1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673
|
#include "fffpy.h"
#include <stdarg.h>
#include <errno.h>
#define COPY_BUFFERS_USING_NUMPY 1
/* This function must be called before the module can work
because PyArray_API is defined static, in order not to share that symbol
within the dso. (import_array() asks the pointer value to the python process)
*/
void* fffpy_import_array(void) {
import_array();
}
/* Static functions */
static npy_intp _PyArray_main_axis(const PyArrayObject* x, int* ok);
static fff_vector* _fff_vector_new_from_buffer(const char* data, npy_intp dim, npy_intp stride, int type, int itemsize);
static fff_vector* _fff_vector_new_from_PyArrayIter(const PyArrayIterObject* it, npy_intp axis);
static void _fff_vector_sync_with_PyArrayIter(fff_vector* y, const PyArrayIterObject* it, npy_intp axis);
/* Routines for copying 1d arrays into contiguous double arrays */
#if COPY_BUFFERS_USING_NUMPY
# define COPY_BUFFER(y, data, stride, type, itemsize) \
fff_vector_fetch_using_NumPy(y, data, stride, type, itemsize);
#else
# define COPY_BUFFER(y, data, stride, type, itemsize) \
fff_vector_fetch(y, (void*)data, fff_datatype_fromNumPy(type), stride/itemsize)
#endif
/*
Copy a buffer using numpy.
Copy buffer x into y assuming that y is contiguous.
*/
void fff_vector_fetch_using_NumPy(fff_vector* y, const char* x, npy_intp stride, int type, int itemsize)
{
npy_intp dim[1] = {(npy_intp)y->size};
npy_intp strides[1] = {stride};
PyArrayObject* X = (PyArrayObject*) PyArray_New(&PyArray_Type, 1, dim, type, strides,
(void*)x, itemsize, NPY_BEHAVED, NULL);
PyArrayObject* Y = (PyArrayObject*) PyArray_SimpleNewFromData(1, dim, NPY_DOUBLE, (void*)y->data);
PyArray_CopyInto(Y, X);
Py_XDECREF(Y);
Py_XDECREF(X);
return;
}
/*
Create a fff_vector from an already allocated buffer. This function
acts as a fff_vector constructor that is compatible with
fff_vector_delete.
*/
static fff_vector* _fff_vector_new_from_buffer(const char* data, npy_intp dim, npy_intp stride, int type, int itemsize)
{
fff_vector* y;
size_t sizeof_double = sizeof(double);
/* If the input array is double and is aligned, just wrap without copying */
if ((type == NPY_DOUBLE) && (itemsize==sizeof_double)) {
y = (fff_vector*)malloc(sizeof(fff_vector));
y->size = (size_t)dim;
y->stride = (size_t)stride/sizeof_double;
y->data = (double*)data;
y->owner = 0;
}
/* Otherwise, output a owner contiguous vector with copied data */
else {
y = fff_vector_new((size_t)dim);
COPY_BUFFER(y, data, stride, type, itemsize);
}
return y;
}
/* Find the axis with largest dimension */
npy_intp _PyArray_main_axis(const PyArrayObject* x, int* ok)
{
npy_intp axis, count, i, dim, ndim = PyArray_NDIM(x);
*ok = 1;
axis = 0;
count = 0;
for(i=0; i<ndim; i++) {
dim = PyArray_DIM(x,i);
if (dim > 1) {
count ++;
axis = i;
}
}
if (count > 1)
*ok = 0;
return axis;
}
fff_vector* fff_vector_fromPyArray(const PyArrayObject* x)
{
fff_vector* y;
int ok;
npy_intp axis = _PyArray_main_axis(x, &ok);
if (!ok) {
FFF_ERROR("Input array is not a vector", EINVAL);
return NULL;
}
y = _fff_vector_new_from_buffer(PyArray_DATA(x),
PyArray_DIM(x, axis),
PyArray_STRIDE(x, axis),
PyArray_TYPE(x),
PyArray_ITEMSIZE(x));
return y;
}
/*
Export a fff_vector to a PyArray, and delete it. This function is a
fff_vector destructor compatible with any either fff_vector_new or
_fff_vector_new_from_buffer.
*/
PyArrayObject* fff_vector_toPyArray(fff_vector* y)
{
PyArrayObject* x;
size_t size;
npy_intp dims[1];
if (y == NULL)
return NULL;
size = y->size;
dims[0] = (npy_intp) size;
/* If the fff_vector is owner (hence contiguous), just pass the
buffer to Python and transfer ownership */
if (y->owner) {
x = (PyArrayObject*) PyArray_SimpleNewFromData(1, dims, NPY_DOUBLE, (void*)y->data);
x->flags = (x->flags) | NPY_OWNDATA;
}
/* Otherwise, create Python array from scratch */
else
x = fff_vector_const_toPyArray(y);
/* Ciao bella */
free(y);
return x;
}
/* Export without deleting */
PyArrayObject* fff_vector_const_toPyArray(const fff_vector* y)
{
PyArrayObject* x;
size_t i, size = y->size, stride = y->stride;
double* data = (double*) malloc(size*sizeof(double));
double* bufX = data;
double* bufY = y->data;
npy_intp dims[1];
dims[0] = (npy_intp) size;
for (i=0; i<size; i++, bufX++, bufY+=stride)
*bufX = *bufY;
x = (PyArrayObject*) PyArray_SimpleNewFromData(1, dims, NPY_DOUBLE, (void*)data);
x->flags = (x->flags) | NPY_OWNDATA;
return x;
}
/*
Get a fff_matrix from an input PyArray. This function acts as a
fff_vector constructor that is compatible with fff_vector_delete.
*/
fff_matrix* fff_matrix_fromPyArray(const PyArrayObject* x)
{
fff_matrix* y;
npy_intp dim[2];
PyArrayObject* xd;
/* Check that the input object is a two-dimensional array */
if (PyArray_NDIM(x) != 2) {
FFF_ERROR("Input array is not a matrix", EINVAL);
return NULL;
}
/* If the PyArray is double, contiguous and aligned just wrap without
copying */
if ((PyArray_TYPE(x) == NPY_DOUBLE) &&
(PyArray_ISCONTIGUOUS(x)) &&
(PyArray_ISALIGNED(x))) {
y = (fff_matrix*) malloc(sizeof(fff_matrix));
y->size1 = (size_t) PyArray_DIM(x,0);
y->size2 = (size_t) PyArray_DIM(x,1);
y->tda = y->size2;
y->data = PyArray_DATA(x);
y->owner = 0;
}
/* Otherwise, output a owner (contiguous) matrix with copied
data */
else {
size_t dim0 = PyArray_DIM(x,0), dim1 = PyArray_DIM(x,1);
y = fff_matrix_new((size_t)dim0, (size_t)dim1);
dim[0] = dim0;
dim[1] = dim1;
xd = (PyArrayObject*) PyArray_SimpleNewFromData(2, dim, NPY_DOUBLE, (void*)y->data);
PyArray_CopyInto(xd, (PyArrayObject*)x);
Py_XDECREF(xd);
}
return y;
}
/*
Export a fff_matrix to a PyArray, and delete it. This function is a
fff_matrix destructor compatible with any of the following
constructors: fff_matrix_new and fff_matrix_fromPyArray.
*/
PyArrayObject* fff_matrix_toPyArray(fff_matrix* y)
{
PyArrayObject* x;
size_t size1;
size_t size2;
size_t tda;
npy_intp dims[2];
if (y == NULL)
return NULL;
size1 = y->size1;
size2 = y->size2;
tda = y->tda;
dims[0] = (npy_intp) size1;
dims[1] = (npy_intp) size2;
/* If the fff_matrix is contiguous and owner, just pass the
buffer to Python and transfer ownership */
if ((tda == size2) && (y->owner)) {
x = (PyArrayObject*) PyArray_SimpleNewFromData(2, dims, NPY_DOUBLE, (void*)y->data);
x->flags = (x->flags) | NPY_OWNDATA;
}
/* Otherwise, create PyArray from scratch. Note, the input
fff_matrix is necessarily in row-major order. */
else
x = fff_matrix_const_toPyArray(y);
/* Ciao bella */
free(y);
return x;
}
/* Export without deleting */
PyArrayObject* fff_matrix_const_toPyArray(const fff_matrix* y)
{
PyArrayObject* x;
size_t size1 = y->size1, size2 = y->size2, tda = y->tda;
size_t i, j, pos;
double* data = (double*) malloc(size1*size2*sizeof(double));
double* bufX = data;
double* bufY = y->data;
npy_intp dims[2];
dims[0] = (npy_intp) size1;
dims[1] = (npy_intp) size2;
for (i=0; i<size1; i++) {
pos = tda*i;
for (j=0; j<size2; j++, bufX++, pos++)
*bufX = bufY[pos];
}
x = (PyArrayObject*) PyArray_SimpleNewFromData(2, dims, NPY_DOUBLE, (void*)data);
x->flags = (x->flags) | NPY_OWNDATA;
return x;
}
/** Static routines **/
/**** Data type conversions *****/
fff_datatype fff_datatype_fromNumPy(int npy_type)
{
fff_datatype fff_type;
switch (npy_type) {
case NPY_UBYTE:
fff_type = FFF_UCHAR;
break;
case NPY_BYTE:
fff_type = FFF_SCHAR;
break;
case NPY_USHORT:
fff_type = FFF_USHORT;
break;
case NPY_SHORT:
fff_type = FFF_SSHORT;
break;
case NPY_UINT:
fff_type = FFF_UINT;
break;
case NPY_INT:
fff_type = FFF_INT;
break;
case NPY_ULONG:
fff_type = FFF_ULONG;
break;
case NPY_LONG:
fff_type = FFF_LONG;
break;
case NPY_FLOAT:
fff_type = FFF_FLOAT;
break;
case NPY_DOUBLE:
fff_type = FFF_DOUBLE;
break;
default:
fff_type = FFF_UNKNOWN_TYPE;
break;
}
/* Return the datatype */
return fff_type;
}
int fff_datatype_toNumPy(fff_datatype fff_type)
{
int npy_type;
switch(fff_type) {
case FFF_UCHAR:
npy_type = NPY_UBYTE;
break;
case FFF_SCHAR:
npy_type = NPY_BYTE;
break;
case FFF_USHORT:
npy_type = NPY_USHORT;
break;
case FFF_SSHORT:
npy_type = NPY_SHORT;
break;
case FFF_UINT:
npy_type = NPY_UINT;
break;
case FFF_INT:
npy_type = NPY_INT;
break;
case FFF_ULONG:
npy_type = NPY_ULONG;
break;
case FFF_LONG:
npy_type = NPY_LONG;
break;
case FFF_FLOAT:
npy_type = NPY_FLOAT;
break;
case FFF_DOUBLE:
npy_type = NPY_DOUBLE;
break;
default:
npy_type = NPY_NOTYPE;
break;
}
return npy_type;
}
/**** fff_array interface ****/
fff_array* fff_array_fromPyArray(const PyArrayObject* x)
{
fff_array* y;
fff_datatype datatype;
unsigned int nbytes;
size_t dimX = 1, dimY = 1, dimZ = 1, dimT = 1;
size_t offX = 0, offY = 0, offZ = 0, offT = 0;
size_t ndims = (size_t)PyArray_NDIM(x);
/* Check that the input array has less than four dimensions */
if (ndims > 4) {
FFF_ERROR("Input array has more than four dimensions", EINVAL);
return NULL;
}
/* Check that the input array is aligned */
if (! PyArray_ISALIGNED(x)) {
FFF_ERROR("Input array is not aligned", EINVAL);
return NULL;
}
/* Match the data type */
datatype = fff_datatype_fromNumPy(PyArray_TYPE(x));
if (datatype == FFF_UNKNOWN_TYPE) {
FFF_ERROR("Unrecognized data type", EINVAL);
return NULL;
}
/* Dimensions and offsets */
nbytes = fff_nbytes(datatype);
dimX = PyArray_DIM(x, 0);
offX = PyArray_STRIDE(x, 0)/nbytes;
if (ndims > 1) {
dimY = PyArray_DIM(x, 1);
offY = PyArray_STRIDE(x, 1)/nbytes;
if (ndims > 2) {
dimZ = PyArray_DIM(x, 2);
offZ = PyArray_STRIDE(x, 2)/nbytes;
if (ndims > 3) {
dimT = PyArray_DIM(x, 3);
offT = PyArray_STRIDE(x, 3)/nbytes;
}
}
}
/* Create array (not owner) */
y = (fff_array*)malloc(sizeof(fff_array));
*y = fff_array_view(datatype,
PyArray_DATA(x),
dimX, dimY, dimZ, dimT,
offX, offY, offZ, offT);
return y;
}
PyArrayObject* fff_array_toPyArray(fff_array* y)
{
PyArrayObject* x;
npy_intp dims[4];
int datatype;
fff_array* yy;
if (y == NULL)
return NULL;
dims[0] = y->dimX;
dims[1] = y->dimY;
dims[2] = y->dimZ;
dims[3] = y->dimT;
/* Match data type */
datatype = fff_datatype_toNumPy(y->datatype);
if (datatype == NPY_NOTYPE) {
FFF_ERROR("Unrecognized data type", EINVAL);
return NULL;
}
/* Make sure the fff array owns its data, which may require a copy */
if (y->owner)
yy = y;
else {
yy = fff_array_new(y->datatype, y->dimX, y->dimY, y->dimZ, y->dimT);
fff_array_copy(yy, y);
}
/*
Create a Python array from the array data (which is contiguous
since it is owner). We can use PyArray_SimpleNewFromData given
that yy is C-contiguous by fff_array_new.
*/
x = (PyArrayObject*) PyArray_SimpleNewFromData(yy->ndims, dims, datatype, (void*)yy->data);
/* Transfer ownership to Python */
x->flags = (x->flags) | NPY_OWNDATA;
/* Dealloc memory if needed */
if (! y->owner)
free(yy);
/* Delete array */
free(y);
return x;
}
/********************************************************************
Multi-iterator object.
********************************************************************/
static int _PyArray_BroadcastAllButAxis (PyArrayMultiIterObject* mit, int axis);
/*
Create a fff multi iterator object.
Involves creating a PyArrayMultiArrayIter instance that lets us
iterate simultaneously on an arbitrary number of numpy arrays
EXCEPT in one common axis.
There does not seem to exist a built-in PyArrayMultiArrayIter
constructor for this usage. If it pops up one day, part of the
following code should be replaced.
Similarly to the default PyArrayMultiArrayIter constructor, we need
to set up broadcasting rules. For now, we simply impose that all
arrays have exactly the same number of dimensions and that all
dimensions be equal except along the "non-iterated" axis.
FIXME: The following code does not perform any checking, and will
surely crash if the arrays do not fulfill the conditions.
*/
fffpy_multi_iterator* fffpy_multi_iterator_new(int narr, int axis, ...)
{
fffpy_multi_iterator* thisone;
va_list va;
fff_vector** vector;
PyArrayMultiIterObject *multi;
PyObject *current, *arr;
int i, err=0;
/* Create new instance */
thisone = (fffpy_multi_iterator*)malloc(sizeof(fffpy_multi_iterator));
/* Static size of PyArrayMultiIterObject.
*
* https://github.com/numpy/numpy/issues/26765#issuecomment-2391737671
*/
multi = PyArray_malloc(PyArrayMultiIter_Type.tp_basicsize);
vector = (fff_vector**)malloc(narr*sizeof(fff_vector*));
/* Initialize the PyArrayMultiIterObject instance from the variadic arguments */
PyObject_Init((PyObject *)multi, &PyArrayMultiIter_Type);
for (i=0; i<narr; i++)
multi->iters[i] = NULL;
multi->numiter = narr;
multi->index = 0;
va_start(va, axis);
for (i=0; i<narr; i++) {
current = va_arg(va, PyObject *);
arr = PyArray_FROM_O(current);
if (arr==NULL) {
err=1; break;
}
else {
multi->iters[i] = (PyArrayIterObject *)PyArray_IterAllButAxis(arr, &axis);
Py_DECREF(arr);
}
}
va_end(va);
/* Test */
if (!err && _PyArray_BroadcastAllButAxis(multi, axis) < 0)
err=1;
if (err) {
FFF_ERROR("Cannot create broadcast object", ENOMEM);
free(thisone);
free(vector);
Py_DECREF(multi);
return NULL;
}
/* Initialize the multi iterator */
PyArray_MultiIter_RESET(multi);
/* Create the fff vectors (views or copies) */
for(i=0; i<narr; i++)
vector[i] = _fff_vector_new_from_PyArrayIter((const PyArrayIterObject*)multi->iters[i], axis);
/* Instantiate fiels */
thisone->narr = narr;
thisone->axis = axis;
thisone->vector = vector;
thisone->multi = multi;
thisone->index = thisone->multi->index;
thisone->size = thisone->multi->size;
return thisone;
}
void fffpy_multi_iterator_delete(fffpy_multi_iterator* thisone)
{
unsigned int i;
Py_DECREF(thisone->multi);
for(i=0; i<thisone->narr; i++)
fff_vector_delete(thisone->vector[i]);
free(thisone->vector);
free(thisone);
return;
}
void fffpy_multi_iterator_update(fffpy_multi_iterator* thisone)
{
unsigned int i;
PyArray_MultiIter_NEXT(thisone->multi);
for(i=0; i<thisone->narr; i++)
_fff_vector_sync_with_PyArrayIter(thisone->vector[i], (const PyArrayIterObject*)thisone->multi->iters[i], thisone->axis);
thisone->index = thisone->multi->index;
return;
}
void fffpy_multi_iterator_reset(fffpy_multi_iterator* thisone)
{
unsigned int i;
PyArray_MultiIter_RESET(thisone->multi);
for(i=0; i<thisone->narr; i++)
_fff_vector_sync_with_PyArrayIter(thisone->vector[i], (const PyArrayIterObject*)thisone->multi->iters[i], thisone->axis);
thisone->index = thisone->multi->index;
return;
}
static int _PyArray_BroadcastAllButAxis (PyArrayMultiIterObject* mit, int axis)
{
int i, nd;
npy_intp size, tmp;
PyArrayIterObject *it;
/* Not very robust */
it = mit->iters[0];
/* Set the dimensions */
nd = it->ao->nd;
mit->nd = nd;
for(i=0, size=1; i<nd; i++) {
tmp = it->ao->dimensions[i];
mit->dimensions[i] = tmp;
if (i!=axis)
size *= tmp;
}
mit->size = size;
/* Not very robust either */
return 0;
}
/* Create an fff_vector from a PyArrayIter object */
fff_vector* _fff_vector_new_from_PyArrayIter(const PyArrayIterObject* it, npy_intp axis)
{
fff_vector* y;
char* data = PyArray_ITER_DATA(it);
PyArrayObject* ao = (PyArrayObject*) it->ao;
npy_intp dim = PyArray_DIM(ao, axis);
npy_intp stride = PyArray_STRIDE(ao, axis);
int type = PyArray_TYPE(ao);
int itemsize = PyArray_ITEMSIZE(ao);
y = _fff_vector_new_from_buffer(data, dim, stride, type, itemsize);
return y;
}
/* Fetch vector data from an iterator (view or copy) */
void _fff_vector_sync_with_PyArrayIter(fff_vector* y, const PyArrayIterObject* it, npy_intp axis)
{
if (y->owner) {
PyArrayObject* ao = (PyArrayObject*) it->ao;
COPY_BUFFER(y, PyArray_ITER_DATA(it), PyArray_STRIDE(ao, axis),
PyArray_TYPE(ao), PyArray_ITEMSIZE(ao));
}
else
y->data = (double*) PyArray_ITER_DATA(it);
return;
}
|