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.

trsm_lncopy_16.c 6.7 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271
  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 "common.h"
  39. #ifndef UNIT
  40. #define INV(a) (ONE / (a))
  41. #else
  42. #define INV(a) (ONE)
  43. #endif
  44. int CNAME(BLASLONG m, BLASLONG n, FLOAT *a, BLASLONG lda, BLASLONG offset, FLOAT *b){
  45. BLASLONG i, ii, j, jj, k;
  46. FLOAT *a1, *a2, *a3, *a4, *a5, *a6, *a7, *a8;
  47. FLOAT *a9, *a10, *a11, *a12, *a13, *a14, *a15, *a16;
  48. jj = offset;
  49. j = (n >> 4);
  50. while (j > 0){
  51. a1 = a + 0 * lda;
  52. a2 = a + 1 * lda;
  53. a3 = a + 2 * lda;
  54. a4 = a + 3 * lda;
  55. a5 = a + 4 * lda;
  56. a6 = a + 5 * lda;
  57. a7 = a + 6 * lda;
  58. a8 = a + 7 * lda;
  59. a9 = a + 8 * lda;
  60. a10 = a + 9 * lda;
  61. a11 = a + 10 * lda;
  62. a12 = a + 11 * lda;
  63. a13 = a + 12 * lda;
  64. a14 = a + 13 * lda;
  65. a15 = a + 14 * lda;
  66. a16 = a + 15 * lda;
  67. a += 16 * lda;
  68. ii = 0;
  69. for (i = 0; i < m; i++) {
  70. if ((ii >= jj ) && (ii - jj < 16)) {
  71. for (k = 0; k < ii - jj; k ++) {
  72. *(b + k) = *(a1 + k * lda);
  73. }
  74. *(b + ii - jj) = INV(*(a1 + (ii - jj) * lda));
  75. }
  76. if (ii - jj >= 16) {
  77. *(b + 0) = *(a1 + 0);
  78. *(b + 1) = *(a2 + 0);
  79. *(b + 2) = *(a3 + 0);
  80. *(b + 3) = *(a4 + 0);
  81. *(b + 4) = *(a5 + 0);
  82. *(b + 5) = *(a6 + 0);
  83. *(b + 6) = *(a7 + 0);
  84. *(b + 7) = *(a8 + 0);
  85. *(b + 8) = *(a9 + 0);
  86. *(b + 9) = *(a10 + 0);
  87. *(b + 10) = *(a11 + 0);
  88. *(b + 11) = *(a12 + 0);
  89. *(b + 12) = *(a13 + 0);
  90. *(b + 13) = *(a14 + 0);
  91. *(b + 14) = *(a15 + 0);
  92. *(b + 15) = *(a16 + 0);
  93. }
  94. a1 ++;
  95. a2 ++;
  96. a3 ++;
  97. a4 ++;
  98. a5 ++;
  99. a6 ++;
  100. a7 ++;
  101. a8 ++;
  102. a9 ++;
  103. a10 ++;
  104. a11 ++;
  105. a12 ++;
  106. a13 ++;
  107. a14 ++;
  108. a15 ++;
  109. a16 ++;
  110. b += 16;
  111. ii ++;
  112. }
  113. jj += 16;
  114. j --;
  115. }
  116. if (n & 8) {
  117. a1 = a + 0 * lda;
  118. a2 = a + 1 * lda;
  119. a3 = a + 2 * lda;
  120. a4 = a + 3 * lda;
  121. a5 = a + 4 * lda;
  122. a6 = a + 5 * lda;
  123. a7 = a + 6 * lda;
  124. a8 = a + 7 * lda;
  125. a += 8 * lda;
  126. ii = 0;
  127. for (i = 0; i < m; i++) {
  128. if ((ii >= jj ) && (ii - jj < 8)) {
  129. for (k = 0; k < ii - jj; k ++) {
  130. *(b + k) = *(a1 + k * lda);
  131. }
  132. *(b + ii - jj) = INV(*(a1 + (ii - jj) * lda));
  133. }
  134. if (ii - jj >= 8) {
  135. *(b + 0) = *(a1 + 0);
  136. *(b + 1) = *(a2 + 0);
  137. *(b + 2) = *(a3 + 0);
  138. *(b + 3) = *(a4 + 0);
  139. *(b + 4) = *(a5 + 0);
  140. *(b + 5) = *(a6 + 0);
  141. *(b + 6) = *(a7 + 0);
  142. *(b + 7) = *(a8 + 0);
  143. }
  144. a1 ++;
  145. a2 ++;
  146. a3 ++;
  147. a4 ++;
  148. a5 ++;
  149. a6 ++;
  150. a7 ++;
  151. a8 ++;
  152. b += 8;
  153. ii ++;
  154. }
  155. jj += 8;
  156. }
  157. if (n & 4) {
  158. a1 = a + 0 * lda;
  159. a2 = a + 1 * lda;
  160. a3 = a + 2 * lda;
  161. a4 = a + 3 * lda;
  162. a += 4 * lda;
  163. ii = 0;
  164. for (i = 0; i < m; i++) {
  165. if ((ii >= jj ) && (ii - jj < 4)) {
  166. for (k = 0; k < ii - jj; k ++) {
  167. *(b + k) = *(a1 + k * lda);
  168. }
  169. *(b + ii - jj) = INV(*(a1 + (ii - jj) * lda));
  170. }
  171. if (ii - jj >= 4) {
  172. *(b + 0) = *(a1 + 0);
  173. *(b + 1) = *(a2 + 0);
  174. *(b + 2) = *(a3 + 0);
  175. *(b + 3) = *(a4 + 0);
  176. }
  177. a1 ++;
  178. a2 ++;
  179. a3 ++;
  180. a4 ++;
  181. b += 4;
  182. ii ++;
  183. }
  184. jj += 4;
  185. }
  186. if (n & 2) {
  187. a1 = a + 0 * lda;
  188. a2 = a + 1 * lda;
  189. a += 2 * lda;
  190. ii = 0;
  191. for (i = 0; i < m; i++) {
  192. if ((ii >= jj ) && (ii - jj < 2)) {
  193. for (k = 0; k < ii - jj; k ++) {
  194. *(b + k) = *(a1 + k * lda);
  195. }
  196. *(b + ii - jj) = INV(*(a1 + (ii - jj) * lda));
  197. }
  198. if (ii - jj >= 2) {
  199. *(b + 0) = *(a1 + 0);
  200. *(b + 1) = *(a2 + 0);
  201. }
  202. a1 ++;
  203. a2 ++;
  204. b += 2;
  205. ii ++;
  206. }
  207. jj += 2;
  208. }
  209. if (n & 1) {
  210. a1 = a + 0 * lda;
  211. ii = 0;
  212. for (i = 0; i < m; i++) {
  213. if ((ii >= jj ) && (ii - jj < 1)) {
  214. for (k = 0; k < ii - jj; k ++) {
  215. *(b + k) = *(a1 + k * lda);
  216. }
  217. *(b + ii - jj) = INV(*(a1 + (ii - jj) * lda));
  218. }
  219. if (ii - jj >= 1) {
  220. *(b + 0) = *(a1 + 0);
  221. }
  222. a1 ++;
  223. b += 1;
  224. ii ++;
  225. }
  226. }
  227. return 0;
  228. }