[CMake] CMake 3.9.x Visual Studio CUDA PDB/compiler flags issues

Thu Sep 7 09:48:45 EDT 2017

Okay I have opened an issue on CMake for handling multiple gencode with
multiple code targets.

On Wed, Sep 6, 2017 at 3:19 PM, Mueller-Roemer, Johannes Sebastian <
Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de> wrote:

> Yes
>
> ________________________________
> From: Robert Maynard [robert.maynard at kitware.com]
> Sent: Wednesday, September 06, 2017 6:58 PM
> To: Mueller-Roemer, Johannes Sebastian
> Cc: cmake at cmake.org
> Subject: Re: [CMake] CMake 3.9.x Visual Studio CUDA PDB/compiler flags
> issues
>
> So it works when they are separate gen-code options, but fails when they
> are combined. Is that correct?
>
>
> On Wed, Sep 6, 2017 at 11:16 AM, Mueller-Roemer, Johannes Sebastian <
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de<mailto:
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de>> wrote:
> The long form fills the Code Generation field with
> compute_20,sm_20;compute_20,compute_20 (the default without gencode
> parameters is compute_20,sm_20) and results in the following command line:
>
> D:\jsroemer\projects\cuda-pdb-test\build\a>"C:\Program Files\NVIDIA GPU
> Computing Toolkit\CUDA\v8.0\bin\nvcc.exe" -gencode=arch=compute_20,code=\"sm_20,compute_20\"
> -gencode=arch=compute_20,code=\"compute_20,compute_20\" --use-local-env
> --cl-version 2015 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio
> 14.0\VC\bin\amd64"  -I"C:\Program Files\NVIDIA GPU Computing
> Toolkit\CUDA\v8.0\include" -I"C:\Program Files\NVIDIA GPU Computing
> Toolkit\CUDA\v8.0\include"  -G   --keep-dir x64\Debug -maxrregcount=0
> --machine 64 --compile -cudart static -Xcompiler="/EHsc -Zi -Ob0" -g
>  -D_WINDOWS -Xcompiler "/EHsc /W3 /nologo /Od /FS /Zi /RTC1 /MDd /GR" -o
> A.dir\Debug\a.cu.obj "D:\jsroemer\projects\cuda-pdb-test\a\a.cu<
> http://a.cu>"
>
> Which is somewhat redundant
>
> MSVC turns compute_20,sm_20 into -gencode=arch=compute_20,code=\"sm_20,compute_20\"
> and compute_20,compute_20 into -gencode=arch=compute_20,code=
> \"compute_20,compute_20\"
>
>
> Fraunhofer-Institut für Graphische Datenverarbeitung IGD
> Fraunhoferstr. 5  |  64283 Darmstadt  |  Germany
> Tel +49 6151 155-606<tel:+49%206151%20155606>  |  Fax +49 6151 155-139
> <tel:+49%206151%20155139>
> johannes.mueller-roemer at igd.fraunhofer.de<mailto:johannes.
> mueller-roemer at igd.fraunhofer.de> | www.igd.fraunhofer.de<http://
> www.igd.fraunhofer.de>
>
> From: Robert Maynard [mailto:robert.maynard at kitware.com<mailto:robert.
> maynard at kitware.com>]
> Sent: Wednesday, September 6, 2017 16:44
>
> To: Mueller-Roemer, Johannes Sebastian <Johannes.Sebastian.Mueller-
> Roemer at igd.fraunhofer.de<mailto:Johannes.Sebastian.
> Mueller-Roemer at igd.fraunhofer.de>>
> Cc: cmake at cmake.org<mailto:cmake at cmake.org>
> Subject: Re: [CMake] CMake 3.9.x Visual Studio CUDA PDB/compiler flags
> issues
>
> Yes I was responding to comment 3.
>
> I am curious, does the longer form signature "-gencode
> arch=compute_20,code=sm_20  -gencode arch=compute_20,code=compute_20"
> work correctly? CMake has to parse the gencode flags and move them to
> special msbuild entries, and I wonder if there is a bug when parsing
> multiple code options.
>
> On Wed, Sep 6, 2017 at 10:36 AM, Mueller-Roemer, Johannes Sebastian <
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de<mailto:
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de>> wrote:
> If the comment is wrt 3., the result is the same no matter if I use a
> space or an equals sign, if I double escape the quotes (e.g., use the value
> as you wrote it in the cmake-gui for CMAKE_CUDA_FLAGS), the value in VS
> becomes compute_20,"sm_20,compute_20" instead of
> compute_20,sm_20,compute_20, which is equally illegal and will cause the
> same error message, also double escaping should not be necessary for a
> subset of flags (Xcompiler works fine without it, even if the default
> values are pointless).
> If the comment is wrt 2., this is the line automatically generated by
> CMake/VS when not passing in any gencode flags explicitly.
>
> Or am I misunderstanding what you mean?
>
> Fraunhofer-Institut für Graphische Datenverarbeitung IGD
> Fraunhoferstr. 5  |  64283 Darmstadt  |  Germany
> Tel +49 6151 155-606<tel:+49%206151%20155606>  |  Fax +49 6151 155-139
> <tel:+49%206151%20155139>
> johannes.mueller-roemer at igd.fraunhofer.de<mailto:johannes.
> mueller-roemer at igd.fraunhofer.de> | www.igd.fraunhofer.de<http://
> www.igd.fraunhofer.de>
>
> From: Robert Maynard [mailto:robert.maynard at kitware.com<mailto:robert.
> maynard at kitware.com>]
> Sent: Wednesday, September 6, 2017 15:38
> To: Mueller-Roemer, Johannes Sebastian <Johannes.Sebastian.Mueller-
> Roemer at igd.fraunhofer.de<mailto:Johannes.Sebastian.
> Mueller-Roemer at igd.fraunhofer.de>>
> Cc: cmake at cmake.org<mailto:cmake at cmake.org>
> Subject: Re: [CMake] CMake 3.9.x Visual Studio CUDA PDB/compiler flags
> issues
>
> As far as gencode goes, your line has errors. With CUDA 8 it should look
> like:
>
> -gencode arch=compute_20,code=\"sm_20,compute_20\"
>
>
>
> On Mon, Sep 4, 2017 at 4:06 AM, Mueller-Roemer, Johannes Sebastian <
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de<mailto:
> Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de>> wrote:
> I’m having two issues with debug symbols for CUDA libraries on a project:
>
>
> 1.       The PDBs are not placed where the linker later expects to find
> them (haven’t been able to replicate this in a minimal project yet)
>
> 2.       Trying to circumnavigate this issue by using “/Z7” instead of
> “/Zi” does not work, Zi is always used (replicable in any project with CUDA)
>
> Replication steps: Replace -Zi or /Zi in CMAKE_*_FLAGS* (CUDA, CXX, C) by
> -Z7 or /Z7. Build and see several warnings like “cl : Command line warning
> D9025: overriding '/Z7' with '/Zi'” in the output window.
> And do the flags in the –Xcompiler make sense at all in VS? The CUDA build
> tools seem to determine these automatically anyways… for example I get the
> following call in the above scenario:
>
> "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin\nvcc.exe"
> -gencode=arch=compute_20,code=\"sm_20,compute_20\" --use-local-env
> --cl-version 2015 -ccbin "C:\Program Files (x86)\Microsoft Visual Studio
> 14.0\VC\bin\amd64"  -I"C:\Program Files\NVIDIA GPU Computing
> Toolkit\CUDA\v8.0\include" -I"C:\Program Files\NVIDIA GPU Computing
> Toolkit\CUDA\v8.0\include"  -G   --keep-dir x64\Debug -maxrregcount=0
> --machine 64 --compile -cudart static -Xcompiler="/EHsc -Z7 -Ob0" -g
>  -D_WINDOWS -Xcompiler "/EHsc /W3 /nologo /Od /FS /Zi /RTC1 /MDd /GR" -o
> A.dir\Debug\a.cu.obj "D:\jsroemer\projects\cuda-pdb-test\a\a.cu<
> http://a.cu>"
>
> 3.       Also, -gencode flags do not behave as expected when given in the
> form -gencode=arch=compute_20,code="sm_20,compute_20", as this causes the
> VS build system to error out:
>
> C:\Program Files (x86)\MSBuild\Microsoft.Cpp\
> v4.0\V140\BuildCustomizations\CUDA 8.0.targets(216,9): error : Item
> '..\..\a\a.cu<http://a.cu>' Code Generation value is not in the expected
> format '[Arch],[Code]'.
>
> The code generation value in this example is set to
> compute_20,sm_20,compute_20 by CMake.
>
> Fraunhofer-Institut für Graphische Datenverarbeitung IGD
> Fraunhoferstr. 5  |  64283 Darmstadt  |  Germany
> Tel +49 6151 155-606<tel:+49%206151%20155606>  |  Fax +49 6151 155-139
> <tel:+49%206151%20155139>
> johannes.mueller-roemer at igd.fraunhofer.de<mailto:johannes.
> mueller-roemer at igd.fraunhofer.de> | www.igd.fraunhofer.de<http://
> www.igd.fraunhofer.de>
>
>
> --
>
> Powered by www.kitware.com<http://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:
> http://public.kitware.com/mailman/listinfo/cmake
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/cmake/attachments/20170907/e90cafbc/attachment-0001.html>