API tools faq
paste
Advertisement
Guest User

Untitled

a guest
Feb 29th, 2016
109
0
Never
Add comment
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 61.14 KB | None | 0 0
raw download clone embed print report
  1. User@User-PC ~/PyFR/euler_vortex_2d-1.3Test
  2. $ export PYFR_DEBUG_OMP_KEEP_LIBS=1
  3. (env4)
  4. User@User-PC ~/PyFR/euler_vortex_2d-1.3Test
  5. $ pyfr run -b openmp -p euler_vortex_2d.pyfrm euler_vortex_2d.ini
  6.  
  7. <pyfr.util.memoize object at 0x6fffe85c278>
  8. --self
  9. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  10. --args
  11.  
  12. --kwargs
  13.  
  14. res cache key bad
  15. [[[-9.93056816 -8.93056816 -7.93056816 ..., 7.06943184 8.06943184
  16. 9.06943184]
  17. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  18. 9.06943184]]
  19.  
  20. [[-9.66999052 -8.66999052 -7.66999052 ..., 7.33000948 8.33000948
  21. 9.33000948]
  22. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  23. 9.06943184]]
  24.  
  25. [[-9.33000948 -8.33000948 -7.33000948 ..., 7.66999052 8.66999052
  26. 9.66999052]
  27. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  28. 9.06943184]]
  29.  
  30. ...,
  31. [[-9.66999052 -8.66999052 -7.66999052 ..., 7.33000948 8.33000948
  32. 9.33000948]
  33. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  34. 9.93056816]]
  35.  
  36. [[-9.33000948 -8.33000948 -7.33000948 ..., 7.66999052 8.66999052
  37. 9.66999052]
  38. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  39. 9.93056816]]
  40.  
  41. [[-9.06943184 -8.06943184 -7.06943184 ..., 7.93056816 8.93056816
  42. 9.93056816]
  43. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  44. 9.93056816]]]
  45. new res cache key
  46.  
  47. <pyfr.util.memoize object at 0x6fffe85c278>
  48. --self
  49. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  50. --args
  51.  
  52. --kwargs
  53.  
  54. [[[-9.93056816 -8.93056816 -7.93056816 ..., 7.06943184 8.06943184
  55. 9.06943184]
  56. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  57. 9.06943184]]
  58.  
  59. [[-9.66999052 -8.66999052 -7.66999052 ..., 7.33000948 8.33000948
  60. 9.33000948]
  61. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  62. 9.06943184]]
  63.  
  64. [[-9.33000948 -8.33000948 -7.33000948 ..., 7.66999052 8.66999052
  65. 9.66999052]
  66. [-9.93056816 -9.93056816 -9.93056816 ..., 9.06943184 9.06943184
  67. 9.06943184]]
  68.  
  69. ...,
  70. [[-9.66999052 -8.66999052 -7.66999052 ..., 7.33000948 8.33000948
  71. 9.33000948]
  72. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  73. 9.93056816]]
  74.  
  75. [[-9.33000948 -8.33000948 -7.33000948 ..., 7.66999052 8.66999052
  76. 9.66999052]
  77. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  78. 9.93056816]]
  79.  
  80. [[-9.06943184 -8.06943184 -7.06943184 ..., 7.93056816 8.93056816
  81. 9.93056816]
  82. [-9.06943184 -9.06943184 -9.06943184 ..., 9.93056816 9.93056816
  83. 9.93056816]]]
  84. res cache key good
  85.  
  86. <pyfr.util.memoize object at 0x6fffe85c278>
  87. --self
  88. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> mpts
  89. --args
  90.  
  91. --kwargs
  92.  
  93. res cache key bad
  94. [[[-10. -9. -8. ..., 7. 8. 9.]
  95. [-10. -10. -10. ..., 9. 9. 9.]]
  96.  
  97. [[ -9. -8. -7. ..., 8. 9. 10.]
  98. [-10. -10. -10. ..., 9. 9. 9.]]
  99.  
  100. [[-10. -9. -8. ..., 7. 8. 9.]
  101. [ -9. -9. -9. ..., 10. 10. 10.]]
  102.  
  103. [[ -9. -8. -7. ..., 8. 9. 10.]
  104. [ -9. -9. -9. ..., 10. 10. 10.]]]
  105. new res cache key
  106.  
  107. <pyfr.util.memoize object at 0x6fffe85c160>
  108. --self
  109. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> M3
  110. --args
  111.  
  112. --kwargs
  113.  
  114. res cache key bad
  115. <pyfr.backends.openmp.types.OpenMPConstMatrix object at 0x6ffffd5ed68>
  116. new res cache key
  117.  
  118. <pyfr.util.memoize object at 0x6fffe7267b8>
  119. --self
  120. <pyfr.backends.openmp.cblas.OpenMPCBLASKernels object at 0x6fffe55f9e8> par_gemm
  121.  
  122. #include <omp.h>
  123. #include <stdlib.h>
  124. #include <tgmath.h>
  125.  
  126. #define PYFR_ALIGN_BYTES 32
  127. #define PYFR_NOINLINE __attribute__ ((noinline))
  128.  
  129. #define min(a, b) ((a) < (b) ? (a) : (b))
  130. #define max(a, b) ((a) > (b) ? (a) : (b))
  131.  
  132. // Typedefs
  133. typedef double fpdtype_t;
  134.  
  135. // OpenMP static loop scheduling functions
  136.  
  137. static inline int
  138. gcd(int a, int b)
  139. {
  140. return (a == 0) ? b : gcd(b % a, a);
  141. }
  142.  
  143. static inline void
  144. loop_sched_1d(int n, int align, int *b, int *e)
  145. {
  146. int tid = omp_get_thread_num();
  147. int nth = omp_get_num_threads();
  148.  
  149. // Round up n to be a multiple of nth
  150. int rn = n + nth - 1 - (n - 1) % nth;
  151.  
  152. // Nominal tile size
  153. int sz = rn / nth;
  154.  
  155. // Handle alignment
  156. sz += align - 1 - (sz - 1) % align;
  157.  
  158. // Assign the starting and ending index
  159. *b = sz * tid;
  160. *e = min(*b + sz, n);
  161.  
  162. // Clamp
  163. if (*b >= n)
  164. *b = *e = 0;
  165. }
  166.  
  167. static inline void
  168. loop_sched_2d(int nrow, int ncol, int colalign,
  169. int *rowb, int *rowe, int *colb, int *cole)
  170. {
  171. int tid = omp_get_thread_num();
  172. int nth = omp_get_num_threads();
  173.  
  174. // Distribute threads
  175. int nrowth = gcd(nrow, nth);
  176. int ncolth = nth / nrowth;
  177.  
  178. // Row and column indices for our thread
  179. int rowix = tid / ncolth;
  180. int colix = tid % ncolth;
  181.  
  182. // Round up ncol to be a multiple of ncolth
  183. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  184.  
  185. // Nominal tile size
  186. int ntilerow = nrow / nrowth;
  187. int ntilecol = rncol / ncolth;
  188.  
  189. // Handle column alignment
  190. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  191.  
  192. // Assign the starting and ending row to each thread
  193. *rowb = ntilerow * rowix;
  194. *rowe = *rowb + ntilerow;
  195.  
  196. // Assign the starting and ending column to each thread
  197. *colb = ntilecol * colix;
  198. *cole = min(*colb + ntilecol, ncol);
  199.  
  200. // Clamp
  201. if (*colb >= ncol)
  202. *colb = *cole = 0;
  203. }
  204.  
  205.  
  206.  
  207.  
  208. // CBLAS GEMM constants
  209. #define ROW_MAJOR 101
  210. #define NO_TRANS 111
  211.  
  212. // CBLAS GEMM prototype
  213. typedef void (*cblas_gemm_t)(int, int, int,
  214. int, int, int,
  215. fpdtype_t, const fpdtype_t *, int,
  216. const fpdtype_t *, int,
  217. fpdtype_t, fpdtype_t *, int);
  218.  
  219. void
  220. par_gemm(cblas_gemm_t gemm, int M, int N, int K,
  221. fpdtype_t alpha, const fpdtype_t *A, int lda,
  222. const fpdtype_t *B, int ldb,
  223. fpdtype_t beta, fpdtype_t *C, int ldc)
  224. {
  225. #pragma omp parallel
  226. {
  227. int begin, end;
  228. loop_sched_1d(N, PYFR_ALIGN_BYTES / sizeof(fpdtype_t), &begin, &end);
  229.  
  230. gemm(ROW_MAJOR, NO_TRANS, NO_TRANS, M, end - begin, K,
  231. alpha, A, lda, B + begin, ldb, beta, C + begin, ldc);
  232. }
  233. }
  234.  
  235. [<class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>]
  236. --args
  237.  
  238. --kwargs
  239.  
  240. res cache key bad
  241. <_FuncPtr object at 0x6fffe5694f8>
  242. new res cache key
  243.  
  244. <pyfr.util.memoize object at 0x6fffe85c240>
  245. --self
  246. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  247. --args
  248.  
  249. --kwargs
  250.  
  251.  
  252. <pyfr.util.memoize object at 0x6fffe85c208>
  253. --self
  254. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  255. --args
  256.  
  257. --kwargs
  258.  
  259. res cache key bad
  260. [[ 4. 4. 4. ..., 4. 4. 4.]
  261. [ 4. 4. 4. ..., 4. 4. 4.]
  262. [ 4. 4. 4. ..., 4. 4. 4.]
  263. ...,
  264. [ 4. 4. 4. ..., 4. 4. 4.]
  265. [ 4. 4. 4. ..., 4. 4. 4.]
  266. [ 4. 4. 4. ..., 4. 4. 4.]]
  267. new res cache key
  268. res cache key bad
  269. <pyfr.backends.openmp.types.OpenMPConstMatrix object at 0x6ffffca9f98>
  270. new res cache key
  271.  
  272. <pyfr.util.memoize object at 0x6fffebbf470>
  273. --self
  274. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> negdivconf pyfr.solvers.baseadvec.kernels.negdivconf {'ndims': 2, 'srcex': ['(0.)', '(0.)', '(0.)', '(0.)'], 'nvars': 4}
  275. --args
  276.  
  277. --kwargs
  278.  
  279. res cache key bad
  280. ('\n\n#include <omp.h>\n#include <stdlib.h>\n#include <tgmath.h>\n\n#define PYFR_ALIGN_BYTES 32\n#define PYFR_NOINLINE __attribute__ ((noinline))\n\n#define min(a, b) ((a) < (b) ? (a) : (b))\n#define max(a, b) ((a) > (b) ? (a) : (b))\n\n// Typedefs\ntypedef double fpdtype_t;\n\n// OpenMP static loop scheduling functions\n\nstatic inline int\ngcd(int a, int b)\n{\n return (a == 0) ? b : gcd(b % a, a);\n}\n\nstatic inline void\nloop_sched_1d(int n, int align, int *b, int *e)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Round up n to be a multiple of nth\n int rn = n + nth - 1 - (n - 1) % nth;\n\n // Nominal tile size\n int sz = rn / nth;\n\n // Handle alignment\n sz += align - 1 - (sz - 1) % align;\n\n // Assign the starting and ending index\n *b = sz * tid;\n *e = min(*b + sz, n);\n\n // Clamp\n if (*b >= n)\n *b = *e = 0;\n}\n\nstatic inline void\nloop_sched_2d(int nrow, int ncol, int colalign,\n int *rowb, int *rowe, int *colb, int *cole)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Distribute threads\n int nrowth = gcd(nrow, nth);\n int ncolth = nth / nrowth;\n\n // Row and column indices for our thread\n int rowix = tid / ncolth;\n int colix = tid % ncolth;\n\n // Round up ncol to be a multiple of ncolth\n int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;\n\n // Nominal tile size\n int ntilerow = nrow / nrowth;\n int ntilecol = rncol / ncolth;\n\n // Handle column alignment\n ntilecol += colalign - 1 - (ntilecol - 1) % colalign;\n\n // Assign the starting and ending row to each thread\n *rowb = ntilerow * rowix;\n *rowe = *rowb + ntilerow;\n\n // Assign the starting and ending column to each thread\n *colb = ntilecol * colix;\n *cole = min(*colb + ntilecol, ncol);\n\n // Clamp\n if (*colb >= ncol)\n *colb = *cole = 0;\n}\n\n\n\n\n\nstatic PYFR_NOINLINE void negdivconf_inner(int _nx, const fpdtype_t *__restrict__ rcpdjac_v, fpdtype_t *__restrict__ tdivtconf_v0, fpdtype_t *__restrict__ tdivtconf_v1, fpdtype_t *__restrict__ tdivtconf_v2, fpdtype_t *__restrict__ tdivtconf_v3)\n {\n for (int _x = 0; _x < _nx; _x++)\n {\n \n tdivtconf_v0[_x] = -rcpdjac_v[_x]*tdivtconf_v0[_x] + (0.);\n tdivtconf_v1[_x] = -rcpdjac_v[_x]*tdivtconf_v1[_x] + (0.);\n tdivtconf_v2[_x] = -rcpdjac_v[_x]*tdivtconf_v2[_x] + (0.);\n tdivtconf_v3[_x] = -rcpdjac_v[_x]*tdivtconf_v3[_x] + (0.);\n\n }\n }\n void negdivconf(int _ny, int _nx, const fpdtype_t* __restrict__ rcpdjac_v, int lsdrcpdjac, fpdtype_t* __restrict__ tdivtconf_v, int lsdtdivtconf)\n {\n #pragma omp parallel\n {\n int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);\n int rb, re, cb, ce;\n loop_sched_2d(_ny, _nx, align, &rb, &re, &cb, &ce);\n for (int _y = rb; _y < re; _y++)\n {\n negdivconf_inner(ce - cb, rcpdjac_v + _y*lsdrcpdjac + cb, tdivtconf_v + (_y*4 + 0)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 1)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 2)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 3)*lsdtdivtconf + cb);\n }\n }\n }\n\n', 2, ['_ny', '_nx', 'rcpdjac', 'tdivtconf'], [[<class 'numpy.int32'>], [<class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int32'>]])
  281. new res cache key
  282.  
  283. <pyfr.util.memoize object at 0x6fffe7267b8>
  284. --self
  285. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> negdivconf
  286.  
  287. #include <omp.h>
  288. #include <stdlib.h>
  289. #include <tgmath.h>
  290.  
  291. #define PYFR_ALIGN_BYTES 32
  292. #define PYFR_NOINLINE __attribute__ ((noinline))
  293.  
  294. #define min(a, b) ((a) < (b) ? (a) : (b))
  295. #define max(a, b) ((a) > (b) ? (a) : (b))
  296.  
  297. // Typedefs
  298. typedef double fpdtype_t;
  299.  
  300. // OpenMP static loop scheduling functions
  301.  
  302. static inline int
  303. gcd(int a, int b)
  304. {
  305. return (a == 0) ? b : gcd(b % a, a);
  306. }
  307.  
  308. static inline void
  309. loop_sched_1d(int n, int align, int *b, int *e)
  310. {
  311. int tid = omp_get_thread_num();
  312. int nth = omp_get_num_threads();
  313.  
  314. // Round up n to be a multiple of nth
  315. int rn = n + nth - 1 - (n - 1) % nth;
  316.  
  317. // Nominal tile size
  318. int sz = rn / nth;
  319.  
  320. // Handle alignment
  321. sz += align - 1 - (sz - 1) % align;
  322.  
  323. // Assign the starting and ending index
  324. *b = sz * tid;
  325. *e = min(*b + sz, n);
  326.  
  327. // Clamp
  328. if (*b >= n)
  329. *b = *e = 0;
  330. }
  331.  
  332. static inline void
  333. loop_sched_2d(int nrow, int ncol, int colalign,
  334. int *rowb, int *rowe, int *colb, int *cole)
  335. {
  336. int tid = omp_get_thread_num();
  337. int nth = omp_get_num_threads();
  338.  
  339. // Distribute threads
  340. int nrowth = gcd(nrow, nth);
  341. int ncolth = nth / nrowth;
  342.  
  343. // Row and column indices for our thread
  344. int rowix = tid / ncolth;
  345. int colix = tid % ncolth;
  346.  
  347. // Round up ncol to be a multiple of ncolth
  348. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  349.  
  350. // Nominal tile size
  351. int ntilerow = nrow / nrowth;
  352. int ntilecol = rncol / ncolth;
  353.  
  354. // Handle column alignment
  355. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  356.  
  357. // Assign the starting and ending row to each thread
  358. *rowb = ntilerow * rowix;
  359. *rowe = *rowb + ntilerow;
  360.  
  361. // Assign the starting and ending column to each thread
  362. *colb = ntilecol * colix;
  363. *cole = min(*colb + ntilecol, ncol);
  364.  
  365. // Clamp
  366. if (*colb >= ncol)
  367. *colb = *cole = 0;
  368. }
  369.  
  370.  
  371.  
  372.  
  373.  
  374. static PYFR_NOINLINE void negdivconf_inner(int _nx, const fpdtype_t *__restrict__ rcpdjac_v, fpdtype_t *__restrict__ tdivtconf_v0, fpdtype_t *__restrict__ tdivtconf_v1, fpdtype_t *__restrict__ tdivtconf_v2, fpdtype_t *__restrict__ tdivtconf_v3)
  375. {
  376. for (int _x = 0; _x < _nx; _x++)
  377. {
  378.  
  379. tdivtconf_v0[_x] = -rcpdjac_v[_x]*tdivtconf_v0[_x] + (0.);
  380. tdivtconf_v1[_x] = -rcpdjac_v[_x]*tdivtconf_v1[_x] + (0.);
  381. tdivtconf_v2[_x] = -rcpdjac_v[_x]*tdivtconf_v2[_x] + (0.);
  382. tdivtconf_v3[_x] = -rcpdjac_v[_x]*tdivtconf_v3[_x] + (0.);
  383.  
  384. }
  385. }
  386. void negdivconf(int _ny, int _nx, const fpdtype_t* __restrict__ rcpdjac_v, int lsdrcpdjac, fpdtype_t* __restrict__ tdivtconf_v, int lsdtdivtconf)
  387. {
  388. #pragma omp parallel
  389. {
  390. int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);
  391. int rb, re, cb, ce;
  392. loop_sched_2d(_ny, _nx, align, &rb, &re, &cb, &ce);
  393. for (int _y = rb; _y < re; _y++)
  394. {
  395. negdivconf_inner(ce - cb, rcpdjac_v + _y*lsdrcpdjac + cb, tdivtconf_v + (_y*4 + 0)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 1)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 2)*lsdtdivtconf + cb, tdivtconf_v + (_y*4 + 3)*lsdtdivtconf + cb);
  396. }
  397. }
  398. }
  399.  
  400. [<class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>]
  401. --args
  402.  
  403. --kwargs
  404.  
  405. res cache key bad
  406. <_FuncPtr object at 0x6fffe5695c0>
  407. new res cache key
  408.  
  409. <pyfr.util.memoize object at 0x6fffe85c160>
  410. --self
  411. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> M1 - M3*M2
  412. --args
  413.  
  414. --kwargs
  415.  
  416. res cache key bad
  417. <pyfr.backends.openmp.types.OpenMPConstMatrix object at 0x6fffe108a20>
  418. new res cache key
  419.  
  420. <pyfr.util.memoize object at 0x6fffe7267b8>
  421. --self
  422. <pyfr.backends.openmp.cblas.OpenMPCBLASKernels object at 0x6fffe55f9e8> par_gemm
  423.  
  424. #include <omp.h>
  425. #include <stdlib.h>
  426. #include <tgmath.h>
  427.  
  428. #define PYFR_ALIGN_BYTES 32
  429. #define PYFR_NOINLINE __attribute__ ((noinline))
  430.  
  431. #define min(a, b) ((a) < (b) ? (a) : (b))
  432. #define max(a, b) ((a) > (b) ? (a) : (b))
  433.  
  434. // Typedefs
  435. typedef double fpdtype_t;
  436.  
  437. // OpenMP static loop scheduling functions
  438.  
  439. static inline int
  440. gcd(int a, int b)
  441. {
  442. return (a == 0) ? b : gcd(b % a, a);
  443. }
  444.  
  445. static inline void
  446. loop_sched_1d(int n, int align, int *b, int *e)
  447. {
  448. int tid = omp_get_thread_num();
  449. int nth = omp_get_num_threads();
  450.  
  451. // Round up n to be a multiple of nth
  452. int rn = n + nth - 1 - (n - 1) % nth;
  453.  
  454. // Nominal tile size
  455. int sz = rn / nth;
  456.  
  457. // Handle alignment
  458. sz += align - 1 - (sz - 1) % align;
  459.  
  460. // Assign the starting and ending index
  461. *b = sz * tid;
  462. *e = min(*b + sz, n);
  463.  
  464. // Clamp
  465. if (*b >= n)
  466. *b = *e = 0;
  467. }
  468.  
  469. static inline void
  470. loop_sched_2d(int nrow, int ncol, int colalign,
  471. int *rowb, int *rowe, int *colb, int *cole)
  472. {
  473. int tid = omp_get_thread_num();
  474. int nth = omp_get_num_threads();
  475.  
  476. // Distribute threads
  477. int nrowth = gcd(nrow, nth);
  478. int ncolth = nth / nrowth;
  479.  
  480. // Row and column indices for our thread
  481. int rowix = tid / ncolth;
  482. int colix = tid % ncolth;
  483.  
  484. // Round up ncol to be a multiple of ncolth
  485. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  486.  
  487. // Nominal tile size
  488. int ntilerow = nrow / nrowth;
  489. int ntilecol = rncol / ncolth;
  490.  
  491. // Handle column alignment
  492. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  493.  
  494. // Assign the starting and ending row to each thread
  495. *rowb = ntilerow * rowix;
  496. *rowe = *rowb + ntilerow;
  497.  
  498. // Assign the starting and ending column to each thread
  499. *colb = ntilecol * colix;
  500. *cole = min(*colb + ntilecol, ncol);
  501.  
  502. // Clamp
  503. if (*colb >= ncol)
  504. *colb = *cole = 0;
  505. }
  506.  
  507.  
  508.  
  509.  
  510. // CBLAS GEMM constants
  511. #define ROW_MAJOR 101
  512. #define NO_TRANS 111
  513.  
  514. // CBLAS GEMM prototype
  515. typedef void (*cblas_gemm_t)(int, int, int,
  516. int, int, int,
  517. fpdtype_t, const fpdtype_t *, int,
  518. const fpdtype_t *, int,
  519. fpdtype_t, fpdtype_t *, int);
  520.  
  521. void
  522. par_gemm(cblas_gemm_t gemm, int M, int N, int K,
  523. fpdtype_t alpha, const fpdtype_t *A, int lda,
  524. const fpdtype_t *B, int ldb,
  525. fpdtype_t beta, fpdtype_t *C, int ldc)
  526. {
  527. #pragma omp parallel
  528. {
  529. int begin, end;
  530. loop_sched_1d(N, PYFR_ALIGN_BYTES / sizeof(fpdtype_t), &begin, &end);
  531.  
  532. gemm(ROW_MAJOR, NO_TRANS, NO_TRANS, M, end - begin, K,
  533. alpha, A, lda, B + begin, ldb, beta, C + begin, ldc);
  534. }
  535. }
  536.  
  537. [<class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>]
  538. --args
  539.  
  540. --kwargs
  541.  
  542. <_FuncPtr object at 0x6fffe5694f8>
  543. res cache key good
  544.  
  545. <pyfr.util.memoize object at 0x6fffe85c160>
  546. --self
  547. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> M0
  548. --args
  549.  
  550. --kwargs
  551.  
  552. res cache key bad
  553. <pyfr.backends.openmp.types.OpenMPConstMatrix object at 0x6fffe108978>
  554. new res cache key
  555.  
  556. <pyfr.util.memoize object at 0x6fffe7267b8>
  557. --self
  558. <pyfr.backends.openmp.cblas.OpenMPCBLASKernels object at 0x6fffe55f9e8> par_gemm
  559.  
  560. #include <omp.h>
  561. #include <stdlib.h>
  562. #include <tgmath.h>
  563.  
  564. #define PYFR_ALIGN_BYTES 32
  565. #define PYFR_NOINLINE __attribute__ ((noinline))
  566.  
  567. #define min(a, b) ((a) < (b) ? (a) : (b))
  568. #define max(a, b) ((a) > (b) ? (a) : (b))
  569.  
  570. // Typedefs
  571. typedef double fpdtype_t;
  572.  
  573. // OpenMP static loop scheduling functions
  574.  
  575. static inline int
  576. gcd(int a, int b)
  577. {
  578. return (a == 0) ? b : gcd(b % a, a);
  579. }
  580.  
  581. static inline void
  582. loop_sched_1d(int n, int align, int *b, int *e)
  583. {
  584. int tid = omp_get_thread_num();
  585. int nth = omp_get_num_threads();
  586.  
  587. // Round up n to be a multiple of nth
  588. int rn = n + nth - 1 - (n - 1) % nth;
  589.  
  590. // Nominal tile size
  591. int sz = rn / nth;
  592.  
  593. // Handle alignment
  594. sz += align - 1 - (sz - 1) % align;
  595.  
  596. // Assign the starting and ending index
  597. *b = sz * tid;
  598. *e = min(*b + sz, n);
  599.  
  600. // Clamp
  601. if (*b >= n)
  602. *b = *e = 0;
  603. }
  604.  
  605. static inline void
  606. loop_sched_2d(int nrow, int ncol, int colalign,
  607. int *rowb, int *rowe, int *colb, int *cole)
  608. {
  609. int tid = omp_get_thread_num();
  610. int nth = omp_get_num_threads();
  611.  
  612. // Distribute threads
  613. int nrowth = gcd(nrow, nth);
  614. int ncolth = nth / nrowth;
  615.  
  616. // Row and column indices for our thread
  617. int rowix = tid / ncolth;
  618. int colix = tid % ncolth;
  619.  
  620. // Round up ncol to be a multiple of ncolth
  621. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  622.  
  623. // Nominal tile size
  624. int ntilerow = nrow / nrowth;
  625. int ntilecol = rncol / ncolth;
  626.  
  627. // Handle column alignment
  628. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  629.  
  630. // Assign the starting and ending row to each thread
  631. *rowb = ntilerow * rowix;
  632. *rowe = *rowb + ntilerow;
  633.  
  634. // Assign the starting and ending column to each thread
  635. *colb = ntilecol * colix;
  636. *cole = min(*colb + ntilecol, ncol);
  637.  
  638. // Clamp
  639. if (*colb >= ncol)
  640. *colb = *cole = 0;
  641. }
  642.  
  643.  
  644.  
  645.  
  646. // CBLAS GEMM constants
  647. #define ROW_MAJOR 101
  648. #define NO_TRANS 111
  649.  
  650. // CBLAS GEMM prototype
  651. typedef void (*cblas_gemm_t)(int, int, int,
  652. int, int, int,
  653. fpdtype_t, const fpdtype_t *, int,
  654. const fpdtype_t *, int,
  655. fpdtype_t, fpdtype_t *, int);
  656.  
  657. void
  658. par_gemm(cblas_gemm_t gemm, int M, int N, int K,
  659. fpdtype_t alpha, const fpdtype_t *A, int lda,
  660. const fpdtype_t *B, int ldb,
  661. fpdtype_t beta, fpdtype_t *C, int ldc)
  662. {
  663. #pragma omp parallel
  664. {
  665. int begin, end;
  666. loop_sched_1d(N, PYFR_ALIGN_BYTES / sizeof(fpdtype_t), &begin, &end);
  667.  
  668. gemm(ROW_MAJOR, NO_TRANS, NO_TRANS, M, end - begin, K,
  669. alpha, A, lda, B + begin, ldb, beta, C + begin, ldc);
  670. }
  671. }
  672.  
  673. [<class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.float64'>, <class 'numpy.int64'>, <class 'numpy.int32'>]
  674. --args
  675.  
  676. --kwargs
  677.  
  678. <_FuncPtr object at 0x6fffe5694f8>
  679. res cache key good
  680.  
  681. <pyfr.util.memoize object at 0x6fffe85c1d0>
  682. --self
  683. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  684. --args
  685.  
  686. --kwargs
  687.  
  688.  
  689. <pyfr.util.memoize object at 0x6fffe85c198>
  690. --self
  691. <pyfr.solvers.euler.elements.EulerElements object at 0x6fffe567630> upts
  692. --args
  693.  
  694. --kwargs
  695.  
  696. res cache key bad
  697. [[[[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  698. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  699. [ -6.16678657e-17 -8.26510554e-16 3.91590947e-15 ...,
  700. 5.16569096e-14 0.00000000e+00 0.00000000e+00]]
  701.  
  702. [[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  703. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  704. [ -2.93107297e-16 -5.95071123e-16 1.86123133e-14 ...,
  705. 3.71919452e-14 0.00000000e+00 0.00000000e+00]]
  706.  
  707. [[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  708. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  709. [ -5.95071123e-16 -2.93107297e-16 3.77870163e-14 ...,
  710. 1.83192061e-14 0.00000000e+00 0.00000000e+00]]
  711.  
  712. ...,
  713. [[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  714. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  715. [ -2.93107297e-16 -5.95071123e-16 1.86123133e-14 ...,
  716. 3.71919452e-14 0.00000000e+00 0.00000000e+00]]
  717.  
  718. [[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  719. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  720. [ -5.95071123e-16 -2.93107297e-16 3.77870163e-14 ...,
  721. 1.83192061e-14 0.00000000e+00 0.00000000e+00]]
  722.  
  723. [[ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  724. 5.00000000e-01 5.00000000e-01 5.00000000e-01]
  725. [ -8.26510554e-16 -6.16678657e-17 5.24834202e-14 ...,
  726. 3.85424160e-15 0.00000000e+00 0.00000000e+00]]]
  727.  
  728.  
  729. [[[ -6.16678657e-17 6.16678657e-17 -1.23335731e-16 ...,
  730. 0.00000000e+00 -8.26510554e-16 8.26510554e-16]
  731. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  732. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]
  733.  
  734. [[ -6.16678657e-17 6.16678657e-17 -1.23335731e-16 ...,
  735. 0.00000000e+00 -8.26510554e-16 8.26510554e-16]
  736. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  737. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]
  738.  
  739. [[ -6.16678657e-17 6.16678657e-17 -1.23335731e-16 ...,
  740. 0.00000000e+00 -8.26510554e-16 8.26510554e-16]
  741. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  742. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]
  743.  
  744. ...,
  745. [[ -8.26510554e-16 8.26510554e-16 -1.65302111e-15 ...,
  746. 0.00000000e+00 -6.16678657e-17 6.16678657e-17]
  747. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  748. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]
  749.  
  750. [[ -8.26510554e-16 8.26510554e-16 -1.65302111e-15 ...,
  751. 0.00000000e+00 -6.16678657e-17 6.16678657e-17]
  752. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  753. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]
  754.  
  755. [[ -8.26510554e-16 8.26510554e-16 -1.65302111e-15 ...,
  756. 0.00000000e+00 -6.16678657e-17 6.16678657e-17]
  757. [ 5.00000000e-01 5.00000000e-01 5.00000000e-01 ...,
  758. 5.00000000e-01 5.00000000e-01 5.00000000e-01]]]]
  759. new res cache key
  760. res cache key bad
  761. <pyfr.backends.openmp.types.OpenMPConstMatrix object at 0x6fffe159b70>
  762. new res cache key
  763.  
  764. <pyfr.util.memoize object at 0x6fffebbf470>
  765. --self
  766. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> tflux pyfr.solvers.euler.kernels.tflux {'ndims': 2, 'c': OrderedDict([('gamma', 1.4), ('S', 13.5), ('M', 0.4), ('R', 1.5)]), 'nvars': 4}
  767. --args
  768.  
  769. --kwargs
  770.  
  771. res cache key bad
  772. ('\n\n#include <omp.h>\n#include <stdlib.h>\n#include <tgmath.h>\n\n#define PYFR_ALIGN_BYTES 32\n#define PYFR_NOINLINE __attribute__ ((noinline))\n\n#define min(a, b) ((a) < (b) ? (a) : (b))\n#define max(a, b) ((a) > (b) ? (a) : (b))\n\n// Typedefs\ntypedef double fpdtype_t;\n\n// OpenMP static loop scheduling functions\n\nstatic inline int\ngcd(int a, int b)\n{\n return (a == 0) ? b : gcd(b % a, a);\n}\n\nstatic inline void\nloop_sched_1d(int n, int align, int *b, int *e)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Round up n to be a multiple of nth\n int rn = n + nth - 1 - (n - 1) % nth;\n\n // Nominal tile size\n int sz = rn / nth;\n\n // Handle alignment\n sz += align - 1 - (sz - 1) % align;\n\n // Assign the starting and ending index\n *b = sz * tid;\n *e = min(*b + sz, n);\n\n // Clamp\n if (*b >= n)\n *b = *e = 0;\n}\n\nstatic inline void\nloop_sched_2d(int nrow, int ncol, int colalign,\n int *rowb, int *rowe, int *colb, int *cole)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Distribute threads\n int nrowth = gcd(nrow, nth);\n int ncolth = nth / nrowth;\n\n // Row and column indices for our thread\n int rowix = tid / ncolth;\n int colix = tid % ncolth;\n\n // Round up ncol to be a multiple of ncolth\n int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;\n\n // Nominal tile size\n int ntilerow = nrow / nrowth;\n int ntilecol = rncol / ncolth;\n\n // Handle column alignment\n ntilecol += colalign - 1 - (ntilecol - 1) % colalign;\n\n // Assign the starting and ending row to each thread\n *rowb = ntilerow * rowix;\n *rowe = *rowb + ntilerow;\n\n // Assign the starting and ending column to each thread\n *colb = ntilecol * colix;\n *cole = min(*colb + ntilecol, ncol);\n\n // Clamp\n if (*colb >= ncol)\n *colb = *cole = 0;\n}\n\n\n\n\n\n\n\n\n\nstatic PYFR_NOINLINE void tflux_inner(int _nx, fpdtype_t *__restrict__ f_v0v0, fpdtype_t *__restrict__ f_v0v1, fpdtype_t *__restrict__ f_v0v2, fpdtype_t *__restrict__ f_v0v3, fpdtype_t *__restrict__ f_v1v0, fpdtype_t *__restrict__ f_v1v1, fpdtype_t *__restrict__ f_v1v2, fpdtype_t *__restrict__ f_v1v3, const fpdtype_t *__restrict__ smats_v0v0, const fpdtype_t *__restrict__ smats_v0v1, const fpdtype_t *__restrict__ smats_v1v0, const fpdtype_t *__restrict__ smats_v1v1, const fpdtype_t *__restrict__ u_v0, const fpdtype_t *__restrict__ u_v1, const fpdtype_t *__restrict__ u_v2, const fpdtype_t *__restrict__ u_v3)\n {\n for (int _x = 0; _x < _nx; _x++)\n {\n \n // Compute the flux\n fpdtype_t ftemp[2][4];\n fpdtype_t p, v[2];\n {\n\n fpdtype_t invrho_ = 1.0/u_v0[_x], E_ = u_v3[_x];\n\n // Compute the velocities\n fpdtype_t rhov_[2];\n rhov_[0] = u_v1[_x];\n v[0] = invrho_*rhov_[0];\n rhov_[1] = u_v2[_x];\n v[1] = invrho_*rhov_[1];\n\n // Compute the pressure\n p = 0.3999999999999999*(E_ - 0.5*invrho_*((rhov_[0])*(rhov_[0]) + (rhov_[1])*(rhov_[1])));\n\n // Density and energy fluxes\n ftemp[0][0] = rhov_[0];\n ftemp[0][3] = (E_ + p)*v[0];\n ftemp[1][0] = rhov_[1];\n ftemp[1][3] = (E_ + p)*v[1];\n\n // Momentum fluxes\n ftemp[0][1] = rhov_[0]*v[0] + p;\n ftemp[0][2] = rhov_[0]*v[1];\n ftemp[1][1] = rhov_[1]*v[0];\n ftemp[1][2] = rhov_[1]*v[1] + p;\n\n};\n\n // Transform the fluxes\n f_v0v0[_x] = smats_v0v0[_x]*ftemp[0][0] + smats_v0v1[_x]*ftemp[1][0];\n f_v0v1[_x] = smats_v0v0[_x]*ftemp[0][1] + smats_v0v1[_x]*ftemp[1][1];\n f_v0v2[_x] = smats_v0v0[_x]*ftemp[0][2] + smats_v0v1[_x]*ftemp[1][2];\n f_v0v3[_x] = smats_v0v0[_x]*ftemp[0][3] + smats_v0v1[_x]*ftemp[1][3];\n f_v1v0[_x] = smats_v1v0[_x]*ftemp[0][0] + smats_v1v1[_x]*ftemp[1][0];\n f_v1v1[_x] = smats_v1v0[_x]*ftemp[0][1] + smats_v1v1[_x]*ftemp[1][1];\n f_v1v2[_x] = smats_v1v0[_x]*ftemp[0][2] + smats_v1v1[_x]*ftemp[1][2];\n f_v1v3[_x] = smats_v1v0[_x]*ftemp[0][3] + smats_v1v1[_x]*ftemp[1][3];\n\n }\n }\n void tflux(int _ny, int _nx, fpdtype_t* __restrict__ f_v, int lsdf, const fpdtype_t* __restrict__ smats_v, int lsdsmats, const fpdtype_t* __restrict__ u_v, int lsdu)\n {\n #pragma omp parallel\n {\n int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);\n int rb, re, cb, ce;\n loop_sched_2d(_ny, _nx, align, &rb, &re, &cb, &ce);\n for (int _y = rb; _y < re; _y++)\n {\n tflux_inner(ce - cb, f_v + ((0*_ny + _y)*4 + 0)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 1)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 2)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 3)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 0)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 1)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 2)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 3)*lsdf + cb, smats_v + ((0*_ny + _y)*2 + 0)*lsdsmats + cb, smats_v + ((0*_ny + _y)*2 + 1)*lsdsmats + cb, smats_v + ((1*_ny + _y)*2 + 0)*lsdsmats + cb, smats_v + ((1*_ny + _y)*2 + 1)*lsdsmats + cb, u_v + (_y*4 + 0)*lsdu + cb, u_v + (_y*4 + 1)*lsdu + cb, u_v + (_y*4 + 2)*lsdu + cb, u_v + (_y*4 + 3)*lsdu + cb);\n }\n }\n }\n\n', 2, ['_ny', '_nx', 'f', 'smats', 'u'], [[<class 'numpy.int32'>], [<class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int32'>]])
  773. new res cache key
  774.  
  775. <pyfr.util.memoize object at 0x6fffe7267b8>
  776. --self
  777. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> tflux
  778.  
  779. #include <omp.h>
  780. #include <stdlib.h>
  781. #include <tgmath.h>
  782.  
  783. #define PYFR_ALIGN_BYTES 32
  784. #define PYFR_NOINLINE __attribute__ ((noinline))
  785.  
  786. #define min(a, b) ((a) < (b) ? (a) : (b))
  787. #define max(a, b) ((a) > (b) ? (a) : (b))
  788.  
  789. // Typedefs
  790. typedef double fpdtype_t;
  791.  
  792. // OpenMP static loop scheduling functions
  793.  
  794. static inline int
  795. gcd(int a, int b)
  796. {
  797. return (a == 0) ? b : gcd(b % a, a);
  798. }
  799.  
  800. static inline void
  801. loop_sched_1d(int n, int align, int *b, int *e)
  802. {
  803. int tid = omp_get_thread_num();
  804. int nth = omp_get_num_threads();
  805.  
  806. // Round up n to be a multiple of nth
  807. int rn = n + nth - 1 - (n - 1) % nth;
  808.  
  809. // Nominal tile size
  810. int sz = rn / nth;
  811.  
  812. // Handle alignment
  813. sz += align - 1 - (sz - 1) % align;
  814.  
  815. // Assign the starting and ending index
  816. *b = sz * tid;
  817. *e = min(*b + sz, n);
  818.  
  819. // Clamp
  820. if (*b >= n)
  821. *b = *e = 0;
  822. }
  823.  
  824. static inline void
  825. loop_sched_2d(int nrow, int ncol, int colalign,
  826. int *rowb, int *rowe, int *colb, int *cole)
  827. {
  828. int tid = omp_get_thread_num();
  829. int nth = omp_get_num_threads();
  830.  
  831. // Distribute threads
  832. int nrowth = gcd(nrow, nth);
  833. int ncolth = nth / nrowth;
  834.  
  835. // Row and column indices for our thread
  836. int rowix = tid / ncolth;
  837. int colix = tid % ncolth;
  838.  
  839. // Round up ncol to be a multiple of ncolth
  840. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  841.  
  842. // Nominal tile size
  843. int ntilerow = nrow / nrowth;
  844. int ntilecol = rncol / ncolth;
  845.  
  846. // Handle column alignment
  847. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  848.  
  849. // Assign the starting and ending row to each thread
  850. *rowb = ntilerow * rowix;
  851. *rowe = *rowb + ntilerow;
  852.  
  853. // Assign the starting and ending column to each thread
  854. *colb = ntilecol * colix;
  855. *cole = min(*colb + ntilecol, ncol);
  856.  
  857. // Clamp
  858. if (*colb >= ncol)
  859. *colb = *cole = 0;
  860. }
  861.  
  862.  
  863.  
  864.  
  865.  
  866.  
  867.  
  868.  
  869.  
  870. static PYFR_NOINLINE void tflux_inner(int _nx, fpdtype_t *__restrict__ f_v0v0, fpdtype_t *__restrict__ f_v0v1, fpdtype_t *__restrict__ f_v0v2, fpdtype_t *__restrict__ f_v0v3, fpdtype_t *__restrict__ f_v1v0, fpdtype_t *__restrict__ f_v1v1, fpdtype_t *__restrict__ f_v1v2, fpdtype_t *__restrict__ f_v1v3, const fpdtype_t *__restrict__ smats_v0v0, const fpdtype_t *__restrict__ smats_v0v1, const fpdtype_t *__restrict__ smats_v1v0, const fpdtype_t *__restrict__ smats_v1v1, const fpdtype_t *__restrict__ u_v0, const fpdtype_t *__restrict__ u_v1, const fpdtype_t *__restrict__ u_v2, const fpdtype_t *__restrict__ u_v3)
  871. {
  872. for (int _x = 0; _x < _nx; _x++)
  873. {
  874.  
  875. // Compute the flux
  876. fpdtype_t ftemp[2][4];
  877. fpdtype_t p, v[2];
  878. {
  879.  
  880. fpdtype_t invrho_ = 1.0/u_v0[_x], E_ = u_v3[_x];
  881.  
  882. // Compute the velocities
  883. fpdtype_t rhov_[2];
  884. rhov_[0] = u_v1[_x];
  885. v[0] = invrho_*rhov_[0];
  886. rhov_[1] = u_v2[_x];
  887. v[1] = invrho_*rhov_[1];
  888.  
  889. // Compute the pressure
  890. p = 0.3999999999999999*(E_ - 0.5*invrho_*((rhov_[0])*(rhov_[0]) + (rhov_[1])*(rhov_[1])));
  891.  
  892. // Density and energy fluxes
  893. ftemp[0][0] = rhov_[0];
  894. ftemp[0][3] = (E_ + p)*v[0];
  895. ftemp[1][0] = rhov_[1];
  896. ftemp[1][3] = (E_ + p)*v[1];
  897.  
  898. // Momentum fluxes
  899. ftemp[0][1] = rhov_[0]*v[0] + p;
  900. ftemp[0][2] = rhov_[0]*v[1];
  901. ftemp[1][1] = rhov_[1]*v[0];
  902. ftemp[1][2] = rhov_[1]*v[1] + p;
  903.  
  904. };
  905.  
  906. // Transform the fluxes
  907. f_v0v0[_x] = smats_v0v0[_x]*ftemp[0][0] + smats_v0v1[_x]*ftemp[1][0];
  908. f_v0v1[_x] = smats_v0v0[_x]*ftemp[0][1] + smats_v0v1[_x]*ftemp[1][1];
  909. f_v0v2[_x] = smats_v0v0[_x]*ftemp[0][2] + smats_v0v1[_x]*ftemp[1][2];
  910. f_v0v3[_x] = smats_v0v0[_x]*ftemp[0][3] + smats_v0v1[_x]*ftemp[1][3];
  911. f_v1v0[_x] = smats_v1v0[_x]*ftemp[0][0] + smats_v1v1[_x]*ftemp[1][0];
  912. f_v1v1[_x] = smats_v1v0[_x]*ftemp[0][1] + smats_v1v1[_x]*ftemp[1][1];
  913. f_v1v2[_x] = smats_v1v0[_x]*ftemp[0][2] + smats_v1v1[_x]*ftemp[1][2];
  914. f_v1v3[_x] = smats_v1v0[_x]*ftemp[0][3] + smats_v1v1[_x]*ftemp[1][3];
  915.  
  916. }
  917. }
  918. void tflux(int _ny, int _nx, fpdtype_t* __restrict__ f_v, int lsdf, const fpdtype_t* __restrict__ smats_v, int lsdsmats, const fpdtype_t* __restrict__ u_v, int lsdu)
  919. {
  920. #pragma omp parallel
  921. {
  922. int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);
  923. int rb, re, cb, ce;
  924. loop_sched_2d(_ny, _nx, align, &rb, &re, &cb, &ce);
  925. for (int _y = rb; _y < re; _y++)
  926. {
  927. tflux_inner(ce - cb, f_v + ((0*_ny + _y)*4 + 0)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 1)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 2)*lsdf + cb, f_v + ((0*_ny + _y)*4 + 3)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 0)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 1)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 2)*lsdf + cb, f_v + ((1*_ny + _y)*4 + 3)*lsdf + cb, smats_v + ((0*_ny + _y)*2 + 0)*lsdsmats + cb, smats_v + ((0*_ny + _y)*2 + 1)*lsdsmats + cb, smats_v + ((1*_ny + _y)*2 + 0)*lsdsmats + cb, smats_v + ((1*_ny + _y)*2 + 1)*lsdsmats + cb, u_v + (_y*4 + 0)*lsdu + cb, u_v + (_y*4 + 1)*lsdu + cb, u_v + (_y*4 + 2)*lsdu + cb, u_v + (_y*4 + 3)*lsdu + cb);
  928. }
  929. }
  930. }
  931.  
  932. [<class 'numpy.int32'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int32'>]
  933. --args
  934.  
  935. --kwargs
  936.  
  937. res cache key bad
  938. <_FuncPtr object at 0x6fffe569688>
  939. new res cache key
  940.  
  941. <pyfr.util.memoize object at 0x6fffebbf470>
  942. --self
  943. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> intcflux pyfr.solvers.euler.kernels.intcflux {'ndims': 2, 'rsolver': 'rusanov', 'c': OrderedDict([('gamma', 1.4), ('S', 13.5), ('M', 0.4), ('R', 1.5)]), 'nvars': 4}
  944. --args
  945.  
  946. --kwargs
  947.  
  948. res cache key bad
  949. ('\n\n#include <omp.h>\n#include <stdlib.h>\n#include <tgmath.h>\n\n#define PYFR_ALIGN_BYTES 32\n#define PYFR_NOINLINE __attribute__ ((noinline))\n\n#define min(a, b) ((a) < (b) ? (a) : (b))\n#define max(a, b) ((a) > (b) ? (a) : (b))\n\n// Typedefs\ntypedef double fpdtype_t;\n\n// OpenMP static loop scheduling functions\n\nstatic inline int\ngcd(int a, int b)\n{\n return (a == 0) ? b : gcd(b % a, a);\n}\n\nstatic inline void\nloop_sched_1d(int n, int align, int *b, int *e)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Round up n to be a multiple of nth\n int rn = n + nth - 1 - (n - 1) % nth;\n\n // Nominal tile size\n int sz = rn / nth;\n\n // Handle alignment\n sz += align - 1 - (sz - 1) % align;\n\n // Assign the starting and ending index\n *b = sz * tid;\n *e = min(*b + sz, n);\n\n // Clamp\n if (*b >= n)\n *b = *e = 0;\n}\n\nstatic inline void\nloop_sched_2d(int nrow, int ncol, int colalign,\n int *rowb, int *rowe, int *colb, int *cole)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Distribute threads\n int nrowth = gcd(nrow, nth);\n int ncolth = nth / nrowth;\n\n // Row and column indices for our thread\n int rowix = tid / ncolth;\n int colix = tid % ncolth;\n\n // Round up ncol to be a multiple of ncolth\n int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;\n\n // Nominal tile size\n int ntilerow = nrow / nrowth;\n int ntilecol = rncol / ncolth;\n\n // Handle column alignment\n ntilecol += colalign - 1 - (ntilecol - 1) % colalign;\n\n // Assign the starting and ending row to each thread\n *rowb = ntilerow * rowix;\n *rowe = *rowb + ntilerow;\n\n // Assign the starting and ending column to each thread\n *colb = ntilecol * colix;\n *cole = min(*colb + ntilecol, ncol);\n\n // Clamp\n if (*colb >= ncol)\n *colb = *cole = 0;\n}\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n void intcflux(int _nx, const fpdtype_t* __restrict__ magnl_v, const fpdtype_t* __restrict__ nl_v, int lsdnl, fpdtype_t* __restrict__ ul_v, const int* __restrict__ ul_vix, const int* __restrict__ ul_vcstri, fpdtype_t* __restrict__ ur_v, const int* __restrict__ ur_vix, const int* __restrict__ ur_vcstri)\n {\n #pragma omp parallel\n {\n int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);\n int cb, ce;\n loop_sched_1d(_nx, align, &cb, &ce);\n for (int _x = cb; _x < ce; _x++)\n {\n \n // Perform the Riemann solve\n fpdtype_t fn[4];\n {\n\n // Compute the left and right fluxes + velocities and pressures\n fpdtype_t fl_[2][4], fr_[2][4];\n fpdtype_t vl_[2], vr_[2];\n fpdtype_t pl_, pr_;\n\n {\n\n fpdtype_t invrho__ = 1.0/ul_v[ul_vix[_x] + ul_vcstri[_x]*0], E__ = ul_v[ul_vix[_x] + ul_vcstri[_x]*3];\n\n // Compute the velocities\n fpdtype_t rhov__[2];\n rhov__[0] = ul_v[ul_vix[_x] + ul_vcstri[_x]*1];\n vl_[0] = invrho__*rhov__[0];\n rhov__[1] = ul_v[ul_vix[_x] + ul_vcstri[_x]*2];\n vl_[1] = invrho__*rhov__[1];\n\n // Compute the pressure\n pl_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));\n\n // Density and energy fluxes\n fl_[0][0] = rhov__[0];\n fl_[0][3] = (E__ + pl_)*vl_[0];\n fl_[1][0] = rhov__[1];\n fl_[1][3] = (E__ + pl_)*vl_[1];\n\n // Momentum fluxes\n fl_[0][1] = rhov__[0]*vl_[0] + pl_;\n fl_[0][2] = rhov__[0]*vl_[1];\n fl_[1][1] = rhov__[1]*vl_[0];\n fl_[1][2] = rhov__[1]*vl_[1] + pl_;\n\n};\n {\n\n fpdtype_t invrho__ = 1.0/ur_v[ur_vix[_x] + ur_vcstri[_x]*0], E__ = ur_v[ur_vix[_x] + ur_vcstri[_x]*3];\n\n // Compute the velocities\n fpdtype_t rhov__[2];\n rhov__[0] = ur_v[ur_vix[_x] + ur_vcstri[_x]*1];\n vr_[0] = invrho__*rhov__[0];\n rhov__[1] = ur_v[ur_vix[_x] + ur_vcstri[_x]*2];\n vr_[1] = invrho__*rhov__[1];\n\n // Compute the pressure\n pr_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));\n\n // Density and energy fluxes\n fr_[0][0] = rhov__[0];\n fr_[0][3] = (E__ + pr_)*vr_[0];\n fr_[1][0] = rhov__[1];\n fr_[1][3] = (E__ + pr_)*vr_[1];\n\n // Momentum fluxes\n fr_[0][1] = rhov__[0]*vr_[0] + pr_;\n fr_[0][2] = rhov__[0]*vr_[1];\n fr_[1][1] = rhov__[1]*vr_[0];\n fr_[1][2] = rhov__[1]*vr_[1] + pr_;\n\n};\n\n // Sum the left and right velocities and take the normal\n fpdtype_t nv_ = ((nl_v[lsdnl*0 + _x])*(vl_[0] + vr_[0]) + (nl_v[lsdnl*1 + _x])*(vl_[1] + vr_[1]));\n\n // Estimate the maximum wave speed / 2\n fpdtype_t a_ = sqrt(0.35*(pl_ + pr_)/(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] + ur_v[ur_vix[_x] + ur_vcstri[_x]*0]))\n + 0.25*fabs(nv_);\n\n // Output\n fn[0] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][0] + fr_[0][0]) + nl_v[lsdnl*1 + _x]*(fl_[1][0] + fr_[1][0]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] - ur_v[ur_vix[_x] + ur_vcstri[_x]*0]);\n fn[1] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][1] + fr_[0][1]) + nl_v[lsdnl*1 + _x]*(fl_[1][1] + fr_[1][1]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*1] - ur_v[ur_vix[_x] + ur_vcstri[_x]*1]);\n fn[2] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][2] + fr_[0][2]) + nl_v[lsdnl*1 + _x]*(fl_[1][2] + fr_[1][2]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*2] - ur_v[ur_vix[_x] + ur_vcstri[_x]*2]);\n fn[3] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][3] + fr_[0][3]) + nl_v[lsdnl*1 + _x]*(fl_[1][3] + fr_[1][3]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*3] - ur_v[ur_vix[_x] + ur_vcstri[_x]*3]);\n\n};\n\n // Scale and write out the common normal fluxes\n ul_v[ul_vix[_x] + ul_vcstri[_x]*0] = magnl_v[_x]*fn[0];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*0] = -magnl_v[_x]*fn[0];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*1] = magnl_v[_x]*fn[1];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*1] = -magnl_v[_x]*fn[1];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*2] = magnl_v[_x]*fn[2];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*2] = -magnl_v[_x]*fn[2];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*3] = magnl_v[_x]*fn[3];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*3] = -magnl_v[_x]*fn[3];\n\n }\n }\n }\n\n', 1, ['_nx', 'magnl', 'nl', 'ul', 'ur'], [[<class 'numpy.int32'>], [<class 'numpy.int64'>], [<class 'numpy.int64'>, <class 'numpy.int32'>], [<class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>], [<class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>]])
  950. new res cache key
  951.  
  952. <pyfr.util.memoize object at 0x6fffe7267b8>
  953. --self
  954. <pyfr.backends.openmp.provider.OpenMPPointwiseKernelProvider object at 0x6fffe826f60> intcflux
  955.  
  956. #include <omp.h>
  957. #include <stdlib.h>
  958. #include <tgmath.h>
  959.  
  960. #define PYFR_ALIGN_BYTES 32
  961. #define PYFR_NOINLINE __attribute__ ((noinline))
  962.  
  963. #define min(a, b) ((a) < (b) ? (a) : (b))
  964. #define max(a, b) ((a) > (b) ? (a) : (b))
  965.  
  966. // Typedefs
  967. typedef double fpdtype_t;
  968.  
  969. // OpenMP static loop scheduling functions
  970.  
  971. static inline int
  972. gcd(int a, int b)
  973. {
  974. return (a == 0) ? b : gcd(b % a, a);
  975. }
  976.  
  977. static inline void
  978. loop_sched_1d(int n, int align, int *b, int *e)
  979. {
  980. int tid = omp_get_thread_num();
  981. int nth = omp_get_num_threads();
  982.  
  983. // Round up n to be a multiple of nth
  984. int rn = n + nth - 1 - (n - 1) % nth;
  985.  
  986. // Nominal tile size
  987. int sz = rn / nth;
  988.  
  989. // Handle alignment
  990. sz += align - 1 - (sz - 1) % align;
  991.  
  992. // Assign the starting and ending index
  993. *b = sz * tid;
  994. *e = min(*b + sz, n);
  995.  
  996. // Clamp
  997. if (*b >= n)
  998. *b = *e = 0;
  999. }
  1000.  
  1001. static inline void
  1002. loop_sched_2d(int nrow, int ncol, int colalign,
  1003. int *rowb, int *rowe, int *colb, int *cole)
  1004. {
  1005. int tid = omp_get_thread_num();
  1006. int nth = omp_get_num_threads();
  1007.  
  1008. // Distribute threads
  1009. int nrowth = gcd(nrow, nth);
  1010. int ncolth = nth / nrowth;
  1011.  
  1012. // Row and column indices for our thread
  1013. int rowix = tid / ncolth;
  1014. int colix = tid % ncolth;
  1015.  
  1016. // Round up ncol to be a multiple of ncolth
  1017. int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;
  1018.  
  1019. // Nominal tile size
  1020. int ntilerow = nrow / nrowth;
  1021. int ntilecol = rncol / ncolth;
  1022.  
  1023. // Handle column alignment
  1024. ntilecol += colalign - 1 - (ntilecol - 1) % colalign;
  1025.  
  1026. // Assign the starting and ending row to each thread
  1027. *rowb = ntilerow * rowix;
  1028. *rowe = *rowb + ntilerow;
  1029.  
  1030. // Assign the starting and ending column to each thread
  1031. *colb = ntilecol * colix;
  1032. *cole = min(*colb + ntilecol, ncol);
  1033.  
  1034. // Clamp
  1035. if (*colb >= ncol)
  1036. *colb = *cole = 0;
  1037. }
  1038.  
  1039.  
  1040.  
  1041.  
  1042.  
  1043.  
  1044.  
  1045.  
  1046.  
  1047.  
  1048.  
  1049.  
  1050.  
  1051.  
  1052.  
  1053. void intcflux(int _nx, const fpdtype_t* __restrict__ magnl_v, const fpdtype_t* __restrict__ nl_v, int lsdnl, fpdtype_t* __restrict__ ul_v, const int* __restrict__ ul_vix, const int* __restrict__ ul_vcstri, fpdtype_t* __restrict__ ur_v, const int* __restrict__ ur_vix, const int* __restrict__ ur_vcstri)
  1054. {
  1055. #pragma omp parallel
  1056. {
  1057. int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);
  1058. int cb, ce;
  1059. loop_sched_1d(_nx, align, &cb, &ce);
  1060. for (int _x = cb; _x < ce; _x++)
  1061. {
  1062.  
  1063. // Perform the Riemann solve
  1064. fpdtype_t fn[4];
  1065. {
  1066.  
  1067. // Compute the left and right fluxes + velocities and pressures
  1068. fpdtype_t fl_[2][4], fr_[2][4];
  1069. fpdtype_t vl_[2], vr_[2];
  1070. fpdtype_t pl_, pr_;
  1071.  
  1072. {
  1073.  
  1074. fpdtype_t invrho__ = 1.0/ul_v[ul_vix[_x] + ul_vcstri[_x]*0], E__ = ul_v[ul_vix[_x] + ul_vcstri[_x]*3];
  1075.  
  1076. // Compute the velocities
  1077. fpdtype_t rhov__[2];
  1078. rhov__[0] = ul_v[ul_vix[_x] + ul_vcstri[_x]*1];
  1079. vl_[0] = invrho__*rhov__[0];
  1080. rhov__[1] = ul_v[ul_vix[_x] + ul_vcstri[_x]*2];
  1081. vl_[1] = invrho__*rhov__[1];
  1082.  
  1083. // Compute the pressure
  1084. pl_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));
  1085.  
  1086. // Density and energy fluxes
  1087. fl_[0][0] = rhov__[0];
  1088. fl_[0][3] = (E__ + pl_)*vl_[0];
  1089. fl_[1][0] = rhov__[1];
  1090. fl_[1][3] = (E__ + pl_)*vl_[1];
  1091.  
  1092. // Momentum fluxes
  1093. fl_[0][1] = rhov__[0]*vl_[0] + pl_;
  1094. fl_[0][2] = rhov__[0]*vl_[1];
  1095. fl_[1][1] = rhov__[1]*vl_[0];
  1096. fl_[1][2] = rhov__[1]*vl_[1] + pl_;
  1097.  
  1098. };
  1099. {
  1100.  
  1101. fpdtype_t invrho__ = 1.0/ur_v[ur_vix[_x] + ur_vcstri[_x]*0], E__ = ur_v[ur_vix[_x] + ur_vcstri[_x]*3];
  1102.  
  1103. // Compute the velocities
  1104. fpdtype_t rhov__[2];
  1105. rhov__[0] = ur_v[ur_vix[_x] + ur_vcstri[_x]*1];
  1106. vr_[0] = invrho__*rhov__[0];
  1107. rhov__[1] = ur_v[ur_vix[_x] + ur_vcstri[_x]*2];
  1108. vr_[1] = invrho__*rhov__[1];
  1109.  
  1110. // Compute the pressure
  1111. pr_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));
  1112.  
  1113. // Density and energy fluxes
  1114. fr_[0][0] = rhov__[0];
  1115. fr_[0][3] = (E__ + pr_)*vr_[0];
  1116. fr_[1][0] = rhov__[1];
  1117. fr_[1][3] = (E__ + pr_)*vr_[1];
  1118.  
  1119. // Momentum fluxes
  1120. fr_[0][1] = rhov__[0]*vr_[0] + pr_;
  1121. fr_[0][2] = rhov__[0]*vr_[1];
  1122. fr_[1][1] = rhov__[1]*vr_[0];
  1123. fr_[1][2] = rhov__[1]*vr_[1] + pr_;
  1124.  
  1125. };
  1126.  
  1127. // Sum the left and right velocities and take the normal
  1128. fpdtype_t nv_ = ((nl_v[lsdnl*0 + _x])*(vl_[0] + vr_[0]) + (nl_v[lsdnl*1 + _x])*(vl_[1] + vr_[1]));
  1129.  
  1130. // Estimate the maximum wave speed / 2
  1131. fpdtype_t a_ = sqrt(0.35*(pl_ + pr_)/(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] + ur_v[ur_vix[_x] + ur_vcstri[_x]*0]))
  1132. + 0.25*fabs(nv_);
  1133.  
  1134. // Output
  1135. fn[0] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][0] + fr_[0][0]) + nl_v[lsdnl*1 + _x]*(fl_[1][0] + fr_[1][0]))
  1136. + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] - ur_v[ur_vix[_x] + ur_vcstri[_x]*0]);
  1137. fn[1] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][1] + fr_[0][1]) + nl_v[lsdnl*1 + _x]*(fl_[1][1] + fr_[1][1]))
  1138. + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*1] - ur_v[ur_vix[_x] + ur_vcstri[_x]*1]);
  1139. fn[2] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][2] + fr_[0][2]) + nl_v[lsdnl*1 + _x]*(fl_[1][2] + fr_[1][2]))
  1140. + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*2] - ur_v[ur_vix[_x] + ur_vcstri[_x]*2]);
  1141. fn[3] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][3] + fr_[0][3]) + nl_v[lsdnl*1 + _x]*(fl_[1][3] + fr_[1][3]))
  1142. + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*3] - ur_v[ur_vix[_x] + ur_vcstri[_x]*3]);
  1143.  
  1144. };
  1145.  
  1146. // Scale and write out the common normal fluxes
  1147. ul_v[ul_vix[_x] + ul_vcstri[_x]*0] = magnl_v[_x]*fn[0];
  1148. ur_v[ur_vix[_x] + ur_vcstri[_x]*0] = -magnl_v[_x]*fn[0];
  1149. ul_v[ul_vix[_x] + ul_vcstri[_x]*1] = magnl_v[_x]*fn[1];
  1150. ur_v[ur_vix[_x] + ur_vcstri[_x]*1] = -magnl_v[_x]*fn[1];
  1151. ul_v[ul_vix[_x] + ul_vcstri[_x]*2] = magnl_v[_x]*fn[2];
  1152. ur_v[ur_vix[_x] + ur_vcstri[_x]*2] = -magnl_v[_x]*fn[2];
  1153. ul_v[ul_vix[_x] + ul_vcstri[_x]*3] = magnl_v[_x]*fn[3];
  1154. ur_v[ur_vix[_x] + ur_vcstri[_x]*3] = -magnl_v[_x]*fn[3];
  1155.  
  1156. }
  1157. }
  1158. }
  1159.  
  1160. [<class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int32'>, <class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>, <class 'numpy.int64'>]
  1161. --args
  1162.  
  1163. --kwargs
  1164.  
  1165. 0 [main] python3 4796 child_info_fork::abort: unable to remap libtmp.so to same address as parent (0x1F0000) - try running rebaseall
  1166. Traceback (most recent call last):
  1167. File "/home/User/PyFR/env4/src/pyfr/pyfr/util.py", line 39, in __call__
  1168. res = cache[key]
  1169. KeyError: (<function OpenMPKernelProvider._build_kernel at 0x6fffe54a6a8>, b'\x80\x03X\x08\x00\x00\x00intcfluxq\x00XZ\x18\x00\x00\n\n#include <omp.h>\n#include <stdlib.h>\n#include <tgmath.h>\n\n#define PYFR_ALIGN_BYTES 32\n#define PYFR_NOINLINE __attribute__ ((noinline))\n\n#define min(a, b) ((a) < (b) ? (a) : (b))\n#define max(a, b) ((a) > (b) ? (a) : (b))\n\n// Typedefs\ntypedef double fpdtype_t;\n\n// OpenMP static loop scheduling functions\n\nstatic inline int\ngcd(int a, int b)\n{\n return (a == 0) ? b : gcd(b % a, a);\n}\n\nstatic inline void\nloop_sched_1d(int n, int align, int *b, int *e)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Round up n to be a multiple of nth\n int rn = n + nth - 1 - (n - 1) % nth;\n\n // Nominal tile size\n int sz = rn / nth;\n\n // Handle alignment\n sz += align - 1 - (sz - 1) % align;\n\n // Assign the starting and ending index\n *b = sz * tid;\n *e = min(*b + sz, n);\n\n // Clamp\n if (*b >= n)\n *b = *e = 0;\n}\n\nstatic inline void\nloop_sched_2d(int nrow, int ncol, int colalign,\n int *rowb, int *rowe, int *colb, int *cole)\n{\n int tid = omp_get_thread_num();\n int nth = omp_get_num_threads();\n\n // Distribute threads\n int nrowth = gcd(nrow, nth);\n int ncolth = nth / nrowth;\n\n // Row and column indices for our thread\n int rowix = tid / ncolth;\n int colix = tid % ncolth;\n\n // Round up ncol to be a multiple of ncolth\n int rncol = ncol + ncolth - 1 - (ncol - 1) % ncolth;\n\n // Nominal tile size\n int ntilerow = nrow / nrowth;\n int ntilecol = rncol / ncolth;\n\n // Handle column alignment\n ntilecol += colalign - 1 - (ntilecol - 1) % colalign;\n\n // Assign the starting and ending row to each thread\n *rowb = ntilerow * rowix;\n *rowe = *rowb + ntilerow;\n\n // Assign the starting and ending column to each thread\n *colb = ntilecol * colix;\n *cole = min(*colb + ntilecol, ncol);\n\n // Clamp\n if (*colb >= ncol)\n *colb = *cole = 0;\n}\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n void intcflux(int _nx, const fpdtype_t* __restrict__ magnl_v, const fpdtype_t* __restrict__ nl_v, int lsdnl, fpdtype_t* __restrict__ ul_v, const int* __restrict__ ul_vix, const int* __restrict__ ul_vcstri, fpdtype_t* __restrict__ ur_v, const int* __restrict__ ur_vix, const int* __restrict__ ur_vcstri)\n {\n #pragma omp parallel\n {\n int align = PYFR_ALIGN_BYTES / sizeof(fpdtype_t);\n int cb, ce;\n loop_sched_1d(_nx, align, &cb, &ce);\n for (int _x = cb; _x < ce; _x++)\n {\n \n // Perform the Riemann solve\n fpdtype_t fn[4];\n {\n\n // Compute the left and right fluxes + velocities and pressures\n fpdtype_t fl_[2][4], fr_[2][4];\n fpdtype_t vl_[2], vr_[2];\n fpdtype_t pl_, pr_;\n\n {\n\n fpdtype_t invrho__ = 1.0/ul_v[ul_vix[_x] + ul_vcstri[_x]*0], E__ = ul_v[ul_vix[_x] + ul_vcstri[_x]*3];\n\n // Compute the velocities\n fpdtype_t rhov__[2];\n rhov__[0] = ul_v[ul_vix[_x] + ul_vcstri[_x]*1];\n vl_[0] = invrho__*rhov__[0];\n rhov__[1] = ul_v[ul_vix[_x] + ul_vcstri[_x]*2];\n vl_[1] = invrho__*rhov__[1];\n\n // Compute the pressure\n pl_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));\n\n // Density and energy fluxes\n fl_[0][0] = rhov__[0];\n fl_[0][3] = (E__ + pl_)*vl_[0];\n fl_[1][0] = rhov__[1];\n fl_[1][3] = (E__ + pl_)*vl_[1];\n\n // Momentum fluxes\n fl_[0][1] = rhov__[0]*vl_[0] + pl_;\n fl_[0][2] = rhov__[0]*vl_[1];\n fl_[1][1] = rhov__[1]*vl_[0];\n fl_[1][2] = rhov__[1]*vl_[1] + pl_;\n\n};\n {\n\n fpdtype_t invrho__ = 1.0/ur_v[ur_vix[_x] + ur_vcstri[_x]*0], E__ = ur_v[ur_vix[_x] + ur_vcstri[_x]*3];\n\n // Compute the velocities\n fpdtype_t rhov__[2];\n rhov__[0] = ur_v[ur_vix[_x] + ur_vcstri[_x]*1];\n vr_[0] = invrho__*rhov__[0];\n rhov__[1] = ur_v[ur_vix[_x] + ur_vcstri[_x]*2];\n vr_[1] = invrho__*rhov__[1];\n\n // Compute the pressure\n pr_ = 0.3999999999999999*(E__ - 0.5*invrho__*((rhov__[0])*(rhov__[0]) + (rhov__[1])*(rhov__[1])));\n\n // Density and energy fluxes\n fr_[0][0] = rhov__[0];\n fr_[0][3] = (E__ + pr_)*vr_[0];\n fr_[1][0] = rhov__[1];\n fr_[1][3] = (E__ + pr_)*vr_[1];\n\n // Momentum fluxes\n fr_[0][1] = rhov__[0]*vr_[0] + pr_;\n fr_[0][2] = rhov__[0]*vr_[1];\n fr_[1][1] = rhov__[1]*vr_[0];\n fr_[1][2] = rhov__[1]*vr_[1] + pr_;\n\n};\n\n // Sum the left and right velocities and take the normal\n fpdtype_t nv_ = ((nl_v[lsdnl*0 + _x])*(vl_[0] + vr_[0]) + (nl_v[lsdnl*1 + _x])*(vl_[1] + vr_[1]));\n\n // Estimate the maximum wave speed / 2\n fpdtype_t a_ = sqrt(0.35*(pl_ + pr_)/(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] + ur_v[ur_vix[_x] + ur_vcstri[_x]*0]))\n + 0.25*fabs(nv_);\n\n // Output\n fn[0] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][0] + fr_[0][0]) + nl_v[lsdnl*1 + _x]*(fl_[1][0] + fr_[1][0]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*0] - ur_v[ur_vix[_x] + ur_vcstri[_x]*0]);\n fn[1] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][1] + fr_[0][1]) + nl_v[lsdnl*1 + _x]*(fl_[1][1] + fr_[1][1]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*1] - ur_v[ur_vix[_x] + ur_vcstri[_x]*1]);\n fn[2] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][2] + fr_[0][2]) + nl_v[lsdnl*1 + _x]*(fl_[1][2] + fr_[1][2]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*2] - ur_v[ur_vix[_x] + ur_vcstri[_x]*2]);\n fn[3] = 0.5*(nl_v[lsdnl*0 + _x]*(fl_[0][3] + fr_[0][3]) + nl_v[lsdnl*1 + _x]*(fl_[1][3] + fr_[1][3]))\n + a_*(ul_v[ul_vix[_x] + ul_vcstri[_x]*3] - ur_v[ur_vix[_x] + ur_vcstri[_x]*3]);\n\n};\n\n // Scale and write out the common normal fluxes\n ul_v[ul_vix[_x] + ul_vcstri[_x]*0] = magnl_v[_x]*fn[0];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*0] = -magnl_v[_x]*fn[0];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*1] = magnl_v[_x]*fn[1];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*1] = -magnl_v[_x]*fn[1];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*2] = magnl_v[_x]*fn[2];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*2] = -magnl_v[_x]*fn[2];\n ul_v[ul_vix[_x] + ul_vcstri[_x]*3] = magnl_v[_x]*fn[3];\n ur_v[ur_vix[_x] + ur_vcstri[_x]*3] = -magnl_v[_x]*fn[3];\n\n }\n }\n }\n\nq\x01]q\x02(cnumpy\nint32\nq\x03cnumpy\nint64\nq\x04h\x04h\x03h\x04h\x04h\x04h\x04h\x04h\x04e\x87q\x05.', b'\x80\x03}q\x00.')
  1170.  
  1171. During handling of the above exception, another exception occurred:
  1172.  
  1173. Traceback (most recent call last):
  1174. File "/home/User/PyFR/env4/lib/python3.4/site-packages/pytools/prefork.py", line 46, in call_capture_output
  1175. popen = Popen(cmdline, cwd=cwd, stdin=PIPE, stdout=PIPE, stderr=PIPE)
  1176. File "/usr/lib/python3.4/subprocess.py", line 859, in __init__
  1177. restore_signals, start_new_session)
  1178. File "/usr/lib/python3.4/subprocess.py", line 1395, in _execute_child
  1179. restore_signals, start_new_session, preexec_fn)
  1180. BlockingIOError: [Errno 11] Resource temporarily unavailable
  1181.  
  1182. During handling of the above exception, another exception occurred:
  1183.  
  1184. Traceback (most recent call last):
  1185. File "/home/User/PyFR/env4/bin/pyfr", line 9, in <module>
  1186. load_entry_point('pyfr==1.3.0', 'console_scripts', 'pyfr')()
  1187. File "/home/User/PyFR/env4/src/pyfr/pyfr/scripts/main.py", line 109, in main
  1188. args.process(args)
  1189. File "/home/User/PyFR/env4/src/pyfr/pyfr/scripts/main.py", line 230, in process_run
  1190. args, NativeReader(args.mesh), None, Inifile.load(args.cfg)
  1191. File "/home/User/PyFR/env4/src/pyfr/pyfr/scripts/main.py", line 214, in _process_common
  1192. solver = get_solver(backend, rallocs, mesh, soln, cfg)
  1193. File "/home/User/PyFR/env4/src/pyfr/pyfr/solvers/__init__.py", line 14, in get_solver
  1194. return get_integrator(backend, systemcls, rallocs, mesh, initsoln, cfg)
  1195. File "/home/User/PyFR/env4/src/pyfr/pyfr/integrators/__init__.py", line 26, in get_integrator
  1196. return integrator(backend, systemcls, rallocs, mesh, initsoln, cfg)
  1197. File "/home/User/PyFR/env4/src/pyfr/pyfr/integrators/controllers.py", line 14, in __init__
  1198. super().__init__(*args, **kwargs)
  1199. File "/home/User/PyFR/env4/src/pyfr/pyfr/integrators/steppers.py", line 9, in __init__
  1200. super().__init__(*args, **kwargs)
  1201. File "/home/User/PyFR/env4/src/pyfr/pyfr/integrators/base.py", line 41, in __init__
  1202. self.system = systemcls(backend, rallocs, mesh, initsoln, nreg, cfg)
  1203. File "/home/User/PyFR/env4/src/pyfr/pyfr/solvers/base/system.py", line 59, in __init__
  1204. self._gen_kernels(eles, int_inters, mpi_inters, bc_inters)
  1205. File "/home/User/PyFR/env4/src/pyfr/pyfr/solvers/base/system.py", line 167, in _gen_kernels
  1206. kernels[pn, kn].append(kgetter())
  1207. File "/home/User/PyFR/env4/src/pyfr/pyfr/solvers/euler/inters.py", line 21, in <lambda>
  1208. magnl=self._mag_pnorm_lhs, nl=self._norm_pnorm_lhs
  1209. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/base/backend.py", line 154, in kernel
  1210. return kern(*args, **kwargs)
  1211. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/base/kernels.py", line 162, in kernel_meth
  1212. fun = self._build_kernel(name, src, list(it.chain(*argt)))
  1213. File "/home/User/PyFR/env4/src/pyfr/pyfr/util.py", line 43, in __call__
  1214. res = cache[key] = self.func(*args, **kwargs)
  1215. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/openmp/provider.py", line 13, in _build_kernel
  1216. mod = GccSourceModule(src, self.backend.cfg)
  1217. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/openmp/compiler.py", line 61, in __init__
  1218. super().__init__(src, cfg)
  1219. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/openmp/compiler.py", line 30, in __init__
  1220. lname = self._build(tmpdir)
  1221. File "/home/User/PyFR/env4/src/pyfr/pyfr/backends/openmp/compiler.py", line 80, in _build
  1222. call_capture_output(cmd + self._cflags, cwd=tmpdir)
  1223. File "/home/User/PyFR/env4/lib/python3.4/site-packages/pytools/prefork.py", line 197, in call_capture_output
  1224. return forker[0].call_capture_output(cmdline, cwd, error_on_nonzero)
  1225. File "/home/User/PyFR/env4/lib/python3.4/site-packages/pytools/prefork.py", line 54, in call_capture_output
  1226. % ( " ".join(cmdline), e))
  1227. pytools.prefork.ExecError: error invoking 'gcc -shared -std=c99 -Ofast -march=native -fopenmp -fPIC -o libtmp.so tmp.c': [Errno 11] Resource temporarily unavailable
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement
Public Pastes
Advertisement
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!