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
|
################################################################################
# Copyright 2014 Advanced Micro Devices, Inc.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
################################################################################
cimport pyclBLAS
import pyopencl
# These are prototypes from clBLAS.h that we wish to call from python
################################################################################
################################################################################
cdef extern from "clBLAS.h":
clblasStatus clblasGetVersion( cl_uint* major, cl_uint* minor, cl_uint* patch )
clblasStatus clblasSetup( )
void clblasTeardown( )
clblasStatus clblasSgemm( clblasOrder order, clblasTranspose transA, clblasTranspose transB,
size_t M, size_t N, size_t K, cl_float alpha, const cl_mem A, size_t offA, size_t lda,
const cl_mem B, size_t offB, size_t ldb, cl_float beta, cl_mem C, size_t offC, size_t ldc,
cl_uint numCommandQueues, cl_command_queue* commandQueues, cl_uint numEventsInWaitList,
const cl_event* eventWaitList, cl_event* events)
################################################################################
################################################################################
# enums to be accessed from python
# TODO: is there a better way to express enums? I like how pyopencl does it,
# they have layers of scoped constants cl.mem_flags.READ_ONLY
# The enums below have global scope
RowMajor = pyclBLAS.clblasRowMajor
ColumnMajor = pyclBLAS.clblasColumnMajor
NoTrans = pyclBLAS.clblasNoTrans
Trans = pyclBLAS.clblasTrans
ConjTrans = pyclBLAS.clblasConjTrans
################################################################################
################################################################################
# The following functions are the python callable wrapper implementations
def Setup( ):
result = clblasSetup( )
if( result != clblasSuccess ):
raise RuntimeError( "clblasSetup( ) failed initialization" )
return result
################################################################################
def Teardown( ):
clblasTeardown( )
return
################################################################################
def GetVersion( ):
cdef pyclBLAS.cl_uint pyMajor
cdef pyclBLAS.cl_uint pyMinor
cdef pyclBLAS.cl_uint pyPatch
result = clblasGetVersion( &pyMajor, &pyMinor, &pyPatch )
if( result != clblasSuccess ):
raise RuntimeError( "clblasGetVersion( ) did not return version information" )
return pyMajor, pyMinor, pyPatch
################################################################################
# TODO: Is there way to template these python callable functions, such that we
# do not need to make a new function for every supported precision?
def Sgemm( clblasOrder order, clblasTranspose transA, clblasTranspose transB,
size_t M, size_t N, size_t K, cl_float alpha, A, size_t offA, size_t lda,
B, size_t offB, size_t ldb, cl_float beta, C, size_t offC, size_t ldc,
cl_uint numCommandQueues, commandQueues, cl_uint numEventsInWaitList,
eventWaitList ):
# Simplify python wrapper to only handle 1 queue at this time
if( numCommandQueues != 1 ):
raise IndexError( "pyblasSgemm( ) requires the number of queues to be 1" )
cdef intptr_t pIntQueue = commandQueues.int_ptr
cdef cl_command_queue pcqQueue = <cl_command_queue>pIntQueue
# This logic does not yet work for numEventsInWaitList > (greater than) 1
# Need to figure out how python & pyopencl pass lists of objects
cdef intptr_t pIntWaitList = 0
cdef cl_event* pWaitList = NULL
if( numEventsInWaitList > 0 ):
if( numEventsInWaitList < 2 ):
pIntWaitList = eventWaitList.int_ptr
pWaitList = <cl_event*>pIntWaitList
else:
raise IndexError( "pyblasSgemm( ) requires numEventsInWaitList to be <= 1" )
# Pyopencl objects contain an int_ptr method to get access to the internally wrapped
# OpenCL object pointers
cdef cl_event outEvent = NULL
cdef intptr_t matA = A.int_ptr
cdef intptr_t matB = B.int_ptr
cdef intptr_t matC = C.int_ptr
# Transition execution to clBLAS
cdef clblasStatus result = clblasSgemm( order, transA, transB, M, N, K, alpha, <const cl_mem>matA, offA, lda,
<const cl_mem>matB, offB, ldb, beta, <cl_mem>matC, offC, ldc,
numCommandQueues, &pcqQueue, numEventsInWaitList,
pWaitList, &outEvent )
if( result != clblasSuccess ):
raise RuntimeError( "clBLAS sgemm call failed" )
# Create a pyopencl Event object from the event returned from clBLAS and return
# it to the user
sgemmEvent = pyopencl.Event.from_int_ptr( <intptr_t>outEvent )
return sgemmEvent
|
