round_wmt.S 4.2 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147
  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 "round_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 round
  59. .text
  60. .align 16,0x90
  61. .globl round
  62. round:
  63. # parameter 1: 8 + %ebp
  64. ..B2.1:
  65. ..L3:
  66. ..B2.2:
  67. pushl %ebp
  68. movl %esp, %ebp
  69. subl $104, %esp
  70. movl %ebx, 40(%esp)
  71. call static_func
  72. movl %eax, %ebx
  73. movsd 112(%esp), %xmm0
  74. pextrw $3, %xmm0, %ecx
  75. movl $32752, %eax
  76. andl %ecx, %eax
  77. cmpl $16352, %eax
  78. jl .L_2TAG_PACKET_0.0.2
  79. cmpl $17200, %eax
  80. jge .L_2TAG_PACKET_1.0.2
  81. movq 16(%ebx), %xmm1
  82. andpd %xmm0, %xmm1
  83. orpd (%ebx), %xmm1
  84. addsd %xmm1, %xmm0
  85. movq 24(%ebx), %xmm1
  86. movq 32(%ebx), %xmm2
  87. andpd %xmm0, %xmm1
  88. psubd %xmm1, %xmm2
  89. psrlq $52, %xmm2
  90. psrlq %xmm2, %xmm0
  91. psllq %xmm2, %xmm0
  92. jmp .L_2TAG_PACKET_2.0.2
  93. .L_2TAG_PACKET_0.0.2:
  94. addsd %xmm0, %xmm0
  95. testl $32768, %ecx
  96. jne .L_2TAG_PACKET_3.0.2
  97. fldz
  98. jmp .L_2TAG_PACKET_4.0.2
  99. .L_2TAG_PACKET_3.0.2:
  100. fldl 40(%ebx)
  101. jmp .L_2TAG_PACKET_4.0.2
  102. .L_2TAG_PACKET_1.0.2:
  103. cmpl $32752, %eax
  104. je .L_2TAG_PACKET_5.0.2
  105. jmp .L_2TAG_PACKET_2.0.2
  106. .L_2TAG_PACKET_5.0.2:
  107. fldl 112(%esp)
  108. fldz
  109. faddp
  110. jmp .L_2TAG_PACKET_4.0.2
  111. .L_2TAG_PACKET_2.0.2:
  112. movsd %xmm0, 24(%esp)
  113. fldl 24(%esp)
  114. .L_2TAG_PACKET_4.0.2:
  115. movl 40(%esp), %ebx
  116. movl %ebp, %esp
  117. popl %ebp
  118. ret
  119. ..B2.3:
  120. .align 16,0x90
  121. .type round,@function
  122. .size round,.-round
  123. .data
  124. # -- End round
  125. .section .rodata, "a"
  126. .align 16
  127. .align 16
  128. static_const_table:
  129. .long 0
  130. .long 1071644672
  131. .long 0
  132. .long 1071644672
  133. .long 0
  134. .long 2147483648
  135. .long 0
  136. .long 2146435072
  137. .long 0
  138. .long 1127219200
  139. .long 0
  140. .long 2147483648
  141. .type static_const_table,@object
  142. .size static_const_table,48
  143. .data
  144. .section .note.GNU-stack, ""
  145. # End