"components/vscode:/vscode.git/clone" did not exist on "97390ac0fb5d560a8a06b1b3005af697d5695af3"
Unverified Commit 18be11fd authored by xjx's avatar xjx Committed by GitHub
Browse files

[BUGFIX]fix CUDA OOM ERROR : invalid argument at cumem_allocator.cpp:119 (#35594)


Signed-off-by: default avatarxjx <493337577@qq.com>
parent 8d8855fd
......@@ -109,16 +109,18 @@ void create_and_map(unsigned long long device, ssize_t size, CUdeviceptr d_mem,
#ifndef USE_ROCM
int flag = 0;
CUDA_CHECK(cuDeviceGetAttribute(
CUresult rdma_result = cuDeviceGetAttribute(
&flag, CU_DEVICE_ATTRIBUTE_GPU_DIRECT_RDMA_WITH_CUDA_VMM_SUPPORTED,
device));
if (flag) { // support GPUDirect RDMA if possible
device);
if (rdma_result == CUDA_SUCCESS &&
flag) { // support GPUDirect RDMA if possible
prop.allocFlags.gpuDirectRDMACapable = 1;
}
int fab_flag = 0;
CUDA_CHECK(cuDeviceGetAttribute(
&fab_flag, CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_FABRIC_SUPPORTED, device));
if (fab_flag) { // support fabric handle if possible
CUresult fab_result = cuDeviceGetAttribute(
&fab_flag, CU_DEVICE_ATTRIBUTE_HANDLE_TYPE_FABRIC_SUPPORTED, device);
if (fab_result == CUDA_SUCCESS &&
fab_flag) { // support fabric handle if possible
prop.requestedHandleTypes = CU_MEM_HANDLE_TYPE_FABRIC;
}
#endif
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment