# Debugging through kernel code

**URL:** <https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762>\
**Category:** CUDA Programming and Performance\
**Created:** [October 14, 2009, 10:23am UTC](https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762 "2009-10-14T10:23:01Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Snowball\_Two](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@Snowball\_Two](https://forums.developer.nvidia.com/u/Snowball_Two)\
**Post date:** [October 14, 2009, 10:23am UTC](https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762/1 "2009-10-14T10:23:01Z")

</div>

hi :)

in the last days i wrote some cuda programs, and i am on a point, on which i just dont know how to get forward

i cant watch over the values of variables which are in device memory (all values seem to be “0” in the debugger), but in reality they are initialized correctly (or not…)  
i cant go through kernel code step by step, to see what actually happens. All i can do is guessing whats wrong, if it doesnt work. (i can barely read assembly code…but don’t wanna…)

at the point of switching inside the kernel, i get a promt to search tmpxft\_000013b4\_00000000-3\_gwbearb\_cuda.cudafe1.stub.c in a directory of my choice…

is there any proper way to watch over device memory and the steps inside the kernel with visual studio 2008?

getting more mad every minute ;(

---

<div class="post-metadata">

**Author:** ![\_teju](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@\_teju](https://forums.developer.nvidia.com/u/_teju)\
**Post date:** [October 14, 2009, 10:35am UTC](https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762/2 "2009-10-14T10:35:58Z")

</div>

Probably, this thread will be of help to you: [url=“[The Official NVIDIA Forums | NVIDIA](http://forums.nvidia.com/index.php?showtopic=108937)”][http://forums.nvidia.com/index.php?showtopic=108937[/url]](http://forums.nvidia.com/index.php?showtopic=108937%5B/url%5D)

---

<div class="post-metadata">

**Author:** ![Snowball\_Two](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@Snowball\_Two](https://forums.developer.nvidia.com/u/Snowball_Two)\
**Post date:** [October 14, 2009, 10:57am UTC](https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762/3 "2009-10-14T10:57:59Z")

</div>

> [@](#):
>
> Probably, this thread will be of help to you: http://forums.nvidia.com/index.php?showtopic=108937

works only for a very small amount of types (int, char ect.)

i’m currently working on a bigger structure which contains mainly floats. :(

---

<div class="post-metadata">

**Author:** ![\_teju](https://developer.download.nvidia.com/images/forums/profile-default-devtalk-84.png) [@\_teju](https://forums.developer.nvidia.com/u/_teju)\
**Post date:** [October 15, 2009, 4:14pm UTC](https://forums.developer.nvidia.com/t/debugging-through-kernel-code/12762/4 "2009-10-15T16:14:20Z")

</div>

Ok… I see the point here… Probably, it’s better to ask the author of this simple logging mechanism on extending the support for floats as well…