zenustech/zpc

zenus parallel computing library for zenus physics-based simulations

Zenus Simulation

Zenus Simulation is the codebase zensim maintained by Zenus Tech, which delivers great parallel computing efficiency for physics-based simulations within a shared-memory heterogeneous architecture through a unified programming interface on multiple compute backends.

This repo is going through rapid changes, and we do not promise ABI compatibility from commit to commit for the moment.

Document

See git wiki page for more build details. See Specification for more usage info.

Compilation

This is a cross-platform C++/CUDA cmake project. The minimum version requirement of cmake is 3.18, yet the latest version is generally recommended. Please install cmake through official website or python3-pip, since the cmake version in apt repo is behind.

When CUDA is enabled, the required CUDA version is 11.4+ (for c++17 and latest cuda utilities).

Currently, supported OS are Ubuntu 20.04+ and Windows 10, and tested compilers includes gcc10.0+, clang-11+, vs2019+.

Build

Before building this framework, please first manually configure these external dependencies, i.e. openvdb if ZS_ENABLE_OPENVDB is set to TRUE. Then pull all dependencies by

git submodule init
git submodule update

If CUDA (>=11.4) is installed and required, be sure to set ZS_ENABLE_CUDA=On first.

Configure the project using the CMake Tools extension in Visual Studio Code (recommended), or follow the build instructions in git wiki.

In addition, make sure to install zlib for building partio when building on linux.

sudo apt install zlib1g

Integration

Directly include the codebase as a submodule. Or install this repo then use find_package(zensim). If the installed package is no longer needed, build the uninstall target as with the install target.

Credits

This framework draws inspirations from Taichi, kokkos, raja, MGMPM, GPU LBVH.

Dependencies

The following libraries are adopted and made fully localized in our project development:

• fmt
• loguru. (pro: "chrono" is not exposed in its header)
• magic_enum
• catch2

For spatial data IO and generation, we use these libraries in addition:

• partio
• openvdb

We import these (less-frequently used) libraries as well:

• function_ref
• rapidjson
• cxxopts
• jitify

zpc's Issues

Compile Error on Win11

cmake

cmake -B build -DZS_ENABLE_CUDA=ON -DZS_ENABLE_OPENVDB=ON -DZS_AUTO_DETECT_CUDA_ARCH=OFF -DCMAKE_CUDA_ARCHITECTURES=86

visual studio 2022 output

9>D:/10yac_workspace2/zpc/zpc/include\zensim/ZpcIterator.hpp(198): error : more than one operator "+" matches these operands:
9>            function template "zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl> zs::operator+(zs::sint_t, zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>)"
9>(391): here
9>            function template "zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl> zs::operator+(zs::detail::infer_difference_type_t<Self>, zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>)"
9>(303): here
9>            operand types are: zs::i64 + zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>
9>          detected during:
9>            instantiation of class "zs::is_ra_iter<Iter, zs::void_t<Iter &, zs::iterator_traits<Iter, void>::difference_type>> [with Iter=zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>]"
9>(204): here
9>            instantiation of "const __nv_bool zs::is_ra_iter_v [with Iter=zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>]"
9>(578): here
9>            instantiation of class "zs::zip_iterator<zs::tuple<Iters...>, zs::index_sequence<Is...>> [with Iters=<zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, zs::LegacyIterator<zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, zs::LegacyIterator<zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, std::_Vector_const_iterator<std::_Vector_val<std::_Simple_types<zs::PropertyTag>>>>, Is=<0ULL, 1ULL, 2ULL, 3ULL>]"
9>(411): here
9>            instantiation of "auto zs::make_iterator<T,Args...>(Args &&...) [with T=zs::zip_iterator, Args=<zs::LegacyIterator<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, zs::LegacyIterator<zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, zs::LegacyIterator<zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>>::iterator_impl>, std::_Vector_const_iterator<std::_Vector_val<std::_Simple_types<zs::PropertyTag>>>>]"
9>(690): here
9>            instantiation of "auto zs::zip(Args &&...) [with Args=<zs::Vector<zs::SmallString, zs::ZSPmrAllocator<false, std::byte>> &, zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>> &, zs::Vector<zs::i32, zs::ZSPmrAllocator<false, std::byte>> &, const std::vector<zs::PropertyTag, std::allocator<zs::PropertyTag>> &>]"
9>D:/10yac_workspace2/zpc/zpc/include\zensim/container/TileVector.hpp(77): here
9>            instantiation of "zs::TileVector<T_, Length, AllocatorT>::TileVector(const zs::TileVector<T_, Length, AllocatorT>::allocator_type &, const std::vector<zs::PropertyTag, std::allocator<zs::PropertyTag>> &, zs::TileVector<T_, Length, AllocatorT>::size_type) [with T_=zs::i32, Length=8ULL, AllocatorT=zs::ZSPmrAllocator<false, std::byte>]"
9>D:/10yac_workspace2/zpc/zpc/include\zensim/container/TileVector.hpp(598): here
9>            instantiation of "void zs::TileVector<T_, Length, AllocatorT>::append_channels(Policy &&, const std::vector<zs::PropertyTag, std::allocator<zs::PropertyTag>> &) [with T_=zs::i32, Length=8ULL, AllocatorT=zs::ZSPmrAllocator<false, std::byte>, Policy=zs::CudaExecutionPolicy &]"
9>D:\10yac_workspace2\zpc\zpc\include\zensim\py_interop\cuda\TileVectorUtility.cpp(19): here

[Doc] Add an instruction to tell the users to install zlib?

It turned out that I need to sudo apt install zlib1g before building zpc :-)

Compile error on windows 11

I got error on windows 11 while compiling zpc “ ptxas fatal : SM version specified by .target is higher than default SM version assumed”，compile process would be ok with parameter ZS_ENABLE_CUDA=OFF.

the error message when building zpc is such：
$ cmake --build . --parallel 8 --config Release
MSBuild version 17.6.3+07e294721 for .NET Framework

zpccore.vcxproj -> D:\github\zpc\build\include\zensim\Release\zpccore.lib
zpcomp.vcxproj -> D:\github\zpc\build\include\zensim\Release\zpcomp.lib
zpccuda.vcxproj -> D:\github\zpc\build\include\zensim\Release\zpccuda.lib
zpc.vcxproj -> D:\github\zpc\build\include\zensim\Release\zpc.lib

D:\github\zpc\build\include\zensim>"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\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"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\bin/crt" -L"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\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 zs_cu_wrangler.dir\Release\VolumeWrangler.ptx zs_cu_wrangler.dir\Release\ParticleWrangler.ptx
ptxas fatal : SM version specified by .target is higher than default SM version assumed
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
VolumeWrangler.ptx
D:\software\Microsoft Visual Studio\2022\Community\MSBuild\Microsoft\VC\v170\BuildCustomizations\CUDA 12.1.targets(908,9): error MSB3721: 命令“"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\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"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\bin/crt" -L"D:\software\NVDIA GPU Computing Toolkit\CUDA\v12.1\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 zs_cu_wrangler.dir\Release\VolumeWrangler.ptx zs_cu_wrangler.dir\Release\ParticleWrangler.ptx”已退出，返回代码为 255。 [D:\github\zpc\build\include\zensim\zs_cu_wrangler.vcxproj]

pastebin

=== ZFX
define f3 @pos
define f3 @vel
define f1 arg0
parname arg0.0
@vel += arg0 * vec3(0,-9.8,0)
@pos += @vel * arg0

=== Parser
(+= vel (* arg0 (vec3 0 (- 9.8) 0)))
(+= pos (* vel arg0))
=== Transcriptor

• #9.8 $_Tm0
  vec3 #0 $_Tm0 #0 $_Tm1

• $arg0 $_Tm1 $_Tm2

• $_Tm2 @vel @vel

• @vel $arg0 $_Tm3

• $_Tm3 @pos @pos
  === InitialPass
  sub $_Tm0 #9.8
  vec3 $_Tm1 #0 $_Tm0 #0
  mul $_Tm2 $arg0 $_Tm1
  add @vel $_Tm2 @vel
  mul $_Tm3 @vel $arg0
  add @pos $_Tm3 @pos
  === UnfuncPass
  define f1 $_Tm0
  define f1 $_Tm1
  define f1 $_Tm2
  define f3 $_Tm3
  define f3 $_Tp0
  define f1 $arg0
  define f3 @pos
  define f3 @vel
  sub $_Tm0 #9.8
  mov $_Tp0.0 #0
  mov $_Tp0.1 $_Tm0
  mov $_Tp0.2 #0
  mov $_Tm1 $_Tp0
  mul $_Tm2 $arg0 $_Tm1
  add @vel $_Tm2 @vel
  mul $_Tm3 @vel $arg0
  add @pos $_Tm3 @pos
  === UnwrapPass
  define f1 $_Tm0
  define f3 $_Tm1
  define f3 $_Tm2
  define f3 $_Tm3
  define f3 $_Tp0
  define f1 $_Tp0.0
  define f1 $_Tp0.1
  define f1 $_Tp0.2
  define f1 $arg0
  define f3 @pos
  define f3 @vel
  sub $_Tm0.0 #9.8
  mov $_Tp0.0 #0
  mov $_Tp0.0 #0
  mov $_Tp0.0 #0
  mov $_Tp0.1 $_Tm0.0
  mov $_Tp0.1 $_Tm0.0
  mov $_Tp0.1 $_Tm0.0
  mov $_Tp0.2 #0
  mov $_Tp0.2 #0
  mov $_Tp0.2 #0
  mov $_Tm1.0 $_Tp0.0
  mov $_Tm1.1 $_Tp0.1
  mov $_Tm1.2 $_Tp0.2
  mul $_Tm2.0 $arg0.0 $_Tm1.0
  mul $_Tm2.1 $arg0.0 $_Tm1.1
  mul $_Tm2.2 $arg0.0 $_Tm1.2
  add @vel.0 $_Tm2.0 @vel.0
  add @vel.1 $_Tm2.1 @vel.1
  add @vel.2 $_Tm2.2 @vel.2
  mul $_Tm3.0 @vel.0 $arg0.0
  mul $_Tm3.1 @vel.1 $arg0.0
  mul $_Tm3.2 @vel.2 $arg0.0
  add @pos.0 $_Tm3.0 @pos.0
  add @pos.1 $_Tm3.1 @pos.1
  add @pos.2 $_Tm3.2 @pos.2
  === ReassignPass
  bind 3 pos.0
  bind 4 pos.1
  bind 5 pos.2
  bind 0 vel.0
  bind 1 vel.1
  bind 2 vel.2
  parm 0 arg0.0
  sub $1 #9.8
  mov $2 #0
  mov $2 #0
  mov $2 #0
  mov $3 $1
  mov $3 $1
  mov $3 $1
  mov $4 #0
  mov $4 #0
  mov $4 #0
  mov $5 $2
  mov $6 $3
  mov $7 $4
  mul $8 $0 $5
  mul $9 $0 $6
  mul $10 $0 $7
  add @0 $8 @0
  add @1 $9 @1
  add @2 $10 @2
  mul $11 @0 $0
  mul $12 @1 $0
  mul $13 @2 $0
  add @3 $11 @3
  add @4 $12 @4
  add @5 $13 @5