AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-339296 - Delay hidden heap allocation till the usage

Browse Source 
Move hidden heap creation to the kernel launch to make sure it's
allocated on the actual first usage.

Change-Id: I1b65a82fc06d9129ed45a69765bf14ea3d945b04
This commit is contained in:
German Andryeyev
2022-06-14 12:18:34 -04:00
parent ecea224bcf
commit 4975f69337
7 changed files with 63 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files
rocclr/device/rocm/rocvirtual.cpp
+8 -2
View File
@@ -2679,8 +2679,10 @@ bool VirtualGPU::createVirtualQueue(uint deviceQueueSize)
return true;
}
bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const amd::Kernel& kernel,
const_address parameters, void* eventHandle, uint32_t sharedMemBytes, amd::NDRangeKernelCommand* vcmd) {
// ================================================================================================
bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes,
const amd::Kernel& kernel, const_address parameters, void* eventHandle,
uint32_t sharedMemBytes, amd::NDRangeKernelCommand* vcmd) {
device::Kernel* devKernel = const_cast<device::Kernel*>(kernel.getDeviceKernel(dev()));
Kernel& gpuKernel = static_cast<Kernel&>(*devKernel);
size_t ldsUsage = gpuKernel.WorkgroupGroupSegmentByteSize();
@@ -2845,6 +2847,10 @@ bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const
break;
}
case amd::KernelParameterDescriptor::HiddenHeap:
// Allocate hidden heap for HIP applications only
if ((amd::IS_HIP) && (dev().HeapBuffer() == nullptr)) {
const_cast<Device&>(dev()).HiddenHeapAlloc();
}
if (dev().HeapBuffer() != nullptr) {
// Add heap pointer to the code
size_t heap_ptr = static_cast<size_t>(dev().HeapBuffer()->virtualAddress());