fdim_gen.S 4.0 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148
  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_gen.c"
  38. .text
  39. ..TXTST0:
  40. # -- Begin fdim
  41. .text
  42. .align 16,0x90
  43. .globl fdim
  44. fdim:
  45. # parameter 1: %xmm0
  46. # parameter 2: %xmm1
  47. ..B1.1:
  48. .cfi_startproc
  49. ..___tag_value_fdim.1:
  50. ..L2:
  51. movsd %xmm0, -16(%rsp)
  52. movsd %xmm1, -8(%rsp)
  53. movzwl -10(%rsp), %edx
  54. andl $32752, %edx
  55. movzwl -2(%rsp), %eax
  56. andl $32752, %eax
  57. shrl $4, %edx
  58. shrl $4, %eax
  59. cmpl $2047, %edx
  60. je ..B1.14
  61. ..B1.2:
  62. cmpl $2047, %eax
  63. je ..B1.11
  64. ..B1.3:
  65. movsd -16(%rsp), %xmm0
  66. movsd -8(%rsp), %xmm2
  67. comisd %xmm2, %xmm0
  68. jbe ..B1.10
  69. ..B1.4:
  70. cmpl $2047, %edx
  71. je ..B1.8
  72. ..B1.5:
  73. cmpl $2047, %eax
  74. je ..B1.8
  75. ..B1.6:
  76. cmpl $2046, %edx
  77. je ..B1.9
  78. ..B1.7:
  79. cmpl $2046, %eax
  80. je ..B1.9
  81. ..B1.8:
  82. subsd %xmm2, %xmm0
  83. ret
  84. ..B1.9:
  85. movsd .L_2il0floatpacket.0(%rip), %xmm1
  86. mulsd %xmm1, %xmm0
  87. mulsd %xmm1, %xmm2
  88. subsd %xmm2, %xmm0
  89. movsd %xmm0, -24(%rsp)
  90. movsd -24(%rsp), %xmm0
  91. mulsd .L_2il0floatpacket.1(%rip), %xmm0
  92. ret
  93. ..B1.10:
  94. lea _zeros(%rip), %rax
  95. movsd (%rax), %xmm0
  96. ret
  97. ..B1.11:
  98. testl $1048575, -4(%rsp)
  99. jne ..B1.13
  100. ..B1.12:
  101. cmpl $0, -8(%rsp)
  102. je ..B1.3
  103. ..B1.13:
  104. movsd -8(%rsp), %xmm0
  105. ret
  106. ..B1.14:
  107. testl $1048575, -12(%rsp)
  108. jne ..B1.16
  109. ..B1.15:
  110. cmpl $0, -16(%rsp)
  111. je ..B1.2
  112. ..B1.16:
  113. movsd -16(%rsp), %xmm0
  114. ret
  115. .align 16,0x90
  116. .cfi_endproc
  117. .type fdim,@function
  118. .size fdim,.-fdim
  119. .data
  120. # -- End fdim
  121. .section .rodata, "a"
  122. .align 8
  123. .align 8
  124. .L_2il0floatpacket.0:
  125. .long 0x00000000,0x3fe00000
  126. .type .L_2il0floatpacket.0,@object
  127. .size .L_2il0floatpacket.0,8
  128. .align 8
  129. .L_2il0floatpacket.1:
  130. .long 0x00000000,0x40000000
  131. .type .L_2il0floatpacket.1,@object
  132. .size .L_2il0floatpacket.1,8
  133. .align 4
  134. _zeros:
  135. .long 0
  136. .long 0
  137. .long 0
  138. .long 2147483648
  139. .type _zeros,@object
  140. .size _zeros,16
  141. .data
  142. .section .note.GNU-stack, ""
  143. // -- Begin DWARF2 SEGMENT .eh_frame
  144. .section .eh_frame,"a",@progbits
  145. .eh_frame_seg:
  146. .align 1
  147. # End