Printing numpy array contents in gdb

kelseyforg · June 17, 2024, 9:01pm

Is there any way to pretty print the contents of numpy arrays whist debugging a numba function using gdb?

I’m attempting to debug a numba function using gdb_init and gdb_breakpoint. My function operates over numpy arrays. As I step through my code, I would like to p(rint) arrays using the gdb console and receive meaningful output.

The output I see now looks like this

(gdb) p current_pair
$1 = {meminfo = 0x5572f1225c40 “\001”, parent = 0x0, nitems = 1, itemsize = 24, data = 0x5572f1225c80, shape = {1}, strides = {24}}

stuartarchibald · June 27, 2024, 11:40am

Hi @kelseyforg,

This is supported if your gdb installation has Python support and the NumPy package is available to that Python. This section of the docs explains how:

github.com

numba/numba/blob/ed19445d09694bdb82ee00e112c027a1685ee546/docs/source/user/troubleshoot.rst?plain=1#L265-L290


      
          Further, the Numba ``gdb`` printing extension can be loaded into ``gdb`` (if the
          ``gdb`` has Python support) to permit the printing of variables as they would be
          in native Python. The extension does this by reinterpreting Numba's memory model
          representations as Python types. Information about the ``gdb`` installation that
          Numba is using, including the path to load the ``gdb`` printing extension, can
          be displayed by using the ``numba -g`` command. For best results ensure that the
          Python that ``gdb`` is using has a NumPy module accessible. An example output
          of the ``gdb`` information follows:
          
          .. code-block:: none
            :emphasize-lines: 1
          
              $ numba -g
              GDB info:
              --------------------------------------------------------------------------------
              Binary location                               : <some path>/gdb
              Print extension location                      : <some python path>/numba/misc/gdb_print_extension.py
              Python version                                : 3.8
              NumPy version                                 : 1.20.0
              Numba printing extension supported            : True

This file has been truncated. show original

Hope this helps?

Topic		Replies	Views
Debug CUDA kernel printing pointers Community Support	2	1309	March 8, 2021
Finite fields (Galois fields) on the GPU using Numba Support: How do I do ...?	1	610	September 9, 2022
Debugger and JIT generated code on CPU Support: How do I do ...?	2	385	August 3, 2020
Inspecting @guvectorize'd Support: How do I do ...?	1	414	March 6, 2023
TypeError: can't unbox array from PyObject into native value Community Support	1	2145	March 3, 2021

Printing numpy array contents in gdb

Related topics