SWDEV-513197 - Improve launch perf for Device Heap kernels

- If any kernel uses device heap, the launch needs to be preceeded by an
  init kernel, Save on the extra barrier packet launch/flush between the
init heap kernel and user kernel

Change-Id: I8ebc6246188200e5f673dc464bc76a53bcb8b7c6


[ROCm/clr commit: ca530c660b]
This commit is contained in:
Saleel Kudchadker committed 2025-02-27 19:17:51 -05:00
1 parent efd1f3f012
commit ef505c7cd8
2 files changed
+6 -6

No files matched your search

+3 -2
View File
@@ -2537,9 +2537,10 @@ bool KernelBlitManager::initHeap(device::Memory* heap_to_initialize, device::Mem
setArgument(kernels_[blitType], 2, sizeof(uint), &heap_size);
setArgument(kernels_[blitType], 3, sizeof(uint), &number_of_initial_blocks);
address parameters = captureArguments(kernels_[blitType]);
result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters, nullptr);
result = gpu().submitKernelInternal(ndrange, *kernels_[blitType], parameters,
nullptr, 0, nullptr, nullptr, true);
releaseArguments(parameters);
gpu().releaseGpuMemoryFence();
gpu().Barriers().WaitCurrent();
return result;
}