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>
#include "pyopencl_ext.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 <thread>
#include <mutex>
#include <condition_variable>
#include <cstdio>
#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) or defined(PYPY_VERSION)
#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)
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
// }}}
// {{{ 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(); \
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
} \
#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) \
{ \
/* 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; \
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
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()
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
#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:
error(const char *routine, cl_int c, const char *msg="")
: std::runtime_error(msg), m_routine(routine), m_code(c)
{
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
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
{
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
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
{
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;
}
// }}}
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
{
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));
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
#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);
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
700
701
702
703
704
705
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
#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:
{
Andreas Klöckner
committed
#if defined(__GNUG__) && !defined(__clang__)
#pragma GCC diagnostic push
// what's being ignored here is an alignment attribute to native size, which
// shouldn't matter on the relevant ABIs that I'm aware of.
#pragma GCC diagnostic ignored "-Wignored-attributes"
#endif
std::vector<cl_device_affinity_domain> result;
Andreas Klöckner
committed
#if defined(__GNUG__) && !defined(__clang__)
#pragma GCC diagnostic pop
#endif
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
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
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
//
// 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);
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
#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)
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
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
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
{
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)
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]))));