Compile error on win11

[TenYac]

Describe the bug (both Chinese and English are OK)

visual studio2022编译报错

output窗口输出: D:\10yac_workspace2\zeno\build\projects\CUDA\zpc\include\zensim>"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin\nvcc.exe" -dlink -o zs_cu_wrangler.dir\Release\zs_cu_wrangler.device-link.obj -Xcompiler "/EHsc /W1 /nologo /O2 /MD /GR" -Xcompiler "/Fdzs_cu_wrangler.dir\Release\zs_cu_wrangler.pdb" -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin/crt" -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\lib\x64" kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib cudart.lib cudadevrt.lib -gencode=arch=compute_89,code=sm_89 zs_cu_wrangler.dir\Release\ParticleWrangler.ptx 67>cl : 命令行 warning D9025: 正在重写“/W1”(用“/w”) 67>cl : 命令行 warning D9025: 正在重写“/W1”(用“/w”) 67>cl : 命令行 warning D9025: 正在重写“/W1”(用“/w”) 65>StringReadUtil.cpp 67>cl : 命令行 warning D9025: 正在重写“/W1”(用“/w”) 68>cl : 命令行 warning D9025: 正在重写“/W1”(用“/w”) 66>kernel32.lib 66>user32.lib 66>gdi32.lib 65>StringWriteUtil.cpp 66>winspool.lib 66>comdlg32.lib 66>advapi32.lib 66>shell32.lib 66>ole32.lib 66>oleaut32.lib 66>uuid.lib 66>odbc32.lib 66>odbccp32.lib 66>cudart.lib 66>cudadevrt.lib 66>ParticleWrangler.ptx 66>CUDALINK : nvlink error : Undefined reference to '_Z16zfx_wrangle_funcPfPKf' in 'C:/Users/10yac/AppData/Local/Temp/tmpxft_00002d58_00000000-10_ParticleWrangler.obj' 66>d:\Program Files\Microsoft Visual Studio\2022\Enterprise\MSBuild\Microsoft\VC\v170\BuildCustomizations\CUDA 12.3.targets(908,9): error MSB3721: 命令“"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin\nvcc.exe" -dlink -o zs_cu_wrangler.dir\Release\zs_cu_wrangler.device-link.obj -Xcompiler "/EHsc /W1 /nologo /O2 /MD /GR" -Xcompiler "/Fdzs_cu_wrangler.dir\Release\zs_cu_wrangler.pdb" -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\bin/crt" -L"C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.3\lib\x64" kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib cudart.lib cudadevrt.lib -gencode=arch=compute_89,code=sm_89 zs_cu_wrangler.dir\Release\ParticleWrangler.ptx”已退出，返回代码为 255。 66>已完成生成项目“zs_cu_wrangler.vcxproj”的操作 - 失败。

How To Reproduce

Please concisely describe the steps before you encounter the bug, e.g.:

1. Cmake:

cmake -B build -DCMAKE_BUILD_TYPE=Release -DQt5_DIR="C:/Qt/Qt5.14.2/5.14.2/msvc2017_64/lib/cmake/Qt5" -DCMAKE_TOOLCHAIN_FILE="D:/10yac_workspace/vcpkg/scripts/buildsystems/vcpkg.cmake" -DZENO_WITH_CUDA:BOOL=ON -DZENO_WITH_ZenoFX:BOOL=ON -DZENO_ENABLE_OPTIX:BOOL=ON -DZENO_SYSTEM_OPENVDB:BOOL=OFF -DZENOFX_ENABLE_OPENVDB:BOOL=ON -DZENOFX_ENABLE_LBVH:BOOL=ON -DZENO_WITH_zenvdb:BOOL=ON -DZENO_WITH_FastFLIP:BOOL=ON -DZENO_WITH_FEM:BOOL=ON -DZENO_WITH_Rigid:BOOL=ON -DZENO_WITH_cgmesh:BOOL=ON -DZENO_WITH_oldzenbase:BOOL=ON -DZENO_WITH_TreeSketch:BOOL=ON -DZENO_WITH_Skinning:BOOL=ON -DZENO_WITH_Euler:BOOL=ON -DZENO_WITH_Functional:BOOL=ON -DZENO_WITH_LSystem:BOOL=ON -DZENO_WITH_mesher:BOOL=ON -DZENO_WITH_Alembic:BOOL=ON -DZENO_WITH_FBX:BOOL=ON -DZENO_WITH_DemBones:BOOL=ON -DZENO_WITH_SampleModel:BOOL=ON -DZENO_WITH_CalcGeometryUV:BOOL=ON -DZENO_WITH_MeshSubdiv:BOOL=ON -DZENO_WITH_Audio:BOOL=ON -DZENO_WITH_PBD:BOOL=ON -DZENO_WITH_GUI:BOOL=ON -DZENO_WITH_ImgCV:BOOL=ON -DZENO_WITH_TOOL_FLIPtools:BOOL=ON -DZENO_WITH_TOOL_cgmeshTools:BOOL=ON -DZENO_WITH_TOOL_BulletTools:BOOL=ON -DZENO_WITH_TOOL_HerculesTools:BOOL=ON -DCMAKE_EXPORT_COMPILE_COMMANDS=ON

2. 由于编译zs_cu_wrangler模块之前会报错'ptxas fatal : SM version specified by .target is higher than default SM version assumed', 所以我尝试在zs_cu_wrangler的项目属性的CUDA C/C++ -> Device -> Code Generation中填入了compute_89,sm_89
3. 如果cmake的ZENO_WITH_CUDA设置为OFF可以编译成功
4. 最新的子模块zpc也可以单独编译成功

Environment

• System: Windows 11
• Zeno Version: 提交哈希(af4e1774398fa254f3f8a957d2549803c236b24f), 最新的release版本(76da4ff44ea8a16d863f4c836b33f1da906d6c75)也尝试过, 不过由于这个版本的zpc模块没有修复该问题(https://github.com/zenustech/zpc/issues/5), 所以我尝试暂时注释掉了ZpcIterator.hpp中的

&& is_same_v<decltype(declval<Iter &>() + declval<diff_t>()), Iter> && is_same_v<decltype(declval<diff_t>() + declval<Iter &>()), Iter>

        虽然这让zpc模块能够成功通过编译, 不过也会出现上述的nvlink error问题.

• Graphics card (for viewport bugs): GTX 4090 驱动版本546.33, CUDA Toolkit 12.3

[teachmain]

@littlemine

[littlemine]

月初调整了下cuda模块构建时architecture设置的逻辑，请问现在问题解决了吗

[TenYac]

有对应的提交版本吗? 我拉下来试一下.