fdim_wmt.S 5.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181
  1. /*
  2. * Math library
  3. *
  4. * Copyright (C) 2016 Intel Corporation. All rights reserved.
  5. *
  6. * Redistribution and use in source and binary forms, with or without
  7. * modification, are permitted provided that the following conditions
  8. * are met:
  9. *
  10. * * Redistributions of source code must retain the above copyright
  11. * notice, this list of conditions and the following disclaimer.
  12. * * Redistributions in binary form must reproduce the above copyright
  13. * notice, this list of conditions and the following disclaimer in
  14. * the documentation and/or other materials provided with the
  15. * distribution.
  16. * * Neither the name of Intel Corporation nor the names of its
  17. * contributors may be used to endorse or promote products derived
  18. * from this software without specific prior written permission.
  19. *
  20. * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  21. * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  22. * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  23. * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  24. * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  25. * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  26. * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  27. * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  28. * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  29. * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  30. * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  31. *
  32. *
  33. * Author Name <jingwei.zhang@intel.com>
  34. * History:
  35. * 03-14-2016 Initial version. numerics svn rev. 12864
  36. */
  37. .file "fdim_wmt.c"
  38. .text
  39. ..TXTST0:
  40. # -- Begin static_func
  41. .text
  42. .align 16,0x90
  43. static_func:
  44. ..B1.1:
  45. ..L1:
  46. call ..L2
  47. ..L2:
  48. popl %eax
  49. lea _GLOBAL_OFFSET_TABLE_+[. - ..L2](%eax), %eax
  50. lea static_const_table@GOTOFF(%eax), %eax
  51. ret
  52. .align 16,0x90
  53. .type static_func,@function
  54. .size static_func,.-static_func
  55. .data
  56. # -- End static_func
  57. .text
  58. # -- Begin fdim
  59. .text
  60. .align 16,0x90
  61. .globl fdim
  62. fdim:
  63. # parameter 1: 8 + %ebp
  64. # parameter 2: 16 + %ebp
  65. ..B2.1:
  66. ..L3:
  67. ..B2.2:
  68. pushl %ebp
  69. movl %esp, %ebp
  70. subl $104, %esp
  71. movl %ebx, 48(%esp)
  72. call static_func
  73. movl %eax, %ebx
  74. stmxcsr 8(%esp)
  75. movsd 112(%esp), %xmm0
  76. movsd 120(%esp), %xmm1
  77. movapd %xmm0, %xmm2
  78. ucomisd %xmm0, %xmm1
  79. movl 8(%esp), %eax
  80. jp .L_2TAG_PACKET_0.0.3
  81. testl $8, %eax
  82. cmpnlesd %xmm1, %xmm0
  83. andpd %xmm0, %xmm1
  84. andpd %xmm2, %xmm0
  85. jne .L_2TAG_PACKET_1.0.3
  86. subsd %xmm1, %xmm0
  87. andpd (%ebx), %xmm0
  88. pextrw $3, %xmm0, %eax
  89. cmpl $32751, %eax
  90. jb .L_2TAG_PACKET_2.0.3
  91. stmxcsr 12(%esp)
  92. movl 12(%esp), %eax
  93. testl $8, %eax
  94. jne .L_2TAG_PACKET_3.0.3
  95. jmp .L_2TAG_PACKET_2.0.3
  96. .L_2TAG_PACKET_1.0.3:
  97. andl $-9, %eax
  98. movl %eax, 12(%esp)
  99. ldmxcsr 12(%esp)
  100. cmpnlesd %xmm1, %xmm0
  101. andpd %xmm0, %xmm1
  102. andpd %xmm2, %xmm0
  103. subsd %xmm1, %xmm0
  104. andpd (%ebx), %xmm0
  105. stmxcsr 12(%esp)
  106. movl 12(%esp), %eax
  107. testl $8, %eax
  108. jne .L_2TAG_PACKET_3.0.3
  109. ldmxcsr 8(%esp)
  110. jmp .L_2TAG_PACKET_2.0.3
  111. .L_2TAG_PACKET_0.0.3:
  112. fldl 112(%esp)
  113. fldl 120(%esp)
  114. faddp %st, %st(1)
  115. jmp .L_2TAG_PACKET_4.0.3
  116. .L_2TAG_PACKET_3.0.3:
  117. movl $196, %edx
  118. .L_2TAG_PACKET_5.0.3:
  119. movsd %xmm0, (%esp)
  120. movsd 112(%esp), %xmm0
  121. movsd 120(%esp), %xmm1
  122. subl $32, %esp
  123. lea 112(%esp), %eax
  124. movl %eax, (%esp)
  125. lea 120(%esp), %eax
  126. movl %eax, 4(%esp)
  127. lea 32(%esp), %eax
  128. movl %eax, 8(%esp)
  129. movl %edx, %eax
  130. movl %eax, 12(%esp)
  131. call __libm_error_support
  132. addl $32, %esp
  133. fldl (%esp)
  134. jmp .L_2TAG_PACKET_4.0.3
  135. .L_2TAG_PACKET_2.0.3:
  136. movsd %xmm0, 32(%esp)
  137. fldl 32(%esp)
  138. .L_2TAG_PACKET_4.0.3:
  139. movl 48(%esp), %ebx
  140. movl %ebp, %esp
  141. popl %ebp
  142. ret
  143. ..B2.3:
  144. .align 16,0x90
  145. .type fdim,@function
  146. .size fdim,.-fdim
  147. .data
  148. # -- End fdim
  149. .text
  150. # -- Begin __libm_error_support
  151. .text
  152. .align 16,0x90
  153. __libm_error_support:
  154. # parameter 1: 4 + %esp
  155. # parameter 2: 8 + %esp
  156. # parameter 3: 12 + %esp
  157. # parameter 4: 16 + %esp
  158. ..B3.1:
  159. ..L4:
  160. ret
  161. .align 16,0x90
  162. .type __libm_error_support,@function
  163. .size __libm_error_support,.-__libm_error_support
  164. .data
  165. # -- End __libm_error_support
  166. .section .rodata, "a"
  167. .align 16
  168. .align 16
  169. static_const_table:
  170. .long 4294967295
  171. .long 2147483647
  172. .long 4294967295
  173. .long 2147483647
  174. .type static_const_table,@object
  175. .size static_const_table,16
  176. .data
  177. .section .note.GNU-stack, ""
  178. # End