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
|
/*
* Copyright (C) 2018-2022 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "shared/source/command_stream/command_stream_receiver_hw.h"
#include "shared/source/command_stream/experimental_command_buffer.h"
#include "shared/source/command_stream/linear_stream.h"
#include "shared/source/execution_environment/root_device_environment.h"
#include "shared/source/helpers/hw_helper.h"
#include "shared/source/helpers/pipe_control_args.h"
#include "shared/source/memory_manager/graphics_allocation.h"
namespace NEO {
template <typename GfxFamily>
void ExperimentalCommandBuffer::injectBufferStart(LinearStream &parentStream, size_t cmdBufferOffset) {
using MI_BATCH_BUFFER_START = typename GfxFamily::MI_BATCH_BUFFER_START;
auto pCmd = parentStream.getSpaceForCmd<MI_BATCH_BUFFER_START>();
auto commandStreamReceiverHw = static_cast<CommandStreamReceiverHw<GfxFamily> *>(commandStreamReceiver);
commandStreamReceiverHw->addBatchBufferStart(pCmd, currentStream->getGraphicsAllocation()->getGpuAddress() + cmdBufferOffset, true);
}
template <typename GfxFamily>
size_t ExperimentalCommandBuffer::getRequiredInjectionSize() noexcept {
using MI_BATCH_BUFFER_START = typename GfxFamily::MI_BATCH_BUFFER_START;
return sizeof(MI_BATCH_BUFFER_START);
}
template <typename GfxFamily>
size_t ExperimentalCommandBuffer::programExperimentalCommandBuffer() {
using MI_BATCH_BUFFER_END = typename GfxFamily::MI_BATCH_BUFFER_END;
getCS(getTotalExperimentalSize<GfxFamily>());
size_t returnOffset = currentStream->getUsed();
//begin timestamp
addTimeStampPipeControl<GfxFamily>();
addExperimentalCommands<GfxFamily>();
//end timestamp
addTimeStampPipeControl<GfxFamily>();
//end
auto pCmd = currentStream->getSpaceForCmd<MI_BATCH_BUFFER_END>();
*pCmd = GfxFamily::cmdInitBatchBufferEnd;
return returnOffset;
}
template <typename GfxFamily>
size_t ExperimentalCommandBuffer::getTotalExperimentalSize() noexcept {
using MI_BATCH_BUFFER_END = typename GfxFamily::MI_BATCH_BUFFER_END;
size_t size = sizeof(MI_BATCH_BUFFER_END) + getTimeStampPipeControlSize<GfxFamily>() + getExperimentalCommandsSize<GfxFamily>();
return size;
}
template <typename GfxFamily>
size_t ExperimentalCommandBuffer::getTimeStampPipeControlSize() noexcept {
// Two P_C for timestamps
return 2 * MemorySynchronizationCommands<GfxFamily>::getSizeForBarrierWithPostSyncOperation(
*commandStreamReceiver->peekExecutionEnvironment().rootDeviceEnvironments[commandStreamReceiver->getRootDeviceIndex()]->getHardwareInfo(), false);
}
template <typename GfxFamily>
void ExperimentalCommandBuffer::addTimeStampPipeControl() {
uint64_t timeStampAddress = timestamps->getGpuAddress() + timestampsOffset;
PipeControlArgs args;
MemorySynchronizationCommands<GfxFamily>::addBarrierWithPostSyncOperation(
*currentStream,
PostSyncMode::Timestamp,
timeStampAddress,
0llu,
*commandStreamReceiver->peekExecutionEnvironment().rootDeviceEnvironments[commandStreamReceiver->getRootDeviceIndex()]->getHardwareInfo(),
args);
//moving to next chunk
timestampsOffset += sizeof(uint64_t);
DEBUG_BREAK_IF(timestamps->getUnderlyingBufferSize() < timestampsOffset);
}
template <typename GfxFamily>
void ExperimentalCommandBuffer::addExperimentalCommands() {
using MI_SEMAPHORE_WAIT = typename GfxFamily::MI_SEMAPHORE_WAIT;
uint32_t *semaphoreData = reinterpret_cast<uint32_t *>(ptrOffset(experimentalAllocation->getUnderlyingBuffer(), experimentalAllocationOffset));
*semaphoreData = 1;
uint64_t gpuAddr = experimentalAllocation->getGpuAddress() + experimentalAllocationOffset;
auto semaphoreCmdSpace = currentStream->getSpaceForCmd<MI_SEMAPHORE_WAIT>();
auto semaphoreCmd = GfxFamily::cmdInitMiSemaphoreWait;
semaphoreCmd.setCompareOperation(MI_SEMAPHORE_WAIT::COMPARE_OPERATION_SAD_EQUAL_SDD);
semaphoreCmd.setSemaphoreDataDword(*semaphoreData);
semaphoreCmd.setSemaphoreGraphicsAddress(gpuAddr);
*semaphoreCmdSpace = semaphoreCmd;
}
template <typename GfxFamily>
size_t ExperimentalCommandBuffer::getExperimentalCommandsSize() noexcept {
using MI_SEMAPHORE_WAIT = typename GfxFamily::MI_SEMAPHORE_WAIT;
return sizeof(MI_SEMAPHORE_WAIT);
}
} // namespace NEO
|
