Merge branch 'master' into OptixBSDF

zenustech · Oct 12, 2023 · ab207be · ab207be
2 parents c1e15f7 + 393f30e
commit ab207be
Show file tree

Hide file tree

Showing 289 changed files with 18,422 additions and 2,341 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -39,6 +39,7 @@ option(ZENO_INSTALL_TARGET "Enable installation targets" OFF)
 option(ZENO_WITH_CUDA "Build ZENO with CUDA support" OFF)
 option(ZENO_MARCH_NATIVE "Build ZENO with -march=native" OFF)
 option(ZENO_USE_FAST_MATH "Build ZENO with -ffast-math" OFF)
+option(ZENO_OPTIX_PROC "Optix with a new proc" OFF)
 
 if (NOT DEFINED CMAKE_POSITION_INDEPENDENT_CODE)
     # Otherwise we can't link .so libs with .a libs

diff --git a/projects/CuLagrange/fem/FleshDynamicStepping.cu b/projects/CuLagrange/fem/FleshDynamicStepping.cu
@@ -2315,7 +2315,7 @@ struct FleshDynamicStepping : INode {
                     auto inds = gh_buffer.pack(dim_c<4>,"inds",ei).reinterpret_bits(int_c);
                     for(int i = 0;i != 4;++i)
                         if(inds[i] < 0 || inds[i] >= vsize)
-                            printf("negative sttemp inds : %d %d %d\n",inds[0],inds[1],inds[2],inds[3]);
+                            printf("negative sttemp inds : %d %d %d, %d\n",inds[0],inds[1],inds[2],inds[3]);
                     auto H = gh_buffer.pack(dim_c<12,12>,"H",ei);
                     update_hessian(spmat,inds,H,true);
             });

diff --git a/projects/CuLagrange/fem/collision_energy/evaluate_collision.hpp b/projects/CuLagrange/fem/collision_energy/evaluate_collision.hpp
@@ -1298,6 +1298,7 @@ void calc_continous_self_PT_collision_impulse(Pol& pol,
         else
             triCCDBvh.build(pol,bvs);
 
+        auto execTag = wrapv<space>{};
         pol(zs::range(verts.size()),[
             invMass = proxy<space>({},invMass),
             xtag = xtag,
@@ -1309,7 +1310,7 @@ void calc_continous_self_PT_collision_impulse(Pol& pol,
             impulse_buffer = proxy<space>(impulse_buffer),
             impulse_count = proxy<space>(impulse_count),
             eps = eps,
-            exec_tag = wrapv<space>{},
+            execTag,
             bvh = proxy<space>(triCCDBvh)] ZS_LAMBDA(int vi) mutable {
                 auto p = verts.pack(dim_c<3>,xtag,vi);
                 auto v = verts.pack(dim_c<3>,vtag,vi);
@@ -1392,9 +1393,9 @@ void calc_continous_self_PT_collision_impulse(Pol& pol,
 
                     for(int i = 0;i != 4;++i) {
                         auto beta = (bary[i] * invMass("minv",inds[i])) / cm;
-                        atomic_add(exec_tag,&impulse_count[inds[i]],1);
+                        atomic_add(execTag,&impulse_count[inds[i]],1);
                         for(int d = 0;d != 3;++d)
-                            atomic_add(exec_tag,&impulse_buffer[inds[i]][d],impulse[d] * beta);
+                            atomic_add(execTag,&impulse_buffer[inds[i]][d],impulse[d] * beta);
                     }
                 };
                 bvh.iter_neighbors(bv,do_close_proximity_detection);
@@ -1435,6 +1436,7 @@ void calc_continous_self_EE_collision_impulse(Pol& pol,
         else
             edgeCCDBvh.build(pol,edgeBvs);
 
+        auto execTag = wrapv<space>{};
         pol(zs::range(edges.size()),[
             xtag = xtag,
             vtag = vtag,
@@ -1445,7 +1447,7 @@ void calc_continous_self_EE_collision_impulse(Pol& pol,
             impulse_count = proxy<space>(impulse_count),
             // thickness = thickness,
             eps = eps,
-            exec_tag = wrapv<space>{},
+            execTag,
             edgeBvs = proxy<space>(edgeBvs),
             bvh = proxy<space>(edgeCCDBvh)] ZS_LAMBDA(int ei) mutable {
                 auto ea = edges.pack(dim_c<2>,"inds",ei,int_c);
@@ -1541,9 +1543,9 @@ void calc_continous_self_EE_collision_impulse(Pol& pol,
                     if(!compute_continous_EE_collision_impulse(ps,vs,minvs,imps))
                         return;    
                     for(int i = 0;i != 4;++i) {
-                        atomic_add(exec_tag,&impulse_count[inds[i]],1);
+                        atomic_add(execTag,&impulse_count[inds[i]],1);
                         for(int d = 0;d != 3;++d)
-                            atomic_add(exec_tag,&impulse_buffer[inds[i]][d],imps[i][d]);
+                            atomic_add(execTag,&impulse_buffer[inds[i]][d],imps[i][d]);
                     }     
 #endif
                 };
@@ -1572,15 +1574,17 @@ void apply_impulse(Pol& pol,
         zs::Vector<int> impulse_count{verts.get_allocator(),verts.size()};
         pol(zs::range(impulse_count),[] ZS_LAMBDA(auto& c) mutable {c = 0;});
 
+        auto execTag = wrapv<space>{};
         pol(zs::range(imminent_collision_buffer.size()),[
             verts = proxy<space>({},verts),
             vtag = zs::SmallString(vtag),
             imminent_collision_buffer = proxy<space>({},imminent_collision_buffer),
-            exec_tag = wrapv<space>{},
+            execTag,
             eps = eps,
             restitution_rate = imminent_restitution_rate,
             impulse_count = proxy<space>(impulse_count),
-            impulse_buffer = proxy<space>(impulse_buffer)] ZS_LAMBDA(auto ci) mutable {
+            impulse_buffer = proxy<space>(impulse_buffer)
+            ] ZS_LAMBDA(auto ci) mutable {
                 auto inds = imminent_collision_buffer.pack(dim_c<4>,"inds",ci,int_c);
                 auto bary = imminent_collision_buffer.pack(dim_c<4>,"bary",ci);
                 auto impulse = imminent_collision_buffer.pack(dim_c<3>,"impulse",ci);
@@ -1594,9 +1598,9 @@ void apply_impulse(Pol& pol,
 
                 for(int i = 0;i != 4;++i) {
                     auto beta = verts("minv",inds[i]) * bary[i] / cminv;
-                    atomic_add(exec_tag,&impulse_count[inds[i]],1);
+                    atomic_add(execTag,&impulse_count[inds[i]],1);
                     for(int d = 0;d != 3;++d)
-                        atomic_add(exec_tag,&impulse_buffer[inds[i]][d],impulse[d] * beta);
+                        atomic_add(execTag,&impulse_buffer[inds[i]][d],impulse[d] * beta);
                 }
 
         });
@@ -1607,14 +1611,14 @@ void apply_impulse(Pol& pol,
             impulse_count = proxy<space>(impulse_count),
             relaxation_rate = imminent_relaxation_rate,
             eps = eps,
-            exec_tag = wrapv<space>{}] ZS_LAMBDA(int vi) mutable {
+            execTag] ZS_LAMBDA(int vi) mutable {
             if(impulse_buffer[vi].norm() < eps || impulse_count[vi] == 0)
                 return;
             auto impulse = relaxation_rate * impulse_buffer[vi] / (T)impulse_count[vi];
             // auto dp = impulse * verts("minv",vi);
 
             for(int i = 0;i != 3;++i)   
-                atomic_add(exec_tag,&verts(vtag,i,vi),impulse[i]);
+                atomic_add(execTag,&verts(vtag,i,vi),impulse[i]);
         }); 
 }
 

diff --git a/projects/CuLagrange/geometry/Topology.cu b/projects/CuLagrange/geometry/Topology.cu
@@ -313,7 +313,7 @@ struct VisualTetrahedraHalfFacet : zeno::INode {
 					auto opposite_hf_idx = zs::reinterpret_bits<int>(halfFacet("opposite_hf",hf_idx));
 					if(opposite_hf_idx >= 0) {
 						if(opposite_hf_idx >= halfFacet.size()) {
-							printf("opposite_hf_idx = %d exceeding size of halfFacet : %d\n",opposite_hf_idx,halfFacet.size());
+							printf("opposite_hf_idx = %d exceeding size of halfFacet : %d\n", (int)opposite_hf_idx, (int)halfFacet.size());
 							return;
 						}
 						auto nti = zs::reinterpret_bits<int>(halfFacet("to_tet",opposite_hf_idx));

diff --git a/projects/CuLagrange/geometry/kernel/geo_math.hpp b/projects/CuLagrange/geometry/kernel/geo_math.hpp
@@ -50,10 +50,10 @@ namespace zeno { namespace LSL_GEO {
         return (costheta / sintheta);
     }
 
-    template<typename DREAL,typename VecT, zs::enable_if_all<VecT::dim == 1, (VecT::extent <= 3), (VecT::extent > 1)> = 0,typename REAL = VecT::value_type>
+    template<typename DREAL,typename VecT, zs::enable_if_all<VecT::dim == 1, (VecT::extent <= 3), (VecT::extent > 1)> = 0,typename REAL = typename VecT::value_type>
     constexpr auto cotTheta(const zs::VecInterface<VecT>& e0,const zs::VecInterface<VecT>& e1){
-        auto de0 = e0.cast<DREAL>();
-        auto de1 = e1.cast<DREAL>();
+        auto de0 = e0.template cast<DREAL>();
+        auto de1 = e1.template cast<DREAL>();
         auto costheta = de0.dot(de1);
         auto sintheta = de0.cross(de1).norm();
         return (REAL)(costheta / sintheta);

diff --git a/projects/CuLagrange/geometry/kernel/intersection.hpp b/projects/CuLagrange/geometry/kernel/intersection.hpp
@@ -1969,7 +1969,8 @@ int do_global_self_intersection_analysis_on_surface_mesh_info(Pol& pol,
                                         r = LSL_GEO::tri_ray_intersect(v1[j],e1s[j],v0[0],v0[1],v0[2]);
                                         if(r < (T)(1.0)) {
                                             e1_indices[nm_e1_its] = j;
-                                            e1_its[nm_e1_its] = v1[i] + e1s[j] * r;
+                                            /// @note both these should be 'j'
+                                            e1_its[nm_e1_its] = v1[j] + e1s[j] * r;
                                             ++nm_e1_its;
                                         } 
                                     }
@@ -3462,7 +3463,7 @@ int do_global_intersection_analysis_with_connected_manifolds(Pol& pol,
                         auto hi = key[0];
                         auto no = halfedges_tab.query(hi);
                         if(no >= closestTriID.size()) {
-                            printf("closestTriID overflow : %d %d\n",no,closestTriID.size());
+                            printf("closestTriID overflow : %d %d\n",(int)no, (int)closestTriID.size());
                             return;
                         }
                         auto cti = closestTriID[no];

diff --git a/projects/CuLagrange/geometry/kernel/topology.hpp b/projects/CuLagrange/geometry/kernel/topology.hpp
@@ -403,7 +403,7 @@ namespace zeno {
         constexpr auto CODIM = VecTi::extent;
         constexpr auto NM_EDGES = (CODIM - 1) * (CODIM) / 2;
 
-        zs::vec<Ti,2> out_edges[NM_EDGES] = {};
+        zs::vec<zs::vec<Ti,2>, NM_EDGES> out_edges{};
         int nm_out_edges = 0;
         for(int i = 0;i != CODIM;++i)
             for(int j = i + 1;j != CODIM;++j)

diff --git a/projects/CuLagrange/mpm/Generation.cpp b/projects/CuLagrange/mpm/Generation.cpp
@@ -73,9 +73,9 @@ struct PoissonDiskSample : INode {
     }
 };
 ZENDEFNODE(PoissonDiskSample, {
-                                  {"VDBGrid", {"float", "dx", "0.1"}, {"float", "ppc", "8"}},
+                                  {"VDBGrid", {"float", "dx", "0.1"}, {"float", "ppc", "8"}, {"string", "path", ""}},
                                   {"prim"},
-                                  {{"string", "path", ""}},
+                                  {},
                                   {"MPM"},
                               });
 
@@ -94,17 +94,15 @@ struct ZSPoissonDiskSample : INode {
             using LsT = RM_CVREF_T(ls);
             if constexpr (is_same_v<LsT, typename ZenoLevelSet::basic_ls_t>) {
                 auto &field = ls._ls;
-                match(
-                    [&sampled = sampled, dx = dx,
-                     ppc = ppc](auto &lsPtr) -> std::enable_if_t<is_spls_v<typename RM_CVREF_T(lsPtr)::element_type>> {
+                match([&sampled = sampled, dx = dx, ppc = ppc](auto &lsPtr) {
+                    if constexpr (is_spls_v<typename RM_CVREF_T(lsPtr)::element_type>) {
                         const auto &ls = *lsPtr;
                         const auto &spls = ls.memspace() != memsrc_e::host ? ls.clone({memsrc_e::host, -1}) : ls;
                         sampled = zs::sample_from_levelset(zs::proxy<zs::execspace_e::openmp>(spls), dx, ppc);
-                    },
-                    [](auto &lsPtr) -> std::enable_if_t<!is_spls_v<typename RM_CVREF_T(lsPtr)::element_type>> {
+                    } else
                         throw std::runtime_error(
                             fmt::format("levelset type [{}] not supported in sampling.", zs::get_var_type_str(lsPtr)));
-                    })(field);
+                })(field);
             } else if constexpr (is_same_v<LsT, typename ZenoLevelSet::const_sdf_vel_ls_t>) {
                 match([&sampled = sampled, dx = dx, ppc = ppc](auto lsv) {
                     sampled = zs::sample_from_levelset(SdfVelFieldView{lsv}, dx, ppc);

diff --git a/projects/CuLagrange/pbd/constraint_function_kernel/constraint.cuh b/projects/CuLagrange/pbd/constraint_function_kernel/constraint.cuh
@@ -133,7 +133,7 @@ namespace zeno { namespace CONSTRAINT {
         return true;
     }
 
-    template<typename VECTOR3d,typename SCALER = VECTOR3d::value_type>
+    template<typename VECTOR3d,typename SCALER = typename VECTOR3d::value_type>
     constexpr bool init_DihedralBendingConstraint(
         const VECTOR3d& p0,
         const VECTOR3d& p1,
@@ -229,7 +229,7 @@ namespace zeno { namespace CONSTRAINT {
         return true;
     }
 
-    template<typename VECTOR3d,typename SCALER = VECTOR3d::value_type>
+    template<typename VECTOR3d,typename SCALER = typename VECTOR3d::value_type>
     constexpr bool solve_DihedralConstraint(
         const VECTOR3d& p0, const SCALER& invMass0,
         const VECTOR3d& p1, const SCALER& invMass1,
@@ -310,24 +310,24 @@ namespace zeno { namespace CONSTRAINT {
         const VECTOR3d x[4] = { p2, p3, p0, p1 };
         // Q = MATRIX4d::uniform(0);
 
-        // const auto e0 = x[1].cast<double>() - x[0].cast<double>();
-        // const auto e1 = x[2].cast<double>() - x[0].cast<double>();
-        // const auto e2 = x[3].cast<double>() - x[0].cast<double>();
-        // const auto e3 = x[2].cast<double>() - x[1].cast<double>();
-        // const auto e4 = x[3].cast<double>() - x[1].cast<double>();
+        // const auto e0 = x[1].template cast<double>() - x[0].template cast<double>();
+        // const auto e1 = x[2].template cast<double>() - x[0].template cast<double>();
+        // const auto e2 = x[3].template cast<double>() - x[0].template cast<double>();
+        // const auto e3 = x[2].template cast<double>() - x[1].template cast<double>();
+        // const auto e4 = x[3].template cast<double>() - x[1].template cast<double>();
 
-        const auto e0 = x[1].cast<double>() - x[0].cast<double>();
-        const auto e1 = x[2].cast<double>() - x[0].cast<double>();
-        const auto e2 = x[3].cast<double>() - x[0].cast<double>();
-        const auto e3 = x[2].cast<double>() - x[1].cast<double>();
-        const auto e4 = x[3].cast<double>() - x[1].cast<double>();
+        const auto e0 = x[1].template cast<double>() - x[0].template cast<double>();
+        const auto e1 = x[2].template cast<double>() - x[0].template cast<double>();
+        const auto e2 = x[3].template cast<double>() - x[0].template cast<double>();
+        const auto e3 = x[2].template cast<double>() - x[1].template cast<double>();
+        const auto e4 = x[3].template cast<double>() - x[1].template cast<double>();
 
 
-        // auto e0 = e0_.cast<double>();
-        // auto e1 = e1_.cast<double>();
-        // auto e2 = e2_.cast<double>();
-        // auto e3 = e3_.cast<double>();
-        // auto e4 = e4_.cast<double>();
+        // auto e0 = e0_.template cast<double>();
+        // auto e1 = e1_.template cast<double>();
+        // auto e2 = e2_.template cast<double>();
+        // auto e3 = e3_.template cast<double>();
+        // auto e4 = e4_.template cast<double>();
 
         // printf("init isometric bending energy : %f %f %f %f\n",
         //     (float)p0.norm(),

diff --git a/projects/FBX/CMakeLists.txt b/projects/FBX/CMakeLists.txt
@@ -43,3 +43,5 @@ endif()
 zeno_add_asset_dir(${DEM_DIR})
 configure_file(zeno_FBX_config.h.in ${CMAKE_CURRENT_BINARY_DIR}/confinc/zeno_FBX_config.h @ONLY)
 target_include_directories(zeno PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/confinc)
+
+execute_process(COMMAND ${CMAKE_COMMAND} -E copy_directory ${CMAKE_CURRENT_SOURCE_DIR}/dem-bones-bin/models ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/models)
diff --git a/projects/FBX/MayaCamera.cpp b/projects/FBX/MayaCamera.cpp
@@ -256,7 +256,15 @@ struct LightNode : INode {
 
         auto exposure = get_input2<float>("exposure");
         auto intensity = get_input2<float>("intensity");
-        intensity *= pow(2.0, exposure);
+
+        auto scaler = powf(2.0f, exposure);
+
+        if (std::isnan(scaler) || std::isinf(scaler) || scaler < 0.0f) {
+            scaler = 1.0f;
+            printf("Light exposure = %f is invalid, fallback to 0.0 \n", exposure);
+        }
+
+        intensity *= scaler;
 
         auto prim = std::make_shared<zeno::PrimitiveObject>();
         auto &verts = prim->verts;
@@ -301,6 +309,14 @@ struct LightNode : INode {
 
         auto &clr = prim->verts.add_attr<zeno::vec3f>("clr");
         auto c = color * intensity;
+
+        for (size_t i=0; i<c.size(); ++i) {
+            if (std::isnan(c[i]) || std::isinf(c[i]) || c[i] < 0.0f) {
+                c[i] = 1.0f;
+                printf("Light color component %llu is invalid, fallback to 1.0 \n", i);
+            }
+        }
+
         for(int i=0; i<verts.size(); i++){
             clr[i] = c;
         }

diff --git a/projects/SampleModel/LoadSampleModel.cpp b/projects/SampleModel/LoadSampleModel.cpp
@@ -7,14 +7,32 @@
 #include <filesystem>
 #include <sstream>
 #include <cctype>
+#ifdef __linux__
+#include <unistd.h>
+#include <stdio.h>
+#endif
 
 namespace zeno {
 namespace {
 
+std::string modelsDir() {
+#ifdef __linux__
+    char path[1024];
+    getcwd(path, sizeof(path));
+    auto cur_path = std::string(path);
+    return cur_path + "/models";
+#else
+    auto cur_path = std::string(_pgmptr);
+    cur_path = cur_path.substr(0, cur_path.find_last_of("\\"));
+    return cur_path + "\\models";
+#endif
+}
+
 struct LoadSampleModel : INode {
     virtual void apply() override {
         auto name = get_input2<std::string>("name");
-        auto dir = getAssetDir(MODELS_DIR, name + ".obj");
+        //auto dir = getAssetDir(MODELS_DIR, name + ".obj");
+        auto dir = getAssetDir(modelsDir(), name + ".obj");  //temp fix for zeno-727 release
         log_info("found sample model path [{}]", dir);
         set_output("prim", getThisGraph()->callTempNode("ReadObjPrim", {
             {"triangulate:", objectFromLiterial(get_input2<bool>("triangulate"))},
@@ -62,7 +80,8 @@ struct LoadStringPrim : INode {
                 std::ostringstream ss;
                 ss << "ascii/" << std::setw(3) << std::setfill('0') << c << ".obj";
                 //printf("asdasd %s\n", ss.str().c_str());
-                auto path = getAssetDir(MODELS_DIR, ss.str());
+                //auto path = getAssetDir(MODELS_DIR, ss.str());
+                auto path = getAssetDir(modelsDir(), ss.str());  //temp fix for zeno-727 release
                 if (!std::filesystem::exists(path)) {
                     zeno::log_warn("LoadStringPrim got ASCII char not printable: {}", c);
                     continue;

diff --git a/projects/ZenoFX/CMakeLists.txt b/projects/ZenoFX/CMakeLists.txt
@@ -4,7 +4,7 @@ target_include_directories(zeno PRIVATE .)
 
 target_link_libraries(zeno PRIVATE $<BUILD_INTERFACE:ZFX>)
 target_sources(zeno PRIVATE
-    nw.cpp pw.cpp pnw.cpp ppw.cpp p2w.cpp pmw.cpp tw.cpp ne.cpp se.cpp FDGather.cpp dbg_printf.h
+    nw.cpp pw.cpp pnw.cpp ppw.cpp p2w.cpp pmw.cpp tw.cpp ne.cpp se.cpp FDGather.cpp refutils.cpp dbg_printf.h
     )
 
 #if (ZENO_WITH_zenvdb)

diff --git a/projects/ZenoFX/ne.cpp b/projects/ZenoFX/ne.cpp
@@ -15,6 +15,8 @@
 #include "dbg_printf.h"
 
 namespace zeno {
+    std::string preApplyRefs(const std::string& code, Graph* pGraph);
+
 namespace {
 static zfx::Compiler compiler;
 static zfx::x64::Assembler assembler;
@@ -115,6 +117,15 @@ struct NumericEval : zeno::INode {
             opts.define_param(key, dim);
         }
 
+        if (1)
+        {
+            // BEGIN 引用预解析：将其他节点参数引用到此处，可能涉及提前对该参数的计算
+            // 方法是: 搜索code里所有ref(...)，然后对于每一个ref(...)，解析ref内部的引用，
+            // 然后将计算结果替换对应ref(...)，相当于预处理操作。
+            code = preApplyRefs(code, getThisGraph());
+            // END 引用预解析
+        }
+
         //开始编译
         if (code.find("@result") == std::string::npos)
             code = "@result = ( " + code + " )";