AI/rocm-systems
2
0
Vork 0
Code Kwesties Pull-aanvragen Actions Paketten Projecten Publicaties Wiki Activiteit

P4 to Git Change 1988234 by kjayapra@3_HIPWS_TXT_ROCM on 2019/08/23 12:12:45

Bladeren bron 
SWDEV-201068 - Reverting back CL 1987582.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/hip/hip_module.cpp#37 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.cpp#84 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rocvirtual.hpp#27 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.cpp#98 edit
... //depot/stg/opencl/drivers/opencl/runtime/platform/command.hpp#96 edit
This commit is contained in:
foreman
2019-08-23 12:15:29 -04:00
bovenliggende a3d3fba77b
commit f684278234
4 gewijzigde bestanden met toevoegingen van 9 en 14 verwijderingen
Download Patch-bestand Download Diff-bestand Expand all files Collapse all files
rocclr/runtime/device/rocm/rocvirtual.cpp
+5 -4
Bestand weergeven
@@ -575,7 +575,7 @@ VirtualGPU::VirtualGPU(Device& device)
kernarg_pool_base_ = nullptr;
kernarg_pool_size_ = 0;
kernarg_pool_cur_offset_ = 0;
aqlHeader_ = kDispatchPacketHeader;
aqlHeader_ = kDispatchPacketHeaderNoSync;
barrier_signal_.handle = 0;
// Note: Virtual GPU device creation must be a thread safe operation
@@ -1980,13 +1980,13 @@ bool VirtualGPU::createVirtualQueue(uint deviceQueueSize)
}
bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const amd::Kernel& kernel,
const_address parameters, void* eventHandle, uint32_t sharedMemBytes, bool cooperativeGroups, bool isSvm) {
const_address parameters, void* eventHandle, uint32_t sharedMemBytes, bool cooperativeGroups) {
device::Kernel* devKernel = const_cast<device::Kernel*>(kernel.getDeviceKernel(dev()));
Kernel& gpuKernel = static_cast<Kernel&>(*devKernel);
size_t ldsUsage = gpuKernel.WorkgroupGroupSegmentByteSize();
// Check memory dependency and SVM objects
if (!isSvm && !processMemObjects(kernel, parameters, ldsUsage, cooperativeGroups)) {
if (!processMemObjects(kernel, parameters, ldsUsage, cooperativeGroups)) {
LogError("Wrong memory objects!");
return false;
}
@@ -2112,6 +2112,7 @@ bool VirtualGPU::submitKernelInternal(const amd::NDRangeContainer& sizes, const
AmdAqlWrap* wrap = reinterpret_cast<AmdAqlWrap*>(reinterpret_cast<uint64_t>(schedulerParam_->getHostMem()) + sizeof(SchedulerParam));
memset(wrap, 0, sizeof(AmdAqlWrap));
wrap->state = AQL_WRAP_DONE;
spVA = reinterpret_cast<uint64_t>(schedulerMem->getDeviceMemory()) + sizeof(SchedulerParam);
}
WriteAqlArgAt(const_cast<address>(parameters), &spVA, it.size_, it.offset_);
@@ -2223,7 +2224,7 @@ void VirtualGPU::submitKernel(amd::NDRangeKernelCommand& vcmd) {
// Submit kernel to HW
if (!queue->submitKernelInternal(vcmd.sizes(), vcmd.kernel(), vcmd.parameters(),
static_cast<void*>(as_cl(&vcmd.event())), vcmd.sharedMemBytes(), vcmd.cooperativeGroups()), vcmd.isSvm()) {
static_cast<void*>(as_cl(&vcmd.event())), vcmd.sharedMemBytes(), vcmd.cooperativeGroups())) {
LogError("AQL dispatch failed!");
vcmd.setStatus(CL_INVALID_OPERATION);
}