Newer
Older
#ifndef _AFJHAYYTA_PYOPENCL_HEADER_SEEN_WRAP_CL_HPP
#define _AFJHAYYTA_PYOPENCL_HEADER_SEEN_WRAP_CL_HPP
// CL 1.2 undecided:
// clSetPrintfCallback
// {{{ includes
#define CL_USE_DEPRECATED_OPENCL_1_1_APIS
// #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION
#ifdef __APPLE__
// Mac ------------------------------------------------------------------------
#include <OpenCL/opencl.h>
#ifdef HAVE_GL
#define PYOPENCL_GL_SHARING_VERSION 1
#include <OpenGL/OpenGL.h>
#include <OpenCL/cl_gl.h>
#include <OpenCL/cl_gl_ext.h>
#endif
#else
// elsewhere ------------------------------------------------------------------
#define CL_TARGET_OPENCL_VERSION 220
#if defined(_WIN32)
#define NOMINMAX
#include <windows.h>
#endif
#ifdef HAVE_GL
#include <GL/gl.h>
#include <CL/cl_gl.h>
#endif
#if defined(cl_khr_gl_sharing) && (cl_khr_gl_sharing >= 1)
#define PYOPENCL_GL_SHARING_VERSION cl_khr_gl_sharing
#endif
#endif
#include <stdexcept>
#include <iostream>
#include <vector>
#include <utility>
#include <numeric>
#include "wrap_helpers.hpp"
#include "numpy_init.hpp"
#include "tools.hpp"
#ifdef PYOPENCL_PRETEND_CL_VERSION
#define PYOPENCL_CL_VERSION PYOPENCL_PRETEND_CL_VERSION
#else
#if defined(CL_VERSION_2_2)
#define PYOPENCL_CL_VERSION 0x2020
#elif defined(CL_VERSION_2_1)
#define PYOPENCL_CL_VERSION 0x2010
#elif defined(CL_VERSION_2_0)
#define PYOPENCL_CL_VERSION 0x2000
#elif defined(CL_VERSION_1_2)
#define PYOPENCL_CL_VERSION 0x1020
#elif defined(CL_VERSION_1_1)
#define PYOPENCL_CL_VERSION 0x1010
#else
#define PYOPENCL_CL_VERSION 0x1000
#endif
#endif
#if PY_VERSION_HEX >= 0x03000000
#define PYOPENCL_USE_NEW_BUFFER_INTERFACE
#define PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(s) std::move(s)
#else
#define PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(s) (s)
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
// }}}
// {{{ tools
#if PY_VERSION_HEX >= 0x02050000
typedef Py_ssize_t PYOPENCL_BUFFER_SIZE_T;
#else
typedef int PYOPENCL_BUFFER_SIZE_T;
#endif
#define PYOPENCL_CAST_BOOL(B) ((B) ? CL_TRUE : CL_FALSE)
#define PYOPENCL_DEPRECATED(WHAT, KILL_VERSION, EXTRA_MSG) \
{ \
PyErr_Warn( \
PyExc_DeprecationWarning, \
WHAT " is deprecated and will stop working in PyOpenCL " KILL_VERSION". " \
EXTRA_MSG); \
}
#if PYOPENCL_CL_VERSION >= 0x1020
#define PYOPENCL_GET_EXT_FUN(PLATFORM, NAME, VAR) \
NAME##_fn VAR \
= (NAME##_fn) \
clGetExtensionFunctionAddressForPlatform(PLATFORM, #NAME); \
\
if (!VAR) \
throw error(#NAME, CL_INVALID_VALUE, #NAME \
"not available");
#else
#define PYOPENCL_GET_EXT_FUN(PLATFORM, NAME, VAR) \
NAME##_fn VAR \
= (NAME##_fn) \
clGetExtensionFunctionAddress(#NAME); \
\
if (!VAR) \
throw error(#NAME, CL_INVALID_VALUE, #NAME \
"not available");
#endif
#define PYOPENCL_PARSE_PY_DEVICES \
std::vector<cl_device_id> devices_vec; \
cl_uint num_devices; \
cl_device_id *devices; \
\
if (py_devices.ptr() == Py_None) \
{ \
num_devices = 0; \
devices = 0; \
} \
else \
{ \
for (py::handle py_dev: py_devices) \
(py_dev).cast<device &>().data()); \
devices = devices_vec.empty( ) ? nullptr : &devices_vec.front(); \
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
} \
#define PYOPENCL_RETRY_RETURN_IF_MEM_ERROR(OPERATION) \
try \
{ \
OPERATION \
} \
catch (pyopencl::error &e) \
{ \
if (!e.is_out_of_memory()) \
throw; \
} \
\
/* If we get here, we got an error from CL.
* We should run the Python GC to try and free up
* some memory references. */ \
run_python_gc(); \
\
/* Now retry the allocation. If it fails again,
* let it fail. */ \
{ \
OPERATION \
}
#define PYOPENCL_RETRY_IF_MEM_ERROR(OPERATION) \
{ \
bool failed_with_mem_error = false; \
try \
{ \
OPERATION \
} \
catch (pyopencl::error &e) \
{ \
failed_with_mem_error = true; \
if (!e.is_out_of_memory()) \
throw; \
} \
\
if (failed_with_mem_error) \
{ \
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
/* If we get here, we got an error from CL.
* We should run the Python GC to try and free up
* some memory references. */ \
run_python_gc(); \
\
/* Now retry the allocation. If it fails again,
* let it fail. */ \
{ \
OPERATION \
} \
} \
}
// }}}
// {{{ tracing and error reporting
#ifdef PYOPENCL_TRACE
#define PYOPENCL_PRINT_CALL_TRACE(NAME) \
std::cerr << NAME << std::endl;
#define PYOPENCL_PRINT_CALL_TRACE_INFO(NAME, EXTRA_INFO) \
std::cerr << NAME << " (" << EXTRA_INFO << ')' << std::endl;
#else
#define PYOPENCL_PRINT_CALL_TRACE(NAME) /*nothing*/
#define PYOPENCL_PRINT_CALL_TRACE_INFO(NAME, EXTRA_INFO) /*nothing*/
#endif
#define PYOPENCL_CALL_GUARDED_THREADED_WITH_TRACE_INFO(NAME, ARGLIST, TRACE_INFO) \
{ \
PYOPENCL_PRINT_CALL_TRACE_INFO(#NAME, TRACE_INFO); \
cl_int status_code; \
{ \
py::gil_scoped_release release; \
if (status_code != CL_SUCCESS) \
throw pyopencl::error(#NAME, status_code);\
}
#define PYOPENCL_CALL_GUARDED_WITH_TRACE_INFO(NAME, ARGLIST, TRACE_INFO) \
{ \
PYOPENCL_PRINT_CALL_TRACE_INFO(#NAME, TRACE_INFO); \
cl_int status_code; \
status_code = NAME ARGLIST; \
if (status_code != CL_SUCCESS) \
throw pyopencl::error(#NAME, status_code);\
}
#define PYOPENCL_CALL_GUARDED_THREADED(NAME, ARGLIST) \
{ \
PYOPENCL_PRINT_CALL_TRACE(#NAME); \
cl_int status_code; \
{ \
py::gil_scoped_release release; \
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
if (status_code != CL_SUCCESS) \
throw pyopencl::error(#NAME, status_code);\
}
#define PYOPENCL_CALL_GUARDED(NAME, ARGLIST) \
{ \
PYOPENCL_PRINT_CALL_TRACE(#NAME); \
cl_int status_code; \
status_code = NAME ARGLIST; \
if (status_code != CL_SUCCESS) \
throw pyopencl::error(#NAME, status_code);\
}
#define PYOPENCL_CALL_GUARDED_CLEANUP(NAME, ARGLIST) \
{ \
PYOPENCL_PRINT_CALL_TRACE(#NAME); \
cl_int status_code; \
status_code = NAME ARGLIST; \
if (status_code != CL_SUCCESS) \
std::cerr \
<< "PyOpenCL WARNING: a clean-up operation failed (dead context maybe?)" \
<< std::endl \
<< #NAME " failed with code " << status_code \
<< std::endl; \
}
// }}}
// {{{ get_info helpers
#define PYOPENCL_GET_OPAQUE_INFO(WHAT, FIRST_ARG, SECOND_ARG, CL_TYPE, TYPE) \
{ \
CL_TYPE param_value; \
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, sizeof(param_value), ¶m_value, 0)); \
if (param_value) \
return py::object(handle_from_new_ptr( \
new TYPE(param_value, /*retain*/ true))); \
else \
}
#define PYOPENCL_GET_VEC_INFO(WHAT, FIRST_ARG, SECOND_ARG, RES_VEC) \
{ \
size_t size; \
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, 0, 0, &size)); \
\
RES_VEC.resize(size / sizeof(RES_VEC.front())); \
\
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, size, \
RES_VEC.empty( ) ? nullptr : &RES_VEC.front(), &size)); \
}
#define PYOPENCL_GET_STR_INFO(WHAT, FIRST_ARG, SECOND_ARG) \
{ \
size_t param_value_size; \
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, 0, 0, ¶m_value_size)); \
\
std::vector<char> param_value(param_value_size); \
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, param_value_size, \
param_value.empty( ) ? nullptr : ¶m_value.front(), ¶m_value_size)); \
param_value.empty( ) ? "" : std::string(¶m_value.front(), param_value_size-1)); \
}
#define PYOPENCL_GET_INTEGRAL_INFO(WHAT, FIRST_ARG, SECOND_ARG, TYPE) \
{ \
TYPE param_value; \
PYOPENCL_CALL_GUARDED(clGet##WHAT##Info, \
(FIRST_ARG, SECOND_ARG, sizeof(param_value), ¶m_value, 0)); \
return py::cast(param_value); \
}
// }}}
// {{{ event helpers --------------------------------------------------------------
#define PYOPENCL_PARSE_WAIT_FOR \
cl_uint num_events_in_wait_list = 0; \
std::vector<cl_event> event_wait_list; \
\
if (py_wait_for.ptr() != Py_None) \
{ \
event_wait_list.resize(len(py_wait_for)); \
event_wait_list[num_events_in_wait_list++] = \
evt.cast<const event &>().data(); \
}
#define PYOPENCL_WAITLIST_ARGS \
num_events_in_wait_list, event_wait_list.empty( ) ? nullptr : &event_wait_list.front()
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
#define PYOPENCL_RETURN_NEW_NANNY_EVENT(evt, obj) \
try \
{ \
return new nanny_event(evt, false, obj); \
} \
catch (...) \
{ \
clReleaseEvent(evt); \
throw; \
}
#define PYOPENCL_RETURN_NEW_EVENT(evt) \
try \
{ \
return new event(evt, false); \
} \
catch (...) \
{ \
clReleaseEvent(evt); \
throw; \
}
// }}}
// {{{ equality testing
#define PYOPENCL_EQUALITY_TESTS(cls) \
bool operator==(cls const &other) const \
{ return data() == other.data(); } \
bool operator!=(cls const &other) const \
{ return data() != other.data(); } \
long hash() const \
{ return (long) (intptr_t) data(); }
// }}}
namespace pyopencl
{
// {{{ error
class error : public std::runtime_error
{
private:
const char *m_routine;
cl_int m_code;
public:
error(const char *rout, cl_int c, const char *msg="")
: std::runtime_error(msg), m_routine(rout), m_code(c)
{ }
const char *routine() const
{
return m_routine;
}
cl_int code() const
{
return m_code;
}
bool is_out_of_memory() const
{
return (code() == CL_MEM_OBJECT_ALLOCATION_FAILURE
|| code() == CL_OUT_OF_RESOURCES
|| code() == CL_OUT_OF_HOST_MEMORY);
}
};
// }}}
// {{{ buffer interface helper
//
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
class py_buffer_wrapper : public noncopyable
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
{
private:
bool m_initialized;
public:
Py_buffer m_buf;
py_buffer_wrapper()
: m_initialized(false)
{}
void get(PyObject *obj, int flags)
{
if (PyObject_GetBuffer(obj, &m_buf, flags))
throw py::error_already_set();
m_initialized = true;
}
virtual ~py_buffer_wrapper()
{
if (m_initialized)
PyBuffer_Release(&m_buf);
}
};
#endif
// }}}
inline
py::tuple get_cl_header_version()
{
return py::make_tuple(
PYOPENCL_CL_VERSION >> (3*4),
(PYOPENCL_CL_VERSION >> (1*4)) & 0xff
);
}
// {{{ platform
class platform : noncopyable
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
{
private:
cl_platform_id m_platform;
public:
platform(cl_platform_id pid)
: m_platform(pid)
{ }
platform(cl_platform_id pid, bool /*retain (ignored)*/)
: m_platform(pid)
{ }
cl_platform_id data() const
{
return m_platform;
}
PYOPENCL_EQUALITY_TESTS(platform);
py::object get_info(cl_platform_info param_name) const
{
switch (param_name)
{
case CL_PLATFORM_PROFILE:
case CL_PLATFORM_VERSION:
case CL_PLATFORM_NAME:
case CL_PLATFORM_VENDOR:
#if !(defined(CL_PLATFORM_NVIDIA) && CL_PLATFORM_NVIDIA == 0x3001)
case CL_PLATFORM_EXTENSIONS:
#endif
PYOPENCL_GET_STR_INFO(Platform, m_platform, param_name);
default:
throw error("Platform.get_info", CL_INVALID_VALUE);
}
}
py::list get_devices(cl_device_type devtype);
};
inline
py::list get_platforms()
{
cl_uint num_platforms = 0;
PYOPENCL_CALL_GUARDED(clGetPlatformIDs, (0, 0, &num_platforms));
std::vector<cl_platform_id> platforms(num_platforms);
PYOPENCL_CALL_GUARDED(clGetPlatformIDs,
(num_platforms, platforms.empty( ) ? nullptr : &platforms.front(), &num_platforms));
for (cl_platform_id pid: platforms)
result.append(handle_from_new_ptr(
new platform(pid)));
return result;
}
// }}}
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
{
public:
enum reference_type_t {
REF_NOT_OWNABLE,
#if PYOPENCL_CL_VERSION >= 0x1020
REF_CL_1_2,
#endif
};
private:
cl_device_id m_device;
reference_type_t m_ref_type;
public:
device(cl_device_id did)
: m_device(did), m_ref_type(REF_NOT_OWNABLE)
{ }
device(cl_device_id did, bool retain, reference_type_t ref_type=REF_NOT_OWNABLE)
: m_device(did), m_ref_type(ref_type)
{
if (retain && ref_type != REF_NOT_OWNABLE)
{
if (false)
{ }
#if PYOPENCL_CL_VERSION >= 0x1020
else if (ref_type == REF_CL_1_2)
{
PYOPENCL_CALL_GUARDED(clRetainDevice, (did));
}
#endif
else
throw error("Device", CL_INVALID_VALUE,
"cannot own references to devices when device fission or CL 1.2 is not available");
}
}
~device()
{
#if PYOPENCL_CL_VERSION >= 0x1020
if (m_ref_type == REF_CL_1_2)
PYOPENCL_CALL_GUARDED_CLEANUP(clReleaseDevice, (m_device));
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
#endif
}
cl_device_id data() const
{
return m_device;
}
PYOPENCL_EQUALITY_TESTS(device);
py::object get_info(cl_device_info param_name) const
{
#define DEV_GET_INT_INF(TYPE) \
PYOPENCL_GET_INTEGRAL_INFO(Device, m_device, param_name, TYPE);
switch (param_name)
{
case CL_DEVICE_TYPE: DEV_GET_INT_INF(cl_device_type);
case CL_DEVICE_VENDOR_ID: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_COMPUTE_UNITS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_WORK_GROUP_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_MAX_WORK_ITEM_SIZES:
{
std::vector<size_t> result;
PYOPENCL_GET_VEC_INFO(Device, m_device, param_name, result);
PYOPENCL_RETURN_VECTOR(size_t, result);
}
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_CHAR: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_SHORT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_CLOCK_FREQUENCY: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_ADDRESS_BITS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_READ_IMAGE_ARGS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_WRITE_IMAGE_ARGS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_MEM_ALLOC_SIZE: DEV_GET_INT_INF(cl_ulong);
case CL_DEVICE_IMAGE2D_MAX_WIDTH: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE2D_MAX_HEIGHT: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE3D_MAX_WIDTH: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE3D_MAX_HEIGHT: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE3D_MAX_DEPTH: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE_SUPPORT: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_MAX_PARAMETER_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_MAX_SAMPLERS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MEM_BASE_ADDR_ALIGN: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MIN_DATA_TYPE_ALIGN_SIZE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_SINGLE_FP_CONFIG: DEV_GET_INT_INF(cl_device_fp_config);
#ifdef CL_DEVICE_DOUBLE_FP_CONFIG
case CL_DEVICE_DOUBLE_FP_CONFIG: DEV_GET_INT_INF(cl_device_fp_config);
#endif
#ifdef CL_DEVICE_HALF_FP_CONFIG
case CL_DEVICE_HALF_FP_CONFIG: DEV_GET_INT_INF(cl_device_fp_config);
#endif
case CL_DEVICE_GLOBAL_MEM_CACHE_TYPE: DEV_GET_INT_INF(cl_device_mem_cache_type);
case CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_GLOBAL_MEM_CACHE_SIZE: DEV_GET_INT_INF(cl_ulong);
case CL_DEVICE_GLOBAL_MEM_SIZE: DEV_GET_INT_INF(cl_ulong);
case CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: DEV_GET_INT_INF(cl_ulong);
case CL_DEVICE_MAX_CONSTANT_ARGS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_LOCAL_MEM_TYPE: DEV_GET_INT_INF(cl_device_local_mem_type);
case CL_DEVICE_LOCAL_MEM_SIZE: DEV_GET_INT_INF(cl_ulong);
case CL_DEVICE_ERROR_CORRECTION_SUPPORT: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_PROFILING_TIMER_RESOLUTION: DEV_GET_INT_INF(size_t);
case CL_DEVICE_ENDIAN_LITTLE: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_AVAILABLE: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_COMPILER_AVAILABLE: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_EXECUTION_CAPABILITIES: DEV_GET_INT_INF(cl_device_exec_capabilities);
#if PYOPENCL_CL_VERSION >= 0x2000
case CL_DEVICE_QUEUE_ON_HOST_PROPERTIES: DEV_GET_INT_INF(cl_command_queue_properties);
#else
case CL_DEVICE_QUEUE_PROPERTIES: DEV_GET_INT_INF(cl_command_queue_properties);
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
case CL_DEVICE_NAME:
case CL_DEVICE_VENDOR:
case CL_DRIVER_VERSION:
case CL_DEVICE_PROFILE:
case CL_DEVICE_VERSION:
case CL_DEVICE_EXTENSIONS:
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
case CL_DEVICE_PLATFORM:
PYOPENCL_GET_OPAQUE_INFO(Device, m_device, param_name, cl_platform_id, platform);
#if PYOPENCL_CL_VERSION >= 0x1010
case CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_CHAR: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_SHORT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_INT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_NATIVE_VECTOR_WIDTH_HALF: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_HOST_UNIFIED_MEMORY: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_OPENCL_C_VERSION:
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
#endif
#ifdef CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV
case CL_DEVICE_COMPUTE_CAPABILITY_MAJOR_NV:
case CL_DEVICE_COMPUTE_CAPABILITY_MINOR_NV:
case CL_DEVICE_REGISTERS_PER_BLOCK_NV:
case CL_DEVICE_WARP_SIZE_NV:
DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_GPU_OVERLAP_NV:
case CL_DEVICE_KERNEL_EXEC_TIMEOUT_NV:
case CL_DEVICE_INTEGRATED_MEMORY_NV:
DEV_GET_INT_INF(cl_bool);
#endif
#ifdef CL_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT_NV
case CL_DEVICE_ATTRIBUTE_ASYNC_ENGINE_COUNT_NV:
DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_PCI_BUS_ID_NV
case CL_DEVICE_PCI_BUS_ID_NV:
DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_PCI_SLOT_ID_NV
case CL_DEVICE_PCI_SLOT_ID_NV:
DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_THREAD_TRACE_SUPPORTED_AMD
case CL_DEVICE_THREAD_TRACE_SUPPORTED_AMD: DEV_GET_INT_INF(cl_bool);
#endif
#ifdef CL_DEVICE_GFXIP_MAJOR_AMD
case CL_DEVICE_GFXIP_MAJOR_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_GFXIP_MINOR_AMD
case CL_DEVICE_GFXIP_MINOR_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_AVAILABLE_ASYNC_QUEUES_AMD
case CL_DEVICE_AVAILABLE_ASYNC_QUEUES_AMD: DEV_GET_INT_INF(cl_uint);
#endif
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
#if PYOPENCL_CL_VERSION >= 0x1020
case CL_DEVICE_LINKER_AVAILABLE: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_BUILT_IN_KERNELS:
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
case CL_DEVICE_IMAGE_MAX_BUFFER_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_IMAGE_MAX_ARRAY_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_PARENT_DEVICE:
PYOPENCL_GET_OPAQUE_INFO(Device, m_device, param_name, cl_device_id, device);
case CL_DEVICE_PARTITION_MAX_SUB_DEVICES: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PARTITION_TYPE:
case CL_DEVICE_PARTITION_PROPERTIES:
{
std::vector<cl_device_partition_property> result;
PYOPENCL_GET_VEC_INFO(Device, m_device, param_name, result);
PYOPENCL_RETURN_VECTOR(cl_device_partition_property, result);
}
case CL_DEVICE_PARTITION_AFFINITY_DOMAIN:
{
std::vector<cl_device_affinity_domain> result;
PYOPENCL_GET_VEC_INFO(Device, m_device, param_name, result);
PYOPENCL_RETURN_VECTOR(cl_device_affinity_domain, result);
}
case CL_DEVICE_REFERENCE_COUNT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_INTEROP_USER_SYNC: DEV_GET_INT_INF(cl_bool);
case CL_DEVICE_PRINTF_BUFFER_SIZE: DEV_GET_INT_INF(cl_bool);
#endif
// {{{ AMD dev attrs cl_amd_device_attribute_query
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
//
// types of AMD dev attrs divined from
// https://www.khronos.org/registry/cl/api/1.2/cl.hpp
#ifdef CL_DEVICE_PROFILING_TIMER_OFFSET_AMD
case CL_DEVICE_PROFILING_TIMER_OFFSET_AMD: DEV_GET_INT_INF(cl_ulong);
#endif
/* FIXME
#ifdef CL_DEVICE_TOPOLOGY_AMD
case CL_DEVICE_TOPOLOGY_AMD:
#endif
*/
#ifdef CL_DEVICE_BOARD_NAME_AMD
case CL_DEVICE_BOARD_NAME_AMD: ;
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
#endif
#ifdef CL_DEVICE_GLOBAL_FREE_MEMORY_AMD
case CL_DEVICE_GLOBAL_FREE_MEMORY_AMD:
{
std::vector<size_t> result;
PYOPENCL_GET_VEC_INFO(Device, m_device, param_name, result);
PYOPENCL_RETURN_VECTOR(size_t, result);
}
#endif
#ifdef CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD
case CL_DEVICE_SIMD_PER_COMPUTE_UNIT_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_GLOBAL_MEM_CHANNELS_AMD
case CL_DEVICE_GLOBAL_MEM_CHANNELS_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_GLOBAL_MEM_CHANNEL_BANKS_AMD
case CL_DEVICE_GLOBAL_MEM_CHANNEL_BANKS_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_GLOBAL_MEM_CHANNEL_BANK_WIDTH_AMD
case CL_DEVICE_GLOBAL_MEM_CHANNEL_BANK_WIDTH_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_LOCAL_MEM_SIZE_PER_COMPUTE_UNIT_AMD
case CL_DEVICE_LOCAL_MEM_SIZE_PER_COMPUTE_UNIT_AMD: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_LOCAL_MEM_BANKS_AMD
case CL_DEVICE_LOCAL_MEM_BANKS_AMD: DEV_GET_INT_INF(cl_uint);
#endif
// }}}
#ifdef CL_DEVICE_MAX_ATOMIC_COUNTERS_EXT
case CL_DEVICE_MAX_ATOMIC_COUNTERS_EXT: DEV_GET_INT_INF(cl_uint);
#endif
#if PYOPENCL_CL_VERSION >= 0x2000
case CL_DEVICE_MAX_READ_WRITE_IMAGE_ARGS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_GLOBAL_VARIABLE_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_QUEUE_ON_DEVICE_PROPERTIES: DEV_GET_INT_INF(cl_command_queue_properties);
case CL_DEVICE_QUEUE_ON_DEVICE_PREFERRED_SIZE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_QUEUE_ON_DEVICE_MAX_SIZE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_ON_DEVICE_QUEUES: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_MAX_ON_DEVICE_EVENTS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_SVM_CAPABILITIES: DEV_GET_INT_INF(cl_device_svm_capabilities);
case CL_DEVICE_GLOBAL_VARIABLE_PREFERRED_TOTAL_SIZE: DEV_GET_INT_INF(size_t);
case CL_DEVICE_MAX_PIPE_ARGS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PIPE_MAX_ACTIVE_RESERVATIONS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PIPE_MAX_PACKET_SIZE: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_PLATFORM_ATOMIC_ALIGNMENT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_GLOBAL_ATOMIC_ALIGNMENT: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_PREFERRED_LOCAL_ATOMIC_ALIGNMENT: DEV_GET_INT_INF(cl_uint);
#endif
#if PYOPENCL_CL_VERSION >= 0x2010
case CL_DEVICE_IL_VERSION:
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
case CL_DEVICE_MAX_NUM_SUB_GROUPS: DEV_GET_INT_INF(cl_uint);
case CL_DEVICE_SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS: DEV_GET_INT_INF(cl_bool);
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
#endif
#ifdef CL_DEVICE_ME_VERSION_INTEL
case CL_DEVICE_ME_VERSION_INTEL: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_EXT_MEM_PADDING_IN_BYTES_QCOM
case CL_DEVICE_EXT_MEM_PADDING_IN_BYTES_QCOM: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_PAGE_SIZE_QCOM
case CL_DEVICE_PAGE_SIZE_QCOM: DEV_GET_INT_INF(cl_uint);
#endif
#ifdef CL_DEVICE_SPIR_VERSIONS
case CL_DEVICE_SPIR_VERSIONS:
PYOPENCL_GET_STR_INFO(Device, m_device, param_name);
#endif
#ifdef CL_DEVICE_CORE_TEMPERATURE_ALTERA
case CL_DEVICE_CORE_TEMPERATURE_ALTERA: DEV_GET_INT_INF(cl_int);
#endif
#ifdef CL_DEVICE_SIMULTANEOUS_INTEROPS_INTEL
case CL_DEVICE_SIMULTANEOUS_INTEROPS_INTEL:
{
std::vector<cl_uint> result;
PYOPENCL_GET_VEC_INFO(Device, m_device, param_name, result);
PYOPENCL_RETURN_VECTOR(cl_uint, result);
}
#endif
#ifdef CL_DEVICE_NUM_SIMULTANEOUS_INTEROPS_INTEL
case CL_DEVICE_NUM_SIMULTANEOUS_INTEROPS_INTEL: DEV_GET_INT_INF(cl_uint);
#endif
default:
throw error("Device.get_info", CL_INVALID_VALUE);
}
}
#if PYOPENCL_CL_VERSION >= 0x1020
py::list create_sub_devices(py::object py_properties)
{
std::vector<cl_device_partition_property> properties;
COPY_PY_LIST(cl_device_partition_property, properties);
properties.push_back(0);
cl_device_partition_property *props_ptr
= properties.empty( ) ? nullptr : &properties.front();
cl_uint num_entries;
PYOPENCL_CALL_GUARDED(clCreateSubDevices,
std::vector<cl_device_id> result;
result.resize(num_entries);
PYOPENCL_CALL_GUARDED(clCreateSubDevices,
(m_device, props_ptr, num_entries, &result.front(), nullptr));
for (cl_device_id did: result)
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
py_result.append(handle_from_new_ptr(
new pyopencl::device(did, /*retain*/true,
device::REF_CL_1_2)));
return py_result;
}
#endif
};
inline py::list platform::get_devices(cl_device_type devtype)
{
cl_uint num_devices = 0;
PYOPENCL_PRINT_CALL_TRACE("clGetDeviceIDs");
{
cl_int status_code;
status_code = clGetDeviceIDs(m_platform, devtype, 0, 0, &num_devices);
if (status_code == CL_DEVICE_NOT_FOUND)
num_devices = 0;
else if (status_code != CL_SUCCESS) \
throw pyopencl::error("clGetDeviceIDs", status_code);
}
if (num_devices == 0)
return py::list();
std::vector<cl_device_id> devices(num_devices);
PYOPENCL_CALL_GUARDED(clGetDeviceIDs,
(m_platform, devtype,
num_devices, devices.empty( ) ? nullptr : &devices.front(), &num_devices));
for (cl_device_id did: devices)
result.append(handle_from_new_ptr(
new device(did)));
return result;
}
// }}}
class context : public noncopyable
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
{
private:
cl_context m_context;
public:
context(cl_context ctx, bool retain)
: m_context(ctx)
{
if (retain)
PYOPENCL_CALL_GUARDED(clRetainContext, (ctx));
}
~context()
{
PYOPENCL_CALL_GUARDED_CLEANUP(clReleaseContext,
(m_context));
}
cl_context data() const
{
return m_context;
}
PYOPENCL_EQUALITY_TESTS(context);
py::object get_info(cl_context_info param_name) const
{
switch (param_name)
{
case CL_CONTEXT_REFERENCE_COUNT:
PYOPENCL_GET_INTEGRAL_INFO(
Context, m_context, param_name, cl_uint);
case CL_CONTEXT_DEVICES:
{
std::vector<cl_device_id> result;
PYOPENCL_GET_VEC_INFO(Context, m_context, param_name, result);
py::list py_result;
for (cl_device_id did: result)
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
py_result.append(handle_from_new_ptr(
new pyopencl::device(did)));
return py_result;
}
case CL_CONTEXT_PROPERTIES:
{
std::vector<cl_context_properties> result;
PYOPENCL_GET_VEC_INFO(Context, m_context, param_name, result);
py::list py_result;
for (size_t i = 0; i < result.size(); i+=2)
{
cl_context_properties key = result[i];
py::object value;
switch (key)
{
case CL_CONTEXT_PLATFORM:
{
value = py::object(
handle_from_new_ptr(new platform(
reinterpret_cast<cl_platform_id>(result[i+1]))));
break;
}
#if defined(PYOPENCL_GL_SHARING_VERSION) && (PYOPENCL_GL_SHARING_VERSION >= 1)
#if defined(__APPLE__) && defined(HAVE_GL)
case CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE:
#else
case CL_GL_CONTEXT_KHR:
case CL_EGL_DISPLAY_KHR:
case CL_GLX_DISPLAY_KHR:
case CL_WGL_HDC_KHR:
case CL_CGL_SHAREGROUP_KHR:
#endif
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
break;
#endif
case 0:
break;
default:
throw error("Context.get_info", CL_INVALID_VALUE,
"unknown context_property key encountered");
}
py_result.append(py::make_tuple(result[i], value));
}
return py_result;
}
#if PYOPENCL_CL_VERSION >= 0x1010
case CL_CONTEXT_NUM_DEVICES:
PYOPENCL_GET_INTEGRAL_INFO(
Context, m_context, param_name, cl_uint);
#endif
default:
throw error("Context.get_info", CL_INVALID_VALUE);
}
}
};
inline
std::vector<cl_context_properties> parse_context_properties(
py::object py_properties)
{
std::vector<cl_context_properties> props;
if (py_properties.ptr() != Py_None)
{
for (py::handle prop_tuple_py: py_properties)
py::tuple prop_tuple(prop_tuple_py.cast<py::tuple>());
if (len(prop_tuple) != 2)
throw error("Context", CL_INVALID_VALUE, "property tuple must have length 2");
cl_context_properties prop = prop_tuple[0].cast<cl_context_properties>();
props.push_back(prop);
if (prop == CL_CONTEXT_PLATFORM)
{
props.push_back(
reinterpret_cast<cl_context_properties>(
prop_tuple[1].cast<const platform &>().data()));
}
#if defined(PYOPENCL_GL_SHARING_VERSION) && (PYOPENCL_GL_SHARING_VERSION >= 1)
#if defined(_WIN32)
else if (prop == CL_WGL_HDC_KHR)
{
// size_t is a stand-in for HANDLE, hopefully has the same size.
size_t hnd = (prop_tuple[1]).cast<size_t>();
props.push_back(hnd);
}
#endif
else if (
#if defined(__APPLE__) && defined(HAVE_GL)
prop == CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE
#else
prop == CL_GL_CONTEXT_KHR
|| prop == CL_EGL_DISPLAY_KHR
|| prop == CL_GLX_DISPLAY_KHR
|| prop == CL_CGL_SHAREGROUP_KHR
#endif
)
{
py::object ctypes = py::module::import("ctypes");
py::object prop = prop_tuple[1], c_void_p = ctypes.attr("c_void_p");
py::object ptr = ctypes.attr("cast")(prop, c_void_p);
props.push_back(ptr.attr("value").cast<cl_context_properties>());
}
#endif
else
throw error("Context", CL_INVALID_VALUE, "invalid context property");
}
props.push_back(0);
}
return props;
}
inline
context *create_context_inner(py::object py_devices, py::object py_properties,
py::object py_dev_type)
{
std::vector<cl_context_properties> props
= parse_context_properties(py_properties);
cl_context_properties *props_ptr
cl_int status_code;
cl_context ctx;
// from device list
if (py_devices.ptr() != Py_None)
{
if (py_dev_type.ptr() != Py_None)
throw error("Context", CL_INVALID_VALUE,
"one of 'devices' or 'dev_type' must be None");
std::vector<cl_device_id> devices;
devices.push_back(py_dev.cast<const device &>().data());
PYOPENCL_PRINT_CALL_TRACE("clCreateContext");
ctx = clCreateContext(
props_ptr,
devices.size(),
0, 0, &status_code);
}
// from dev_type
else
{
cl_device_type dev_type = CL_DEVICE_TYPE_DEFAULT;
if (py_dev_type.ptr() != Py_None)
dev_type = py_dev_type.cast<cl_device_type>();
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
PYOPENCL_PRINT_CALL_TRACE("clCreateContextFromType");
ctx = clCreateContextFromType(props_ptr, dev_type, 0, 0, &status_code);
}
if (status_code != CL_SUCCESS)
throw pyopencl::error("Context", status_code);
try
{
return new context(ctx, false);
}
catch (...)
{
PYOPENCL_CALL_GUARDED(clReleaseContext, (ctx));
throw;
}
}
inline
context *create_context(py::object py_devices, py::object py_properties,
py::object py_dev_type)
{
PYOPENCL_RETRY_RETURN_IF_MEM_ERROR(
return create_context_inner(py_devices, py_properties, py_dev_type);
)
}
// }}}
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
class command_queue
{
private:
cl_command_queue m_queue;
public:
command_queue(cl_command_queue q, bool retain)
: m_queue(q)
{
if (retain)
PYOPENCL_CALL_GUARDED(clRetainCommandQueue, (q));
}
command_queue(command_queue const &src)
: m_queue(src.m_queue)
{
PYOPENCL_CALL_GUARDED(clRetainCommandQueue, (m_queue));
}
command_queue(
const context &ctx,
const device *py_dev=0,
cl_command_queue_properties props=0)
{
cl_device_id dev;
if (py_dev)
dev = py_dev->data();
else
{
std::vector<cl_device_id> devs;
PYOPENCL_GET_VEC_INFO(Context, ctx.data(), CL_CONTEXT_DEVICES, devs);
if (devs.size() == 0)
throw pyopencl::error("CommandQueue", CL_INVALID_VALUE,
"context doesn't have any devices? -- don't know which one to default to");
dev = devs[0];
}
cl_int status_code;
PYOPENCL_PRINT_CALL_TRACE("clCreateCommandQueue");
m_queue = clCreateCommandQueue(
ctx.data(), dev, props, &status_code);
if (status_code != CL_SUCCESS)
throw pyopencl::error("CommandQueue", status_code);
}
~command_queue()
{
PYOPENCL_CALL_GUARDED_CLEANUP(clReleaseCommandQueue,
(m_queue));
}
const cl_command_queue data() const
{ return m_queue; }
PYOPENCL_EQUALITY_TESTS(command_queue);
py::object get_info(cl_command_queue_info param_name) const
{
switch (param_name)
{
case CL_QUEUE_CONTEXT:
PYOPENCL_GET_OPAQUE_INFO(CommandQueue, m_queue, param_name,
cl_context, context);
case CL_QUEUE_DEVICE:
PYOPENCL_GET_OPAQUE_INFO(CommandQueue, m_queue, param_name,
cl_device_id, device);
case CL_QUEUE_REFERENCE_COUNT:
PYOPENCL_GET_INTEGRAL_INFO(CommandQueue, m_queue, param_name,
cl_uint);
case CL_QUEUE_PROPERTIES:
PYOPENCL_GET_INTEGRAL_INFO(CommandQueue, m_queue, param_name,
cl_command_queue_properties);
default:
throw error("CommandQueue.get_info", CL_INVALID_VALUE);
}
}
std::unique_ptr<context> get_context() const
{
cl_context param_value;
PYOPENCL_CALL_GUARDED(clGetCommandQueueInfo,
(m_queue, CL_QUEUE_CONTEXT, sizeof(param_value), ¶m_value, 0));
return std::unique_ptr<context>(
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
new context(param_value, /*retain*/ true));
}
#if PYOPENCL_CL_VERSION < 0x1010
cl_command_queue_properties set_property(
cl_command_queue_properties prop,
bool enable)
{
cl_command_queue_properties old_prop;
PYOPENCL_CALL_GUARDED(clSetCommandQueueProperty,
(m_queue, prop, PYOPENCL_CAST_BOOL(enable), &old_prop));
return old_prop;
}
#endif
void flush()
{ PYOPENCL_CALL_GUARDED(clFlush, (m_queue)); }
void finish()
{ PYOPENCL_CALL_GUARDED_THREADED(clFinish, (m_queue)); }
};
// }}}
class event : noncopyable
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
{
private:
cl_event m_event;
public:
event(cl_event event, bool retain)
: m_event(event)
{
if (retain)
PYOPENCL_CALL_GUARDED(clRetainEvent, (event));
}
event(event const &src)
: m_event(src.m_event)
{ PYOPENCL_CALL_GUARDED(clRetainEvent, (m_event)); }
virtual ~event()
{
PYOPENCL_CALL_GUARDED_CLEANUP(clReleaseEvent,
(m_event));
}
const cl_event data() const
{ return m_event; }
PYOPENCL_EQUALITY_TESTS(event);
py::object get_info(cl_event_info param_name) const
{
switch (param_name)
{
case CL_EVENT_COMMAND_QUEUE:
PYOPENCL_GET_OPAQUE_INFO(Event, m_event, param_name,
cl_command_queue, command_queue);
case CL_EVENT_COMMAND_TYPE:
PYOPENCL_GET_INTEGRAL_INFO(Event, m_event, param_name,
cl_command_type);
case CL_EVENT_COMMAND_EXECUTION_STATUS:
PYOPENCL_GET_INTEGRAL_INFO(Event, m_event, param_name,
cl_int);
case CL_EVENT_REFERENCE_COUNT:
PYOPENCL_GET_INTEGRAL_INFO(Event, m_event, param_name,
cl_uint);
#if PYOPENCL_CL_VERSION >= 0x1010
case CL_EVENT_CONTEXT:
PYOPENCL_GET_OPAQUE_INFO(Event, m_event, param_name,
cl_context, context);
#endif
default:
throw error("Event.get_info", CL_INVALID_VALUE);
}
}
py::object get_profiling_info(cl_profiling_info param_name) const
{
switch (param_name)
{
case CL_PROFILING_COMMAND_QUEUED:
case CL_PROFILING_COMMAND_SUBMIT:
case CL_PROFILING_COMMAND_START:
case CL_PROFILING_COMMAND_END:
#if PYOPENCL_CL_VERSION >= 0x2000
case CL_PROFILING_COMMAND_COMPLETE:
#endif
PYOPENCL_GET_INTEGRAL_INFO(EventProfiling, m_event, param_name,
cl_ulong);
default:
throw error("Event.get_profiling_info", CL_INVALID_VALUE);
}
}
virtual void wait()
{
PYOPENCL_CALL_GUARDED_THREADED(clWaitForEvents, (1, &m_event));
}
};
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
class nanny_event : public event
{
// In addition to everything an event does, the nanny event holds a reference
// to a Python object and waits for its own completion upon destruction.
protected:
std::unique_ptr<py_buffer_wrapper> m_ward;
nanny_event(cl_event evt, bool retain, std::unique_ptr<py_buffer_wrapper> &ward)
: event(evt, retain), m_ward(std::move(ward))
{ }
~nanny_event()
{ wait(); }
py::object get_ward() const
{
if (m_ward.get())
{
return py::reinterpret_borrow<py::object>(m_ward->m_buf.obj);
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
}
virtual void wait()
{
event::wait();
m_ward.reset();
}
};
#else
class nanny_event : public event
{
// In addition to everything an event does, the nanny event holds a reference
// to a Python object and waits for its own completion upon destruction.
protected:
py::object m_ward;
public:
nanny_event(cl_event evt, bool retain, py::object ward)
: event(evt, retain), m_ward(ward)
{ }
nanny_event(nanny_event const &src)
: event(src), m_ward(src.m_ward)
{ }
~nanny_event()
{ wait(); }
py::object get_ward() const
{ return m_ward; }
virtual void wait()
{
event::wait();
}
};
#endif
inline
void wait_for_events(py::object events)
{
cl_uint num_events_in_wait_list = 0;
std::vector<cl_event> event_wait_list(len(events));
event_wait_list[num_events_in_wait_list++] =
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
PYOPENCL_CALL_GUARDED_THREADED(clWaitForEvents, (
PYOPENCL_WAITLIST_ARGS));
}
#if PYOPENCL_CL_VERSION >= 0x1020
inline
event *enqueue_marker_with_wait_list(command_queue &cq,
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
cl_event evt;
PYOPENCL_CALL_GUARDED(clEnqueueMarkerWithWaitList, (
cq.data(), PYOPENCL_WAITLIST_ARGS, &evt));
PYOPENCL_RETURN_NEW_EVENT(evt);
}
inline
event *enqueue_barrier_with_wait_list(command_queue &cq,
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
cl_event evt;
PYOPENCL_CALL_GUARDED(clEnqueueBarrierWithWaitList,
(cq.data(), PYOPENCL_WAITLIST_ARGS, &evt));
PYOPENCL_RETURN_NEW_EVENT(evt);
}
#endif
// {{{ used internally for pre-OpenCL-1.2 contexts
inline
event *enqueue_marker(command_queue &cq)
{
cl_event evt;
PYOPENCL_CALL_GUARDED(clEnqueueMarker, (
cq.data(), &evt));
PYOPENCL_RETURN_NEW_EVENT(evt);
}
inline
void enqueue_wait_for_events(command_queue &cq, py::object py_events)
{
cl_uint num_events = 0;
std::vector<cl_event> event_list(len(py_events));
for (py::handle py_evt: py_events)
event_list[num_events++] = py_evt.cast<event &>().data();
PYOPENCL_CALL_GUARDED(clEnqueueWaitForEvents, (
cq.data(), num_events, event_list.empty( ) ? nullptr : &event_list.front()));
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
}
inline
void enqueue_barrier(command_queue &cq)
{
PYOPENCL_CALL_GUARDED(clEnqueueBarrier, (cq.data()));
}
// }}}
#if PYOPENCL_CL_VERSION >= 0x1010
class user_event : public event
{
public:
user_event(cl_event evt, bool retain)
: event(evt, retain)
{ }
void set_status(cl_int execution_status)
{
PYOPENCL_CALL_GUARDED(clSetUserEventStatus, (data(), execution_status));
}
};
inline
user_event *create_user_event(context &ctx)
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
{
cl_int status_code;
PYOPENCL_PRINT_CALL_TRACE("clCreateUserEvent");
cl_event evt = clCreateUserEvent(ctx.data(), &status_code);
if (status_code != CL_SUCCESS)
throw pyopencl::error("UserEvent", status_code);
try
{
return new user_event(evt, false);
}
catch (...)
{
clReleaseEvent(evt);
throw;
}
}
#endif
// }}}
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
// {{{ memory_object
py::object create_mem_object_wrapper(cl_mem mem);
class memory_object_holder
{
public:
virtual const cl_mem data() const = 0;
PYOPENCL_EQUALITY_TESTS(memory_object_holder);
size_t size() const
{
size_t param_value;
PYOPENCL_CALL_GUARDED(clGetMemObjectInfo,
(data(), CL_MEM_SIZE, sizeof(param_value), ¶m_value, 0));
return param_value;
}
py::object get_info(cl_mem_info param_name) const;
};
class memory_object : noncopyable, public memory_object_holder
{
public:
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
typedef std::unique_ptr<py_buffer_wrapper> hostbuf_t;
#else
typedef py::object hostbuf_t;
#endif
private:
bool m_valid;
cl_mem m_mem;
hostbuf_t m_hostbuf;
public:
memory_object(cl_mem mem, bool retain, hostbuf_t hostbuf=hostbuf_t())
: m_valid(true), m_mem(mem)
{
if (retain)
PYOPENCL_CALL_GUARDED(clRetainMemObject, (mem));
m_hostbuf = PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(hostbuf);
}
memory_object(memory_object &src)
m_hostbuf(PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(src.m_hostbuf))
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
{
PYOPENCL_CALL_GUARDED(clRetainMemObject, (m_mem));
}
memory_object(memory_object_holder const &src)
: m_valid(true), m_mem(src.data())
{
PYOPENCL_CALL_GUARDED(clRetainMemObject, (m_mem));
}
void release()
{
if (!m_valid)
throw error("MemoryObject.free", CL_INVALID_VALUE,
"trying to double-unref mem object");
PYOPENCL_CALL_GUARDED_CLEANUP(clReleaseMemObject, (m_mem));
m_valid = false;
}
virtual ~memory_object()
{
if (m_valid)
release();
}
py::object hostbuf()
{
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
if (m_hostbuf.get())
return py::reinterpret_borrow<py::object>(m_hostbuf->m_buf.obj);
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
#else
return m_hostbuf;
#endif
}
const cl_mem data() const
{ return m_mem; }
};
#if PYOPENCL_CL_VERSION >= 0x1020
inline
event *enqueue_migrate_mem_objects(
command_queue &cq,
py::object py_mem_objects,
cl_mem_migration_flags flags,
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
std::vector<cl_mem> mem_objects;
for (py::handle mo: py_mem_objects)
mem_objects.push_back(mo.cast<const memory_object &>().data());
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED(clEnqueueMigrateMemObjects, (
cq.data(),
mem_objects.size(), mem_objects.empty( ) ? nullptr : &mem_objects.front(),
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
flags,
PYOPENCL_WAITLIST_ARGS, &evt
));
);
PYOPENCL_RETURN_NEW_EVENT(evt);
}
#endif
#ifdef cl_ext_migrate_memobject
inline
event *enqueue_migrate_mem_object_ext(
command_queue &cq,
py::object py_mem_objects,
cl_mem_migration_flags_ext flags,
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
#if PYOPENCL_CL_VERSION >= 0x1020
// {{{ get platform
cl_device_id dev;
PYOPENCL_CALL_GUARDED(clGetCommandQueueInfo, (cq.data(), CL_QUEUE_DEVICE,
cl_platform_id plat;
PYOPENCL_CALL_GUARDED(clGetDeviceInfo, (dev, CL_DEVICE_PLATFORM,
// }}}
#endif
PYOPENCL_GET_EXT_FUN(plat,
clEnqueueMigrateMemObjectEXT, enqueue_migrate_fn);
std::vector<cl_mem> mem_objects;
for (py::handle mo: py_mem_objects)
mem_objects.push_back(mo.cast<memory_object &>().data());
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED(enqueue_migrate_fn, (
cq.data(),
mem_objects.size(), mem_objects.empty( ) ? nullptr : &mem_objects.front(),
flags,
PYOPENCL_WAITLIST_ARGS, &evt
));
);
PYOPENCL_RETURN_NEW_EVENT(evt);
}
#endif
// }}}
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
// {{{ buffer
inline cl_mem create_buffer(
cl_context ctx,
cl_mem_flags flags,
size_t size,
void *host_ptr)
{
cl_int status_code;
PYOPENCL_PRINT_CALL_TRACE("clCreateBuffer");
cl_mem mem = clCreateBuffer(ctx, flags, size, host_ptr, &status_code);
if (status_code != CL_SUCCESS)
throw pyopencl::error("create_buffer", status_code);
return mem;
}
inline cl_mem create_buffer_gc(
cl_context ctx,
cl_mem_flags flags,
size_t size,
void *host_ptr)
{
PYOPENCL_RETRY_RETURN_IF_MEM_ERROR(
return create_buffer(ctx, flags, size, host_ptr);
);
}
#if PYOPENCL_CL_VERSION >= 0x1010
inline cl_mem create_sub_buffer(
cl_mem buffer, cl_mem_flags flags, cl_buffer_create_type bct,
const void *buffer_create_info)
{
cl_int status_code;
PYOPENCL_PRINT_CALL_TRACE("clCreateSubBuffer");
cl_mem mem = clCreateSubBuffer(buffer, flags,
bct, buffer_create_info, &status_code);
if (status_code != CL_SUCCESS)
throw pyopencl::error("clCreateSubBuffer", status_code);
return mem;
}
inline cl_mem create_sub_buffer_gc(
cl_mem buffer, cl_mem_flags flags, cl_buffer_create_type bct,
const void *buffer_create_info)
{
PYOPENCL_RETRY_RETURN_IF_MEM_ERROR(
return create_sub_buffer(buffer, flags, bct, buffer_create_info);
);
}
#endif
class buffer : public memory_object
{
public:
buffer(cl_mem mem, bool retain, hostbuf_t hostbuf=hostbuf_t())
: memory_object(mem, retain, PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(hostbuf))
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
{ }
#if PYOPENCL_CL_VERSION >= 0x1010
buffer *get_sub_region(
size_t origin, size_t size, cl_mem_flags flags) const
{
cl_buffer_region region = { origin, size};
cl_mem mem = create_sub_buffer_gc(
data(), flags, CL_BUFFER_CREATE_TYPE_REGION, ®ion);
try
{
return new buffer(mem, false);
}
catch (...)
{
PYOPENCL_CALL_GUARDED(clReleaseMemObject, (mem));
throw;
}
}
buffer *getitem(py::slice slc) const
{
PYOPENCL_BUFFER_SIZE_T start, end, stride, length;
size_t my_length;
PYOPENCL_CALL_GUARDED(clGetMemObjectInfo,
(data(), CL_MEM_SIZE, sizeof(my_length), &my_length, 0));
#if PY_VERSION_HEX >= 0x03020000
if (PySlice_GetIndicesEx(slc.ptr(),
#else
if (PySlice_GetIndicesEx(reinterpret_cast<PySliceObject *>(slc.ptr()),
#endif
my_length, &start, &end, &stride, &length) != 0)
throw py::error_already_set();
if (stride != 1)
throw pyopencl::error("Buffer.__getitem__", CL_INVALID_VALUE,
"Buffer slice must have stride 1");
cl_mem_flags my_flags;
PYOPENCL_CALL_GUARDED(clGetMemObjectInfo,
(data(), CL_MEM_FLAGS, sizeof(my_flags), &my_flags, 0));
my_flags &= ~CL_MEM_COPY_HOST_PTR;
if (end <= start)
throw pyopencl::error("Buffer.__getitem__", CL_INVALID_VALUE,
"Buffer slice have end > start");
return get_sub_region(start, end-start, my_flags);
}
#endif
};
// {{{ buffer creation
inline
buffer *create_buffer_py(
context &ctx,
cl_mem_flags flags,
size_t size,
py::object py_hostbuf
)
{
if (py_hostbuf.ptr() != Py_None &&
!(flags & (CL_MEM_USE_HOST_PTR | CL_MEM_COPY_HOST_PTR)))
PyErr_Warn(PyExc_UserWarning, "'hostbuf' was passed, "
"but no memory flags to make use of it.");
void *buf = 0;
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
std::unique_ptr<py_buffer_wrapper> retained_buf_obj;
if (py_hostbuf.ptr() != Py_None)
{
retained_buf_obj = std::unique_ptr<py_buffer_wrapper>(new py_buffer_wrapper);
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
int py_buf_flags = PyBUF_ANY_CONTIGUOUS;
if ((flags & CL_MEM_USE_HOST_PTR)
&& ((flags & CL_MEM_READ_WRITE)
|| (flags & CL_MEM_WRITE_ONLY)))
py_buf_flags |= PyBUF_WRITABLE;
retained_buf_obj->get(py_hostbuf.ptr(), py_buf_flags);
buf = retained_buf_obj->m_buf.buf;
if (size > size_t(retained_buf_obj->m_buf.len))
throw pyopencl::error("Buffer", CL_INVALID_VALUE,
"specified size is greater than host buffer size");
if (size == 0)
size = retained_buf_obj->m_buf.len;
}
#else
py::object retained_buf_obj;
if (py_hostbuf.ptr() != Py_None)
{
PYOPENCL_BUFFER_SIZE_T len;
if ((flags & CL_MEM_USE_HOST_PTR)
&& ((flags & CL_MEM_READ_WRITE)
|| (flags & CL_MEM_WRITE_ONLY)))
{
if (PyObject_AsWriteBuffer(py_hostbuf.ptr(), &buf, &len))
throw py::error_already_set();
}
else
{
if (PyObject_AsReadBuffer(
py_hostbuf.ptr(), const_cast<const void **>(&buf), &len))
throw py::error_already_set();
}
if (flags & CL_MEM_USE_HOST_PTR)
retained_buf_obj = py_hostbuf;
if (size > size_t(len))
throw pyopencl::error("Buffer", CL_INVALID_VALUE,
"specified size is greater than host buffer size");
if (size == 0)
size = len;
}
#endif
cl_mem mem = create_buffer_gc(ctx.data(), flags, size, buf);
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
if (!(flags & CL_MEM_USE_HOST_PTR))
retained_buf_obj.reset();
#endif
try
{
return new buffer(mem, false, PYOPENCL_STD_MOVE_IF_NEW_BUF_INTF(retained_buf_obj));
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
}
catch (...)
{
PYOPENCL_CALL_GUARDED(clReleaseMemObject, (mem));
throw;
}
}
// }}}
// {{{ buffer transfers
// {{{ byte-for-byte transfers
inline
event *enqueue_read_buffer(
command_queue &cq,
memory_object_holder &mem,
py::object buffer,
size_t device_offset,
py::object py_wait_for,
bool is_blocking)
{
PYOPENCL_PARSE_WAIT_FOR;
void *buf;
PYOPENCL_BUFFER_SIZE_T len;
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
std::unique_ptr<py_buffer_wrapper> ward(new py_buffer_wrapper);
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
ward->get(buffer.ptr(), PyBUF_ANY_CONTIGUOUS | PyBUF_WRITABLE);
buf = ward->m_buf.buf;
len = ward->m_buf.len;
#else
py::object ward = buffer;
if (PyObject_AsWriteBuffer(buffer.ptr(), &buf, &len))
throw py::error_already_set();
#endif
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED_THREADED(clEnqueueReadBuffer, (
cq.data(),
mem.data(),
PYOPENCL_CAST_BOOL(is_blocking),
device_offset, len, buf,
PYOPENCL_WAITLIST_ARGS, &evt
))
);
PYOPENCL_RETURN_NEW_NANNY_EVENT(evt, ward);
}
inline
event *enqueue_write_buffer(
command_queue &cq,
memory_object_holder &mem,
py::object buffer,
size_t device_offset,
py::object py_wait_for,
bool is_blocking)
{
PYOPENCL_PARSE_WAIT_FOR;
const void *buf;
PYOPENCL_BUFFER_SIZE_T len;
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
std::unique_ptr<py_buffer_wrapper> ward(new py_buffer_wrapper);
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
ward->get(buffer.ptr(), PyBUF_ANY_CONTIGUOUS);
buf = ward->m_buf.buf;
len = ward->m_buf.len;
#else
py::object ward = buffer;
if (PyObject_AsReadBuffer(buffer.ptr(), &buf, &len))
throw py::error_already_set();
#endif
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED_THREADED(clEnqueueWriteBuffer, (
cq.data(),
mem.data(),
PYOPENCL_CAST_BOOL(is_blocking),
device_offset, len, buf,
PYOPENCL_WAITLIST_ARGS, &evt
))
);
PYOPENCL_RETURN_NEW_NANNY_EVENT(evt, ward);
}
inline
event *enqueue_copy_buffer(
command_queue &cq,
memory_object_holder &src,
memory_object_holder &dst,
ptrdiff_t byte_count,
size_t src_offset,
size_t dst_offset,
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
if (byte_count < 0)
{
size_t byte_count_src = 0;
size_t byte_count_dst = 0;
PYOPENCL_CALL_GUARDED(clGetMemObjectInfo,
(src.data(), CL_MEM_SIZE, sizeof(byte_count), &byte_count_src, 0));
PYOPENCL_CALL_GUARDED(clGetMemObjectInfo,
(src.data(), CL_MEM_SIZE, sizeof(byte_count), &byte_count_dst, 0));
byte_count = std::min(byte_count_src, byte_count_dst);
}
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED(clEnqueueCopyBuffer, (
cq.data(),
src.data(), dst.data(),
src_offset, dst_offset,
byte_count,
PYOPENCL_WAITLIST_ARGS,
&evt
))
);
PYOPENCL_RETURN_NEW_EVENT(evt);
}
// }}}
// {{{ rectangular transfers
#if PYOPENCL_CL_VERSION >= 0x1010
inline
event *enqueue_read_buffer_rect(
command_queue &cq,
memory_object_holder &mem,
py::object buffer,
py::object py_buffer_origin,
py::object py_host_origin,
py::object py_region,
py::sequence py_buffer_pitches,
py::sequence py_host_pitches,
py::object py_wait_for,
bool is_blocking
)
{
PYOPENCL_PARSE_WAIT_FOR;
COPY_PY_COORD_TRIPLE(buffer_origin);
COPY_PY_COORD_TRIPLE(host_origin);
COPY_PY_REGION_TRIPLE(region);
COPY_PY_PITCH_TUPLE(buffer_pitches);
COPY_PY_PITCH_TUPLE(host_pitches);
void *buf;
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
std::unique_ptr<py_buffer_wrapper> ward(new py_buffer_wrapper);
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
ward->get(buffer.ptr(), PyBUF_ANY_CONTIGUOUS | PyBUF_WRITABLE);
buf = ward->m_buf.buf;
#else
py::object ward = buffer;
PYOPENCL_BUFFER_SIZE_T len;
if (PyObject_AsWriteBuffer(buffer.ptr(), &buf, &len))
throw py::error_already_set();
#endif
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED_THREADED(clEnqueueReadBufferRect, (
cq.data(),
mem.data(),
PYOPENCL_CAST_BOOL(is_blocking),
buffer_origin, host_origin, region,
buffer_pitches[0], buffer_pitches[1],
host_pitches[0], host_pitches[1],
buf,
PYOPENCL_WAITLIST_ARGS, &evt
))
);
PYOPENCL_RETURN_NEW_NANNY_EVENT(evt, ward);
}
inline
event *enqueue_write_buffer_rect(
command_queue &cq,
memory_object_holder &mem,
py::object buffer,
py::object py_buffer_origin,
py::object py_host_origin,
py::object py_region,
py::sequence py_buffer_pitches,
py::sequence py_host_pitches,
py::object py_wait_for,
bool is_blocking
)
{
PYOPENCL_PARSE_WAIT_FOR;
COPY_PY_COORD_TRIPLE(buffer_origin);
COPY_PY_COORD_TRIPLE(host_origin);
COPY_PY_REGION_TRIPLE(region);
COPY_PY_PITCH_TUPLE(buffer_pitches);
COPY_PY_PITCH_TUPLE(host_pitches);
const void *buf;
#ifdef PYOPENCL_USE_NEW_BUFFER_INTERFACE
std::unique_ptr<py_buffer_wrapper> ward(new py_buffer_wrapper);
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
ward->get(buffer.ptr(), PyBUF_ANY_CONTIGUOUS);
buf = ward->m_buf.buf;
#else
py::object ward = buffer;
PYOPENCL_BUFFER_SIZE_T len;
if (PyObject_AsReadBuffer(buffer.ptr(), &buf, &len))
throw py::error_already_set();
#endif
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED_THREADED(clEnqueueWriteBufferRect, (
cq.data(),
mem.data(),
PYOPENCL_CAST_BOOL(is_blocking),
buffer_origin, host_origin, region,
buffer_pitches[0], buffer_pitches[1],
host_pitches[0], host_pitches[1],
buf,
PYOPENCL_WAITLIST_ARGS, &evt
))
);
PYOPENCL_RETURN_NEW_NANNY_EVENT(evt, ward);
}
inline
event *enqueue_copy_buffer_rect(
command_queue &cq,
memory_object_holder &src,
memory_object_holder &dst,
py::object py_src_origin,
py::object py_dst_origin,
py::object py_region,
py::sequence py_src_pitches,
py::sequence py_dst_pitches,
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
py::object py_wait_for)
{
PYOPENCL_PARSE_WAIT_FOR;
COPY_PY_COORD_TRIPLE(src_origin);
COPY_PY_COORD_TRIPLE(dst_origin);
COPY_PY_REGION_TRIPLE(region);
COPY_PY_PITCH_TUPLE(src_pitches);
COPY_PY_PITCH_TUPLE(dst_pitches);
cl_event evt;
PYOPENCL_RETRY_IF_MEM_ERROR(
PYOPENCL_CALL_GUARDED(clEnqueueCopyBufferRect, (
cq.data(),
src.data(), dst.data(),
src_origin, dst_origin, region,
src_pitches[0], src_pitches[1],
dst_pitches[0], dst_pitches[1],
PYOPENCL_WAITLIST_ARGS,
&evt
))
);
PYOPENCL_RETURN_NEW_EVENT(evt);
Loading
Loading full blame...