Merge remote-tracking branch 'master' into trace

Author: manuschneider

.github/workflows/conda_build.yml

@@ -16,7 +16,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-13, macos-latest]
+        os: [ubuntu-latest, macos-15-intel, macos-latest]
     defaults:
       run:
         shell: bash -l {0}

.github/workflows/conda_build_release.yml

@@ -13,7 +13,7 @@ jobs:
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        os: [ubuntu-latest, macos-13, macos-latest]
+        os: [ubuntu-latest, macos-15-intel, macos-latest]
 
     steps:
     - uses: actions/checkout@v4

CMakeLists.txt

@@ -317,14 +317,14 @@ endif()
 # #######
 IF(BUILD_PYTHON)
   find_package (Python COMPONENTS Interpreter Development)
-  find_package(pybind11 REQUIRED)
+  find_package(pybind11 3.0.0 QUIET)
 
   if(NOT pybind11_FOUND)
     include(FetchContent)
     FetchContent_Declare(
       pybind11_sources
       GIT_REPOSITORY https://github.com/pybind/pybind11.git
-      GIT_TAG v2.11.1
+      GIT_TAG v3.0.1
     )
 
     FetchContent_GetProperties(pybind11_sources)

CytnxBKNDCMakeLists.cmake

@@ -126,6 +126,33 @@ if(USE_CUDA)
     #      -gencode=arch=compute_75,code=compute_75 ")
     target_compile_definitions(cytnx PUBLIC UNI_GPU)
     target_include_directories(cytnx PRIVATE ${CUDAToolkit_INCLUDE_DIRS})
+    # CUDA 12+/13 may place Thrust/CUB headers under include/cccl.
+    set(_cytnx_cccl_candidates)
+    if(DEFINED CUDAToolkit_TARGET_DIR AND NOT "${CUDAToolkit_TARGET_DIR}" STREQUAL "")
+      list(APPEND _cytnx_cccl_candidates "${CUDAToolkit_TARGET_DIR}/include/cccl")
+    endif()
+    foreach(_cuda_inc IN LISTS CUDAToolkit_INCLUDE_DIRS)
+      list(APPEND _cytnx_cccl_candidates
+        "${_cuda_inc}/cccl"
+        "${_cuda_inc}/../include/cccl"
+        "${_cuda_inc}/../../include/cccl"
+        "${_cuda_inc}/../../../include/cccl")
+    endforeach()
+    list(REMOVE_DUPLICATES _cytnx_cccl_candidates)
+
+    set(_cytnx_cccl_dir "")
+    foreach(_cccl_candidate IN LISTS _cytnx_cccl_candidates)
+      get_filename_component(_cccl_candidate_abs "${_cccl_candidate}" ABSOLUTE)
+      if(EXISTS "${_cccl_candidate_abs}")
+        set(_cytnx_cccl_dir "${_cccl_candidate_abs}")
+        break()
+      endif()
+    endforeach()
+    if(NOT "${_cytnx_cccl_dir}" STREQUAL "")
+      target_include_directories(cytnx PRIVATE "${_cytnx_cccl_dir}")
+      message(STATUS "Detected CCCL headers at: ${_cytnx_cccl_dir}")
+    endif()
+
     target_link_libraries(cytnx PUBLIC CUDA::toolkit)
     target_link_libraries(cytnx PUBLIC CUDA::cudart CUDA::cublas CUDA::cusparse CUDA::curand CUDA::cusolver)
     target_link_libraries(cytnx PUBLIC -lcudadevrt)

Readme.md

@@ -153,8 +153,8 @@ Tensor A({3,4,5},Type.Double);
 UniTensor tA = UniTensor(A); // convert directly.
 UniTensor tB = UniTensor({Bond(3),Bond(4),Bond(5)},{}); // init from scratch.
 // Relabel the tensor and then contract.
-tA.relabels_({"common_1", "common_2", "out_a"});
-tB.relabels_({"common_1", "common_2", "out_b"});
+tA.relabel_({"common_1", "common_2", "out_a"});
+tB.relabel_({"common_1", "common_2", "out_b"});
 UniTensor out = cytnx::Contract(tA,tB);
 tA.print_diagram();
 tB.print_diagram();

benchmarks/linalg/Arnoldi_bm.cpp

@@ -11,7 +11,7 @@ namespace BMTest_Arnoldi {
          const unsigned int& dtype = Type.Double, const int& device = Device.cpu);
     UniTensor matvec(const UniTensor& l) override {
       auto tmp = Contracts({A, l, B}, "", true);
-      tmp.relabels_(l.labels()).set_rowrank(l.rowrank());
+      tmp.relabel_(l.labels()).set_rowrank(l.rowrank());
       return tmp;
     }
   };
@@ -22,15 +22,15 @@ namespace BMTest_Arnoldi {
     std::vector<Bond> bonds = {Bond(D), Bond(d), Bond(D)};
     A = UniTensor(bonds, {}, -1, in_dtype, in_device)
           .set_name("A")
-          .relabels_({"al", "phys", "ar"})
+          .relabel_({"al", "phys", "ar"})
           .set_rowrank(2);
     B = UniTensor(bonds, {}, -1, in_dtype, in_device)
           .set_name("B")
-          .relabels_({"bl", "phys", "br"})
+          .relabel_({"bl", "phys", "br"})
           .set_rowrank(2);
     T_init = UniTensor({Bond(D), Bond(D)}, {}, -1, in_dtype, in_device)
                .set_name("l")
-               .relabels_({"al", "bl"})
+               .relabel_({"al", "bl"})
                .set_rowrank(1);
     if (Type.is_float(this->dtype())) {
       double low = -1.0, high = 1.0;

benchmarks/linalg/Lanczos_bm.cpp

@@ -27,7 +27,7 @@ namespace BMTest_Lanczos {
      *         | |         +          | |             "pi"
      *         |_|--"vol" "po" "vor"--|_|
      *
-     * Then relabels ["vil", "pi", "vir"] -> ["vol", "po", "vor"]
+     * Then relabel ["vil", "pi", "vir"] -> ["vol", "po", "vor"]
      *
      * "vil":virtual in bond left
      * "po":physical out bond

benchmarks/linalg/Svd_bm.cpp

@@ -51,7 +51,7 @@ namespace BMTest_Svd {
     auto UT = UniTensor(bonds, labels, row_rank, dtype, device, is_diag);
     double low = -1000, high = 1000;
     int rnd_seed = 0;
-    random::Make_uniform(UT, low = -1000, high = 1000, rnd_seed);
+    random::uniform_(UT, low = -1000, high = 1000, rnd_seed);
     return UT;
   }
 
@@ -122,7 +122,7 @@ namespace BMTest_Svd {
     auto UT = UniTensor(bonds, labels, row_rank, dtype, Device.cpu, is_diag = false);
     double l_bd, h_bd;
     int rnd_seed;
-    random::Make_uniform(UT, l_bd = -1000, h_bd = 1000, rnd_seed);
+    random::uniform_(UT, l_bd = -1000, h_bd = 1000, rnd_seed);
     return UT;
   }
   static void BM_bkUT_U1_Svd_F64(benchmark::State& state) {

benchmarks/linalg/Svd_truncate_bm.cpp

@@ -17,7 +17,7 @@ namespace BMTest_Svd_truncate {
     auto UT = UniTensor(bonds, labels, row_rank, dtype, device, is_diag);
     double low = -1000, high = 1000;
     int rnd_seed = 0;
-    random::Make_uniform(UT, low = -1000, high = 1000, rnd_seed);
+    random::uniform_(UT, low = -1000, high = 1000, rnd_seed);
     return UT;
   }
 
@@ -88,7 +88,7 @@ namespace BMTest_Svd_truncate {
     auto UT = UniTensor(bonds, labels, row_rank, dtype, Device.cpu, is_diag = false);
     double l_bd, h_bd;
     int rnd_seed;
-    random::Make_uniform(UT, l_bd = -1000, h_bd = 1000, rnd_seed);
+    random::uniform_(UT, l_bd = -1000, h_bd = 1000, rnd_seed);
     return UT;
   }
 

developer_tools/dev_test.py

@@ -7,7 +7,7 @@
 
 # uT = cytnx.UniTensor.arange(2*3*4, name="tensor uT")
 # uT.reshape_(2,3,4);
-# uT.relabels_(["a","b","c"])
+# uT.relabel_(["a","b","c"])
 
 # T = cytnx.random.uniform([4,4], low=-1., high=1.)
 # print(T)