ENH: Implement methods for uint-alignment

Implements IsAligned, IsUintAligned, npy_uint_alignment

7 files changed, 78 insertions, 62 deletions

diff --git a/numpy/core/src/common/array_assign.c b/numpy/core/src/common/array_assign.c
index e4cb651d6..ac3fdbef7 100644
--- a/numpy/core/src/common/array_assign.c
+++ b/numpy/core/src/common/array_assign.c
@@ -130,6 +130,23 @@ raw_array_is_aligned(int ndim, npy_intp *shape,
     }
 }
 
+NPY_NO_EXPORT int
+IsAligned(PyArrayObject *ap)
+{
+    return raw_array_is_aligned(PyArray_NDIM(ap), PyArray_DIMS(ap),
+                                PyArray_DATA(ap), PyArray_STRIDES(ap),
+                                PyArray_DESCR(ap)->alignment);
+}
+
+NPY_NO_EXPORT int
+IsUintAligned(PyArrayObject *ap)
+{
+    return raw_array_is_aligned(PyArray_NDIM(ap), PyArray_DIMS(ap),
+                                PyArray_DATA(ap), PyArray_STRIDES(ap),
+                                npy_uint_alignment(PyArray_DESCR(ap)->elsize));
+}
+
+
 
 /* Returns 1 if the arrays have overlapping data, 0 otherwise */
 NPY_NO_EXPORT int
diff --git a/numpy/core/src/common/array_assign.h b/numpy/core/src/common/array_assign.h
index fdc36d3a2..07438c5e8 100644
--- a/numpy/core/src/common/array_assign.h
+++ b/numpy/core/src/common/array_assign.h
@@ -94,6 +94,20 @@ NPY_NO_EXPORT int
 raw_array_is_aligned(int ndim, npy_intp *shape,
                      char *data, npy_intp *strides, int alignment);
 
+/*
+ * Checks if an array is aligned to its "true alignment"
+ * given by dtype->alignment.
+ */
+NPY_NO_EXPORT int
+IsAligned(PyArrayObject *ap);
+
+/*
+ * Checks if an array is aligned to its "uint alignment"
+ * given by npy_uint_alignment(dtype->elsize).
+ */
+NPY_NO_EXPORT int
+IsUintAligned(PyArrayObject *ap);
+
 /* Returns 1 if the arrays have overlapping data, 0 otherwise */
 NPY_NO_EXPORT int
 arrays_overlap(PyArrayObject *arr1, PyArrayObject *arr2);
diff --git a/numpy/core/src/multiarray/common.c b/numpy/core/src/multiarray/common.c
index 4f695fdc7..5b4611e8a 100644
--- a/numpy/core/src/multiarray/common.c
+++ b/numpy/core/src/multiarray/common.c
@@ -587,50 +587,6 @@ _zerofill(PyArrayObject *ret)
     return 0;
 }
 
-NPY_NO_EXPORT int
-_IsAligned(PyArrayObject *ap)
-{
-    int i;
-    npy_uintp aligned;
-    npy_uintp alignment = PyArray_DESCR(ap)->alignment;
-
-    /* alignment 1 types should have a efficient alignment for copy loops */
-    if (PyArray_ISFLEXIBLE(ap) || PyArray_ISSTRING(ap)) {
-        npy_intp itemsize = PyArray_ITEMSIZE(ap);
-        /* power of two sizes may be loaded in larger moves */
-        if (((itemsize & (itemsize - 1)) == 0)) {
-            alignment = itemsize > NPY_MAX_COPY_ALIGNMENT ?
-                NPY_MAX_COPY_ALIGNMENT : itemsize;
-        }
-        else {
-            /* if not power of two it will be accessed bytewise */
-            alignment = 1;
-        }
-    }
-
-    if (alignment == 1) {
-        return 1;
-    }
-    aligned = (npy_uintp)PyArray_DATA(ap);
-
-    for (i = 0; i < PyArray_NDIM(ap); i++) {
-#if NPY_RELAXED_STRIDES_CHECKING
-        /* skip dim == 1 as it is not required to have stride 0 */
-        if (PyArray_DIM(ap, i) > 1) {
-            /* if shape[i] == 1, the stride is never used */
-            aligned |= (npy_uintp)PyArray_STRIDES(ap)[i];
-        }
-        else if (PyArray_DIM(ap, i) == 0) {
-            /* an array with zero elements is always aligned */
-            return 1;
-        }
-#else /* not NPY_RELAXED_STRIDES_CHECKING */
-        aligned |= (npy_uintp)PyArray_STRIDES(ap)[i];
-#endif /* not NPY_RELAXED_STRIDES_CHECKING */
-    }
-    return npy_is_aligned((void *)aligned, alignment);
-}
-
 NPY_NO_EXPORT npy_bool
 _IsWriteable(PyArrayObject *ap)
 {
diff --git a/numpy/core/src/multiarray/common.h b/numpy/core/src/multiarray/common.h
index db0a49920..13876f17b 100644
--- a/numpy/core/src/multiarray/common.h
+++ b/numpy/core/src/multiarray/common.h
@@ -1,5 +1,6 @@
 #ifndef _NPY_PRIVATE_COMMON_H_
 #define _NPY_PRIVATE_COMMON_H_
+#include "structmember.h"
 #include <numpy/npy_common.h>
 #include <numpy/npy_cpu.h>
 #include <numpy/ndarraytypes.h>
@@ -56,9 +57,6 @@ index2ptr(PyArrayObject *mp, npy_intp i);
 NPY_NO_EXPORT int
 _zerofill(PyArrayObject *ret);
 
-NPY_NO_EXPORT int
-_IsAligned(PyArrayObject *ap);
-
 NPY_NO_EXPORT npy_bool
 _IsWriteable(PyArrayObject *ap);
 
@@ -182,6 +180,15 @@ check_and_adjust_axis(int *axis, int ndim)
     return check_and_adjust_axis_msg(axis, ndim, Py_None);
 }
 
+/* used for some alignment checks */
+#define _ALIGN(type) offsetof(struct {char c; type v;}, v)
+/*
+ * Disable harmless compiler warning "4116: unnamed type definition in
+ * parentheses" which is caused by the _ALIGN macro.
+ */
+#if defined(_MSC_VER)
+#pragma warning(disable:4116)
+#endif
 
 /*
  * return true if pointer is aligned to 'alignment'
@@ -201,6 +208,37 @@ npy_is_aligned(const void * p, const npy_uintp alignment)
     }
 }
 
+/* Get equivalent "uint" alignment given an itemsize, for use in copy code */
+static NPY_INLINE int
+npy_uint_alignment(int itemsize)
+{
+    npy_uintp alignment = 0; /* return value of 0 means unaligned */
+
+    switch(itemsize){
+        case 1:
+            return 1;
+        case 2:
+            alignment = _ALIGN(npy_uint16);
+            break;
+        case 4:
+            alignment = _ALIGN(npy_uint32);
+            break;
+        case 8:
+            alignment = _ALIGN(npy_uint64);
+            break;
+        case 16:
+            /*
+             * 16 byte types are copied using 2 uint64 assignments.
+             * See the strided copy function in lowlevel_strided_loops.c.
+             */
+            alignment = _ALIGN(npy_uint64);
+            break;
+        default:
+    }
+    
+    return alignment;
+}
+
 /*
  * memchr with stride and invert argument
  * intended for small searches where a call out to libc memchr is costly.
diff --git a/numpy/core/src/multiarray/flagsobject.c b/numpy/core/src/multiarray/flagsobject.c
index a78bedccb..85ea49fb4 100644
--- a/numpy/core/src/multiarray/flagsobject.c
+++ b/numpy/core/src/multiarray/flagsobject.c
@@ -12,6 +12,7 @@
 #include "npy_config.h"
 
 #include "npy_pycompat.h"
+#include "array_assign.h"
 
 #include "common.h"
 
@@ -64,7 +65,7 @@ PyArray_UpdateFlags(PyArrayObject *ret, int flagmask)
         _UpdateContiguousFlags(ret);
     }
     if (flagmask & NPY_ARRAY_ALIGNED) {
-        if (_IsAligned(ret)) {
+        if (IsAligned(ret)) {
             PyArray_ENABLEFLAGS(ret, NPY_ARRAY_ALIGNED);
         }
         else {
diff --git a/numpy/core/src/multiarray/lowlevel_strided_loops.c.src b/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
index b25b4a8b6..51b1a273d 100644
--- a/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
+++ b/numpy/core/src/multiarray/lowlevel_strided_loops.c.src
@@ -10,7 +10,6 @@
 
 #define PY_SSIZE_T_CLEAN
 #include "Python.h"
-#include "structmember.h"
 
 #define NPY_NO_DEPRECATED_API NPY_API_VERSION
 #define _MULTIARRAYMODULE
@@ -20,16 +19,6 @@
 
 #include "lowlevel_strided_loops.h"
 
-/* used for some alignment checks */
-#define _ALIGN(type) offsetof(struct {char c; type v;}, v)
-/*
- * Disable harmless compiler warning "4116: unnamed type definition in
- * parentheses" which is caused by the _ALIGN macro.
- */
-#if defined(_MSC_VER)
-#pragma warning(disable:4116)
-#endif
-
 
 /*
  * x86 platform works with unaligned access but the compiler is allowed to
diff --git a/numpy/core/src/multiarray/methods.c b/numpy/core/src/multiarray/methods.c
index 6317d6a16..cb63c7f74 100644
--- a/numpy/core/src/multiarray/methods.c
+++ b/numpy/core/src/multiarray/methods.c
@@ -21,6 +21,7 @@
 #include "conversion_utils.h"
 #include "shape.h"
 #include "strfuncs.h"
+#include "array_assign.h"
 
 #include "methods.h"
 #include "alloc.h"
@@ -1785,11 +1786,11 @@ array_setstate(PyArrayObject *self, PyObject *args)
         fa->data = datastr;
 #ifndef NPY_PY3K
         /* Check that the string is not interned */
-        if (!_IsAligned(self) || swap || PyString_CHECK_INTERNED(rawdata)) {
+        if (!IsAligned(self) || swap || PyString_CHECK_INTERNED(rawdata)) {
 #else
         /* Bytes should always be considered immutable, but we just grab the
          * pointer if they are large, to save memory. */
-        if (!_IsAligned(self) || swap || (len <= 1000)) {
+        if (!IsAligned(self) || swap || (len <= 1000)) {
 #endif
             npy_intp num = PyArray_NBYTES(self);
             fa->data = PyDataMem_NEW(num);
@@ -2281,7 +2282,7 @@ array_setflags(PyArrayObject *self, PyObject *args, PyObject *kwds)
         if (PyObject_Not(align_flag)) {
             PyArray_CLEARFLAGS(self, NPY_ARRAY_ALIGNED);
         }
-        else if (_IsAligned(self)) {
+        else if (IsAligned(self)) {
             PyArray_ENABLEFLAGS(self, NPY_ARRAY_ALIGNED);
         }
         else {