intel/compute-runtime
1
0
Fork 0
mirror of https://github.com/intel/compute-runtime.git synced 2025-12-24 21:18:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

performance(ocl): program barrier pc in taskStream

Browse Source 
Program barrier to task stream, before next enqueue kernel.
This will reduce the number of batch buffer starts for sequences of
enqueue, barrier, enqueue, ... .

Related-To: NEO-8147

Signed-off-by: Dominik Dabek <dominik.dabek@intel.com>
This commit is contained in:
Dominik Dabek
2023-09-12 14:17:52 +00:00
committed by Compute-Runtime-Automation
parent e08d46085b
commit 1b7e178b25
23 changed files with 224 additions and 98 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
opencl/test/unit_test/mem_obj/buffer_bcs_tests.cpp
View File

@@ -811,7 +811,7 @@ HWTEST_TEMPLATED_F(BcsBufferTests, givenPipeControlRequestWhenDispatchingBlitEnq
}
}
HWTEST_TEMPLATED_F(BcsBufferTests, givenBarrierWithEmptyWaitlistWhenReleasingMultipleBlockedEnqueuesThenProgramBarrierOnce) {
HWTEST_TEMPLATED_F(BcsBufferTests, givenStallingCommandsOnNextFlushWhenReleasingMultipleBlockedEnqueuesThenProgramBarrierOnce) {
DebugManager.flags.OptimizeIoqBarriersHandling.set(0);
using PIPE_CONTROL = typename FamilyType::PIPE_CONTROL;
@@ -827,7 +827,7 @@ HWTEST_TEMPLATED_F(BcsBufferTests, givenBarrierWithEmptyWaitlistWhenReleasingMul
cl_event waitlist0[] = {&userEvent0};
cl_event waitlist1[] = {&userEvent1};
cmdQ->enqueueBarrierWithWaitList(0, nullptr, nullptr);
cmdQ->setStallingCommandsOnNextFlush(true);
cmdQ->enqueueWriteBuffer(buffer.get(), false, 0, 1, hostPtr, nullptr, 1, waitlist0, nullptr);
cmdQ->enqueueWriteBuffer(buffer.get(), false, 0, 1, hostPtr, nullptr, 1, waitlist1, nullptr);