This file is indexed.

/usr/lib/python3/dist-packages/pyopencl/clrandom.py is in python3-pyopencl 2017.2.2-1.

This file is owned by root:root, with mode 0o644.

The actual contents of the file can be viewed below.

  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
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
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
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
750
751
752
753
754
755
756
757
758
759
# encoding: utf8
from __future__ import division, absolute_import

__copyright__ = "Copyright (C) 2009-16 Andreas Kloeckner"

__license__ = """
Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in
all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
THE SOFTWARE.
"""


# {{{ documentation

__doc__ = u"""
PyOpenCL now includes and uses some of the `Random123 random number generators
<https://www.deshawresearch.com/resources_random123.html>`_ by D.E. Shaw
Research.  In addition to being usable through the convenience functions above,
they are available in any piece of code compiled through PyOpenCL by::

    #include <pyopencl-random123/philox.cl>
    #include <pyopencl-random123/threefry.cl>

See the `Philox source
<https://github.com/inducer/pyopencl/blob/master/pyopencl/cl/pyopencl-random123/philox.cl>`_
and the `Threefry source
<https://github.com/inducer/pyopencl/blob/master/pyopencl/cl/pyopencl-random123/threefry.cl>`_
for some documentation if you're planning on using Random123 directly.

.. note::

    PyOpenCL previously had documented support for the `RANLUXCL random number
    generator <https://bitbucket.org/ivarun/ranluxcl/>`_ by Ivar Ursin
    Nikolaisen.  This support is now deprecated because of the general slowness
    of these generators and will be removed from PyOpenCL in the 2018.x series.
    All users are encouraged to switch to one of the Random123 genrators,
    :class:`PhiloxGenerator` or :class:`ThreefryGenerator`.

.. autoclass:: PhiloxGenerator

.. autoclass:: ThreefryGenerator

.. autofunction:: rand
.. autofunction:: fill_rand

"""

# }}}

import pyopencl as cl
import pyopencl.array as cl_array
import pyopencl.cltypes as cltypes
from pyopencl.tools import first_arg_dependent_memoize
from pytools import memoize_method

import numpy as np


# {{{ RanluxGenerator (deprecated)

class RanluxGenerator(object):
    """
    .. warning::

        This class is deprecated, to be removed in PyOpenCL 2018.x.

    .. versionadded:: 2011.2

    .. attribute:: state

        A :class:`pyopencl.array.Array` containing the state of the generator.

    .. attribute:: nskip

        nskip is an integer which can (optionally) be defined in the kernel
        code as RANLUXCL_NSKIP. If this is done the generator will be faster
        for luxury setting 0 and 1, or when the p-value is manually set to a
        multiple of 24.
    """

    def __init__(self, queue, num_work_items=None,
            luxury=None, seed=None, no_warmup=False,
            use_legacy_init=False, max_work_items=None):
        """
        :param queue: :class:`pyopencl.CommandQueue`, only used for initialization
        :param luxury: the "luxury value" of the generator, and should be 0-4,
            where 0 is fastest and 4 produces the best numbers. It can also be
            >=24, in which case it directly sets the p-value of RANLUXCL.
        :param num_work_items: is the number of generators to initialize,
            usually corresponding to the number of work-items in the NDRange
            RANLUXCL will be used with.  May be `None`, in which case a default
            value is used.
        :param max_work_items: should reflect the maximum number of work-items
            that will be used on any parallel instance of RANLUXCL. So for
            instance if we are launching 5120 work-items on GPU1 and 10240
            work-items on GPU2, GPU1's RANLUXCLTab would be generated by
            calling ranluxcl_intialization with numWorkitems = 5120 while
            GPU2's RANLUXCLTab would use numWorkitems = 10240. However
            maxWorkitems must be at least 10240 for both GPU1 and GPU2, and it
            must be set to the same value for both. (may be `None`)

        .. versionchanged:: 2013.1
            Added default value for `num_work_items`.
        """

        if luxury is None:
            luxury = 4

        if num_work_items is None:
            if queue.device.type & cl.device_type.CPU:
                num_work_items = 8 * queue.device.max_compute_units
            else:
                num_work_items = 64 * queue.device.max_compute_units

        if seed is None:
            from time import time
            seed = int(time()*1e6) % 2 << 30

        self.context = queue.context
        self.luxury = luxury
        self.num_work_items = num_work_items

        from pyopencl.characterize import has_double_support
        self.support_double = has_double_support(queue.device)

        self.no_warmup = no_warmup
        self.use_legacy_init = use_legacy_init
        self.max_work_items = max_work_items

        src = """
            %(defines)s

            #include <pyopencl-ranluxcl.cl>

            kernel void init_ranlux(unsigned seeds,
                global ranluxcl_state_t *ranluxcltab)
            {
              if (get_global_id(0) < %(num_work_items)d)
                ranluxcl_initialization(seeds, ranluxcltab);
            }
            """ % {
                    "defines": self.generate_settings_defines(),
                    "num_work_items": num_work_items
                }
        prg = cl.Program(queue.context, src).build()

        # {{{ compute work group size

        wg_size = None

        import sys
        import platform
        if ("darwin" in sys.platform
                and "Apple" in queue.device.platform.vendor
                and platform.mac_ver()[0].startswith("10.7")
                and queue.device.type & cl.device_type.CPU):
            wg_size = (1,)

        self.wg_size = wg_size

        # }}}

        self.state = cl_array.empty(queue, (num_work_items, 112), dtype=np.uint8)
        self.state.fill(17)

        prg.init_ranlux(queue, (num_work_items,), self.wg_size, np.uint32(seed),
                self.state.data)

    def generate_settings_defines(self, include_double_pragma=True):
        lines = []
        if include_double_pragma and self.support_double:
            lines.append("""
                #if __OPENCL_C_VERSION__ < 120
                #pragma OPENCL EXTENSION cl_khr_fp64: enable
                #endif
                """)

        lines.append("#define RANLUXCL_LUX %d" % self.luxury)

        if self.no_warmup:
            lines.append("#define RANLUXCL_NO_WARMUP")

        if self.support_double:
            lines.append("#define RANLUXCL_SUPPORT_DOUBLE")

        if self.use_legacy_init:
            lines.append("#define RANLUXCL_USE_LEGACY_INITIALIZATION")

            if self.max_work_items:
                lines.append(
                        "#define RANLUXCL_MAXWORKITEMS %d" % self.max_work_items)

        return "\n".join(lines)

    @memoize_method
    def get_gen_kernel(self, dtype, distribution="uniform"):
        size_multiplier = 1
        arg_dtype = dtype

        if dtype == np.float64:
            bits = 64
            c_type = "double"
            rng_expr = "(shift + scale * gen)"
        elif dtype == np.float32:
            bits = 32
            c_type = "float"
            rng_expr = "(shift + scale * gen)"
        elif dtype == cltypes.float2:
            bits = 32
            c_type = "float"
            rng_expr = "(shift + scale * gen)"
            size_multiplier = 2
            arg_dtype = np.float32
        elif dtype in [cltypes.float3, cltypes.float4]:
            bits = 32
            c_type = "float"
            rng_expr = "(shift + scale * gen)"
            size_multiplier = 4
            arg_dtype = np.float32
        elif dtype == np.int32:
            assert distribution == "uniform"
            bits = 32
            c_type = "int"
            rng_expr = ("(shift "
                    "+ convert_int4((float) scale * gen) "
                    "+ convert_int4(((float) scale / (1<<24)) * gen))")

        elif dtype == np.int64:
            assert distribution == "uniform"
            if self.support_double:
                bits = 64
            else:
                bits = 32
            c_type = "long"
            rng_expr = ("(shift "
                    "+ convert_long4((float) scale * gen) "
                    "+ convert_long4(((float) scale / (1l<<24)) * gen)"
                    "+ convert_long4(((float) scale / (1l<<48)) * gen)"
                    ")")

        else:
            raise TypeError("unsupported RNG data type '%s'" % dtype)

        rl_flavor = "%d%s" % (bits, {
                "uniform": "",
                "normal": "norm"
                }[distribution])

        src = """//CL//
            %(defines)s

            #include <pyopencl-ranluxcl.cl>

            typedef %(output_t)s output_t;
            typedef %(output_t)s4 output_vec_t;
            #define NUM_WORKITEMS %(num_work_items)d
            #define RANLUX_FUNC ranluxcl%(rlflavor)s
            #define GET_RANDOM_NUM(gen) %(rng_expr)s

            kernel void generate(
                global ranluxcl_state_t *ranluxcltab,
                global output_t *output,
                unsigned long out_size,
                output_t scale,
                output_t shift)
            {

              ranluxcl_state_t ranluxclstate;
              ranluxcl_download_seed(&ranluxclstate, ranluxcltab);

              // output bulk
              unsigned long idx = get_global_id(0)*4;
              while (idx + 4 < out_size)
              {
                  *(global output_vec_t *) (output + idx) =
                      GET_RANDOM_NUM(RANLUX_FUNC(&ranluxclstate));
                  idx += 4*NUM_WORKITEMS;
              }

              // output tail
              output_vec_t tail_ran = GET_RANDOM_NUM(RANLUX_FUNC(&ranluxclstate));
              if (idx < out_size)
                output[idx] = tail_ran.x;
              if (idx+1 < out_size)
                output[idx+1] = tail_ran.y;
              if (idx+2 < out_size)
                output[idx+2] = tail_ran.z;
              if (idx+3 < out_size)
                output[idx+3] = tail_ran.w;

              ranluxcl_upload_seed(&ranluxclstate, ranluxcltab);
            }
            """ % {
                "defines": self.generate_settings_defines(),
                "rlflavor": rl_flavor,
                "output_t": c_type,
                "num_work_items": self.num_work_items,
                "rng_expr": rng_expr
                }

        prg = cl.Program(self.context, src).build()
        knl = prg.generate
        knl.set_scalar_arg_dtypes([None, None, np.uint64, arg_dtype, arg_dtype])

        return knl, size_multiplier

    def fill_uniform(self, ary, a=0, b=1, queue=None):
        """Fill *ary* with uniformly distributed random numbers in the interval
        *(a, b)*, endpoints excluded.

        :return: a :class:`pyopencl.Event`

        .. versionchanged:: 2014.1.1

            Added return value.
        """

        if queue is None:
            queue = ary.queue

        knl, size_multiplier = self.get_gen_kernel(ary.dtype, "uniform")
        return knl(queue,
                (self.num_work_items,), None,
                self.state.data, ary.data, ary.size*size_multiplier,
                b-a, a)

    def uniform(self, *args, **kwargs):
        """Make a new empty array, apply :meth:`fill_uniform` to it.
        """
        a = kwargs.pop("a", 0)
        b = kwargs.pop("b", 1)

        result = cl_array.empty(*args, **kwargs)

        result.add_event(
                self.fill_uniform(result, queue=result.queue, a=a, b=b))
        return result

    def fill_normal(self, ary, mu=0, sigma=1, queue=None):
        """Fill *ary* with normally distributed numbers with mean *mu* and
        standard deviation *sigma*.

        .. versionchanged:: 2014.1.1

            Added return value.
        """

        if queue is None:
            queue = ary.queue

        knl, size_multiplier = self.get_gen_kernel(ary.dtype, "normal")
        return knl(queue,
                (self.num_work_items,), self.wg_size,
                self.state.data, ary.data, ary.size*size_multiplier, sigma, mu)

    def normal(self, *args, **kwargs):
        """Make a new empty array, apply :meth:`fill_normal` to it.
        """
        mu = kwargs.pop("mu", 0)
        sigma = kwargs.pop("sigma", 1)

        result = cl_array.empty(*args, **kwargs)

        result.add_event(
                self.fill_normal(result, queue=result.queue, mu=mu, sigma=sigma))
        return result

    @memoize_method
    def get_sync_kernel(self):
        src = """//CL//
            %(defines)s

            #include <pyopencl-ranluxcl.cl>

            kernel void sync(
                global ranluxcl_state_t *ranluxcltab)
            {
              ranluxcl_state_t ranluxclstate;
              ranluxcl_download_seed(&ranluxclstate, ranluxcltab);
              ranluxcl_synchronize(&ranluxclstate);
              ranluxcl_upload_seed(&ranluxclstate, ranluxcltab);
            }
            """ % {
                "defines": self.generate_settings_defines(),
                }
        prg = cl.Program(self.context, src).build()
        return prg.sync

    def synchronize(self, queue):
        """The generator gets inefficient when different work items invoke the
        generator a differing number of times. This function ensures
        efficiency.
        """

        self.get_sync_kernel()(queue, (self.num_work_items,),
                self.wg_size, self.state.data)

# }}}


# {{{ Random123 generators

class Random123GeneratorBase(object):
    """
    .. versionadded:: 2016.2

    .. automethod:: fill_uniform
    .. automethod:: uniform
    .. automethod:: fill_normal
    .. automethod:: normal
    """

    def __init__(self, context, key=None, counter=None, seed=None):
        int32_info = np.iinfo(np.int32)
        from random import Random

        rng = Random(seed)

        if key is not None and counter is not None and seed is not None:
            raise TypeError("seed is unused and may not be specified "
                    "if both counter and key are given")

        if key is None:
            key = [
                    rng.randrange(
                        int(int32_info.min), int(int32_info.max)+1)
                    for i in range(self.key_length-1)]
        if counter is None:
            counter = [
                    rng.randrange(
                        int(int32_info.min), int(int32_info.max)+1)
                    for i in range(4)]

        self.context = context
        self.key = key
        self.counter = counter

        self.counter_max = int32_info.max

    @memoize_method
    def get_gen_kernel(self, dtype, distribution):
        size_multiplier = 1
        arg_dtype = dtype

        rng_key = (distribution, dtype)

        if rng_key in [("uniform", np.float64), ("normal", np.float64)]:
            c_type = "double"
            scale1_const = "((double) %r)" % (1/2**32)
            scale2_const = "((double) %r)" % (1/2**64)
            if distribution == "normal":
                transform = "box_muller"
            else:
                transform = ""

            rng_expr = (
                    "shift + scale * "
                    "%s( %s * convert_double4(gen)"
                    "+ %s * convert_double4(gen))"
                    % (transform, scale1_const, scale2_const))

            counter_multiplier = 2

        elif rng_key in [(dist, cmp_dtype)
                for dist in ["normal", "uniform"]
                for cmp_dtype in [
                    np.float32,
                    cltypes.float2,
                    cltypes.float3,
                    cltypes.float4,
                    ]]:
            c_type = "float"
            scale_const = "((float) %r)" % (1/2**32)

            if distribution == "normal":
                transform = "box_muller"
            else:
                transform = ""

            rng_expr = (
                    "shift + scale * %s(%s * convert_float4(gen))"
                    % (transform, scale_const))
            counter_multiplier = 1
            arg_dtype = np.float32
            try:
                _, size_multiplier = cltypes.vec_type_to_scalar_and_count[dtype]
            except KeyError:
                pass

        elif rng_key == ("uniform", np.int32):
            c_type = "int"
            rng_expr = (
                    "shift + convert_int4((convert_long4(gen) * scale) / %s)"
                    % (str(2**32)+"l")
                    )
            counter_multiplier = 1

        elif rng_key == ("uniform", np.int64):
            c_type = "long"
            rng_expr = (
                    "shift"
                    "+ convert_long4(gen) * (scale/two32) "
                    "+ ((convert_long4(gen) * scale) / two32)"
                    .replace("two32", (str(2**32)+"l")))
            counter_multiplier = 2

        else:
            raise TypeError(
                    "unsupported RNG distribution/data type combination '%s/%s'"
                    % rng_key)

        kernel_name = "rng_gen_%s_%s" % (self.generator_name, distribution)
        src = """//CL//
            #include <%(header_name)s>

            #ifndef M_PI
            #ifdef M_PI_F
            #define M_PI M_PI_F
            #else
            #define M_PI 3.14159265359f
            #endif
            #endif

            typedef %(output_t)s output_t;
            typedef %(output_t)s4 output_vec_t;
            typedef %(gen_name)s_ctr_t ctr_t;
            typedef %(gen_name)s_key_t key_t;

            uint4 gen_bits(key_t *key, ctr_t *ctr)
            {
                union {
                    ctr_t ctr_el;
                    uint4 vec_el;
                } u;

                u.ctr_el = %(gen_name)s(*ctr, *key);
                if (++ctr->v[0] == 0)
                    if (++ctr->v[1] == 0)
                        ++ctr->v[2];

                return u.vec_el;
            }

            #if %(include_box_muller)s
            output_vec_t box_muller(output_vec_t x)
            {
                #define BOX_MULLER(I, COMPA, COMPB) \
                    output_t r##I = sqrt(-2*log(x.COMPA)); \
                    output_t c##I; \
                    output_t s##I = sincos((output_t) (2*M_PI) * x.COMPB, &c##I);

                BOX_MULLER(0, x, y);
                BOX_MULLER(1, z, w);
                return (output_vec_t) (r0*c0, r0*s0, r1*c1, r1*s1);
            }
            #endif

            #define GET_RANDOM_NUM(gen) %(rng_expr)s

            kernel void %(kernel_name)s(
                int k1,
                #if %(key_length)s > 2
                int k2, int k3,
                #endif
                int c0, int c1, int c2, int c3,
                global output_t *output,
                long out_size,
                output_t scale,
                output_t shift)
            {
                #if %(key_length)s == 2
                key_t k = {{get_global_id(0), k1}};
                #else
                key_t k = {{get_global_id(0), k1, k2, k3}};
                #endif

                ctr_t c = {{c0, c1, c2, c3}};

                // output bulk
                unsigned long idx = get_global_id(0)*4;
                while (idx + 4 < out_size)
                {
                    *(global output_vec_t *) (output + idx) =
                        GET_RANDOM_NUM(gen_bits(&k, &c));
                    idx += 4*get_global_size(0);
                }

                // output tail
                output_vec_t tail_ran = GET_RANDOM_NUM(gen_bits(&k, &c));
                if (idx < out_size)
                  output[idx] = tail_ran.x;
                if (idx+1 < out_size)
                  output[idx+1] = tail_ran.y;
                if (idx+2 < out_size)
                  output[idx+2] = tail_ran.z;
                if (idx+3 < out_size)
                  output[idx+3] = tail_ran.w;
            }
            """ % {
                "kernel_name": kernel_name,
                "gen_name": self.generator_name,
                "header_name": self.header_name,
                "output_t": c_type,
                "key_length": self.key_length,
                "include_box_muller": int(distribution == "normal"),
                "rng_expr": rng_expr
                }

        prg = cl.Program(self.context, src).build()
        knl = getattr(prg, kernel_name)
        knl.set_scalar_arg_dtypes(
                [np.int32] * (self.key_length - 1 + 4)
                + [None, np.int64, arg_dtype, arg_dtype])

        return knl, counter_multiplier, size_multiplier

    def _fill(self, distribution, ary, scale, shift, queue=None):
        """Fill *ary* with uniformly distributed random numbers in the interval
        *(a, b)*, endpoints excluded.

        :return: a :class:`pyopencl.Event`
        """

        if queue is None:
            queue = ary.queue

        knl, counter_multiplier, size_multiplier = \
                self.get_gen_kernel(ary.dtype, distribution)

        args = self.key + self.counter + [
                ary.data, ary.size*size_multiplier,
                scale, shift]

        n = ary.size
        from pyopencl.array import splay
        gsize, lsize = splay(queue, ary.size)

        evt = knl(queue, gsize, lsize, *args)

        self.counter[0] += n * counter_multiplier
        c1_incr, self.counter[0] = divmod(self.counter[0], self.counter_max)
        if c1_incr:
            self.counter[1] += c1_incr
            c2_incr, self.counter[1] = divmod(self.counter[1], self.counter_max)
            self.counter[2] += c2_incr

        return evt

    def fill_uniform(self, ary, a=0, b=1, queue=None):
        return self._fill("uniform", ary,
                scale=(b-a), shift=a, queue=queue)

    def uniform(self, *args, **kwargs):
        """Make a new empty array, apply :meth:`fill_uniform` to it.
        """
        a = kwargs.pop("a", 0)
        b = kwargs.pop("b", 1)

        result = cl_array.empty(*args, **kwargs)

        result.add_event(
                self.fill_uniform(result, queue=result.queue, a=a, b=b))
        return result

    def fill_normal(self, ary, mu=0, sigma=1, queue=None):
        """Fill *ary* with normally distributed numbers with mean *mu* and
        standard deviation *sigma*.
        """

        return self._fill("normal", ary, scale=sigma, shift=mu, queue=queue)

    def normal(self, *args, **kwargs):
        """Make a new empty array, apply :meth:`fill_normal` to it.
        """
        mu = kwargs.pop("mu", 0)
        sigma = kwargs.pop("sigma", 1)

        result = cl_array.empty(*args, **kwargs)

        result.add_event(
                self.fill_normal(result, queue=result.queue, mu=mu, sigma=sigma))
        return result


class PhiloxGenerator(Random123GeneratorBase):
    __doc__ = Random123GeneratorBase.__doc__

    header_name = "pyopencl-random123/philox.cl"
    generator_name = "philox4x32"
    key_length = 2


class ThreefryGenerator(Random123GeneratorBase):
    __doc__ = Random123GeneratorBase.__doc__

    header_name = "pyopencl-random123/threefry.cl"
    generator_name = "threefry4x32"
    key_length = 4

# }}}


@first_arg_dependent_memoize
def _get_generator(context):
    if context.devices[0].type & cl.device_type.CPU:
        gen = PhiloxGenerator(context)
    else:
        gen = ThreefryGenerator(context)

    return gen


def fill_rand(result, queue=None, luxury=None, a=0, b=1):
    """Fill *result* with random values of `dtype` in the range [0,1).
    """
    if luxury is not None:
        from warnings import warn
        warn("Specifying the 'luxury' argument is deprecated and will stop being "
                "supported in PyOpenCL 2018.x", stacklevel=2)

    if queue is None:
        queue = result.queue
    gen = _get_generator(queue.context)
    gen.fill_uniform(result, a=a, b=b)


def rand(queue, shape, dtype, luxury=None, a=0, b=1):
    """Return an array of `shape` filled with random values of `dtype`
    in the range [a,b).
    """

    if luxury is not None:
        from warnings import warn
        warn("Specifying the 'luxury' argument is deprecated and will stop being "
                "supported in PyOpenCL 2018.x", stacklevel=2)

    from pyopencl.array import Array
    gen = _get_generator(queue.context)
    result = Array(queue, shape, dtype)
    result.add_event(
            gen.fill_uniform(result, a=a, b=b))
    return result


# vim: filetype=pyopencl:foldmethod=marker