API tools faq
paste
matheus__serpa

xeon_phi

matheus__serpa
Oct 3rd, 2014
217
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
C 2.82 KB | None | 0 0
raw download clone embed print report
  1. // compile for host-based OpenMP
  2. // icc -mkl -Ofast -no-offload -openmp -Wno-unknown-pragmas -std=c99 -vec-report3 matrix.c -o matrix.omp
  3.  
  4. // compile for offload mode
  5. //icc -mkl -Ofast -offload-build -Wno-unknown-pragmas -std=c99 -vec-report3 matrix.c -o matrix.off
  6.  
  7. // compile to run natively on the Xeon Phi
  8. //icc -mkl -Ofast -mmic -openmp -L  /opt/intel/lib/mic -Wno-unknown-pragmas -std=c99 -vec-report3 matrix.c -o matrix.mic -liomp5
  9.  
  10. // export PHI_OMP_NUM_THREADS=240
  11. // export PHI_KMP_AFFINITY="granularity=thread,balanced"
  12. // export LD_LIBRARY_PATH=/tmp
  13.  
  14. float (*restrict A)[size] = malloc(sizeof(float)*size*size);
  15.  
  16. void doMult(int size, float (* restrict A)[size], float (* restrict B)[size], float (* restrict C)[size]){
  17.  
  18. #pragma offload target(mic:0) in(A:length(size*size)) in( B:length(size*size)) out(C:length(size*size))
  19.   {
  20. #pragma omp parallel for default(none) shared(A, B, C, size)
  21.     for (int i = 0; i < size; ++i)
  22.       for (int k = 0; k < size; ++k)
  23.         for (int j = 0; j < size; ++j)
  24.           C[i][j] += A[i][k] * B[k][j];
  25.   }
  26.  
  27. }
  28.  
  29.  
  30. // nowait_example
  31. #pragma omp parallel
  32. {
  33. #pragma omp for schedule(static) nowait
  34. for (i=0; i<n; i++)
  35.  c[i] = (a[i] + b[i]) / 2.0f;
  36.  
  37. #pragma omp for schedule(static) nowait
  38. for (i=0; i<n; i++)
  39.  z[i] = sqrtf(c[i]);
  40.  
  41. #pragma omp for schedule(static) nowait
  42. for (i=1; i<=n; i++)
  43.  y[i] = z[i-1] + a[i];
  44. }
  45.  
  46. // collapse
  47.  #pragma omp parallel
  48.  {
  49.  #pragma omp for collapse(2) lastprivate(jlast, klast)
  50.  for (k=1; k<=2; k++)
  51.   for (j=1; j<=3; j++){
  52.    jlast=j;
  53.    klast=k;
  54.   }
  55.  #pragma omp single
  56.  printf("%d %d\n", klast, jlast);
  57.  }
  58.  
  59. // section
  60.  #pragma omp parallel sections
  61.  {
  62.  #pragma omp section
  63.  XAXIS();
  64.  #pragma omp section
  65.  YAXIS();
  66.  #pragma omp section
  67.  ZAXIS();
  68.  }
  69.  
  70.  
  71. // task 1
  72. void postorder_traverse( struct node *p ) {
  73.  if (p->left)
  74.   #pragma omp task // p is firstprivate by default
  75.    postorder_traverse(p->left);
  76.  if (p->right)
  77.   #pragma omp task // p is firstprivate by default
  78.    postorder_traverse(p->right);
  79.  #pragma omp taskwait
  80.  process(p);
  81. }
  82.  
  83.  
  84.  
  85. // task 2
  86. void process(node * p)
  87. {
  88.  /* do work here */
  89. }
  90.  
  91. void increment_list_items(node * head)
  92. {
  93.  #pragma omp parallel
  94.  {
  95.   #pragma omp single
  96.   {
  97.    node * p = head;
  98.    while (p) {
  99.     #pragma omp task
  100.     // p is firstprivate by default
  101.      process(p);
  102.      p = p->next;
  103.   }
  104.  }
  105. }
  106.  
  107.  
  108. // task 3
  109.  
  110. int fib(int n) {
  111.  int i, j;
  112.  if (n < 2)
  113.   return n;
  114.  else {
  115.   #pragma omp task shared(i)
  116.   i = fib(n-1);
  117.   #pragma omp task shared(j)
  118.   j = fib(n-2);
  119.   #pragma omp taskwait
  120.   return i+j;
  121.   }
  122. }
  123.  
  124. // Reduction
  125.  #pragma omp parallel for private(i) shared(x, y, n) reduction(+:a) reduction(^:b) reduction(min:c) reduction(max:d)
  126.  for (i=0; i<n; i++) {
  127.   a += x[i];
  128.   b ^= y[i];
  129.   if (c > y[i])
  130.    c = y[i];
  131.   d = fmaxf(d,x[i]);
  132. }
Advertisement
Add Comment
Please, Sign In to add comment
Public Pastes
We use cookies for various purposes including analytics. By continuing to use Pastebin, you agree to our use of cookies as described in the Cookies Policy.  OK, I Understand
Not a member of Pastebin yet?
Sign Up, it unlocks many cool features!