# Debugging in CUDAnative

**URL:** <https://discourse.julialang.org/t/debugging-in-cudanative/34193>\
**Category:** GPU\
**Tags:** cudanative\
**Created:** [February 5, 2020, 12:13am UTC](https://discourse.julialang.org/t/debugging-in-cudanative/34193 "2020-02-05T00:13:26Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![Michael\_Benton](https://avatars.discourse-cdn.com/v4/letter/m/e68b1a/32.png) [@Michael\_Benton](https://discourse.julialang.org/u/Michael_Benton)\
**Post date:** [February 5, 2020, 12:13am UTC](https://discourse.julialang.org/t/debugging-in-cudanative/34193/1 "2020-02-05T00:13:27Z")

</div>

struggling to find the source of problems in some kernels I write due to I don’t know of a julia CUDA debugger, I’ve found some for C.

I get:  
CUDA error: unspecified launch failure (code 719, ERROR\_LAUNCH\_FAILED)  
and that means I have to restart Julia, which makes debugging blind exceptionally slow

---

<div class="post-metadata">

**Author:** ![maleadt](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/maleadt/32/10097_2.png) [@maleadt](https://discourse.julialang.org/u/maleadt)\
**Post date:** [February 5, 2020, 7:01am UTC](https://discourse.julialang.org/t/debugging-in-cudanative/34193/2 "2020-02-05T07:01:51Z")

</div>

Yes, CUDA is very unfriendly wrt. invalid parameters and throwing its hands up in the air. Not much we can do about this, and there are no debuggers that make CUDA report errors in a better way. Sometimes running under `cuda-memcheck` gives slightly better errors. Best suggestion is to modify the sources of CUDAdrv and surround the API call arguments (here `cuLaunchKernel`) with `@show`. Then see if they are valid, comparing against the CUDA documentation. If you run into this error with high-level functionality (i.e. CuArrays.jl) you can also file a bug, as this behavior is unexpected.