intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-21 09:14:47 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add implementation for reading logical subDevice from builtin

Browse Source 
- add function for resolving builtins using relocations
- add relocation for __SubDeviceID symbol

Related-To: NEO-6258
Signed-off-by: Kamil Diedrich <kamil.diedrich@intel.com>
This commit is contained in:
Kamil Diedrich
2021-12-08 22:01:47 +00:00
committed by Compute-Runtime-Automation
parent d07c76c237
commit 24c19509b3
5 changed files with 110 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
shared/source/command_stream/command_stream_receiver.cpp
View File

@@ -592,15 +592,16 @@ bool CommandStreamReceiver::createWorkPartitionAllocation(const Device &device)
return false;
}
uint32_t logicalId = 0;
for (uint32_t deviceIndex = 0; deviceIndex < deviceBitfield.size(); deviceIndex++) {
if (!deviceBitfield.test(deviceIndex)) {
continue;
}
const uint32_t copySrc = deviceIndex;
const uint32_t copySrc[2] = {logicalId++, deviceIndex};
DeviceBitfield copyBitfield{};
copyBitfield.set(deviceIndex);
auto copySuccess = MemoryTransferHelper::transferMemoryToAllocationBanks(device, workPartitionAllocation, 0, &copySrc, sizeof(copySrc), copyBitfield);
auto copySuccess = MemoryTransferHelper::transferMemoryToAllocationBanks(device, workPartitionAllocation, 0, copySrc, sizeof(copySrc), copyBitfield);
if (!copySuccess) {
return false;

18
shared/source/compiler_interface/linker.cpp
View File

@@ -7,6 +7,7 @@
#include "shared/source/compiler_interface/linker.h"
#include "shared/source/command_stream/command_stream_receiver.h"
#include "shared/source/compiler_interface/linker.inl"
#include "shared/source/device/device.h"
#include "shared/source/device_binary_format/elf/zebin_elf.h"
@@ -481,4 +482,21 @@ void Linker::resolveImplicitArgs(const KernelDescriptorsT &kernelDescriptors, De
}
}
}
void Linker::resolveBuiltins(Device *pDevice, UnresolvedExternals &outUnresolvedExternals, const std::vector<PatchableSegment> &instructionsSegments) {
int vecIndex = static_cast<int>(outUnresolvedExternals.size() - 1u);
for (; vecIndex >= 0; --vecIndex) {
if (outUnresolvedExternals[vecIndex].unresolvedRelocation.symbolName == subDeviceID) {
RelocatedSymbol symbol;
symbol.gpuAddress = static_cast<uintptr_t>(pDevice->getDefaultEngine().commandStreamReceiver->getWorkPartitionAllocationGpuAddress());
auto relocAddress = ptrOffset(instructionsSegments[outUnresolvedExternals[vecIndex].instructionsSegmentId].hostPointer,
static_cast<uintptr_t>(outUnresolvedExternals[vecIndex].unresolvedRelocation.offset));
NEO::Linker::patchAddress(relocAddress, symbol, outUnresolvedExternals[vecIndex].unresolvedRelocation);
outUnresolvedExternals[vecIndex] = outUnresolvedExternals[outUnresolvedExternals.size() - 1u];
outUnresolvedExternals.resize(outUnresolvedExternals.size() - 1u);
}
}
}
} // namespace NEO

5
shared/source/compiler_interface/linker.h
View File

@@ -152,6 +152,8 @@ struct LinkerInput {
};
struct Linker {
static constexpr std::string_view subDeviceID{"__SubDeviceID"};
using RelocationInfo = LinkerInput::RelocationInfo;
struct SegmentInfo {
@@ -197,7 +199,7 @@ struct Linker {
patchDataSegments(globalVariablesSegInfo, globalConstantsSegInfo, globalVariablesSeg, globalConstantsSeg,
outUnresolvedExternals, pDevice, constantsInitData, variablesInitData);
resolveImplicitArgs(kernelDescriptors, pDevice);
resolveBuiltins(pDevice, outUnresolvedExternals, instructionsSegments);
if (initialUnresolvedExternalsCount < outUnresolvedExternals.size()) {
return LinkingStatus::LinkedPartially;
}
@@ -227,6 +229,7 @@ struct Linker {
const void *constantsInitData, const void *variablesInitData);
void resolveImplicitArgs(const KernelDescriptorsT &kernelDescriptors, Device *pDevice);
void resolveBuiltins(Device *pDevice, UnresolvedExternals &outUnresolvedExternals, const std::vector<PatchableSegment> &instructionsSegments);
template <typename PatchSizeT>
void patchIncrement(Device *pDevice, GraphicsAllocation *dstAllocation, size_t relocationOffset, const void *initData, uint64_t incrementValue);