Remove thread-synchronization from arrays, which was way too slow, and

only really needed for dmnsn_progress anyway.

7 files changed, 108 insertions, 223 deletions

diff --git a/doc/libdimension.texi b/doc/libdimension.texi
index 2329feb..371a5dd 100644
--- a/doc/libdimension.texi
+++ b/doc/libdimension.texi
@@ -121,35 +121,19 @@ void dmnsn_array_pop(dmnsn_array *array, void *obj);
 void dmnsn_array_get(const dmnsn_array *array, size_t i, void *obj);
 @findex dmnsn_array_set()
 void dmnsn_array_set(dmnsn_array *array, size_t i, const void *obj);
+@findex dmnsn_array_at()
+void *dmnsn_array_at(dmnsn_array *array, size_t i);
 
 @findex dmnsn_array_size()
 size_t dmnsn_array_size(const dmnsn_array *array);
 @findex dmnsn_array_resize()
 void   dmnsn_array_resize(dmnsn_array *array, size_t length);
-
-/* Non-atomic operations for manual locking */
-@findex dmnsn_array_at()
-void *dmnsn_array_at(dmnsn_array *array, size_t i);
-@findex dmnsn_array_size_unlocked()
-size_t dmnsn_array_size_unlocked(const dmnsn_array *array);
-@findex dmnsn_array_resize_unlocked()
-void dmnsn_array_resize_unlocked(dmnsn_array *array, size_t length);
-
-/* Manual locking */
-@findex dmnsn_array_rdlock()
-void dmnsn_array_rdlock(const dmnsn_array *array);
-@findex dmnsn_array_wrlock()
-void dmnsn_array_wrlock(dmnsn_array *array);
-@findex dmnsn_array_unlock()
-void dmnsn_array_unlock(const dmnsn_array *array);
 @end example
 
 @cindex array
 The Dimension Library often has cause to work with adjustable-size arrays.  It provides an interface for dynamically-allocated arrays of arbitrary objects.  Arrays are allocated with the @code{dmnsn_new_array()} function, and freed with the @code{dmnsn_delete_array()} function, a pattern common in libdimension.  @code{dmnsn_new_array()} takes the size of the new array's elements as its parameter.  Unlike other allocation functions throughout libdimension, @code{dmnsn_new_array()} cannot fail if it returns (other allocators may return NULL).  In fact, all array operations are guaranteed to either succeed or report a fatal error, which may happen if memory allocation fails.
 
-Arrays in libdimension are thread-safe; every individual operation is atomic.  libdimension provides push, pop, get, and set operations for arrays, taking a pointer to an object to read or write as their last parameter, the array index when applicable as the second-last parameter, and the @code{dmnsn_array *} as the first parameter.  The array's length may be queried with @code{dmnsn_array_size()}, or set with @code{dmnsn_array_resize()}.
-
-When sets of array operations must be atomic, one may use the manual locking array interface.  Implemented as a read-write lock, applications should call @code{dmnsn_array_rdlock()} to set a read-lock, @code{dmnsn_array_wrlock()} to set a write-lock, and @code{dmnsn_array_unlock()} to unlock the array.  While the array is locked, one may call @code{dmnsn_array_at()} to get a pointer to the @code{i}'th element of the array, or the @code{*_unlocked} suffixed versions of the @code{..._size()} and @code{..._resize()} functions to get or set the size, respectively.
+Arrays support the push, pop, get, and set operations for arrays, taking a pointer to an object to read or write as their last parameter, the array index when applicable as the second-last parameter, and the @code{dmnsn_array *} as the first parameter.  The get operation is bounds-checked; other operations dynamically resize the array as needed.  The array's length may be queried with @code{dmnsn_array_size()}, or set with @code{dmnsn_array_resize()}, and a pointer to the @code{i}'th object may be obtained with @code{dmnsn_array_at()}.
 
 
 @node Asynchronicity
diff --git a/libdimension/array.c b/libdimension/array.c
index 82588aa..a22b450 100644
--- a/libdimension/array.c
+++ b/libdimension/array.c
@@ -22,10 +22,6 @@
 #include <pthread.h>
 #include <string.h> /* For memcpy */
 
-/* The raw implementations, which don't do any thread synchronicity */
-static void dmnsn_array_get_impl(const dmnsn_array *array, size_t i, void *obj);
-static void dmnsn_array_set_impl(dmnsn_array *array, size_t i, const void *obj);
-
 /* Allocate a new array - guaranteed not to fail if it returns */
 dmnsn_array *
 dmnsn_new_array(size_t obj_size)
@@ -41,12 +37,6 @@ dmnsn_new_array(size_t obj_size)
     if (!array->ptr) {
       dmnsn_error(DMNSN_SEVERITY_HIGH, "Array allocation failed.");
     }
-
-    /* Allocate the read-write lock */
-    array->rwlock = malloc(sizeof(pthread_rwlock_t));
-    if (!array->rwlock || pthread_rwlock_init(array->rwlock, NULL) != 0) {
-      dmnsn_error(DMNSN_SEVERITY_HIGH, "Array rwlock allocation failed.");
-    }
   }
 
   return array;
@@ -55,98 +45,70 @@ dmnsn_new_array(size_t obj_size)
 /* Delete the array */
 void dmnsn_delete_array(dmnsn_array *array) {
   if (array) {
-    if (pthread_rwlock_destroy(array->rwlock) != 0) {
-      dmnsn_error(DMNSN_SEVERITY_LOW, "Leaking rwlock in array deallocation.");
-    }
-
-    free(array->rwlock);
     free(array->ptr);
     free(array);
   }
 }
 
-/* Push obj to the end of the array, atomically */
+/* Push obj to the end of the array */
 void
 dmnsn_array_push(dmnsn_array *array, const void *obj)
 {
-  dmnsn_array_wrlock(array);
-    dmnsn_array_set_impl(array, dmnsn_array_size_unlocked(array), obj);
-  dmnsn_array_unlock(array);
+  dmnsn_array_set(array, dmnsn_array_size(array), obj);
 }
 
-/* Pop obj from the end of the array, atomically */
+/* Pop obj from the end of the array */
 void
 dmnsn_array_pop(dmnsn_array *array, void *obj)
 {
-  size_t size;
-
-  dmnsn_array_wrlock(array);
-    size = dmnsn_array_size_unlocked(array);
-    dmnsn_array_get_impl(array, size - 1, obj);   /* Copy the object */
-    dmnsn_array_resize_unlocked(array, size - 1); /* Shrink the array */
-  dmnsn_array_unlock(array);
+  size_t size = dmnsn_array_size(array);
+  dmnsn_array_get(array, size - 1, obj); /* Copy the object */
+  dmnsn_array_resize(array, size - 1);   /* Shrink the array */
 }
 
 /* Get the i'th object, bailing out if i is out of range */
 void
 dmnsn_array_get(const dmnsn_array *array, size_t i, void *obj)
 {
-  dmnsn_array_rdlock(array);
-    dmnsn_array_get_impl(array, i, obj);
-  dmnsn_array_unlock(array);
+  if (i >= dmnsn_array_size(array)) {
+    /* Range check failed */
+    dmnsn_error(DMNSN_SEVERITY_HIGH, "Array index out of bounds.");
+  }
+  memcpy(obj, (char *)array->ptr + array->obj_size*i, array->obj_size);
 }
 
 /* Set the i'th object, expanding the array if necessary */
 void
 dmnsn_array_set(dmnsn_array *array, size_t i, const void *obj)
 {
-  dmnsn_array_wrlock(array);
-    dmnsn_array_set_impl(array, i, obj);
-  dmnsn_array_unlock(array);
-}
-
-/* Get the size of the array, atomically */
-size_t
-dmnsn_array_size(const dmnsn_array *array)
-{
-  size_t size;
-
-  dmnsn_array_rdlock(array);
-    size = dmnsn_array_size_unlocked(array);
-  dmnsn_array_unlock(array);
-
-  return size;
-}
-
-/* Set the size of the array, atomically */
-void
-dmnsn_array_resize(dmnsn_array *array, size_t length)
-{
-  dmnsn_array_wrlock(array);
-    dmnsn_array_resize_unlocked(array, length);
-  dmnsn_array_unlock(array);
+  if (i >= dmnsn_array_size(array)) {
+    /* Resize if i is out of range */
+    dmnsn_array_resize(array, i + 1);
+  }
+  memcpy((char *)array->ptr + array->obj_size*i, obj, array->obj_size);
 }
 
-/* Thread-unsafe range-checked element access */
+/* Element access */
 void *
 dmnsn_array_at(dmnsn_array *array, size_t i)
 {
-  if (i >= dmnsn_array_size_unlocked(array)) {
-    dmnsn_error(DMNSN_SEVERITY_HIGH, "Array index out of bounds.");
+  if (i >= dmnsn_array_size(array)) {
+    /* Resize if i is out of range */
+    dmnsn_array_resize(array, i + 1);
   }
   return (char *)array->ptr + array->obj_size*i;
 }
 
-/* Get the size non-atomically, for manual locking */
+/* Get the size of the array */
 size_t
-dmnsn_array_size_unlocked(const dmnsn_array *array)
+dmnsn_array_size(const dmnsn_array *array)
 {
   return array->length;
 }
 
-/* Set the size non-atomically, for manual locking */
+/* Set the size of the array, atomically */
 void
-dmnsn_array_resize_unlocked(dmnsn_array *array, size_t length)
+dmnsn_array_resize(dmnsn_array *array, size_t length)
 {
   if (length > array->capacity) {
     /* Resize if we don't have enough capacity */
@@ -159,60 +121,3 @@ dmnsn_array_resize_unlocked(dmnsn_array *array, size_t length)
 
   array->length = length;
 }
-
-/* Set a manual read-lock */
-void
-dmnsn_array_rdlock(const dmnsn_array *array)
-{
-  if (pthread_rwlock_rdlock(array->rwlock) != 0) {
-    /* Medium severity, because undefined behaviour is pretty likely if our
-       reads and writes aren't synced */
-    dmnsn_error(DMNSN_SEVERITY_MEDIUM,
-                "Couldn't acquire read-lock for array.");
-  }
-}
-
-/* Set a manual write-lock */
-void
-dmnsn_array_wrlock(dmnsn_array *array)
-{
-  if (pthread_rwlock_wrlock(array->rwlock) != 0) {
-    /* Medium severity, because undefined behaviour is pretty likely if our
-       reads and writes aren't synced */
-    dmnsn_error(DMNSN_SEVERITY_MEDIUM,
-                "Couldn't acquire write-lock for array.");
-  }
-}
-
-/* Unset a manual lock */
-void
-dmnsn_array_unlock(const dmnsn_array *array)
-{
-  if (pthread_rwlock_unlock(array->rwlock) != 0) {
-    /* Medium severity, because if the array is locked, we're likely to hang
-       the next time we try to read or write it */
-    dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't unlock array.");
-  }
-}
-
-/* Actual "get" implementation */
-static void
-dmnsn_array_get_impl(const dmnsn_array *array, size_t i, void *obj)
-{
-  if (i >= dmnsn_array_size_unlocked(array)) {
-    /* Range check failed */
-    dmnsn_error(DMNSN_SEVERITY_HIGH, "Array index out of bounds.");
-  }
-  memcpy(obj, array->ptr + array->obj_size*i, array->obj_size);
-}
-
-/* Actual "set" implementation */
-static void
-dmnsn_array_set_impl(dmnsn_array *array, size_t i, const void *obj)
-{
-  if (i >= dmnsn_array_size_unlocked(array)) {
-    /* Resize if i is out of range */
-    dmnsn_array_resize_unlocked(array, i + 1);
-  }
-  memcpy(array->ptr + array->obj_size*i, obj, array->obj_size);
-}
diff --git a/libdimension/dimension/array.h b/libdimension/dimension/array.h
index 1c8f601..5d5e50e 100644
--- a/libdimension/dimension/array.h
+++ b/libdimension/dimension/array.h
@@ -32,9 +32,6 @@
 typedef struct {
   void *ptr;
   size_t obj_size, length, capacity;
-
-  /* Synchronicity control (pointer so it's not const) */
-  pthread_rwlock_t *rwlock;
 } dmnsn_array;
 
 /* Array allocation never returns NULL - if dmnsn_new_array returns, it
@@ -42,24 +39,13 @@ typedef struct {
 dmnsn_array *dmnsn_new_array(size_t obj_size);
 void dmnsn_delete_array(dmnsn_array *array);
 
-/* Thread-safe atomic array access */
-
 void dmnsn_array_push(dmnsn_array *array, const void *obj);
 void dmnsn_array_pop(dmnsn_array *array, void *obj);
 void dmnsn_array_get(const dmnsn_array *array, size_t i, void *obj);
 void dmnsn_array_set(dmnsn_array *array, size_t i, const void *obj);
+void *dmnsn_array_at(dmnsn_array *array, size_t i);
 
 size_t dmnsn_array_size(const dmnsn_array *array);
 void   dmnsn_array_resize(dmnsn_array *array, size_t length);
 
-/* Non-atomic operations for manual locking */
-void *dmnsn_array_at(dmnsn_array *array, size_t i);
-size_t dmnsn_array_size_unlocked(const dmnsn_array *array);
-void dmnsn_array_resize_unlocked(dmnsn_array *array, size_t length);
-
-/* Manual locking */
-void dmnsn_array_rdlock(const dmnsn_array *array);
-void dmnsn_array_wrlock(dmnsn_array *array);
-void dmnsn_array_unlock(const dmnsn_array *array);
-
 #endif /* DIMENSION_ARRAY_H */
diff --git a/libdimension/dimension/progress.h b/libdimension/dimension/progress.h
index 7958ab9..57a38b6 100644
--- a/libdimension/dimension/progress.h
+++ b/libdimension/dimension/progress.h
@@ -44,6 +44,9 @@ typedef struct {
   /* The worker thread */
   pthread_t thread;
 
+  /* Read-write synchronization */
+  pthread_rwlock_t *rwlock;
+
   /* Condition variable for waiting for a particular amount of progress */
   pthread_cond_t  *cond;
   pthread_mutex_t *mutex;
diff --git a/libdimension/progress.c b/libdimension/progress.c
index f25af74..1f50e4c 100644
--- a/libdimension/progress.c
+++ b/libdimension/progress.c
@@ -22,6 +22,11 @@
 #include <pthread.h>
 #include <stdlib.h> /* For malloc */
 
+/* For thread synchronization */
+static void dmnsn_progress_rdlock(const dmnsn_progress *progress);
+static void dmnsn_progress_wrlock(dmnsn_progress *progress);
+static void dmnsn_progress_unlock(const dmnsn_progress *progress);
+
 /* Allocate a new dmnsn_progress*, returning NULL on failure */
 dmnsn_progress *
 dmnsn_new_progress()
@@ -32,10 +37,18 @@ dmnsn_new_progress()
     progress->elements = dmnsn_new_array(sizeof(dmnsn_progress_element));
     dmnsn_array_push(progress->elements, &element);
 
-    /* Allocate space for the condition variable and mutex */
+    /* Allocate space for the rwlock, condition variable, and mutex */
+
+    progress->rwlock = malloc(sizeof(pthread_rwlock_t));
+    if (!progress->rwlock) {
+      dmnsn_delete_array(progress->elements);
+      free(progress);
+      return NULL;
+    }
 
     progress->cond = malloc(sizeof(pthread_cond_t));
     if (!progress->cond) {
+      free(progress->rwlock);
       dmnsn_delete_array(progress->elements);
       free(progress);
       return NULL;
@@ -43,6 +56,16 @@ dmnsn_new_progress()
 
     progress->mutex = malloc(sizeof(pthread_mutex_t));
     if (!progress->mutex) {
+      free(progress->rwlock);
+      free(progress->cond);
+      dmnsn_delete_array(progress->elements);
+      free(progress);
+      return NULL;
+    }
+
+    if (pthread_rwlock_init(progress->rwlock, NULL) != 0) {
+      free(progress->rwlock);
+      free(progress->mutex);
       free(progress->cond);
       dmnsn_delete_array(progress->elements);
       free(progress);
@@ -50,6 +73,11 @@ dmnsn_new_progress()
     }
 
     if (pthread_cond_init(progress->cond, NULL) != 0) {
+      if (pthread_rwlock_destroy(progress->rwlock) != 0) {
+        dmnsn_error(DMNSN_SEVERITY_LOW,
+                    "Leaking rwlock in failed allocation.");
+      }
+      free(progress->rwlock);
       free(progress->mutex);
       free(progress->cond);
       dmnsn_delete_array(progress->elements);
@@ -57,10 +85,15 @@ dmnsn_new_progress()
       return NULL;
     }
     if (pthread_mutex_init(progress->mutex, NULL) != 0) {
+      if (pthread_rwlock_destroy(progress->rwlock) != 0) {
+        dmnsn_error(DMNSN_SEVERITY_LOW,
+                    "Leaking rwlock in failed allocation.");
+      }
       if (pthread_cond_destroy(progress->cond) != 0) {
         dmnsn_error(DMNSN_SEVERITY_LOW,
                     "Leaking condition variable in failed allocation.");
       }
+      free(progress->rwlock);
       free(progress->mutex);
       free(progress->cond);
       dmnsn_delete_array(progress->elements);
@@ -77,6 +110,9 @@ void
 dmnsn_delete_progress(dmnsn_progress *progress)
 {
   if (progress) {
+    if (pthread_rwlock_destroy(progress->rwlock) != 0) {
+      dmnsn_error(DMNSN_SEVERITY_LOW, "Leaking rwlock.");
+    }
     if (pthread_mutex_destroy(progress->mutex) != 0) {
       dmnsn_error(DMNSN_SEVERITY_LOW, "Leaking mutex.");
     }
@@ -124,14 +160,14 @@ dmnsn_get_progress(const dmnsn_progress *progress)
   double prog = 0.0;
   unsigned int i, size;
 
-  dmnsn_array_rdlock(progress->elements);
-    size = dmnsn_array_size_unlocked(progress->elements);
+  dmnsn_progress_rdlock(progress);
+    size = dmnsn_array_size(progress->elements);
     for (i = 0; i < size; ++i) {
       element = dmnsn_array_at(progress->elements, size - i - 1);
       prog += element->progress;
       prog /= element->total;
     }
-  dmnsn_array_unlock(progress->elements);
+  dmnsn_progress_unlock(progress);
 
   return prog;
 }
@@ -172,15 +208,15 @@ dmnsn_increment_progress(dmnsn_progress *progress)
   dmnsn_progress_element *element;
   size_t size;
 
-  dmnsn_array_wrlock(progress->elements);
-    size = dmnsn_array_size_unlocked(progress->elements);
+  dmnsn_progress_wrlock(progress);
+    size = dmnsn_array_size(progress->elements);
     element = dmnsn_array_at(progress->elements, size - 1);
     ++element->progress; /* Increment the last element */
 
     while (element->progress >= element->total && size > 1) {
       /* As long as the last element is complete, pop it */
       --size;
-      dmnsn_array_resize_unlocked(progress->elements, size);
+      dmnsn_array_resize(progress->elements, size);
       element = dmnsn_array_at(progress->elements, size - 1);
       ++element->progress; /* Increment the next element */
     }
@@ -188,7 +224,7 @@ dmnsn_increment_progress(dmnsn_progress *progress)
     if (pthread_cond_broadcast(progress->cond) != 0) {
       dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't signal condition variable.");
     }
-  dmnsn_array_unlock(progress->elements);
+  dmnsn_progress_unlock(progress);
 }
 
 /* Immediately set to 100% completion */
@@ -197,13 +233,39 @@ dmnsn_done_progress(dmnsn_progress *progress)
 {
   dmnsn_progress_element *element;
 
-  dmnsn_array_wrlock(progress->elements);
-    dmnsn_array_resize_unlocked(progress->elements, 1);
+  dmnsn_progress_wrlock(progress);
+    dmnsn_array_resize(progress->elements, 1);
     element = dmnsn_array_at(progress->elements, 0);
     element->progress = element->total;
 
     if (pthread_cond_broadcast(progress->cond) != 0) {
       dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't signal condition variable.");
     }
-  dmnsn_array_unlock(progress->elements);
+  dmnsn_progress_unlock(progress);
+}
+
+/* Thread synchronization */
+
+static void
+dmnsn_progress_rdlock(const dmnsn_progress *progress)
+{
+  if (pthread_rwlock_rdlock(progress->rwlock) != 0) {
+    dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't acquire read-lock.");
+  }
+}
+
+static void
+dmnsn_progress_wrlock(dmnsn_progress *progress)
+{
+  if (pthread_rwlock_wrlock(progress->rwlock) != 0) {
+    dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't acquire write-lock.");
+  }
+}
+
+static void
+dmnsn_progress_unlock(const dmnsn_progress *progress)
+{
+  if (pthread_rwlock_unlock(progress->rwlock) != 0) {
+    dmnsn_error(DMNSN_SEVERITY_MEDIUM, "Couldn't unlock read-write lock.");
+  }
 }
diff --git a/libdimension/raytrace.c b/libdimension/raytrace.c
index d634747..4af0a75 100644
--- a/libdimension/raytrace.c
+++ b/libdimension/raytrace.c
@@ -101,7 +101,7 @@ dmnsn_raytrace_scene_multithread(dmnsn_raytrace_payload *payload)
   dmnsn_raytrace_payload *payloads;
   pthread_t *threads;
 
-  /* TODO: do this portably */
+  /* Find the number of processors/cores running (TODO: do this portably) */
   nthreads = sysconf(_SC_NPROCESSORS_ONLN);
   if (nthreads < 1) {
     nthreads = 1;
diff --git a/libdimensionxx/dimensionxx/array.hpp b/libdimensionxx/dimensionxx/array.hpp
index 735e17a..4804de4 100644
--- a/libdimensionxx/dimensionxx/array.hpp
+++ b/libdimensionxx/dimensionxx/array.hpp
@@ -28,40 +28,6 @@
 
 namespace Dimension
 {
-  // RAII scoped read-lock
-  class Array_Read_Lock
-  {
-  public:
-    explicit Array_Read_Lock(const dmnsn_array* array)
-      : m_array(new const dmnsn_array*(array)) { dmnsn_array_rdlock(*m_array); }
-    // Array_Read_Lock(const Array_Read_Lock& lock);
-    ~Array_Read_Lock()
-    { if (m_array.unique()) { dmnsn_array_unlock(*m_array); } }
-
-  private:
-    // Copy assignment prohibited
-    Array_Read_Lock& operator=(const Array_Read_Lock&);
-
-    std::tr1::shared_ptr<const dmnsn_array*> m_array;
-  };
-
-  // RAII scoped write-lock
-  class Array_Write_Lock
-  {
-  public:
-    explicit Array_Write_Lock(dmnsn_array* array)
-      : m_array(new dmnsn_array*(array)) { dmnsn_array_wrlock(*m_array); }
-    // Array_Write_Lock(const Array_Write_Lock& lock);
-    ~Array_Write_Lock()
-    { if (m_array.unique()) { dmnsn_array_unlock(*m_array); } }
-
-  private:
-    // Copy assignment prohibited
-    Array_Write_Lock& operator=(const Array_Write_Lock&);
-
-    std::tr1::shared_ptr<dmnsn_array*> m_array;
-  };
-
   // Array template class, wraps a dmnsn_array*.  Copying is possible, but
   // copies refer to the same object, which is reference counted.  T must be
   // a POD type.
@@ -77,25 +43,13 @@ namespace Dimension
 
     // Array& operator=(const Array& a);
 
-    inline T at(std::size_t i) const;
-    void set(std::size_t i, T object) { dmnsn_array_set(dmnsn(), i, &object); }
-
-    std::size_t size() const { return dmnsn_array_size(dmnsn()); }
-    void resize(std::size_t size) { dmnsn_array_resize(dmnsn(), size); }
-
-    // For manual locking
-
-    Array_Read_Lock read_lock() const { return Array_Read_Lock(dmnsn()); }
-    Array_Write_Lock write_lock() { return Array_Write_Lock(dmnsn()); }
-
     T& operator[](std::size_t i)
     { return *reinterpret_cast<T*>(dmnsn_array_at(dmnsn(), i)); }
     const T& operator[](std::size_t i) const
     { return *reinterpret_cast<const T*>(dmnsn_array_at(dmnsn(), i)); }
-    std::size_t size_unlocked() const
-    { return dmnsn_array_size_unlocked(dmnsn()); }
-    void resize_unlocked(std::size_t size)
-    { dmnsn_array_resize_unlocked(dmnsn(), size); }
+
+    std::size_t size() const { return dmnsn_array_size(dmnsn()); }
+    void resize(std::size_t size) { dmnsn_array_resize(dmnsn(), size); }
 
     // Access the wrapped C object.
     dmnsn_array*       dmnsn();
@@ -139,15 +93,6 @@ namespace Dimension
     static_cast<void>(constraint); // Silence unused variable warning
   }
 
-  template <typename T>
-  inline T
-  Array<T>::at(std::size_t i) const
-  {
-    T ret;
-    dmnsn_array_get(dmnsn(), i, &ret);
-    return ret;
-  }
-
   // Access the underlying dmnsn_array*
 
   template <typename T>