/docs/MyDocs

To get this branch, use:
bzr branch http://darksoft.org/webbzr/docs/MyDocs

« back to all changes in this revision

Viewing changes to Development/debugging/benchmarking/cuda.txt

  • Committer: Suren A. Chilingaryan
  • Date: 2015-08-21 03:52:00 UTC
  • Revision ID: csa@suren.me-20150821035200-xu1zh22cqlk2omcq
Profiling

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
Block all calls:
2
 
    CUDA_LAUNCH_BLOCKING=1 ./cudaapp
3
 
 
4
 
Select GPUs to use:
5
 
    CUDA_VISIBLE_DEVICES="0,2"
6
 
 
7
 
Binding GPU to CPU cores:
8
 
    numactl --cpunodebind=
9
 
 
10
 
GPU Direct:
11
 
    To force cudaHostAlloc to allocate memory compatible with GPUDirect (Infiniband) under cuda 4.0. 
12
 
    With 4.1 it should work automatically: 
13
 
        CUDA_NIC_INTEROP=1
14
 
    GPUDirect is not working over IOH   
15
 
 
16
 
Compile to PTX:
17
 
    nvcc --ptx hst_cuda.cu -I .. -I /opt/cuda/sdk/C/common/inc/ -I /usr/include/glib-2.0/ -I /usr/lib/glib-2.0/include/
18