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.

ztrmm_lncopy_2.c 5.9 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230
  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;
  42. BLASLONG X;
  43. FLOAT data01, data02, data03, data04;
  44. FLOAT data05, data06, data07, data08;
  45. FLOAT *ao1, *ao2;
  46. lda += lda;
  47. js = (n >> 1);
  48. if (js > 0){
  49. do {
  50. X = posX;
  51. if (posX <= posY) {
  52. ao1 = a + posY * 2 + (posX + 0) * lda;
  53. ao2 = a + posY * 2 + (posX + 1) * lda;
  54. } else {
  55. ao1 = a + posX * 2 + (posY + 0) * lda;
  56. ao2 = a + posX * 2 + (posY + 1) * lda;
  57. }
  58. i = (m >> 1);
  59. if (i > 0) {
  60. do {
  61. if (X > posY) {
  62. data01 = *(ao1 + 0);
  63. data02 = *(ao1 + 1);
  64. data03 = *(ao1 + 2);
  65. data04 = *(ao1 + 3);
  66. data05 = *(ao2 + 0);
  67. data06 = *(ao2 + 1);
  68. data07 = *(ao2 + 2);
  69. data08 = *(ao2 + 3);
  70. b[ 0] = data01;
  71. b[ 1] = data02;
  72. b[ 2] = data05;
  73. b[ 3] = data06;
  74. b[ 4] = data03;
  75. b[ 5] = data04;
  76. b[ 6] = data07;
  77. b[ 7] = data08;
  78. ao1 += 4;
  79. ao2 += 4;
  80. b += 8;
  81. } else
  82. if (X < posY) {
  83. ao1 += 2 * lda;
  84. ao2 += 2 * lda;
  85. b += 8;
  86. } else {
  87. #ifdef UNIT
  88. data03 = *(ao1 + 2);
  89. data04 = *(ao1 + 3);
  90. b[ 0] = ONE;
  91. b[ 1] = ZERO;
  92. b[ 2] = ZERO;
  93. b[ 3] = ZERO;
  94. b[ 4] = data03;
  95. b[ 5] = data04;
  96. b[ 6] = ONE;
  97. b[ 7] = ZERO;
  98. #else
  99. data01 = *(ao1 + 0);
  100. data02 = *(ao1 + 1);
  101. data03 = *(ao1 + 2);
  102. data04 = *(ao1 + 3);
  103. data07 = *(ao2 + 2);
  104. data08 = *(ao2 + 3);
  105. b[ 0] = data01;
  106. b[ 1] = data02;
  107. b[ 2] = ZERO;
  108. b[ 3] = ZERO;
  109. b[ 4] = data03;
  110. b[ 5] = data04;
  111. b[ 6] = data07;
  112. b[ 7] = data08;
  113. #endif
  114. ao1 += 4;
  115. ao2 += 4;
  116. b += 8;
  117. }
  118. X += 2;
  119. i --;
  120. } while (i > 0);
  121. }
  122. if (m & 1) {
  123. if (X > posY) {
  124. data01 = *(ao1 + 0);
  125. data02 = *(ao1 + 1);
  126. data03 = *(ao2 + 0);
  127. data04 = *(ao2 + 1);
  128. b[ 0] = data01;
  129. b[ 1] = data02;
  130. b[ 2] = data03;
  131. b[ 3] = data04;
  132. /* ao1 += 2;
  133. ao2 += 2; */
  134. b += 4;
  135. } else
  136. if (X < posY) {
  137. // ao1 += lda;
  138. b += 4;
  139. } else {
  140. #ifdef UNIT
  141. b[ 0] = ONE;
  142. b[ 1] = ZERO;
  143. b[ 0] = ZERO;
  144. b[ 1] = ZERO;
  145. #else
  146. data01 = *(ao1 + 0);
  147. data02 = *(ao1 + 1);
  148. b[ 0] = data01;
  149. b[ 1] = data02;
  150. b[ 2] = ZERO;
  151. b[ 3] = ZERO;
  152. #endif
  153. b += 4;
  154. }
  155. }
  156. posY += 2;
  157. js --;
  158. } while (js > 0);
  159. } /* End of main loop */
  160. if (n & 1){
  161. X = posX;
  162. if (posX <= posY) {
  163. ao1 = a + posY * 2 + (posX + 0) * lda;
  164. } else {
  165. ao1 = a + posX * 2 + (posY + 0) * lda;
  166. }
  167. i = m;
  168. if (i > 0) {
  169. do {
  170. if (X > posY) {
  171. data01 = *(ao1 + 0);
  172. data02 = *(ao1 + 1);
  173. b[ 0] = data01;
  174. b[ 1] = data02;
  175. b += 2;
  176. ao1 += 2;
  177. } else
  178. if (X < posY) {
  179. b += 2;
  180. ao1 += lda;
  181. } else {
  182. #ifdef UNIT
  183. b[ 0] = ONE;
  184. b[ 1] = ZERO;
  185. #else
  186. data01 = *(ao1 + 0);
  187. data02 = *(ao1 + 1);
  188. b[ 0] = data01;
  189. b[ 1] = data02;
  190. #endif
  191. b += 2;
  192. ao1 += 2;
  193. }
  194. X ++;
  195. i --;
  196. } while (i > 0);
  197. }
  198. // posY += 1;
  199. }
  200. return 0;
  201. }