nvprof - TX2 CPU Profiling Support

Hi All,

I am currently trying to profile an application on my TX2, but I am having problems with CPU profiling. Specifically, I run the following command:

nvprof --analysis-metrics --cpu-profiling on --cpu-thread-tracing on -o profile.nvvp ./program

Which gives the following warnings:

======== Warning: Metric "dram_write_throughput" cannot be found on device 0.
======== Warning: Metric "dram_read_throughput" cannot be found on device 0.
======== Warning: Metric "l1_shared_utilization" cannot be found on device 0.
======== Warning: Metric "sysmem_utilization" cannot be found on device 0.
======== Warning: Metric "dram_write_transactions" cannot be found on device 0.
======== Warning: Metric "dram_read_transactions" cannot be found on device 0.
======== Warning: Metric "alu_fu_utilization" cannot be found on device 0.
======== Warning: Metric "ecc_throughput" cannot be found on device 0.
======== Warning: Metric "ecc_transactions" cannot be found on device 0.
======== Warning: Metric "l2_l1_read_transactions" cannot be found on device 0.
======== Warning: Metric "l2_l1_write_transactions" cannot be found on device 0.
======== Warning: Metric "nc_l2_read_transactions" cannot be found on device 0.
======== Warning: Metric "l2_l1_read_throughput" cannot be found on device 0.
======== Warning: Metric "l2_l1_write_throughput" cannot be found on device 0.
======== Warning: Metric "nc_l2_read_throughput" cannot be found on device 0.
======== Warning: Metric "atomic_throughput" cannot be found on device 0.
======== Warning: CPU Profiling is not supported on the underlying platform.

The program is able to execute in its entirety, and I am able to load the nvvp file in the Nvidia Visual Profiler, but I do not see any information given in the CPU Details section of the tool. This result would make sense if CPU profiling truly isn’t supported on the TX2 as the last warning suggests.

With that being said, my question is this: Is CPU Profiling really not supported on the TX2, or is there a problem with my setup that is preventing me from profiling the CPU?

Setup Details (deviceQuery output)

CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "NVIDIA Tegra X2"
  CUDA Driver Version / Runtime Version          8.0 / 8.0
  CUDA Capability Major/Minor version number:    6.2
  Total amount of global memory:                 7851 MBytes (8232062976 bytes)
  ( 2) Multiprocessors, (128) CUDA Cores/MP:     256 CUDA Cores
  GPU Max Clock rate:                            1301 MHz (1.30 GHz)
  Memory Clock rate:                             1600 Mhz
  Memory Bus Width:                              128-bit
  L2 Cache Size:                                 524288 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total number of registers available per block: 32768
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  2048
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 1 copy engine(s)
  Run time limit on kernels:                     No
  Integrated GPU sharing Host Memory:            Yes
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 0 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 8.0, CUDA Runtime Version = 8.0, NumDevs = 1, Device0 = NVIDIA Tegra X2
Result = PASS

Hi, angelowhvs6

CPU Profiling is not supported on Jetson TX2 just as the warning said.

Hey veraj,

Okay, is there some documentation that mentions this? Just from looking at the home page for the NVIDIA System Profiler (https://developer.nvidia.com/embedded/nvidia-system-profiler) it seems like there should be CPU profiling support. In fact, the video on that page shows that the CPU was profiled in order to obtain activity for each core.

Hi, angelowhvs6

NVIDIA system is another tool, and its function not supported in Visual Profiler.
I have no idea about the tool, you can ask in that section.

Well in this case NVIDIA Visual Profiler is just the tool that I will use to analyze the results that nvprof produces.

With that said, do you have some documentation that says that CPU profiling isn’t supported on the TX2? I would expect such a limitation to appear in this section of the documentation detailing CPU Sampling Limitations:

https://docs.nvidia.com/cuda/archive/8.0/profiler-users-guide/index.html#cpu-sampling-limitations

Hi, sangelowhvs6

Sorry that the info is not described in document and only show as warning

======== Warning: CPU Profiling is not supported on the underlying platform.

Will let developer know about this. Thanks for your suggestion !