Does enable_language(CUDA) still have the problem with reconfiguring?

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

Does enable_language(CUDA) still have the problem with reconfiguring?

JR Cary
Re the thread below,

Does the newer, enable_languagedo a better job of not having the
repeated reconfigurations and full project rebuilds as described at

https://cmake.org/pipermail/cmake/2011-January/042173.html

and many times after?

John



On 3/30/19 6:45 AM, [hidden email] wrote:

> -----------------------------
> Message: 4
> Date: Sat, 30 Mar 2019 12:45:07 +0000
> From: Luis Caro Campos <[hidden email]>
>
> I believe the closest equivalent to settinga CUDA_TOOLKIT_ROOT_DIR when
> using enable_language(CUDA) is to set either the environment variable
> CUDACXX or the CMake cache entry CMAKE_CUDA_COMPILER to point to the
> location of the nvcc compiler.
>
> However I think there are still some "features" of FindCUDA that
> enable_language don't support, e.g. variables and/or import targets with
> locations, see https://gitlab.kitware.com/cmake/cmake/issues/17816
>
> Kind regards,
> Luis
>
> On Fri, Mar 29, 2019 at 4:57 PM Dustyn Blasig <[hidden email]> wrote:
>
>> "we should not try to combine enable_language(CUDA) with
>> find_package(CUDA). They do not work together, either use one or another."
>>
>> Absolutely, my goal is to move to the new built-in language support. I'm
>> having trouble doing that because I can't find any documentation on it. For
>> instance, what is the new equivalent CUDA_TOOLKIT_ROOT_DIR? Without
>> find_package(CUDA), I don't see anything set that have the equivalent.
>> Also, in some cases the CUDA include directory is added to targets, and in
>> other cases it isn't, even if the target depends on source with .cu. What
>> is the documented behavior for this?
>>
>> I'm sure there has to be a page somewhere on this, but going to page 4 on
>> Google search didn't uncover anything, and the first 2 pages all point to
>> various versions of FindCUDA : )
>>
>> On Fri, Mar 29, 2019 at 11:28 AM Dmitry Mikushin <
>> [hidden email]> wrote:
>>
>>> Hi,
>>>
>>> That was my confusion as well: to my understanding, we should not try to
>>> combine enable_language(CUDA) with find_package(CUDA). They do not work
>>> together, either use one or another.
>>>
>>> Kind regards,
>>> - Dmitry.
>>>
>>> ??, 29 ???. 2019 ?. ? 19:58, Dustyn Blasig <[hidden email]>:
>>>
>>>> Hi All,
>>>>
>>>> I can't find any documentation on the new-ish "native" CUDA support. I
>>>> need to know all the variables that we can use, and (for instance) whether
>>>> checking the CUDA version is now supported. When searching online, I'm
>>>> always directed to the old FindCUDA pages which don't seem to match what is
>>>> available with the native language support.
>>>>
>>>> Currently, I'm "hacking" my way through things by using both flows. I
>>>> add CUDA as a language, use the found CUDA compiler to construct a
>>>> CUDA_TOOLKIT_ROOT_DIR and then call find_package(CUDA) to get the old
>>>> environment variables set up.
>>>>
>>>> Cheers!
>>>>

--

Powered by www.kitware.com

Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ

Kitware offers various services to support the CMake community. For more information on each offering, please visit:

CMake Support: http://cmake.org/cmake/help/support.html
CMake Consulting: http://cmake.org/cmake/help/consulting.html
CMake Training Courses: http://cmake.org/cmake/help/training.html

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Follow this link to subscribe/unsubscribe:
https://cmake.org/mailman/listinfo/cmake
Reply | Threaded
Open this post in threaded view
|

Re: Does enable_language(CUDA) still have the problem with reconfiguring?

CMake mailing list
It doesn't have this behavior.

For Makefile and Ninja generators we get the proper header
dependencies at compile time just like we do for C or C++.
For MSBuild we rely on the CUDA extensions written by nvidia which do
proper dependency tracking.

On Sat, Mar 30, 2019 at 4:12 PM JR Cary <[hidden email]> wrote:

>
> Re the thread below,
>
> Does the newer, enable_languagedo a better job of not having the
> repeated reconfigurations and full project rebuilds as described at
>
> https://cmake.org/pipermail/cmake/2011-January/042173.html
>
> and many times after?
>
> John
>
>
>
> On 3/30/19 6:45 AM, [hidden email] wrote:
> > -----------------------------
> > Message: 4
> > Date: Sat, 30 Mar 2019 12:45:07 +0000
> > From: Luis Caro Campos <[hidden email]>
> >
> > I believe the closest equivalent to settinga CUDA_TOOLKIT_ROOT_DIR when
> > using enable_language(CUDA) is to set either the environment variable
> > CUDACXX or the CMake cache entry CMAKE_CUDA_COMPILER to point to the
> > location of the nvcc compiler.
> >
> > However I think there are still some "features" of FindCUDA that
> > enable_language don't support, e.g. variables and/or import targets with
> > locations, see https://gitlab.kitware.com/cmake/cmake/issues/17816
> >
> > Kind regards,
> > Luis
> >
> > On Fri, Mar 29, 2019 at 4:57 PM Dustyn Blasig <[hidden email]> wrote:
> >
> >> "we should not try to combine enable_language(CUDA) with
> >> find_package(CUDA). They do not work together, either use one or another."
> >>
> >> Absolutely, my goal is to move to the new built-in language support. I'm
> >> having trouble doing that because I can't find any documentation on it. For
> >> instance, what is the new equivalent CUDA_TOOLKIT_ROOT_DIR? Without
> >> find_package(CUDA), I don't see anything set that have the equivalent.
> >> Also, in some cases the CUDA include directory is added to targets, and in
> >> other cases it isn't, even if the target depends on source with .cu. What
> >> is the documented behavior for this?
> >>
> >> I'm sure there has to be a page somewhere on this, but going to page 4 on
> >> Google search didn't uncover anything, and the first 2 pages all point to
> >> various versions of FindCUDA : )
> >>
> >> On Fri, Mar 29, 2019 at 11:28 AM Dmitry Mikushin <
> >> [hidden email]> wrote:
> >>
> >>> Hi,
> >>>
> >>> That was my confusion as well: to my understanding, we should not try to
> >>> combine enable_language(CUDA) with find_package(CUDA). They do not work
> >>> together, either use one or another.
> >>>
> >>> Kind regards,
> >>> - Dmitry.
> >>>
> >>> ??, 29 ???. 2019 ?. ? 19:58, Dustyn Blasig <[hidden email]>:
> >>>
> >>>> Hi All,
> >>>>
> >>>> I can't find any documentation on the new-ish "native" CUDA support. I
> >>>> need to know all the variables that we can use, and (for instance) whether
> >>>> checking the CUDA version is now supported. When searching online, I'm
> >>>> always directed to the old FindCUDA pages which don't seem to match what is
> >>>> available with the native language support.
> >>>>
> >>>> Currently, I'm "hacking" my way through things by using both flows. I
> >>>> add CUDA as a language, use the found CUDA compiler to construct a
> >>>> CUDA_TOOLKIT_ROOT_DIR and then call find_package(CUDA) to get the old
> >>>> environment variables set up.
> >>>>
> >>>> Cheers!
> >>>>
>
> --
>
> Powered by www.kitware.com
>
> Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ
>
> Kitware offers various services to support the CMake community. For more information on each offering, please visit:
>
> CMake Support: http://cmake.org/cmake/help/support.html
> CMake Consulting: http://cmake.org/cmake/help/consulting.html
> CMake Training Courses: http://cmake.org/cmake/help/training.html
>
> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html
>
> Follow this link to subscribe/unsubscribe:
> https://cmake.org/mailman/listinfo/cmake
--

Powered by www.kitware.com

Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ

Kitware offers various services to support the CMake community. For more information on each offering, please visit:

CMake Support: http://cmake.org/cmake/help/support.html
CMake Consulting: http://cmake.org/cmake/help/consulting.html
CMake Training Courses: http://cmake.org/cmake/help/training.html

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Follow this link to subscribe/unsubscribe:
https://cmake.org/mailman/listinfo/cmake
Reply | Threaded
Open this post in threaded view
|

Re: Does enable_language(CUDA) still have the problem with reconfiguring?

Mojca Miklavec
On Mon, 1 Apr 2019 at 22:11, Robert Maynard via CMake wrote:
>
> For MSBuild we rely on the CUDA extensions written by nvidia which do
> proper dependency tracking.

Except when they don't. It drove me nearly crazy that any given
trivial change had to be followed by a complete rebuild of the
project.

But I see that someone commented an hour ago:
    https://stackoverflow.com/questions/48183845/visual-studio-2017-not-detecting-change-in-cu-cuda-files
saying that the fix has finally arrived (in March).

Mojca
--

Powered by www.kitware.com

Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ

Kitware offers various services to support the CMake community. For more information on each offering, please visit:

CMake Support: http://cmake.org/cmake/help/support.html
CMake Consulting: http://cmake.org/cmake/help/consulting.html
CMake Training Courses: http://cmake.org/cmake/help/training.html

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Follow this link to subscribe/unsubscribe:
https://cmake.org/mailman/listinfo/cmake
Reply | Threaded
Open this post in threaded view
|

Re: Does enable_language(CUDA) still have the problem with reconfiguring?

CMake mailing list
Yes, we are dependent on the CUDA extensions for ms-build working correctly.

On Mon, Apr 1, 2019 at 5:02 PM Mojca Miklavec
<[hidden email]> wrote:

>
> On Mon, 1 Apr 2019 at 22:11, Robert Maynard via CMake wrote:
> >
> > For MSBuild we rely on the CUDA extensions written by nvidia which do
> > proper dependency tracking.
>
> Except when they don't. It drove me nearly crazy that any given
> trivial change had to be followed by a complete rebuild of the
> project.
>
> But I see that someone commented an hour ago:
>     https://stackoverflow.com/questions/48183845/visual-studio-2017-not-detecting-change-in-cu-cuda-files
> saying that the fix has finally arrived (in March).
>
> Mojca
--

Powered by www.kitware.com

Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ

Kitware offers various services to support the CMake community. For more information on each offering, please visit:

CMake Support: http://cmake.org/cmake/help/support.html
CMake Consulting: http://cmake.org/cmake/help/consulting.html
CMake Training Courses: http://cmake.org/cmake/help/training.html

Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html

Follow this link to subscribe/unsubscribe:
https://cmake.org/mailman/listinfo/cmake