einsum support for generalized product involving tensor-of-tensor and regular tensor

.pre-commit-config.yaml

@@ -38,5 +38,5 @@ repos:
                 name: Format C/C++ code using clang-format.
                 language: system
                 files: \.(c|cc|cxx|cpp|h|hpp|hxx)$
-                entry: clang-format -i
-                args: [--style=file]
+                entry: bin/admin/clang-format.sh
+                args: [--style=file -i]

CMakeLists.txt

@@ -165,6 +165,10 @@ if(TA_ENABLE_TILE_OPS_LOGGING AND NOT DEFINED TA_TILE_OPS_LOG_LEVEL)
   set(TA_TILE_OPS_LOG_LEVEL 1)
 endif(TA_ENABLE_TILE_OPS_LOGGING AND NOT DEFINED TA_TILE_OPS_LOG_LEVEL)
 
+option(TA_TRACE_GLOBAL_COMM_STATS "Enable tracing of communication stats of global objects (DistEval's and DIstributedStorage) TiledArray" OFF)
+add_feature_info(TASK_TRACE_DEBUG TA_TRACE_GLOBAL_COMM_STATS "Debug communication stats of global objects (DistEval's and DIstributedStorage) TiledArray")
+set(TILEDARRAY_ENABLE_GLOBAL_COMM_STATS_TRACE ${TA_TRACE_GLOBAL_COMM_STATS})
+
 option(TA_RANGEV3 "Enable Range-V3 library" OFF)
 add_feature_info(TA_RANGEV3 TA_RANGEV3 "Range-V3 ranges library")
 

INSTALL.md

@@ -42,7 +42,7 @@ Both methods are supported. However, for most users we _strongly_ recommend to b
   - Boost.Range: header-only, *only used for unit testing*
 - [BTAS](http://github.com/ValeevGroup/BTAS), tag bf0c376d5cdd6f668174b2a4c67b19634d1c0da7 . If usable BTAS installation is not found, TiledArray will download and compile
   BTAS from source. *This is the recommended way to compile BTAS for all users*.
-- [MADNESS](https://github.com/m-a-d-n-e-s-s/madness), tag c0c4ea543439c740e3ee848fdd055c633a47f6c5 .
+- [MADNESS](https://github.com/m-a-d-n-e-s-s/madness), tag b1f1c39c497b86ab3ef4e560a686de63eb555cc4 .
   Only the MADworld runtime and BLAS/LAPACK C API component of MADNESS is used by TiledArray.
   If usable MADNESS installation is not found, TiledArray will download and compile
   MADNESS from source. *This is the recommended way to compile MADNESS for all users*.

bin/admin/clang-format.sh

@@ -0,0 +1,94 @@
+#!/bin/bash
+
+# these are the versions of clang-format that are supported required
+# should be ordered from oldest to newest to make sure the newest is picked
+supported_clang_format_versions="16 17"
+preferred_clang_format_version=""  # prefer most recent supported clang-format version
+for v in $supported_clang_format_versions; do
+  preferred_clang_format_version=$v
+done
+
+# append common locations of clang-format to PATH
+unameOut="$(uname -s)"
+case "${unameOut}" in
+    Darwin*)
+      extra_path=""
+      # this prefers more recent versions
+      for v in $supported_clang_format_versions; do
+        extra_path=/opt/homebrew/opt/llvm@$v/bin:/opt/homebrew/opt/clang-format@$v/bin:$extra_path
+      done
+      # prepend paths
+      export PATH=$extra_path:$PATH:/opt/homebrew/bin
+    ;;
+esac
+
+path_to_clang_format=`which clang-format`
+have_supported_clang_format_version=0
+if [[ "X$path_to_clang_format" != "X" ]]; then
+
+  # check clang-format version
+  clang_format_version=`clang-format --version | sed 's/.* version //' | awk -F'[.]' '{print $1}'`
+
+  #echo "supported_clang_format_versions=\"$supported_clang_format_versions\" clang_format_version=$clang_format_version"
+
+  # if found clang-format, but wrong version, check if docker is available
+  for v in $supported_clang_format_versions; do
+    if [[ $clang_format_version -eq $v ]]; then
+      have_supported_clang_format_version=1
+      break
+    fi
+  done
+fi
+
+if [[ $have_supported_clang_format_version -eq 0 ]]; then
+  echo "WARNING: found clang-format with unsupported version $clang_format_version (supported versions: $supported_clang_format_versions)"
+
+  # look for docker
+  path_to_docker=`which docker`
+  if [[ "X$path_to_docker" = "X" ]]; then
+    echo "ERROR: docker is not found either, PATH=$PATH, install one of supported clang-format versions (any of these: $supported_clang_format_versions) or install docker"
+    exit 1
+  fi
+
+  # if docker up?
+  docker info >/dev/null 2>&1
+  if [[ $? -ne 0 ]]; then
+    echo "ERROR: docker is found but not running, start it"
+    exit 1
+  fi
+
+  # use docker to run clang-format
+  mount_path=$(readlink -f "$HOME")
+
+  # convert file names in the arguments to relative paths
+  args=""
+  for i in "$@"; do
+    # skip options
+    if [[ "$i" == -* ]]; then
+      args="$args $i"
+      continue
+    fi
+    abs_file_path=$(readlink -f "$i")
+    if [[ "X$abs_file_path" = "X" ]]; then
+      echo "ERROR: given file $i is not found"
+      exit 1
+    fi
+
+    dir=$(dirname $abs_file_path)
+    file_path_relative_to_project_root=$(basename $abs_file_path)
+    while [[ "$dir" != "$mount_path" && "$dir" != "/" ]]; do
+      file_path_relative_to_project_root="$(basename $dir)/$file_path_relative_to_project_root"
+      dir=$(dirname $dir)
+      #echo "dir=$dir file_path_relative_to_project_root=$file_path_relative_to_project_root"
+    done
+    if [[ "$dir" == "/" ]]; then
+      echo "ERROR: given file $i (absolute path $abs_file_path) is not under \$HOME=$mount_path, cannot use docker-based clang-format in this case"
+      exit 1
+    fi
+    args="$args /hostHOME/$file_path_relative_to_project_root"
+  done
+  docker run --platform linux/x86_64 -v $mount_path:/hostHOME xianpengshen/clang-tools:$preferred_clang_format_version clang-format $args
+else
+  #echo "found $path_to_clang_format with required version $clang_format_version"
+  clang-format $*
+fi

external/umpire.cmake

@@ -87,6 +87,7 @@ else()
         -DENABLE_EXAMPLES=OFF
         -DENABLE_LOGGING=OFF
         -DENABLE_ASSERTS=${enable_umpire_asserts}
+        -DENABLE_CLANGFORMAT=OFF
         )
 
     # caveat: on recent Ubuntu default libstdc++ provides filesystem, but if using older gcc (gcc-8) must link against

external/versions.cmake

@@ -19,8 +19,8 @@ set(TA_INSTALL_EIGEN_PREVIOUS_VERSION 3.3.7)
 set(TA_INSTALL_EIGEN_URL_HASH SHA256=b4c198460eba6f28d34894e3a5710998818515104d6e74e5cc331ce31e46e626)
 set(TA_INSTALL_EIGEN_PREVIOUS_URL_HASH MD5=b9e98a200d2455f06db9c661c5610496)
 
-set(TA_TRACKED_MADNESS_TAG c0c4ea543439c740e3ee848fdd055c633a47f6c5)
-set(TA_TRACKED_MADNESS_PREVIOUS_TAG 03c82cf2780d9e96298cc9140ac128c73eacd3b1)
+set(TA_TRACKED_MADNESS_TAG b1f1c39c497b86ab3ef4e560a686de63eb555cc4)
+set(TA_TRACKED_MADNESS_PREVIOUS_TAG cf3c98053453329f35b775c8b9f561301f6a997e)
 set(TA_TRACKED_MADNESS_VERSION 0.10.1)
 set(TA_TRACKED_MADNESS_PREVIOUS_VERSION 0.10.1)
 

src/CMakeLists.txt

@@ -100,7 +100,6 @@ TiledArray/dist_eval/contraction_eval.h
 TiledArray/dist_eval/dist_eval.h
 TiledArray/dist_eval/unary_eval.h
 TiledArray/einsum/index.h
-TiledArray/einsum/index.cpp
 TiledArray/einsum/range.h
 TiledArray/einsum/string.h
 TiledArray/expressions/add_engine.h
@@ -195,13 +194,10 @@ TiledArray/util/bug.h
 TiledArray/util/function.h
 TiledArray/util/initializer_list.h
 TiledArray/util/logger.h
-TiledArray/util/ptr_registry.cpp
 TiledArray/util/ptr_registry.h
-TiledArray/util/random.cpp
 TiledArray/util/random.h
 TiledArray/util/singleton.h
 TiledArray/util/threads.h
-TiledArray/util/threads.cpp
 TiledArray/util/thread_specific.h
 TiledArray/util/time.h
 TiledArray/util/vector.h
@@ -243,10 +239,15 @@ TiledArray/tensor_impl.cpp
 TiledArray/array_impl.cpp
 TiledArray/dist_array.cpp
 TiledArray/version.cpp
-TiledArray/util/backtrace.cpp
-TiledArray/util/bug.cpp
+TiledArray/einsum/index.cpp
+TiledArray/expressions/permopt.cpp
 TiledArray/math/linalg/basic.cpp
 TiledArray/math/linalg/rank-local.cpp
+TiledArray/util/backtrace.cpp
+TiledArray/util/bug.cpp
+TiledArray/util/ptr_registry.cpp
+TiledArray/util/random.cpp
+TiledArray/util/threads.cpp
 )
 # feed TILEDARRAY_GIT_REVISION and TILEDARRAY_GIT_DESCRIPTION to TiledArray/version.cpp only to avoid recompiling everything
 set_source_files_properties(

src/TiledArray/array_impl.h

@@ -636,7 +636,18 @@ class ArrayImpl : public TensorImpl<Policy> {
   /// DistributedStorage
 
   /// @return const reference to the atomic counter of live DelayedSet requests
-  const madness::AtomicInt& num_live_ds() const { return data_.num_live_ds(); }
+  const std::atomic<std::size_t>& num_live_ds() const {
+    return data_.num_live_ds();
+  }
+
+  /// Reports the number of live DelayedForward requests for this object's
+  /// DistributedStorage
+
+  /// @return const reference to the atomic counter of live DelayedForward
+  /// requests
+  const std::atomic<std::size_t>& num_live_df() const {
+    return data_.num_live_df();
+  }
 
 };  // class ArrayImpl
 

src/TiledArray/config.h.in

@@ -174,6 +174,9 @@
 #cmakedefine TA_ENABLE_TILE_OPS_LOGGING 1
 #define TA_TILE_OPS_LOG_LEVEL 0@TA_TILE_OPS_LOG_LEVEL@
 
+/* Enables collection of communication statistics for global objects (DistEval and DistributedStorage) */
+#cmakedefine TILEDARRAY_ENABLE_GLOBAL_COMM_STATS_TRACE 1
+
 /* ----------- pragma helpers ---------------*/
 #define TILEDARRAY_PRAGMA(x) _Pragma(#x)
 /* same as TILEDARRAY_PRAGMA(x), but expands x */