7,r6
  114. blt cr6,4f
  115. ld r0,8(r4)
  116. # s1<< in r8, d0=(s0<<|s1>>) in r7, s3 in r0, s2 in r9, nix in r6 & r12
  117. b 2f
  118. 0: ld r0,0(r4) # 4+2n loads, 3+2n stores
  119. ldu r9,8(r4)
  120. sld r8,r0,r10
  121. addi r3,r3,-8
  122. blt cr6,5f
  123. ld r0,8(r4)
  124. srd r12,r9,r11
  125. sld r6,r9,r10
  126. ldu r9,16(r4)
  127. or r12,r8,r12
  128. srd r7,r0,r11
  129. sld r8,r0,r10
  130. addi r3,r3,16
  131. beq cr6,3f
  132. # d0=(s0<<|s1>>) in r12, s1<< in r6, s2>> in r7, s2<< in r8, s3 in r9
  133. 1: or r7,r7,r6
  134. ld r0,8(r4)
  135. std r12,8(r3)
  136. 2: srd r12,r9,r11
  137. sld r6,r9,r10
  138. ldu r9,16(r4)
  139. or r12,r8,r12
  140. stdu r7,16(r3)
  141. srd r7,r0,r11
  142. sld r8,r0,r10
  143. bdnz 1b
  144. 3: std r12,8(r3)
  145. or r7,r7,r6
  146. 4: std r7,16(r3)
  147. 5: srd r12,r9,r11
  148. or r12,r8,r12
  149. std r12,24(r3)
  150. beq 4f
  151. cmpwi cr1,r5,8
  152. addi r3,r3,32
  153. sld r9,r9,r10
  154. ble cr1,6f
  155. ld r0,8(r4)
  156. srd r7,r0,r11
  157. or r9,r7,r9
  158. 6:
  159. bf cr7*4+1,1f
  160. rotldi r9,r9,32
  161. stw r9,0(r3)
  162. addi r3,r3,4
  163. 1: bf cr7*4+2,2f
  164. rotldi r9,r9,16
  165. sth r9,0(r3)
  166. addi r3,r3,2
  167. 2: bf cr7*4+3,3f
  168. rotldi r9,r9,8
  169. stb r9,0(r3)
  170. 3: ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1) /* return dest pointer */
  171. blr
  172. .Ldst_unaligned:
  173. PPC_MTOCRF(0x01,r6) # put #bytes to 8B bdry into cr7
  174. subf r5,r6,r5
  175. li r7,0
  176. cmpldi cr1,r5,16
  177. bf cr7*4+3,1f
  178. lbz r0,0(r4)
  179. stb r0,0(r3)
  180. addi r7,r7,1
  181. 1: bf cr7*4+2,2f
  182. lhzx r0,r7,r4
  183. sthx r0,r7,r3
  184. addi r7,r7,2
  185. 2: bf cr7*4+1,3f
  186. lwzx r0,r7,r4
  187. stwx r0,r7,r3
  188. 3: PPC_MTOCRF(0x01,r5)
  189. add r4,r6,r4
  190. add r3,r6,r3
  191. b .Ldst_aligned
  192. .Lshort_copy:
  193. bf cr7*4+0,1f
  194. lwz r0,0(r4)
  195. lwz r9,4(r4)
  196. addi r4,r4,8
  197. stw r0,0(r3)
  198. stw r9,4(r3)
  199. addi r3,r3,8
  200. 1: bf cr7*4+1,2f
  201. lwz r0,0(r4)
  202. addi r4,r4,4
  203. stw r0,0(r3)
  204. addi r3,r3,4
  205. 2: bf cr7*4+2,3f
  206. lhz r0,0(r4)
  207. addi r4,r4,2
  208. sth r0,0(r3)
  209. addi r3,r3,2
  210. 3: bf cr7*4+3,4f
  211. lbz r0,0(r4)
  212. stb r0,0(r3)
  213. 4: ld r3,-STACKFRAMESIZE+STK_REG(R31)(r1) /* return dest pointer */
  214. blr
  215. #endif
  216. EXPORT_SYMBOL(memcpy)