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

Report owning agent with pointer info block information.

Browse Source 
Physical owning agent may not be visible to the current process
due to RVD.

Change-Id: Ib463336a5ed73a479f3aa74eb140932b9e0435fb


[ROCm/ROCR-Runtime commit: 247606c455]
This commit is contained in:
Sean Keely
2022-04-29 15:38:56 -05:00
parent 289a86785b
commit b757b209ad
2 changed files with 14 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocr-runtime/runtime/hsa-runtime/core/inc/runtime.h
+1
View File
@@ -281,6 +281,7 @@ class Runtime {
struct PtrInfoBlockData { struct PtrInfoBlockData {
void* base; void* base;
size_t length; size_t length;
core::Agent* agentOwner;
}; };
hsa_status_t PtrInfo(const void* ptr, hsa_amd_pointer_info_t* info, void* (*alloc)(size_t), hsa_status_t PtrInfo(const void* ptr, hsa_amd_pointer_info_t* info, void* (*alloc)(size_t),
projects/rocr-runtime/runtime/hsa-runtime/core/runtime/runtime.cpp
+13 -4
View File
@@ -803,6 +803,11 @@ hsa_status_t Runtime::PtrInfo(const void* ptr, hsa_amd_pointer_info_t* info, voi
assert((retInfo.hostBaseAddress || retInfo.agentBaseAddress) && "Thunk pointer info returned no base address."); assert((retInfo.hostBaseAddress || retInfo.agentBaseAddress) && "Thunk pointer info returned no base address.");
block_info->base = (retInfo.hostBaseAddress ? retInfo.hostBaseAddress : retInfo.agentBaseAddress); block_info->base = (retInfo.hostBaseAddress ? retInfo.hostBaseAddress : retInfo.agentBaseAddress);
block_info->length = retInfo.sizeInBytes; block_info->length = retInfo.sizeInBytes;
// Report the owning agent, even if such an agent is not usable in the process.
auto nodeAgents = agents_by_node_.find(thunkInfo.Node);
assert(nodeAgents != agents_by_node_.end() && "Node id not found!");
block_info->agentOwner = nodeAgents->second[0];
} }
auto fragment = allocation_map_.upper_bound(ptr); auto fragment = allocation_map_.upper_bound(ptr);
if (fragment != allocation_map_.begin()) { if (fragment != allocation_map_.begin()) {
@@ -831,11 +836,15 @@ hsa_status_t Runtime::PtrInfo(const void* ptr, hsa_amd_pointer_info_t* info, voi
// IPC and Graphics memory may come from a node that does not have an agent in this process. // IPC and Graphics memory may come from a node that does not have an agent in this process.
// Ex. ROCR_VISIBLE_DEVICES or peer GPU is not supported by ROCm. // Ex. ROCR_VISIBLE_DEVICES or peer GPU is not supported by ROCm.
retInfo.agentOwner.handle = 0;
auto nodeAgents = agents_by_node_.find(thunkInfo.Node); auto nodeAgents = agents_by_node_.find(thunkInfo.Node);
if (nodeAgents != agents_by_node_.end()) assert(nodeAgents != agents_by_node_.end() && "Node id not found!");
retInfo.agentOwner = nodeAgents->second[0]->public_handle(); for (auto agent : nodeAgents->second) {
else if (agent->Enabled()) {
retInfo.agentOwner.handle = 0; retInfo.agentOwner = agent->public_handle();
break;
}
}
// Correct agentOwner for locked memory. Thunk reports the GPU that owns the // Correct agentOwner for locked memory. Thunk reports the GPU that owns the
// alias but users are expecting to see a CPU when the memory is system. // alias but users are expecting to see a CPU when the memory is system.