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
|
/*
* Copyright (C) 2020-2024 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "shared/source/command_container/command_encoder.h"
#include "shared/source/command_container/command_encoder.inl"
#include "shared/source/command_container/command_encoder_bdw_and_later.inl"
#include "shared/source/command_container/command_encoder_pre_xe2_hpg_core.inl"
#include "shared/source/command_stream/stream_properties.h"
#include "shared/source/gen12lp/hw_cmds_base.h"
#include "shared/source/gen12lp/reg_configs.h"
#include "shared/source/gmm_helper/gmm_helper.h"
#include "shared/source/helpers/preamble.h"
#include "shared/source/release_helper/release_helper.h"
#include "encode_surface_state_args.h"
using Family = NEO::Gen12LpFamily;
#include "shared/source/command_container/command_encoder_heap_addressing.inl"
#include "shared/source/command_container/command_encoder_tgllp_and_later.inl"
#include "shared/source/command_container/encode_compute_mode_tgllp_and_later.inl"
#include "shared/source/command_container/image_surface_state/compression_params_bdw_and_later.inl"
#include "shared/source/command_container/image_surface_state/compression_params_tgllp_and_later.inl"
#include "shared/source/command_stream/command_stream_receiver.h"
namespace NEO {
template <>
size_t EncodeWA<Family>::getAdditionalPipelineSelectSize(Device &device, bool isRcs) {
size_t size = 0;
const auto &productHelper = device.getProductHelper();
if (isRcs && productHelper.is3DPipelineSelectWARequired()) {
size += 2 * PreambleHelper<Family>::getCmdSizeForPipelineSelect(device.getRootDeviceEnvironment());
}
return size;
}
template <>
void EncodeComputeMode<Family>::programComputeModeCommand(LinearStream &csr, StateComputeModeProperties &properties, const RootDeviceEnvironment &rootDeviceEnvironment) {
using STATE_COMPUTE_MODE = typename Family::STATE_COMPUTE_MODE;
using FORCE_NON_COHERENT = typename STATE_COMPUTE_MODE::FORCE_NON_COHERENT;
STATE_COMPUTE_MODE stateComputeMode = Family::cmdInitStateComputeMode;
auto maskBits = stateComputeMode.getMaskBits();
FORCE_NON_COHERENT coherencyValue = (properties.isCoherencyRequired.value == 1) ? FORCE_NON_COHERENT::FORCE_NON_COHERENT_FORCE_DISABLED
: FORCE_NON_COHERENT::FORCE_NON_COHERENT_FORCE_GPU_NON_COHERENT;
stateComputeMode.setForceNonCoherent(coherencyValue);
maskBits |= Family::stateComputeModeForceNonCoherentMask;
stateComputeMode.setMaskBits(maskBits);
auto buffer = csr.getSpace(sizeof(STATE_COMPUTE_MODE));
*reinterpret_cast<STATE_COMPUTE_MODE *>(buffer) = stateComputeMode;
}
template <>
void EncodeWA<Family>::encodeAdditionalPipelineSelect(LinearStream &stream, const PipelineSelectArgs &args,
bool is3DPipeline, const RootDeviceEnvironment &rootDeviceEnvironment, bool isRcs) {
const auto &productHelper = rootDeviceEnvironment.getHelper<ProductHelper>();
if (productHelper.is3DPipelineSelectWARequired() && isRcs) {
PipelineSelectArgs pipelineSelectArgs = args;
pipelineSelectArgs.is3DPipelineRequired = is3DPipeline;
PreambleHelper<Family>::programPipelineSelect(&stream, pipelineSelectArgs, rootDeviceEnvironment);
}
}
template <>
void EncodeSurfaceState<Family>::encodeExtraBufferParams(EncodeSurfaceStateArgs &args) {
auto surfaceState = reinterpret_cast<R_SURFACE_STATE *>(args.outMemory);
const bool isL3Allowed = surfaceState->getMemoryObjectControlState() == args.gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER);
if (isL3Allowed) {
const bool isConstantSurface = args.allocation && args.allocation->getAllocationType() == AllocationType::constantSurface;
bool useL1 = args.isReadOnly || isConstantSurface;
if (debugManager.flags.ForceL1Caching.get() != 1) {
useL1 = false;
}
if (useL1) {
surfaceState->setMemoryObjectControlState(args.gmmHelper->getMOCS(GMM_RESOURCE_USAGE_OCL_BUFFER_CONST));
}
}
}
template <>
bool EncodeSurfaceState<Family>::isBindingTablePrefetchPreferred() {
return false;
}
template <>
void EncodeL3State<Family>::encode(CommandContainer &container, bool enableSLM) {
}
template <>
void EncodeStoreMMIO<Family>::appendFlags(MI_STORE_REGISTER_MEM *storeRegMem, bool workloadPartition) {
storeRegMem->setMmioRemapEnable(true);
}
template <>
void EncodeComputeMode<Family>::adjustPipelineSelect(CommandContainer &container, const NEO::KernelDescriptor &kernelDescriptor) {
PipelineSelectArgs pipelineSelectArgs;
pipelineSelectArgs.systolicPipelineSelectMode = kernelDescriptor.kernelAttributes.flags.usesSystolicPipelineSelectMode;
pipelineSelectArgs.systolicPipelineSelectSupport = container.systolicModeSupportRef();
PreambleHelper<Family>::programPipelineSelect(container.getCommandStream(),
pipelineSelectArgs,
container.getDevice()->getRootDeviceEnvironment());
}
} // namespace NEO
#include "shared/source/command_container/command_encoder_enablers.inl"
namespace NEO {
template struct EncodeL3State<Family>;
template void InOrderPatchCommandHelpers::PatchCmd<Family>::patchComputeWalker(uint64_t appendCounterValue);
} // namespace NEO
#include "shared/source/command_container/implicit_scaling_before_xe_hp.inl"
|
