intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2026-01-03 23:03:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

Add method translating attention bitmask to threads

Browse Source 
Related-To: NEO-4558

Signed-off-by: Mateusz Hoppe <mateusz.hoppe@intel.com>
This commit is contained in:
Mateusz Hoppe
2021-06-25 17:43:56 +02:00
committed by Compute-Runtime-Automation
parent 1d3d32cf39
commit 8526521f63
3 changed files with 148 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

33
level_zero/core/source/debugger/debugger_l0.cpp
View File

@@ -167,4 +167,37 @@ void DebuggerL0::getAttentionBitmaskForThread(uint32_t slice, uint32_t subslice,
}
}
std::vector<ze_device_thread_t> DebuggerL0::getThreadsFromAttentionBitmask(const NEO::HardwareInfo &hwInfo, const uint8_t *bitmask, const size_t bitmaskSize) {
const uint32_t numSubslicesPerSlice = hwInfo.gtSystemInfo.MaxSubSlicesSupported / hwInfo.gtSystemInfo.MaxSlicesSupported;
const uint32_t numEuPerSubslice = hwInfo.gtSystemInfo.MaxEuPerSubSlice;
const uint32_t numThreadsPerEu = (hwInfo.gtSystemInfo.ThreadCount / hwInfo.gtSystemInfo.EUCount);
const uint32_t bytesPerEu = alignUp(numThreadsPerEu, 8) / 8;
const uint32_t threadsSizePerSlice = numSubslicesPerSlice * numEuPerSubslice * bytesPerEu;
const uint32_t threadsSizePerSubSlice = numEuPerSubslice * bytesPerEu;
UNRECOVERABLE_IF(bytesPerEu != 1);
std::vector<ze_device_thread_t> threads;
for (uint32_t slice = 0; slice < hwInfo.gtSystemInfo.MaxSlicesSupported; slice++) {
for (uint32_t subslice = 0; subslice < numSubslicesPerSlice; subslice++) {
for (uint32_t eu = 0; eu < hwInfo.gtSystemInfo.MaxEuPerSubSlice; eu++) {
size_t offset = slice * threadsSizePerSlice + subslice * threadsSizePerSubSlice + eu * bytesPerEu;
if (offset >= bitmaskSize) {
return threads;
}
std::bitset<8> bits(bitmask[offset]);
for (uint32_t i = 0; i < 8; i++) {
if (bits.test(i)) {
threads.emplace_back(ze_device_thread_t{slice, subslice, eu, i});
}
}
}
}
}
return threads;
}
} // namespace L0

3
level_zero/core/source/debugger/debugger_l0.h
View File

@@ -10,6 +10,8 @@
#include "shared/source/helpers/non_copyable_or_moveable.h"
#include "shared/source/memory_manager/memory_manager.h"
#include <level_zero/ze_api.h>
#include <memory>
#include <unordered_map>
@@ -89,6 +91,7 @@ class DebuggerL0 : public NEO::Debugger, NEO::NonCopyableOrMovableClass {
virtual void programSbaTrackingCommands(NEO::LinearStream &cmdStream, const SbaAddresses &sba) = 0;
static void getAttentionBitmaskForThread(uint32_t slice, uint32_t subslice, uint32_t eu, uint32_t thread, const NEO::HardwareInfo &hwInfo, std::unique_ptr<uint8_t[]> &bitmask, size_t &bitmaskSize);
static std::vector<ze_device_thread_t> getThreadsFromAttentionBitmask(const NEO::HardwareInfo &hwInfo, const uint8_t *bitmask, const size_t bitmaskSize);
protected:
static bool isAnyTrackedAddressChanged(SbaAddresses sba) {