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
|
/*
* Copyright (C) 2017-2020 Intel Corporation
*
* SPDX-License-Identifier: MIT
*
*/
#include "opencl/source/helpers/per_thread_data.h"
#include "shared/source/command_stream/linear_stream.h"
#include "shared/source/helpers/debug_helpers.h"
#include <array>
namespace NEO {
size_t PerThreadDataHelper::sendPerThreadData(
LinearStream &indirectHeap,
uint32_t simd,
uint32_t grfSize,
uint32_t numChannels,
const size_t localWorkSizes[3],
const std::array<uint8_t, 3> &workgroupWalkOrder,
bool hasKernelOnlyImages) {
auto offsetPerThreadData = indirectHeap.getUsed();
if (numChannels) {
auto localWorkSize = localWorkSizes[0] * localWorkSizes[1] * localWorkSizes[2];
auto sizePerThreadDataTotal = getPerThreadDataSizeTotal(simd, grfSize, numChannels, localWorkSize);
auto pDest = indirectHeap.getSpace(sizePerThreadDataTotal);
// Generate local IDs
DEBUG_BREAK_IF(numChannels != 3);
generateLocalIDs(pDest, static_cast<uint16_t>(simd),
std::array<uint16_t, 3>{{static_cast<uint16_t>(localWorkSizes[0]),
static_cast<uint16_t>(localWorkSizes[1]),
static_cast<uint16_t>(localWorkSizes[2])}},
std::array<uint8_t, 3>{{workgroupWalkOrder[0], workgroupWalkOrder[1], workgroupWalkOrder[2]}},
hasKernelOnlyImages, grfSize);
}
return offsetPerThreadData;
}
uint32_t PerThreadDataHelper::getThreadPayloadSize(const iOpenCL::SPatchThreadPayload &threadPayload, uint32_t simd, uint32_t grfSize) {
uint32_t multiplier = static_cast<uint32_t>(getGRFsPerThread(simd, grfSize));
uint32_t threadPayloadSize = 0;
threadPayloadSize = getNumLocalIdChannels(threadPayload) * multiplier * grfSize;
threadPayloadSize += (threadPayload.HeaderPresent) ? grfSize : 0;
threadPayloadSize += (threadPayload.LocalIDFlattenedPresent) ? (grfSize * multiplier) : 0;
threadPayloadSize += (threadPayload.UnusedPerThreadConstantPresent) ? grfSize : 0;
return threadPayloadSize;
}
} // namespace NEO
|
