msDebug (Operator Debugging)

msDebug can debug all Ascend operators. You can use different functions as required. For example, you can set breakpoints, print variables and memory, perform single-step debugging, stop running, and switch cores.

1. Install the NPU driver and firmware. For details, see Overview.
2. Run the following command in the ${git_clone_path}/samples/operator/ascendc/0_introduction/1_add_frameworklaunch directory to generate a custom operator project and implement the operator on the host and kernel:

   bash install.sh -v Ascendxxxyy    # xxxyy indicates the processor type used by the user.

3. In the CMakePresets.json file under the ${git_clone_path}/samples/operator/ascendc/0_introduction/1_add_frameworklaunch/CustomOp directory, set the cacheVariables configuration item from Release to Debug.

   "cacheVariables": {
       "CMAKE_BUILD_TYPE": {
         "type": "STRING",
         "value": "Debug"
       },
       ...
     }

4. Run the following commands to build and deploy the operator again:
   1. Run the following command in the ${git_clone_path}/samples/operator/ascendc/0_introduction/1_add_frameworklaunch/CustomOp directory to build and deploy the operator again:

      bash build.sh
      ./build_out/custom_opp_<target_os>_<target_architecture>.run   // Name of the .run package in the current directory

   2. Switch to the ${git_clone_path}/samples/operator/ascendc/0_introduction/1_add_frameworklaunch/AclNNInvocation directory and run the following command to generate the executable file execute_add_op in the ./output directory:

      bash run.sh
      cd  ./output

5. Before debugging, configure the following environment variables, specify the operator loading path, and import the debugging information. An example is provided as follows:

   export LAUNCH_KERNEL_PATH=${INSTALL_DIR}/opp/vendors/customize/op_impl/ai_core/tbe/kernel/${soc_version}/add_custom/AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b.o   //soc_version is the chip name of the Ascend AI Processor and must be in lowercase.

6. Specify the path of the dynamic library on which the operator depends and load the .so file of the dynamic library.

   export LD_LIBRARY_PATH=${INSTALL_DIR}/opp/vendors/customize/op_api/lib:$LD_LIBRARY_PATH

7. Run the msdebug execute_add_op command in the executable file directory to access msDebug.

   msdebug execute_add_op

8. Set a breakpoint.
   1. Set a breakpoint.

      (msdebug) b add_custom.cpp:55

   2. The following command output indicates that the breakpoint is successfully added:

      Breakpoint 1: where = AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b.o`KernelAdd::Compute(int) (.vector) + 68 at add_custom.cpp:55:9, address = 0x00000000000014f4

9. Enter the r command to run the operator program and wait until the breakpoint is hit.

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

   (msdebug) r Process 1454802 launched: '${INSTALL_DIR}/add_cus/AclNNInvocation/output/execute_add_op' (aarch64) [INFO] Set device[0] success [INFO] Get RunMode[1] success [INFO] Init resource success [INFO] Set input success [INFO] Copy input[0] success [INFO] Copy input[1] success [INFO] Create stream success [INFO] Execute aclnnAddCustomGetWorkspaceSize success, workspace size 0 [Launch of Kernel AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b on Device 0] [INFO] Execute aclnnAddCustom success Process 1454802 stopped [Switching to focus on Kernel AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b, CoreId 39, Type aiv] * thread #1, name = 'execute_add_op', stop reason = breakpoint 1.1 frame #0: 0x00000000000014f4 AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b.o`KernelAdd::Compute(this=0x00000000003078a8, progress=0) (.vector) at add_custom.cpp:55:9 52 __aicore__ inline void Compute(int32_t progress) 53 { 54 LocalTensor<DTYPE_X> xLocal = inQueueX.DeQue<DTYPE_X>(); -> 55 LocalTensor<DTYPE_Y> yLocal = inQueueY.DeQue<DTYPE_Y>(); //The line number at the breakpoint is correct. Other information is subject to the actual situation. 56 LocalTensor<DTYPE_Z> zLocal = outQueueZ.AllocTensor<DTYPE_Z>(); 57 Add(zLocal, xLocal, yLocal, this->tileLength); 58 outQueueZ.EnQue<DTYPE_Z>(zLocal);

10. Keep running.
    1. Enter the following command to continue the running:

       (msdebug) c

    2. The following shows that the program hits the breakpoint again:

       1 2 3 4 5 6 7 8 9 10 11 12

       Process 1454802 resuming Process 1454802 stopped [Switching to focus on Kernel AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b, CoreId 39, Type aiv] * thread #1, name = 'execute_add_op', stop reason = breakpoint 1.1 frame #0: 0x00000000000014f4 AddCustom_1e04ee05ab491cc5ae9c3d5c9ee8950b.o`KernelAdd::Compute(this=0x00000000003078a8, progress=0) (.vector) at add_custom.cpp:55:9 52 __aicore__ inline void Compute(int32_t progress) 53 { 54 LocalTensor<DTYPE_X> xLocal = inQueueX.DeQue<DTYPE_X>(); -> 55 LocalTensor<DTYPE_Y> yLocal = inQueueY.DeQue<DTYPE_Y>(); //The line number at the breakpoint is correct. Other information is subject to the actual situation. 56 LocalTensor<DTYPE_Z> zLocal = outQueueZ.AllocTensor<DTYPE_Z>(); 57 Add(zLocal, xLocal, yLocal, this->tileLength); 58 outQueueZ.EnQue<DTYPE_Z>(zLocal);

11. Stop debugging.

    (msdebug) q