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.

ztrsm_uncopy_8.c 6.2 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228
  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 offset, FLOAT *b){
  41. BLASLONG i, ii, j, jj, k;
  42. FLOAT *a1, *a2, *a3, *a4, *a5, *a6, *a7, *a8;
  43. FLOAT data1, data2;
  44. lda *= 2;
  45. jj = offset;
  46. j = (n >> 3);
  47. while (j > 0){
  48. a1 = a + 0 * lda;
  49. a2 = a + 1 * lda;
  50. a3 = a + 2 * lda;
  51. a4 = a + 3 * lda;
  52. a5 = a + 4 * lda;
  53. a6 = a + 5 * lda;
  54. a7 = a + 6 * lda;
  55. a8 = a + 7 * lda;
  56. a += 8 * lda;
  57. ii = 0;
  58. for (i = 0; i < m; i++) {
  59. if ((ii >= jj ) && (ii - jj < 8)) {
  60. data1 = *(a1 + (ii - jj) * lda + 0);
  61. data2 = *(a1 + (ii - jj) * lda + 1);
  62. compinv(b + (ii - jj) * 2, data1, data2);
  63. for (k = ii - jj + 1; k < 8; k ++) {
  64. *(b + k * 2 + 0) = *(a1 + k * lda + 0);
  65. *(b + k * 2 + 1) = *(a1 + k * lda + 1);
  66. }
  67. }
  68. if (ii - jj < 0) {
  69. *(b + 0) = *(a1 + 0);
  70. *(b + 1) = *(a1 + 1);
  71. *(b + 2) = *(a2 + 0);
  72. *(b + 3) = *(a2 + 1);
  73. *(b + 4) = *(a3 + 0);
  74. *(b + 5) = *(a3 + 1);
  75. *(b + 6) = *(a4 + 0);
  76. *(b + 7) = *(a4 + 1);
  77. *(b + 8) = *(a5 + 0);
  78. *(b + 9) = *(a5 + 1);
  79. *(b + 10) = *(a6 + 0);
  80. *(b + 11) = *(a6 + 1);
  81. *(b + 12) = *(a7 + 0);
  82. *(b + 13) = *(a7 + 1);
  83. *(b + 14) = *(a8 + 0);
  84. *(b + 15) = *(a8 + 1);
  85. }
  86. a1 += 2;
  87. a2 += 2;
  88. a3 += 2;
  89. a4 += 2;
  90. a5 += 2;
  91. a6 += 2;
  92. a7 += 2;
  93. a8 += 2;
  94. b += 16;
  95. ii ++;
  96. }
  97. jj += 8;
  98. j --;
  99. }
  100. if (n & 4) {
  101. a1 = a + 0 * lda;
  102. a2 = a + 1 * lda;
  103. a3 = a + 2 * lda;
  104. a4 = a + 3 * lda;
  105. a += 4 * lda;
  106. ii = 0;
  107. for (i = 0; i < m; i++) {
  108. if ((ii >= jj ) && (ii - jj < 4)) {
  109. data1 = *(a1 + (ii - jj) * lda + 0);
  110. data2 = *(a1 + (ii - jj) * lda + 1);
  111. compinv(b + (ii - jj) * 2, data1, data2);
  112. for (k = ii - jj + 1; k < 4; k ++) {
  113. *(b + k * 2 + 0) = *(a1 + k * lda + 0);
  114. *(b + k * 2 + 1) = *(a1 + k * lda + 1);
  115. }
  116. }
  117. if (ii - jj < 0) {
  118. *(b + 0) = *(a1 + 0);
  119. *(b + 1) = *(a1 + 1);
  120. *(b + 2) = *(a2 + 0);
  121. *(b + 3) = *(a2 + 1);
  122. *(b + 4) = *(a3 + 0);
  123. *(b + 5) = *(a3 + 1);
  124. *(b + 6) = *(a4 + 0);
  125. *(b + 7) = *(a4 + 1);
  126. }
  127. a1 += 2;
  128. a2 += 2;
  129. a3 += 2;
  130. a4 += 2;
  131. b += 8;
  132. ii ++;
  133. }
  134. jj += 4;
  135. }
  136. if (n & 2) {
  137. a1 = a + 0 * lda;
  138. a2 = a + 1 * lda;
  139. a += 2 * lda;
  140. ii = 0;
  141. for (i = 0; i < m; i++) {
  142. if ((ii >= jj ) && (ii - jj < 2)) {
  143. data1 = *(a1 + (ii - jj) * lda + 0);
  144. data2 = *(a1 + (ii - jj) * lda + 1);
  145. compinv(b + (ii - jj) * 2, data1, data2);
  146. for (k = ii - jj + 1; k < 2; k ++) {
  147. *(b + k * 2 + 0) = *(a1 + k * lda + 0);
  148. *(b + k * 2 + 1) = *(a1 + k * lda + 1);
  149. }
  150. }
  151. if (ii - jj < 0) {
  152. *(b + 0) = *(a1 + 0);
  153. *(b + 1) = *(a1 + 1);
  154. *(b + 2) = *(a2 + 0);
  155. *(b + 3) = *(a2 + 1);
  156. }
  157. a1 += 2;
  158. a2 += 2;
  159. b += 4;
  160. ii ++;
  161. }
  162. jj += 2;
  163. }
  164. if (n & 1) {
  165. a1 = a + 0 * lda;
  166. ii = 0;
  167. for (i = 0; i < m; i++) {
  168. if ((ii >= jj ) && (ii - jj < 1)) {
  169. data1 = *(a1 + (ii - jj) * lda + 0);
  170. data2 = *(a1 + (ii - jj) * lda + 1);
  171. compinv(b + (ii - jj) * 2, data1, data2);
  172. for (k = ii - jj + 1; k < 1; k ++) {
  173. *(b + k * 2 + 0) = *(a1 + k * lda + 0);
  174. *(b + k * 2 + 1) = *(a1 + k * lda + 1);
  175. }
  176. }
  177. if (ii - jj < 0) {
  178. *(b + 0) = *(a1 + 0);
  179. *(b + 1) = *(a1 + 1);
  180. }
  181. a1 += 2;
  182. b += 2;
  183. ii ++;
  184. }
  185. }
  186. return 0;
  187. }