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
|
/*
* Copyright (C) 2017-2020 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "opencl/source/command_queue/command_queue.h"
#include "opencl/source/event/event.h"
#include "opencl/test/unit_test/fixtures/hello_world_fixture.h"
#include "gtest/gtest.h"
using namespace NEO;
typedef HelloWorldTest<HelloWorldFixtureFactory> EventTests;
TEST_F(EventTests, WhenEnqueingKernelThenCorrectEventIsReturned) {
cl_event event = nullptr;
auto retVal = callOneWorkItemNDRKernel(nullptr, 0, &event);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event);
auto pEvent = (Event *)event;
EXPECT_EQ(pCmdQ->taskLevel, pEvent->taskLevel);
// Check CL_EVENT_COMMAND_TYPE
{
cl_command_type cmdType = 0;
size_t sizeReturned = 0;
auto result = clGetEventInfo(pEvent, CL_EVENT_COMMAND_TYPE, sizeof(cmdType), &cmdType, &sizeReturned);
ASSERT_EQ(CL_SUCCESS, result);
EXPECT_EQ(static_cast<cl_command_type>(CL_COMMAND_NDRANGE_KERNEL), cmdType);
EXPECT_EQ(sizeof(cl_command_type), sizeReturned);
}
delete pEvent;
}
TEST_F(EventTests, WhenEnqueingKernelThenEventReturnedShouldBeMaxOfInputEventsAndCmdQPlus1) {
uint32_t taskLevelCmdQ = 17;
pCmdQ->taskLevel = taskLevelCmdQ;
uint32_t taskLevelEvent1 = 8;
uint32_t taskLevelEvent2 = 19;
Event event1(pCmdQ, CL_COMMAND_NDRANGE_KERNEL, taskLevelEvent1, 15);
Event event2(pCmdQ, CL_COMMAND_NDRANGE_KERNEL, taskLevelEvent2, 16);
cl_event eventWaitList[] =
{
&event1,
&event2};
cl_uint numEventsInWaitList = sizeof(eventWaitList) / sizeof(eventWaitList[0]);
cl_event event = nullptr;
auto retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event);
auto pEvent = (Event *)event;
EXPECT_EQ(19u + 1u, pEvent->taskLevel);
delete pEvent;
}
TEST_F(EventTests, WhenWaitingForEventThenPipeControlIsNotInserted) {
cl_uint numEventsInWaitList = 0;
cl_event *eventWaitList = nullptr;
cl_event event = nullptr;
auto &csr = pCmdQ->getGpgpuCommandStreamReceiver();
auto retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event);
auto pEvent = (Event *)event;
EXPECT_EQ(pCmdQ->taskLevel, pEvent->taskLevel);
retVal = Event::waitForEvents(1, &event);
EXPECT_EQ(CL_SUCCESS, retVal);
//we expect event is completed
uint32_t taskCountOfEvent = pEvent->peekTaskCount();
EXPECT_LE(taskCountOfEvent, pCmdQ->getHwTag());
// no more tasks after WFE, no need to write PC
EXPECT_EQ(pEvent->taskLevel + 1, csr.peekTaskLevel());
pCmdQ->finish();
// Check CL_EVENT_COMMAND_TYPE
{
cl_command_type cmdType = 0;
size_t sizeReturned = 0;
auto result = clGetEventInfo(pEvent, CL_EVENT_COMMAND_TYPE, sizeof(cmdType), &cmdType, &sizeReturned);
ASSERT_EQ(CL_SUCCESS, result);
EXPECT_EQ(static_cast<cl_command_type>(CL_COMMAND_NDRANGE_KERNEL), cmdType);
EXPECT_EQ(sizeof(cl_command_type), sizeReturned);
}
delete pEvent;
}
TEST_F(EventTests, GivenTwoEnqueuesWhenWaitingForBothEventsThenTaskLevelIsCorrect) {
cl_uint numEventsInWaitList = 0;
cl_event *eventWaitList = nullptr;
cl_event event[2] = {};
auto &csr = pCmdQ->getGpgpuCommandStreamReceiver();
auto retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event[0]);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event[0]);
auto pEvent0 = castToObject<Event>(event[0]);
EXPECT_EQ(pCmdQ->taskLevel, pEvent0->taskLevel);
retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event[1]);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event[1]);
auto pEvent1 = castToObject<Event>(event[1]);
EXPECT_EQ(pCmdQ->taskLevel, pEvent1->taskLevel);
EXPECT_GT(pEvent1->taskLevel, pEvent0->taskLevel);
retVal = Event::waitForEvents(2, event);
EXPECT_EQ(CL_SUCCESS, retVal);
EXPECT_EQ(pEvent1->taskLevel + 1, csr.peekTaskLevel());
pCmdQ->finish();
EXPECT_EQ(pEvent1->taskLevel + 1, csr.peekTaskLevel());
// Check CL_EVENT_COMMAND_TYPE
{
cl_command_type cmdType = 0;
size_t sizeReturned = 0;
auto result = clGetEventInfo(pEvent0, CL_EVENT_COMMAND_TYPE, sizeof(cmdType), &cmdType, &sizeReturned);
ASSERT_EQ(CL_SUCCESS, result);
EXPECT_EQ(static_cast<cl_command_type>(CL_COMMAND_NDRANGE_KERNEL), cmdType);
EXPECT_EQ(sizeof(cl_command_type), sizeReturned);
}
delete pEvent0;
delete pEvent1;
}
TEST_F(EventTests, GivenNoEventsWhenEnqueuingKernelThenTaskLevelIsIncremented) {
cl_uint numEventsInWaitList = 0;
cl_event *eventWaitList = nullptr;
cl_event event = nullptr;
auto &csr = pCmdQ->getGpgpuCommandStreamReceiver();
auto retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event);
auto pEvent = (Event *)event;
EXPECT_EQ(pCmdQ->taskLevel, pEvent->taskLevel);
retVal = Event::waitForEvents(1, &event);
EXPECT_EQ(CL_SUCCESS, retVal);
EXPECT_EQ(pEvent->taskLevel + 1, csr.peekTaskLevel());
retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, nullptr);
ASSERT_EQ(CL_SUCCESS, retVal);
EXPECT_EQ(pEvent->taskLevel + 2, csr.peekTaskLevel());
pCmdQ->finish();
EXPECT_EQ(pEvent->taskLevel + 2, csr.peekTaskLevel());
// Check CL_EVENT_COMMAND_TYPE
{
cl_command_type cmdType = 0;
size_t sizeReturned = 0;
auto result = clGetEventInfo(pEvent, CL_EVENT_COMMAND_TYPE, sizeof(cmdType), &cmdType, &sizeReturned);
ASSERT_EQ(CL_SUCCESS, result);
EXPECT_EQ(static_cast<cl_command_type>(CL_COMMAND_NDRANGE_KERNEL), cmdType);
EXPECT_EQ(sizeof(cl_command_type), sizeReturned);
}
delete pEvent;
}
TEST_F(EventTests, WhenEnqueuingMarkerThenPassedEventHasTheSameLevelAsPreviousCommand) {
cl_uint numEventsInWaitList = 0;
cl_event *eventWaitList = nullptr;
cl_event event = nullptr;
auto &csr = pCmdQ->getGpgpuCommandStreamReceiver();
auto retVal = callOneWorkItemNDRKernel(eventWaitList, numEventsInWaitList, &event);
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event);
auto pEvent = (Event *)event;
EXPECT_EQ(pEvent->taskLevel + 1, csr.peekTaskLevel());
cl_event event2 = nullptr;
retVal = clEnqueueMarkerWithWaitList(pCmdQ, 1, &event, &event2);
auto pEvent2 = castToObject<Event>(event2);
if (csr.peekTimestampPacketWriteEnabled()) {
EXPECT_EQ(pEvent2->taskLevel, pEvent->taskLevel + 1);
} else {
EXPECT_EQ(pEvent2->taskLevel, pEvent->taskLevel);
}
ASSERT_EQ(CL_SUCCESS, retVal);
ASSERT_NE(nullptr, event2);
retVal = clWaitForEvents(1, &event2);
ASSERT_EQ(CL_SUCCESS, retVal);
if (csr.peekTimestampPacketWriteEnabled()) {
EXPECT_EQ(csr.peekTaskLevel(), pCmdQ->taskLevel + 1);
} else {
EXPECT_EQ(csr.peekTaskLevel(), pEvent->taskLevel + 1);
}
clReleaseEvent(event);
clReleaseEvent(event2);
}
|
