Media
/
compute-runtime
mirror of https://github.com/intel/compute-runtime.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

performance: Signal inOrder counter with pipe control, part 8 

Allow wait for in order allocation when latest submit does not have
optmized counter based event.

Resolves: HSD-18041399091
Related-To: NEO-13441

Signed-off-by: Lukasz Jobczyk <lukasz.jobczyk@intel.com>
This commit is contained in:
Lukasz Jobczyk 2025-03-03 10:54:04 +00:00 committed by Compute-Runtime-Automation
parent f4f779f5b2
commit c1a986c71c
2 changed files with 10 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
level_zero/core/source/cmdlist/cmdlist_hw_immediate.inl
View File

@ -1104,7 +1104,7 @@ ze_result_t CommandListCoreFamilyImmediate<gfxCoreFamily>::hostSynchronize(uint6
auto tempAllocsCleanupRequired = handlePostWaitOperations && (mainStorageCleanupNeeded || copyOffloadStorageCleanupNeeded);
bool inOrderWaitAllowed = (isInOrderExecutionEnabled() && !tempAllocsCleanupRequired && this->latestFlushIsHostVisible && this->heaplessModeEnabled);
bool inOrderWaitAllowed = (isInOrderExecutionEnabled() && !tempAllocsCleanupRequired && this->latestFlushIsHostVisible && (this->heaplessModeEnabled || !this->latestOperationHasOptimizedCbEvent));
uint64_t inOrderSyncValue = this->inOrderExecInfo.get() ? inOrderExecInfo->getCounterValue() : 0;

18
level_zero/core/test/unit_tests/sources/cmdlist/test_in_order_cmdlist_1.cpp
View File

@ -294,17 +294,17 @@ HWTEST2_F(InOrderCmdListTests, givenCmdListsWhenDispatchingThenUseInternalTaskCo
CmdListMemoryCopyParams copyParams = {};
immCmdList0->appendMemoryCopy(deviceAlloc, &hostCopyData, 1, nullptr, 0, nullptr, copyParams);
auto expectedLatestTaskCount = immCmdList0->dcFlushSupport || !heapless ? 1u : 2u;
auto expectedLatestTaskCount = immCmdList0->dcFlushSupport || (!heapless && immCmdList0->latestOperationHasOptimizedCbEvent) ? 1u : 2u;
expectedLatestTaskCount += (heapless ? 1u : 0u);
EXPECT_EQ(expectedLatestTaskCount, ultCsr->latestWaitForCompletionWithTimeoutTaskCount.load());
EXPECT_EQ(immCmdList0->dcFlushSupport || !heapless ? 3u : 2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled.load());
EXPECT_EQ(immCmdList0->dcFlushSupport || (!heapless && immCmdList0->latestOperationHasOptimizedCbEvent) ? 3u : 2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled.load());
immCmdList1->appendMemoryCopy(deviceAlloc, &hostCopyData, 1, nullptr, 0, nullptr, copyParams);
expectedLatestTaskCount = 2u;
expectedLatestTaskCount += (heapless ? 1u : 0u);
EXPECT_EQ(expectedLatestTaskCount, ultCsr->latestWaitForCompletionWithTimeoutTaskCount.load());
EXPECT_EQ(immCmdList0->dcFlushSupport || !heapless ? 4u : 2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled.load());
EXPECT_EQ(immCmdList0->dcFlushSupport || (!heapless && immCmdList0->latestOperationHasOptimizedCbEvent) ? 4u : 2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled.load());
context->freeMem(deviceAlloc);
}
@ -3148,7 +3148,7 @@ HWTEST2_F(InOrderCmdListTests, givenHostVisibleEventOnLatestFlushWhenCallingSync
immCmdList->hostSynchronize(0, false);
if (immCmdList->dcFlushSupport || !immCmdList->isHeaplessModeEnabled()) {
if (immCmdList->dcFlushSupport || (!immCmdList->isHeaplessModeEnabled() && immCmdList->latestOperationHasOptimizedCbEvent)) {
EXPECT_EQ(0u, immCmdList->synchronizeInOrderExecutionCalled);
EXPECT_EQ(1u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled);
} else {
@ -3162,7 +3162,7 @@ HWTEST2_F(InOrderCmdListTests, givenHostVisibleEventOnLatestFlushWhenCallingSync
immCmdList->hostSynchronize(0, false);
if (!immCmdList->isHeaplessModeEnabled()) {
if (!immCmdList->isHeaplessModeEnabled() && immCmdList->latestOperationHasOptimizedCbEvent) {
EXPECT_EQ(0u, immCmdList->synchronizeInOrderExecutionCalled);
EXPECT_EQ(2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled);
} else if (immCmdList->dcFlushSupport) {
@ -3176,7 +3176,7 @@ HWTEST2_F(InOrderCmdListTests, givenHostVisibleEventOnLatestFlushWhenCallingSync
// handle post sync operations
immCmdList->hostSynchronize(0, true);
if (!immCmdList->isHeaplessModeEnabled()) {
if (!immCmdList->isHeaplessModeEnabled() && immCmdList->latestOperationHasOptimizedCbEvent) {
EXPECT_EQ(0u, immCmdList->synchronizeInOrderExecutionCalled);
EXPECT_EQ(3u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled);
} else if (immCmdList->dcFlushSupport) {
@ -3208,7 +3208,7 @@ HWTEST2_F(InOrderCmdListTests, givenEmptyTempAllocationsStorageWhenCallingSynchr
immCmdList->hostSynchronize(0, true);
if (!immCmdList->isHeaplessModeEnabled()) {
if (!immCmdList->isHeaplessModeEnabled() && immCmdList->latestOperationHasOptimizedCbEvent) {
EXPECT_EQ(0u, immCmdList->synchronizeInOrderExecutionCalled);
EXPECT_EQ(1u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled);
} else {
@ -3220,7 +3220,7 @@ HWTEST2_F(InOrderCmdListTests, givenEmptyTempAllocationsStorageWhenCallingSynchr
immCmdList->hostSynchronize(0, true);
if (!immCmdList->isHeaplessModeEnabled()) {
if (!immCmdList->isHeaplessModeEnabled() && immCmdList->latestOperationHasOptimizedCbEvent) {
EXPECT_EQ(0u, immCmdList->synchronizeInOrderExecutionCalled);
EXPECT_EQ(2u, ultCsr->waitForCompletionWithTimeoutTaskCountCalled);
} else {
@ -5122,7 +5122,7 @@ HWTEST2_F(InOrderCmdListTests, givenAubModeWhenSyncCalledAlwaysPollForCompletion
immCmdList->hostSynchronize(0, false);
auto expectPollForCompletion = immCmdList->isHeaplessModeEnabled() ? 1u : 0u;
auto expectPollForCompletion = (immCmdList->isHeaplessModeEnabled() || !immCmdList->latestOperationHasOptimizedCbEvent) ? 1u : 0u;
EXPECT_EQ(expectPollForCompletion++, ultCsr->pollForAubCompletionCalled);
events[0]->hostSynchronize(std::numeric_limits<uint64_t>::max());