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.

zsymm_lcopy_8.c 7.5 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224
  1. /*********************************************************************/
  2. /* Copyright 2009, 2010 The University of Texas at Austin. */
  3. /* All rights reserved. */
  4. /* */
  5. /* Redistribution and use in source and binary forms, with or */
  6. /* without modification, are permitted provided that the following */
  7. /* conditions are met: */
  8. /* */
  9. /* 1. Redistributions of source code must retain the above */
  10. /* copyright notice, this list of conditions and the following */
  11. /* disclaimer. */
  12. /* */
  13. /* 2. Redistributions in binary form must reproduce the above */
  14. /* copyright notice, this list of conditions and the following */
  15. /* disclaimer in the documentation and/or other materials */
  16. /* provided with the distribution. */
  17. /* */
  18. /* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
  19. /* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
  20. /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
  21. /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
  22. /* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
  23. /* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
  24. /* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
  25. /* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
  26. /* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
  27. /* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
  28. /* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
  29. /* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
  30. /* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
  31. /* POSSIBILITY OF SUCH DAMAGE. */
  32. /* */
  33. /* The views and conclusions contained in the software and */
  34. /* documentation are those of the authors and should not be */
  35. /* interpreted as representing official policies, either expressed */
  36. /* or implied, of The University of Texas at Austin. */
  37. /*********************************************************************/
  38. #include <stdio.h>
  39. #include "common.h"
  40. int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG posX, BLASLONG posY, FLOAT *b){
  41. BLASLONG i, js, offset;
  42. FLOAT data01, data02, data03, data04, data05, data06, data07, data08;
  43. FLOAT data09, data10, data11, data12, data13, data14, data15, data16;
  44. FLOAT *ao1, *ao2, *ao3, *ao4, *ao5, *ao6, *ao7, *ao8;
  45. lda *= 2;
  46. js = (n >> 3);
  47. while (js > 0){
  48. offset = posX - posY;
  49. if (offset > 0) ao1 = a + (posX + 0) * 2 + posY * lda; else ao1 = a + posY * 2 + (posX + 0) * lda;
  50. if (offset > -1) ao2 = a + (posX + 1) * 2 + posY * lda; else ao2 = a + posY * 2 + (posX + 1) * lda;
  51. if (offset > -2) ao3 = a + (posX + 2) * 2 + posY * lda; else ao3 = a + posY * 2 + (posX + 2) * lda;
  52. if (offset > -3) ao4 = a + (posX + 3) * 2 + posY * lda; else ao4 = a + posY * 2 + (posX + 3) * lda;
  53. if (offset > -4) ao5 = a + (posX + 4) * 2 + posY * lda; else ao5 = a + posY * 2 + (posX + 4) * lda;
  54. if (offset > -5) ao6 = a + (posX + 5) * 2 + posY * lda; else ao6 = a + posY * 2 + (posX + 5) * lda;
  55. if (offset > -6) ao7 = a + (posX + 6) * 2 + posY * lda; else ao7 = a + posY * 2 + (posX + 6) * lda;
  56. if (offset > -7) ao8 = a + (posX + 7) * 2 + posY * lda; else ao8 = a + posY * 2 + (posX + 7) * lda;
  57. i = m;
  58. while (i > 0) {
  59. data01 = *(ao1 + 0);
  60. data02 = *(ao1 + 1);
  61. data03 = *(ao2 + 0);
  62. data04 = *(ao2 + 1);
  63. data05 = *(ao3 + 0);
  64. data06 = *(ao3 + 1);
  65. data07 = *(ao4 + 0);
  66. data08 = *(ao4 + 1);
  67. data09 = *(ao5 + 0);
  68. data10 = *(ao5 + 1);
  69. data11 = *(ao6 + 0);
  70. data12 = *(ao6 + 1);
  71. data13 = *(ao7 + 0);
  72. data14 = *(ao7 + 1);
  73. data15 = *(ao8 + 0);
  74. data16 = *(ao8 + 1);
  75. if (offset > 0) ao1 += lda; else ao1 += 2;
  76. if (offset > -1) ao2 += lda; else ao2 += 2;
  77. if (offset > -2) ao3 += lda; else ao3 += 2;
  78. if (offset > -3) ao4 += lda; else ao4 += 2;
  79. if (offset > -4) ao5 += lda; else ao5 += 2;
  80. if (offset > -5) ao6 += lda; else ao6 += 2;
  81. if (offset > -6) ao7 += lda; else ao7 += 2;
  82. if (offset > -7) ao8 += lda; else ao8 += 2;
  83. b[ 0] = data01;
  84. b[ 1] = data02;
  85. b[ 2] = data03;
  86. b[ 3] = data04;
  87. b[ 4] = data05;
  88. b[ 5] = data06;
  89. b[ 6] = data07;
  90. b[ 7] = data08;
  91. b[ 8] = data09;
  92. b[ 9] = data10;
  93. b[10] = data11;
  94. b[11] = data12;
  95. b[12] = data13;
  96. b[13] = data14;
  97. b[14] = data15;
  98. b[15] = data16;
  99. b += 16;
  100. offset --;
  101. i --;
  102. }
  103. posX += 8;
  104. js --;
  105. }
  106. if (n & 4) {
  107. offset = posX - posY;
  108. if (offset > 0) ao1 = a + (posX + 0) * 2 + posY * lda; else ao1 = a + posY * 2 + (posX + 0) * lda;
  109. if (offset > -1) ao2 = a + (posX + 1) * 2 + posY * lda; else ao2 = a + posY * 2 + (posX + 1) * lda;
  110. if (offset > -2) ao3 = a + (posX + 2) * 2 + posY * lda; else ao3 = a + posY * 2 + (posX + 2) * lda;
  111. if (offset > -3) ao4 = a + (posX + 3) * 2 + posY * lda; else ao4 = a + posY * 2 + (posX + 3) * lda;
  112. i = m;
  113. while (i > 0) {
  114. data01 = *(ao1 + 0);
  115. data02 = *(ao1 + 1);
  116. data03 = *(ao2 + 0);
  117. data04 = *(ao2 + 1);
  118. data05 = *(ao3 + 0);
  119. data06 = *(ao3 + 1);
  120. data07 = *(ao4 + 0);
  121. data08 = *(ao4 + 1);
  122. if (offset > 0) ao1 += lda; else ao1 += 2;
  123. if (offset > -1) ao2 += lda; else ao2 += 2;
  124. if (offset > -2) ao3 += lda; else ao3 += 2;
  125. if (offset > -3) ao4 += lda; else ao4 += 2;
  126. b[ 0] = data01;
  127. b[ 1] = data02;
  128. b[ 2] = data03;
  129. b[ 3] = data04;
  130. b[ 4] = data05;
  131. b[ 5] = data06;
  132. b[ 6] = data07;
  133. b[ 7] = data08;
  134. b += 8;
  135. offset --;
  136. i --;
  137. }
  138. posX += 4;
  139. }
  140. if (n & 2) {
  141. offset = posX - posY;
  142. if (offset > 0) ao1 = a + (posX + 0) * 2 + posY * lda; else ao1 = a + posY * 2 + (posX + 0) * lda;
  143. if (offset > -1) ao2 = a + (posX + 1) * 2 + posY * lda; else ao2 = a + posY * 2 + (posX + 1) * lda;
  144. i = m;
  145. while (i > 0) {
  146. data01 = *(ao1 + 0);
  147. data02 = *(ao1 + 1);
  148. data03 = *(ao2 + 0);
  149. data04 = *(ao2 + 1);
  150. if (offset > 0) ao1 += lda; else ao1 += 2;
  151. if (offset > -1) ao2 += lda; else ao2 += 2;
  152. b[ 0] = data01;
  153. b[ 1] = data02;
  154. b[ 2] = data03;
  155. b[ 3] = data04;
  156. b += 4;
  157. offset --;
  158. i --;
  159. }
  160. posX += 2;
  161. }
  162. if (n & 1) {
  163. offset = posX - posY;
  164. if (offset > 0) ao1 = a + (posX + 0) * 2 + posY * lda; else ao1 = a + posY * 2 + (posX + 0) * lda;
  165. i = m;
  166. while (i > 0) {
  167. data01 = *(ao1 + 0);
  168. data02 = *(ao1 + 1);
  169. if (offset > 0) ao1 += lda; else ao1 += 2;
  170. b[ 0] = data01;
  171. b[ 1] = data02;
  172. b += 2;
  173. offset --;
  174. i --;
  175. }
  176. }
  177. return 0;
  178. }