Implement typed storage for tensor. (#7429)

* Add `Tensor` class.
* Add elementwise kernel for CPU and GPU.
* Add unravel index.
* Move some computation to compile time.

include/xgboost/linalg.h

@@ -6,12 +6,16 @@
 #ifndef XGBOOST_LINALG_H_
 #define XGBOOST_LINALG_H_
 
+#include <dmlc/endian.h>
 #include <xgboost/base.h>
-#include <xgboost/generic_parameters.h>
+#include <xgboost/host_device_vector.h>
+#include <xgboost/json.h>
 #include <xgboost/span.h>
 
 #include <algorithm>
 #include <cassert>
+#include <limits>
+#include <string>
 #include <type_traits>
 #include <utility>
 #include <vector>
@@ -19,16 +23,35 @@
 namespace xgboost {
 namespace linalg {
 namespace detail {
-template <typename S, typename Head, size_t D>
+
-constexpr size_t Offset(S (&strides)[D], size_t n, size_t dim, Head head) {
+struct ArrayInterfaceHandler {
-  assert(dim < D);
+  template <typename T>
+  static constexpr char TypeChar() {
+    return (std::is_floating_point<T>::value
+                ? 'f'
+                : (std::is_integral<T>::value ? (std::is_signed<T>::value ? 'i' : 'u') : '\0'));
+  }
+};
+
+template <size_t dim, typename S, typename Head, size_t D>
+constexpr size_t Offset(S (&strides)[D], size_t n, Head head) {
+  static_assert(dim < D, "");
   return n + head * strides[dim];
 }
 
-template <typename S, size_t D, typename Head, typename... Tail>
+template <size_t dim, typename S, size_t D, typename Head, typename... Tail>
-constexpr size_t Offset(S (&strides)[D], size_t n, size_t dim, Head head, Tail &&...rest) {
+constexpr std::enable_if_t<sizeof...(Tail) != 0, size_t> Offset(S (&strides)[D], size_t n,
-  assert(dim < D);
+                                                                Head head, Tail &&...rest) {
-  return Offset(strides, n + (head * strides[dim]), dim + 1, rest...);
+  static_assert(dim < D, "");
+  return Offset<dim + 1>(strides, n + (head * strides[dim]), std::forward<Tail>(rest)...);
+}
+
+template <int32_t D>
+constexpr void CalcStride(size_t (&shape)[D], size_t (&stride)[D]) {
+  stride[D - 1] = 1;
+  for (int32_t s = D - 2; s >= 0; --s) {
+    stride[s] = shape[s + 1] * stride[s + 1];
+  }
 }
 
 struct AllTag {};
@@ -71,6 +94,101 @@ XGBOOST_DEVICE constexpr auto UnrollLoop(Fn fn) {
     fn(i);
   }
 }
+
+template <typename T>
+int32_t NativePopc(T v) {
+  int c = 0;
+  for (; v != 0; v &= v - 1) c++;
+  return c;
+}
+
+inline XGBOOST_DEVICE int Popc(uint32_t v) {
+#if defined(__CUDA_ARCH__)
+  return __popc(v);
+#elif defined(__GNUC__) || defined(__clang__)
+  return __builtin_popcount(v);
+#elif defined(_MSC_VER)
+  return __popcnt(v);
+#else
+  return NativePopc(v);
+#endif  // compiler
+}
+
+inline XGBOOST_DEVICE int Popc(uint64_t v) {
+#if defined(__CUDA_ARCH__)
+  return __popcll(v);
+#elif defined(__GNUC__) || defined(__clang__)
+  return __builtin_popcountll(v);
+#elif defined(_MSC_VER)
+  return __popcnt64(v);
+#else
+  return NativePopc(v);
+#endif  // compiler
+}
+
+template <class T, std::size_t N, std::size_t... Idx>
+constexpr auto Arr2Tup(T (&arr)[N], std::index_sequence<Idx...>) {
+  return std::make_tuple(arr[Idx]...);
+}
+
+template <class T, std::size_t N>
+constexpr auto Arr2Tup(T (&arr)[N]) {
+  return Arr2Tup(arr, std::make_index_sequence<N>{});
+}
+
+// uint division optimization inspired by the CIndexer in cupy.  Division operation is
+// slow on both CPU and GPU, especially 64 bit integer.  So here we first try to avoid 64
+// bit when the index is smaller, then try to avoid division when it's exp of 2.
+template <typename I, int32_t D>
+XGBOOST_DEVICE auto UnravelImpl(I idx, common::Span<size_t const, D> shape) {
+  size_t index[D]{0};
+  static_assert(std::is_signed<decltype(D)>::value,
+                "Don't change the type without changing the for loop.");
+  for (int32_t dim = D; --dim > 0;) {
+    auto s = static_cast<std::remove_const_t<std::remove_reference_t<I>>>(shape[dim]);
+    if (s & (s - 1)) {
+      auto t = idx / s;
+      index[dim] = idx - t * s;
+      idx = t;
+    } else {  // exp of 2
+      index[dim] = idx & (s - 1);
+      idx >>= Popc(s - 1);
+    }
+  }
+  index[0] = idx;
+  return Arr2Tup(index);
+}
+
+template <size_t dim, typename I, int32_t D>
+void ReshapeImpl(size_t (&out_shape)[D], I s) {
+  static_assert(dim < D, "");
+  out_shape[dim] = s;
+}
+
+template <size_t dim, int32_t D, typename... S, typename I,
+          std::enable_if_t<sizeof...(S) != 0> * = nullptr>
+void ReshapeImpl(size_t (&out_shape)[D], I &&s, S &&...rest) {
+  static_assert(dim < D, "");
+  out_shape[dim] = s;
+  ReshapeImpl<dim + 1>(out_shape, std::forward<S>(rest)...);
+}
+
+template <typename Fn, typename Tup, size_t... I>
+XGBOOST_DEVICE decltype(auto) constexpr Apply(Fn &&f, Tup &&t, std::index_sequence<I...>) {
+  return f(std::get<I>(t)...);
+}
+
+/**
+ * C++ 17 style apply.
+ *
+ * \param f function to apply
+ * \param t tuple of arguments
+ */
+template <typename Fn, typename Tup>
+XGBOOST_DEVICE decltype(auto) constexpr Apply(Fn &&f, Tup &&t) {
+  constexpr auto kSize = std::tuple_size<Tup>::value;
+  return Apply(std::forward<Fn>(f), std::forward<Tup>(t), std::make_index_sequence<kSize>{});
+}
 }  // namespace detail
 
 /**
@@ -85,6 +203,11 @@ constexpr detail::AllTag All() { return {}; }
  * much linear algebra routines, this class is a helper intended to ease some high level
  * operations like indexing into prediction tensor or gradient matrix.  It can be passed
  * into CUDA kernel as normal argument for GPU algorithms.
+ *
+ * Ideally we should add a template parameter `bool on_host` so that the compiler can
+ * prevent passing/accessing the wrong view, but inheritance is heavily used in XGBoost so
+ * some functions expect data types that can be used in everywhere (update prediction
+ * cache for example).
  */
 template <typename T, int32_t kDim = 5>
 class TensorView {
@@ -96,7 +219,7 @@ class TensorView {
   StrideT stride_{1};
   ShapeT shape_{0};
   common::Span<T> data_;
-  T* ptr_{nullptr};  // pointer of data_ to avoid bound check.
+  T *ptr_{nullptr};  // pointer of data_ to avoid bound check.
 
   size_t size_{0};
   int32_t device_{-1};
@@ -110,42 +233,45 @@ class TensorView {
     }
   }
 
-  struct SliceHelper {
+  template <size_t old_dim, size_t new_dim, int32_t D, typename... S>
-    size_t old_dim;
+  XGBOOST_DEVICE size_t MakeSliceDim(size_t new_shape[D], size_t new_stride[D],
-    size_t new_dim;
+                                     detail::AllTag) const {
-    size_t offset;
+    static_assert(new_dim < D, "");
-  };
+    static_assert(old_dim < kDim, "");
-
-  template <int32_t D, typename... S>
-  XGBOOST_DEVICE SliceHelper MakeSliceDim(size_t old_dim, size_t new_dim, size_t new_shape[D],
-                                          size_t new_stride[D], detail::AllTag) const {
     new_stride[new_dim] = stride_[old_dim];
     new_shape[new_dim] = shape_[old_dim];
-    return {old_dim + 1, new_dim + 1, 0};
+    return 0;
   }
-
+  /**
-  template <int32_t D, typename... S>
+   * \brief Slice dimension for All tag.
-  XGBOOST_DEVICE SliceHelper MakeSliceDim(size_t old_dim, size_t new_dim, size_t new_shape[D],
+   */
-                                          size_t new_stride[D], detail::AllTag,
+  template <size_t old_dim, size_t new_dim, int32_t D, typename... S>
-                                          S &&...slices) const {
+  XGBOOST_DEVICE size_t MakeSliceDim(size_t new_shape[D], size_t new_stride[D], detail::AllTag,
+                                     S &&...slices) const {
+    static_assert(new_dim < D, "");
+    static_assert(old_dim < kDim, "");
     new_stride[new_dim] = stride_[old_dim];
     new_shape[new_dim] = shape_[old_dim];
-    return MakeSliceDim<D>(old_dim + 1, new_dim + 1, new_shape, new_stride, slices...);
+    return MakeSliceDim<old_dim + 1, new_dim + 1, D>(new_shape, new_stride,
+                                                     std::forward<S>(slices)...);
   }
 
-  template <int32_t D, typename Index>
+  template <size_t old_dim, size_t new_dim, int32_t D, typename Index>
-  XGBOOST_DEVICE SliceHelper MakeSliceDim(size_t old_dim, size_t new_dim, size_t new_shape[D],
+  XGBOOST_DEVICE size_t MakeSliceDim(size_t new_shape[D], size_t new_stride[D], Index i) const {
-                                          size_t new_stride[D], Index i) const {
+    static_assert(old_dim < kDim, "");
-    return {old_dim + 1, new_dim, stride_[old_dim] * i};
+    return stride_[old_dim] * i;
   }
-
+  /**
-  template <int32_t D, typename Index, typename... S>
+   * \brief Slice dimension for Index tag.
-  XGBOOST_DEVICE std::enable_if_t<std::is_integral<Index>::value, SliceHelper> MakeSliceDim(
+   */
-      size_t old_dim, size_t new_dim, size_t new_shape[D], size_t new_stride[D], Index i,
+  template <size_t old_dim, size_t new_dim, int32_t D, typename Index, typename... S>
-      S &&...slices) const {
+  XGBOOST_DEVICE std::enable_if_t<std::is_integral<Index>::value, size_t> MakeSliceDim(
+      size_t new_shape[D], size_t new_stride[D], Index i, S &&...slices) const {
+    static_assert(old_dim < kDim, "");
     auto offset = stride_[old_dim] * i;
-    auto res = MakeSliceDim<D>(old_dim + 1, new_dim, new_shape, new_stride, slices...);
+    auto res =
-    return {res.old_dim, res.new_dim, res.offset + offset};
+        MakeSliceDim<old_dim + 1, new_dim, D>(new_shape, new_stride, std::forward<S>(slices)...);
+    return res + offset;
   }
 
  public:
@@ -174,12 +300,10 @@ class TensorView {
       shape_[i] = 1;
     }
     // stride
-    stride_[kDim - 1] = 1;
+    detail::CalcStride(shape_, stride_);
-    for (int32_t s = kDim - 2; s >= 0; --s) {
+    // size
-      stride_[s] = shape_[s + 1] * stride_[s + 1];
-    }
     this->CalcSize();
-  };
+  }
 
   /**
    * \brief Create a tensor with data, shape and strides.  Don't use this constructor if
@@ -195,7 +319,7 @@ class TensorView {
       stride_[i] = stride[i];
     });
     this->CalcSize();
-  };
+  }
 
   XGBOOST_DEVICE TensorView(TensorView const &that)
       : data_{that.data_}, ptr_{data_.data()}, size_{that.size_}, device_{that.device_} {
@@ -221,7 +345,8 @@ class TensorView {
   template <typename... Index>
   XGBOOST_DEVICE T &operator()(Index &&...index) {
     static_assert(sizeof...(index) <= kDim, "Invalid index.");
-    size_t offset = detail::Offset(stride_, 0ul, 0ul, index...);
+    size_t offset = detail::Offset<0ul>(stride_, 0ul, std::forward<Index>(index)...);
+    assert(offset < data_.size() && "Out of bound access.");
     return ptr_[offset];
   }
   /**
@@ -230,12 +355,14 @@ class TensorView {
   template <typename... Index>
   XGBOOST_DEVICE T const &operator()(Index &&...index) const {
     static_assert(sizeof...(index) <= kDim, "Invalid index.");
-    size_t offset = detail::Offset(stride_, 0ul, 0ul, index...);
+    size_t offset = detail::Offset<0ul>(stride_, 0ul, std::forward<Index>(index)...);
+    assert(offset < data_.size() && "Out of bound access.");
     return ptr_[offset];
   }
 
   /**
-   * \brief Slice the tensor.  The returned tensor has inferred dim and shape.
+   * \brief Slice the tensor.  The returned tensor has inferred dim and shape.  Scalar
+   *        result is not supported.
    *
    * \code
    *
@@ -252,10 +379,10 @@ class TensorView {
     int32_t constexpr kNewDim{detail::CalcSliceDim<detail::IndexToTag<S>...>()};
     size_t new_shape[kNewDim];
     size_t new_stride[kNewDim];
-    auto res = MakeSliceDim<kNewDim>(size_t(0), size_t(0), new_shape, new_stride, slices...);
+    auto offset = MakeSliceDim<0, 0, kNewDim>(new_shape, new_stride, std::forward<S>(slices)...);
     // ret is a different type due to changed dimension, so we can not access its private
     // fields.
-    TensorView<T, kNewDim> ret{data_.subspan(data_.empty() ? 0 : res.offset), new_shape, new_stride,
+    TensorView<T, kNewDim> ret{data_.subspan(data_.empty() ? 0 : offset), new_shape, new_stride,
                                device_};
     return ret;
   }
@@ -275,12 +402,91 @@ class TensorView {
   XGBOOST_DEVICE auto cend() const { return data_.cend(); }      // NOLINT
   XGBOOST_DEVICE auto begin() { return data_.begin(); }          // NOLINT
   XGBOOST_DEVICE auto end() { return data_.end(); }              // NOLINT
-
+  /**
+   * \brief Number of items in the tensor.
+   */
   XGBOOST_DEVICE size_t Size() const { return size_; }
+  /**
+   * \brief Whether it's a contiguous array. (c and f contiguous are both contiguous)
+   */
+  XGBOOST_DEVICE bool Contiguous() const { return size_ == data_.size(); }
+  /**
+   * \brief Obtain the raw data.
+   */
   XGBOOST_DEVICE auto Values() const { return data_; }
+  /**
+   * \brief Obtain the CUDA device ordinal.
+   */
   XGBOOST_DEVICE auto DeviceIdx() const { return device_; }
+
+  /**
+   * \brief Array Interface defined by
+   * <a href="https://numpy.org/doc/stable/reference/arrays.interface.html">numpy</a>.
+   *
+   * `stream` is optionally included when data is on CUDA device.
+   */
+  Json ArrayInterface() const {
+    Json array_interface{Object{}};
+    array_interface["data"] = std::vector<Json>(2);
+    array_interface["data"][0] = Integer(reinterpret_cast<int64_t>(data_.data()));
+    array_interface["data"][1] = Boolean{true};
+    if (this->DeviceIdx() >= 0) {
+      // Change this once we have different CUDA stream.
+      array_interface["stream"] = Null{};
+    }
+    std::vector<Json> shape(Shape().size());
+    std::vector<Json> stride(Stride().size());
+    for (size_t i = 0; i < Shape().size(); ++i) {
+      shape[i] = Integer(Shape(i));
+      stride[i] = Integer(Stride(i) * sizeof(T));
+    }
+    array_interface["shape"] = Array{shape};
+    array_interface["strides"] = Array{stride};
+    array_interface["version"] = 3;
+
+    char constexpr kT = detail::ArrayInterfaceHandler::TypeChar<T>();
+    static_assert(kT != '\0', "");
+    if (DMLC_LITTLE_ENDIAN) {
+      array_interface["typestr"] = String{"<" + (kT + std::to_string(sizeof(T)))};
+    } else {
+      array_interface["typestr"] = String{">" + (kT + std::to_string(sizeof(T)))};
+    }
+    return array_interface;
+  }
+  /**
+   * \brief Same as const version, but returns non-readonly data pointer.
+   */
+  Json ArrayInterface() {
+    auto const &as_const = *this;
+    auto res = as_const.ArrayInterface();
+    res["data"][1] = Boolean{false};
+    return res;
+  }
+
+  auto ArrayInterfaceStr() const {
+    std::string str;
+    Json::Dump(this->ArrayInterface(), &str);
+    return str;
+  }
+  auto ArrayInterfaceStr() {
+    std::string str;
+    Json::Dump(this->ArrayInterface(), &str);
+    return str;
+  }
 };
 
+/**
+ * \brief Turns linear index into multi-dimension index.  Similar to numpy unravel.
+ */
+template <size_t D>
+XGBOOST_DEVICE auto UnravelIndex(size_t idx, common::Span<size_t const, D> shape) {
+  if (idx > std::numeric_limits<uint32_t>::max()) {
+    return detail::UnravelImpl<uint64_t, D>(static_cast<uint64_t>(idx), shape);
+  } else {
+    return detail::UnravelImpl<uint32_t, D>(static_cast<uint32_t>(idx), shape);
+  }
+}
+
 /**
  * \brief A view over a vector, specialization of Tensor
  *
@@ -289,6 +495,19 @@ class TensorView {
 template <typename T>
 using VectorView = TensorView<T, 1>;
 
+/**
+ * \brief Create a vector view from contigious memory.
+ *
+ * \param ptr Pointer to the contigious memory.
+ * \param s   Size of the vector.
+ * \param device (optional) Device ordinal, default to be host.
+ */
+template <typename T>
+auto MakeVec(T *ptr, size_t s, int32_t device = -1) {
+  using U = std::remove_const_t<std::remove_pointer_t<decltype(ptr)>> const;
+  return linalg::TensorView<U, 1>{{ptr, s}, {s}, device};
+}
+
 /**
  * \brief A view over a matrix, specialization of Tensor.
  *
@@ -296,6 +515,163 @@ using VectorView = TensorView<T, 1>;
  */
 template <typename T>
 using MatrixView = TensorView<T, 2>;
+
+/**
+ * \brief A tensor storage. To use it for other functionality like slicing one needs to
+ *        obtain a view first.  This way we can use it on both host and device.
+ */
+template <typename T, int32_t kDim = 5>
+class Tensor {
+ public:
+  using ShapeT = size_t[kDim];
+  using StrideT = ShapeT;
+
+ private:
+  HostDeviceVector<T> data_;
+  ShapeT shape_{0};
+
+ public:
+  Tensor() = default;
+
+  /**
+   * \brief Create a tensor with shape and device ordinal.  The storage is initialized
+   *        automatically.
+   *
+   * See \ref TensorView for parameters of this constructor.
+   */
+  template <typename I, int32_t D>
+  explicit Tensor(I const (&shape)[D], int32_t device) {
+    // No device unroll as this is a host only function.
+    std::copy(shape, shape + D, shape_);
+    for (auto i = D; i < kDim; ++i) {
+      shape_[i] = 1;
+    }
+    auto size = detail::CalcSize(shape_);
+    if (device >= 0) {
+      data_.SetDevice(device);
+    }
+    data_.Resize(size);
+    if (device >= 0) {
+      data_.DevicePointer();  // Pull to device
+    }
+  }
+  /**
+   * Initialize from 2 host iterators.
+   */
+  template <typename It, typename I, int32_t D>
+  explicit Tensor(It begin, It end, I const (&shape)[D], int32_t device) {
+    // shape
+    static_assert(D <= kDim, "Invalid shape.");
+    std::copy(shape, shape + D, shape_);
+    for (auto i = D; i < kDim; ++i) {
+      shape_[i] = 1;
+    }
+    auto &h_vec = data_.HostVector();
+    h_vec.insert(h_vec.begin(), begin, end);
+    if (device >= 0) {
+      data_.SetDevice(device);
+      data_.DevicePointer();  // Pull to device;
+    }
+    CHECK_EQ(data_.Size(), detail::CalcSize(shape_));
+  }
+  /**
+   * \brief Get a \ref TensorView for this tensor.
+   */
+  TensorView<T, kDim> View(int32_t device) {
+    if (device >= 0) {
+      data_.SetDevice(device);
+      auto span = data_.DeviceSpan();
+      return {span, shape_, device};
+    } else {
+      auto span = data_.HostSpan();
+      return {span, shape_, device};
+    }
+  }
+  TensorView<T const, kDim> View(int32_t device) const {
+    if (device >= 0) {
+      data_.SetDevice(device);
+      auto span = data_.ConstDeviceSpan();
+      return {span, shape_, device};
+    } else {
+      auto span = data_.ConstHostSpan();
+      return {span, shape_, device};
+    }
+  }
+
+  size_t Size() const { return data_.Size(); }
+  auto Shape() const { return common::Span<size_t const, kDim>{shape_}; }
+  auto Shape(size_t i) const { return shape_[i]; }
+
+  HostDeviceVector<T> *Data() { return &data_; }
+  HostDeviceVector<T> const *Data() const { return &data_; }
+
+  /**
+   * \brief Visitor function for modification that changes shape and data.
+   *
+   * \tparam Fn function that takes a pointer to `HostDeviceVector` and a static sized
+   *         span as parameters.
+   */
+  template <typename Fn>
+  void ModifyInplace(Fn &&fn) {
+    fn(this->Data(), common::Span<size_t, kDim>{this->shape_});
+    CHECK_EQ(this->Data()->Size(), detail::CalcSize(this->shape_))
+        << "Inconsistent size after modification.";
+  }
+
+  /**
+   * \brief Reshape the tensor.
+   *
+   *    If the total size is changed, then data in this tensor is no longer valid.
+   */
+  template <typename... S>
+  void Reshape(S &&...s) {
+    static_assert(sizeof...(S) <= kDim, "Invalid shape.");
+    detail::ReshapeImpl<0>(shape_, std::forward<S>(s)...);
+    auto constexpr kEnd = sizeof...(S);
+    static_assert(kEnd <= kDim, "Invalid shape.");
+    std::fill(shape_ + kEnd, shape_ + kDim, 1);
+    auto n = detail::CalcSize(shape_);
+    data_.Resize(n);
+  }
+
+  /**
+   * \brief Reshape the tensor.
+   *
+   *    If the total size is changed, then data in this tensor is no longer valid.
+   */
+  template <int32_t D>
+  void Reshape(size_t (&shape)[D]) {
+    static_assert(D <= kDim, "Invalid shape.");
+    std::copy(shape, shape + D, this->shape_);
+    std::fill(shape_ + D, shape_ + kDim, 1);
+    auto n = detail::CalcSize(shape_);
+    data_.Resize(n);
+  }
+
+  /**
+   * \brief Set device ordinal for this tensor.
+   */
+  void SetDevice(int32_t device) { data_.SetDevice(device); }
+};
+
+// Only first axis is supported for now.
+template <typename T, int32_t D>
+void Stack(Tensor<T, D> *l, Tensor<T, D> const &r) {
+  if (r.Data()->DeviceIdx() >= 0) {
+    l->Data()->SetDevice(r.Data()->DeviceIdx());
+  }
+  l->ModifyInplace([&](HostDeviceVector<T> *data, common::Span<size_t, D> shape) {
+    for (size_t i = 1; i < D; ++i) {
+      if (shape[i] == 0) {
+        shape[i] = r.Shape(i);
+      } else {
+        CHECK_EQ(shape[i], r.Shape(i));
+      }
+    }
+    data->Extend(*r.Data());
+    shape[0] = l->Shape(0) + r.Shape(0);
+  });
+}
 }  // namespace linalg
 }  // namespace xgboost
 #endif  // XGBOOST_LINALG_H_

src/common/host_device_vector.cc

@@ -170,6 +170,7 @@ void HostDeviceVector<T>::SetDevice(int) const {}
 
 // explicit instantiations are required, as HostDeviceVector isn't header-only
 template class HostDeviceVector<bst_float>;
+template class HostDeviceVector<double>;
 template class HostDeviceVector<GradientPair>;
 template class HostDeviceVector<int32_t>;   // bst_node_t
 template class HostDeviceVector<uint8_t>;

src/common/host_device_vector.cu

@@ -398,6 +398,7 @@ void HostDeviceVector<T>::Resize(size_t new_size, T v) {
 
 // explicit instantiations are required, as HostDeviceVector isn't header-only
 template class HostDeviceVector<bst_float>;
+template class HostDeviceVector<double>;
 template class HostDeviceVector<GradientPair>;
 template class HostDeviceVector<int32_t>;   // bst_node_t
 template class HostDeviceVector<uint8_t>;

src/common/linalg_op.cuh

@@ -0,0 +1,25 @@
+/*!
+ * Copyright 2021 by XGBoost Contributors
+ */
+#ifndef XGBOOST_COMMON_LINALG_OP_CUH_
+#define XGBOOST_COMMON_LINALG_OP_CUH_
+#include "device_helpers.cuh"
+#include "xgboost/linalg.h"
+
+namespace xgboost {
+namespace linalg {
+template <typename T, int32_t D, typename Fn>
+void ElementWiseKernelDevice(linalg::TensorView<T, D> t, Fn&& fn, cudaStream_t s = nullptr) {
+  if (t.Contiguous()) {
+    auto ptr = t.Values().data();
+    dh::LaunchN(t.Size(), s, [=] __device__(size_t i) { ptr[i] = fn(i, ptr[i]); });
+  } else {
+    dh::LaunchN(t.Size(), s, [=] __device__(size_t i) mutable {
+      T& v = detail::Apply(t, linalg::UnravelIndex(i, t.Shape()));
+      v = fn(i, v);
+    });
+  }
+}
+}  // namespace linalg
+}  // namespace xgboost
+#endif  // XGBOOST_COMMON_LINALG_OP_CUH_

src/common/linalg_op.h

@@ -0,0 +1,25 @@
+/*!
+ * Copyright 2021 by XGBoost Contributors
+ */
+#ifndef XGBOOST_COMMON_LINALG_OP_H_
+#define XGBOOST_COMMON_LINALG_OP_H_
+#include "threading_utils.h"
+#include "xgboost/linalg.h"
+
+namespace xgboost {
+namespace linalg {
+template <typename T, int32_t D, typename Fn>
+void ElementWiseKernelHost(linalg::TensorView<T, D> t, int32_t n_threads, Fn&& fn) {
+  if (t.Contiguous()) {
+    auto ptr = t.Values().data();
+    common::ParallelFor(t.Size(), n_threads, [&](size_t i) { ptr[i] = fn(i, ptr[i]); });
+  } else {
+    common::ParallelFor(t.Size(), n_threads, [&](size_t i) {
+      auto& v = detail::Apply(t, linalg::UnravelIndex(i, t.Shape()));
+      v = fn(i, v);
+    });
+  }
+}
+}  // namespace linalg
+}  // namespace xgboost
+#endif  // XGBOOST_COMMON_LINALG_OP_H_

tests/cpp/common/test_linalg.cc

@@ -1,11 +1,21 @@
+/*!
+ * Copyright 2021 by XGBoost Contributors
+ */
 #include <gtest/gtest.h>
+#include <xgboost/generic_parameters.h>
 #include <xgboost/host_device_vector.h>
 #include <xgboost/linalg.h>
 
 #include <numeric>
 
+#include "../../../src/common/linalg_op.h"
+
 namespace xgboost {
 namespace linalg {
+namespace {
+auto kCpuId = GenericParameter::kCpuId;
+}
+
 auto MakeMatrixFromTest(HostDeviceVector<float> *storage, size_t n_rows, size_t n_cols) {
   storage->Resize(n_rows * n_cols);
   auto &h_storage = storage->HostVector();
@@ -16,16 +26,16 @@ auto MakeMatrixFromTest(HostDeviceVector<float> *storage, size_t n_rows, size_t
   return m;
 }
 
-TEST(Linalg, Matrix) {
+TEST(Linalg, MatrixView) {
   size_t kRows = 31, kCols = 77;
   HostDeviceVector<float> storage;
   auto m = MakeMatrixFromTest(&storage, kRows, kCols);
-  ASSERT_EQ(m.DeviceIdx(), GenericParameter::kCpuId);
+  ASSERT_EQ(m.DeviceIdx(), kCpuId);
   ASSERT_EQ(m(0, 0), 0);
   ASSERT_EQ(m(kRows - 1, kCols - 1), storage.Size() - 1);
 }
 
-TEST(Linalg, Vector) {
+TEST(Linalg, VectorView) {
   size_t kRows = 31, kCols = 77;
   HostDeviceVector<float> storage;
   auto m = MakeMatrixFromTest(&storage, kRows, kCols);
@@ -37,7 +47,7 @@ TEST(Linalg, Vector) {
   ASSERT_EQ(v(0), 3);
 }
 
-TEST(Linalg, Tensor) {
+TEST(Linalg, TensorView) {
   std::vector<double> data(2 * 3 * 4, 0);
   std::iota(data.begin(), data.end(), 0);
 
@@ -99,14 +109,123 @@ TEST(Linalg, Tensor) {
   }
 }
 
-TEST(Linalg, Empty) {
+TEST(Linalg, Tensor) {
-  auto t = TensorView<double, 2>{{}, {0, 3}, GenericParameter::kCpuId};
+  {
-  for (int32_t i : {0, 1, 2}) {
+    Tensor<float, 3> t{{2, 3, 4}, kCpuId};
-    auto s = t.Slice(All(), i);
+    auto view = t.View(kCpuId);
-    ASSERT_EQ(s.Size(), 0);
+
-    ASSERT_EQ(s.Shape().size(), 1);
+    auto const &as_const = t;
-    ASSERT_EQ(s.Shape(0), 0);
+    auto k_view = as_const.View(kCpuId);
+
+    size_t n = 2 * 3 * 4;
+    ASSERT_EQ(t.Size(), n);
+    ASSERT_TRUE(std::equal(k_view.cbegin(), k_view.cbegin(), view.begin()));
+
+    Tensor<float, 3> t_0{std::move(t)};
+    ASSERT_EQ(t_0.Size(), n);
+    ASSERT_EQ(t_0.Shape(0), 2);
+    ASSERT_EQ(t_0.Shape(1), 3);
+    ASSERT_EQ(t_0.Shape(2), 4);
   }
+  {
+    // Reshape
+    Tensor<float, 3> t{{2, 3, 4}, kCpuId};
+    t.Reshape(4, 3, 2);
+    ASSERT_EQ(t.Size(), 24);
+    ASSERT_EQ(t.Shape(2), 2);
+    t.Reshape(1);
+    ASSERT_EQ(t.Size(), 1);
+    t.Reshape(0, 0, 0);
+    ASSERT_EQ(t.Size(), 0);
+    t.Reshape(0, 3, 0);
+    ASSERT_EQ(t.Size(), 0);
+    ASSERT_EQ(t.Shape(1), 3);
+    t.Reshape(3, 3, 3);
+    ASSERT_EQ(t.Size(), 27);
+  }
+}
+
+TEST(Linalg, Empty) {
+  {
+    auto t = TensorView<double, 2>{{}, {0, 3}, kCpuId};
+    for (int32_t i : {0, 1, 2}) {
+      auto s = t.Slice(All(), i);
+      ASSERT_EQ(s.Size(), 0);
+      ASSERT_EQ(s.Shape().size(), 1);
+      ASSERT_EQ(s.Shape(0), 0);
+    }
+  }
+  {
+    auto t = Tensor<double, 2>{{0, 3}, kCpuId};
+    ASSERT_EQ(t.Size(), 0);
+    auto view = t.View(kCpuId);
+
+    for (int32_t i : {0, 1, 2}) {
+      auto s = view.Slice(All(), i);
+      ASSERT_EQ(s.Size(), 0);
+      ASSERT_EQ(s.Shape().size(), 1);
+      ASSERT_EQ(s.Shape(0), 0);
+    }
+  }
+}
+
+TEST(Linalg, ArrayInterface) {
+  auto cpu = kCpuId;
+  auto t = Tensor<double, 2>{{3, 3}, cpu};
+  auto v = t.View(cpu);
+  std::iota(v.begin(), v.end(), 0);
+  auto arr = Json::Load(StringView{v.ArrayInterfaceStr()});
+  ASSERT_EQ(get<Integer>(arr["shape"][0]), 3);
+  ASSERT_EQ(get<Integer>(arr["strides"][0]), 3 * sizeof(double));
+
+  ASSERT_FALSE(get<Boolean>(arr["data"][1]));
+  ASSERT_EQ(reinterpret_cast<double *>(get<Integer>(arr["data"][0])), v.Values().data());
+}
+
+TEST(Linalg, Popc) {
+  {
+    uint32_t v{0};
+    ASSERT_EQ(detail::NativePopc(v), 0);
+    ASSERT_EQ(detail::Popc(v), 0);
+    v = 1;
+    ASSERT_EQ(detail::NativePopc(v), 1);
+    ASSERT_EQ(detail::Popc(v), 1);
+    v = 0xffffffff;
+    ASSERT_EQ(detail::NativePopc(v), 32);
+    ASSERT_EQ(detail::Popc(v), 32);
+  }
+  {
+    uint64_t v{0};
+    ASSERT_EQ(detail::NativePopc(v), 0);
+    ASSERT_EQ(detail::Popc(v), 0);
+    v = 1;
+    ASSERT_EQ(detail::NativePopc(v), 1);
+    ASSERT_EQ(detail::Popc(v), 1);
+    v = 0xffffffff;
+    ASSERT_EQ(detail::NativePopc(v), 32);
+    ASSERT_EQ(detail::Popc(v), 32);
+    v = 0xffffffffffffffff;
+    ASSERT_EQ(detail::NativePopc(v), 64);
+    ASSERT_EQ(detail::Popc(v), 64);
+  }
+}
+
+TEST(Linalg, Stack) {
+  Tensor<float, 3> l{{2, 3, 4}, kCpuId};
+  ElementWiseKernelHost(l.View(kCpuId), omp_get_max_threads(),
+                        [=](size_t i, float v) { return i; });
+  Tensor<float, 3> r_0{{2, 3, 4}, kCpuId};
+  ElementWiseKernelHost(r_0.View(kCpuId), omp_get_max_threads(),
+                        [=](size_t i, float v) { return i; });
+
+  Stack(&l, r_0);
+
+  Tensor<float, 3> r_1{{0, 3, 4}, kCpuId};
+  Stack(&l, r_1);
+  ASSERT_EQ(l.Shape(0), 4);
+
+  Stack(&r_1, l);
+  ASSERT_EQ(r_1.Shape(0), l.Shape(0));
 }
 }  // namespace linalg
 }  // namespace xgboost

tests/cpp/common/test_linalg.cu

@@ -0,0 +1,62 @@
+/*!
+ * Copyright 2021 by XGBoost Contributors
+ */
+#include <gtest/gtest.h>
+
+#include "../../../src/common/linalg_op.cuh"
+#include "xgboost/generic_parameters.h"
+#include "xgboost/linalg.h"
+
+namespace xgboost {
+namespace linalg {
+namespace {
+void TestElementWiseKernel() {
+  Tensor<float, 3> l{{2, 3, 4}, 0};
+  {
+    /**
+     * Non-contiguous
+     */
+    // GPU view
+    auto t = l.View(0).Slice(linalg::All(), 1, linalg::All());
+    ASSERT_FALSE(t.Contiguous());
+    ElementWiseKernelDevice(t, [] __device__(size_t i, float) { return i; });
+    // CPU view
+    t = l.View(GenericParameter::kCpuId).Slice(linalg::All(), 1, linalg::All());
+    size_t k = 0;
+    for (size_t i = 0; i < l.Shape(0); ++i) {
+      for (size_t j = 0; j < l.Shape(2); ++j) {
+        ASSERT_EQ(k++, t(i, j));
+      }
+    }
+
+    t = l.View(0).Slice(linalg::All(), 1, linalg::All());
+    ElementWiseKernelDevice(t, [] __device__(size_t i, float v) {
+      SPAN_CHECK(v == i);
+      return v;
+    });
+  }
+
+  {
+    /**
+     * Contiguous
+     */
+    auto t = l.View(0);
+    ElementWiseKernelDevice(t, [] __device__(size_t i, float) { return i; });
+    ASSERT_TRUE(t.Contiguous());
+    // CPU view
+    t = l.View(GenericParameter::kCpuId);
+
+    size_t ind = 0;
+    for (size_t i = 0; i < l.Shape(0); ++i) {
+      for (size_t j = 0; j < l.Shape(1); ++j) {
+        for (size_t k = 0; k < l.Shape(2); ++k) {
+          ASSERT_EQ(ind++, t(i, j, k));
+        }
+      }
+    }
+  }
+}
+}  // anonymous namespace
+TEST(Linalg, GPUElementWise) { TestElementWiseKernel(); }
+}  // namespace linalg
+}  // namespace xgboost