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.

zhemm_utcopy_4.c 6.7 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242
  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 *ao1, *ao2, *ao3, *ao4;
  44. lda *= 2;
  45. js = (n >> 2);
  46. while (js > 0){
  47. offset = posX - posY;
  48. if (offset > 0) ao1 = a + posY * 2 + (posX + 0) * lda; else ao1 = a + (posX + 0) * 2 + posY * lda;
  49. if (offset > -1) ao2 = a + posY * 2 + (posX + 1) * lda; else ao2 = a + (posX + 1) * 2 + posY * lda;
  50. if (offset > -2) ao3 = a + posY * 2 + (posX + 2) * lda; else ao3 = a + (posX + 2) * 2 + posY * lda;
  51. if (offset > -3) ao4 = a + posY * 2 + (posX + 3) * lda; else ao4 = a + (posX + 3) * 2 + posY * lda;
  52. i = m;
  53. while (i > 0) {
  54. data01 = *(ao1 + 0);
  55. data02 = *(ao1 + 1);
  56. data03 = *(ao2 + 0);
  57. data04 = *(ao2 + 1);
  58. data05 = *(ao3 + 0);
  59. data06 = *(ao3 + 1);
  60. data07 = *(ao4 + 0);
  61. data08 = *(ao4 + 1);
  62. if (offset > 0) ao1 += 2; else ao1 += lda;
  63. if (offset > -1) ao2 += 2; else ao2 += lda;
  64. if (offset > -2) ao3 += 2; else ao3 += lda;
  65. if (offset > -3) ao4 += 2; else ao4 += lda;
  66. if (offset > 0) {
  67. b[ 0] = data01;
  68. b[ 1] = -data02;
  69. b[ 2] = data03;
  70. b[ 3] = -data04;
  71. b[ 4] = data05;
  72. b[ 5] = -data06;
  73. b[ 6] = data07;
  74. b[ 7] = -data08;
  75. } else
  76. if (offset < -3) {
  77. b[ 0] = data01;
  78. b[ 1] = data02;
  79. b[ 2] = data03;
  80. b[ 3] = data04;
  81. b[ 4] = data05;
  82. b[ 5] = data06;
  83. b[ 6] = data07;
  84. b[ 7] = data08;
  85. } else {
  86. switch (offset) {
  87. case 0 :
  88. b[ 0] = data01;
  89. b[ 1] = ZERO;
  90. b[ 2] = data03;
  91. b[ 3] = -data04;
  92. b[ 4] = data05;
  93. b[ 5] = -data06;
  94. b[ 6] = data07;
  95. b[ 7] = -data08;
  96. break;
  97. case -1 :
  98. b[ 0] = data01;
  99. b[ 1] = data02;
  100. b[ 2] = data03;
  101. b[ 3] = ZERO;
  102. b[ 4] = data05;
  103. b[ 5] = -data06;
  104. b[ 6] = data07;
  105. b[ 7] = -data08;
  106. break;
  107. case -2 :
  108. b[ 0] = data01;
  109. b[ 1] = data02;
  110. b[ 2] = data03;
  111. b[ 3] = data04;
  112. b[ 4] = data05;
  113. b[ 5] = ZERO;
  114. b[ 6] = data07;
  115. b[ 7] = -data08;
  116. break;
  117. case -3 :
  118. b[ 0] = data01;
  119. b[ 1] = data02;
  120. b[ 2] = data03;
  121. b[ 3] = data04;
  122. b[ 4] = data05;
  123. b[ 5] = data06;
  124. b[ 6] = data07;
  125. b[ 7] = ZERO;
  126. break;
  127. }
  128. }
  129. b += 8;
  130. offset --;
  131. i --;
  132. }
  133. posX += 4;
  134. js --;
  135. }
  136. if (n & 2) {
  137. offset = posX - posY;
  138. if (offset > 0) ao1 = a + posY * 2 + (posX + 0) * lda; else ao1 = a + (posX + 0) * 2 + posY * lda;
  139. if (offset > -1) ao2 = a + posY * 2 + (posX + 1) * lda; else ao2 = a + (posX + 1) * 2 + posY * lda;
  140. i = m;
  141. while (i > 0) {
  142. data01 = *(ao1 + 0);
  143. data02 = *(ao1 + 1);
  144. data03 = *(ao2 + 0);
  145. data04 = *(ao2 + 1);
  146. if (offset > 0) ao1 += 2; else ao1 += lda;
  147. if (offset > -1) ao2 += 2; else ao2 += lda;
  148. if (offset > 0) {
  149. b[ 0] = data01;
  150. b[ 1] = -data02;
  151. b[ 2] = data03;
  152. b[ 3] = -data04;
  153. } else
  154. if (offset < -1) {
  155. b[ 0] = data01;
  156. b[ 1] = data02;
  157. b[ 2] = data03;
  158. b[ 3] = data04;
  159. } else {
  160. switch (offset) {
  161. case 0 :
  162. b[ 0] = data01;
  163. b[ 1] = ZERO;
  164. b[ 2] = data03;
  165. b[ 3] = -data04;
  166. break;
  167. case -1 :
  168. b[ 0] = data01;
  169. b[ 1] = data02;
  170. b[ 2] = data03;
  171. b[ 3] = ZERO;
  172. break;
  173. }
  174. }
  175. b += 4;
  176. offset --;
  177. i --;
  178. }
  179. posX += 2;
  180. }
  181. if (n & 1) {
  182. offset = posX - posY;
  183. if (offset > 0) ao1 = a + posY * 2 + (posX + 0) * lda; else ao1 = a + (posX + 0) * 2 + posY * lda;
  184. i = m;
  185. while (i > 0) {
  186. data01 = *(ao1 + 0);
  187. data02 = *(ao1 + 1);
  188. if (offset > 0) ao1 += 2; else ao1 += lda;
  189. if (offset > 0) {
  190. b[ 0] = data01;
  191. b[ 1] = -data02;
  192. } else
  193. if (offset < 0) {
  194. b[ 0] = data01;
  195. b[ 1] = data02;
  196. } else {
  197. b[ 0] = data01;
  198. b[ 1] = ZERO;
  199. }
  200. b += 2;
  201. offset --;
  202. i --;
  203. }
  204. }
  205. return 0;
  206. }