P4 to Git Change 1766264 by cpaquot@cpaquot-ocl-lc-lnx on 2019/04/05 11:38:18

SWDEV-145570 - [HIP] Use a context with all devices in system for host register
	hipHostRegister and hipMemcpy 0x10 and 0x20 fail in mGPU systems because
	we only register the memory on the current device. But in HIP, the registering
	needs to happen on all devices.

Affected files ...

... //depot/stg/opencl/drivers/opencl/api/hip/hip_context.cpp#17 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_internal.hpp#26 edit
... //depot/stg/opencl/drivers/opencl/api/hip/hip_memory.cpp#50 edit


[ROCm/hip commit: 8753616c97]

projects/hip/api/hip/hip_context.cpp

@@ -34,6 +34,7 @@ thread_local amd::Context* g_context = nullptr;
 thread_local std::stack<amd::Context*> g_ctxtStack;
 thread_local hipError_t g_lastError = hipSuccess;
 std::once_flag g_ihipInitialized;
+amd::Context* host_context = nullptr;
 
 std::map<amd::Context*, amd::HostQueue*> g_nullStreams;
 
@@ -56,6 +57,13 @@ void init() {
       g_devices.push_back(context);
     }
   }
+
+  host_context = new amd::Context(devices, amd::Context::Info());
+  if (!host_context) return;
+
+  if (host_context && CL_SUCCESS != host_context->create(nullptr)) {
+    host_context->release();
+  }
 }
 
 amd::Context* getCurrentContext() {

projects/hip/api/hip/hip_internal.hpp

@@ -65,6 +65,7 @@ namespace hip {
   extern std::once_flag g_ihipInitialized;
   extern thread_local amd::Context* g_context;
   extern thread_local hipError_t g_lastError;
+  extern amd::Context* host_context;
 
   extern void init();
 

projects/hip/api/hip/hip_memory.cpp

@@ -62,11 +62,14 @@ hipError_t ihipMalloc(void** ptr, size_t sizeBytes, unsigned int flags)
     return hipErrorInvalidValue;
   }
 
-  if (hip::getCurrentContext()->devices()[0]->info().maxMemAllocSize_ < sizeBytes) {
+  amd::Context* amdContext = ((flags & CL_MEM_SVM_FINE_GRAIN_BUFFER) != 0)?
+    hip::host_context : hip::getCurrentContext();
+
+  if (amdContext->devices()[0]->info().maxMemAllocSize_ < sizeBytes) {
     return hipErrorMemoryAllocation;
   }
 
-  *ptr = amd::SvmBuffer::malloc(*hip::getCurrentContext(), flags, sizeBytes, hip::getCurrentContext()->devices()[0]->info().memBaseAddrAlign_);
+  *ptr = amd::SvmBuffer::malloc(*amdContext, flags, sizeBytes, amdContext->devices()[0]->info().memBaseAddrAlign_);
   if (*ptr == nullptr) {
     return hipErrorOutOfMemory;
   }
@@ -518,8 +521,7 @@ hipError_t hipHostGetFlags(unsigned int* flagsPtr, void* hostPtr) {
 hipError_t hipHostRegister(void* hostPtr, size_t sizeBytes, unsigned int flags) {
   HIP_INIT_API(hostPtr, sizeBytes, flags);
   if(hostPtr != nullptr) {
-    amd::Context *amdContext = hip::getCurrentContext();
-    amd::Memory* mem = new (*amdContext) amd::Buffer(*amdContext, CL_MEM_USE_HOST_PTR, sizeBytes);
+    amd::Memory* mem = new (*hip::host_context) amd::Buffer(*hip::host_context, CL_MEM_USE_HOST_PTR, sizeBytes);
 
     if (!mem->create(hostPtr)) {
       mem->release();
@@ -538,7 +540,7 @@ hipError_t hipHostUnregister(void* hostPtr) {
   if (amd::SvmBuffer::malloced(hostPtr)) {
     hip::syncStreams();
     hip::getNullStream()->finish();
-    amd::SvmBuffer::free(*hip::getCurrentContext(), hostPtr);
+    amd::SvmBuffer::free(*hip::host_context, hostPtr);
     HIP_RETURN(hipSuccess);
   } else {
     size_t offset = 0;