xref: /openbsd-src/gnu/llvm/clang/docs/MatrixTypes.rst (revision 12c855180aad702bbcca06e0398d774beeafb155)
1==================
2Matrix Types
3==================
4
5.. contents::
6   :local:
7
8.. _matrixtypes:
9
10Clang provides a C/C++ language extension that allows users to directly express
11fixed-size 2-dimensional matrices as language values and perform arithmetic on
12them.
13
14This feature is currently experimental, and both its design and its
15implementation are in flux.
16
17Draft Specification
18===================
19
20Matrix Type
21-----------
22
23A matrix type is a scalar type with an underlying *element type*, a constant
24number of *rows*, and a constant number of *columns*. Matrix types with the same
25element type, rows, and columns are the same type. A value of a matrix type
26includes storage for ``rows * columns`` values of the *element type*. The
27internal layout, overall size and alignment are implementation-defined.
28
29The maximum of the product of the number of rows and columns is
30implementation-defined. If that implementation-defined limit is exceeded, the
31program is ill-formed.
32
33Currently, the element type of a matrix is only permitted to be one of the
34following types:
35
36* an integer type (as in C2x 6.2.5p19), but excluding enumerated types and ``_Bool``
37* the standard floating types ``float`` or ``double``
38* a half-precision floating point type, if one is supported on the target
39
40Other types may be supported in the future.
41
42Matrix Type Attribute
43---------------------
44
45Matrix types can be declared by adding the ``matrix_type`` attribute to the
46declaration of a *typedef* (or a C++ alias declaration). The underlying type
47of the *typedef* must be a valid matrix element type. The
48attribute takes two arguments, both of which must be integer constant
49expressions that evaluate to a value greater than zero. The first specifies the
50number of rows, and the second specifies the number of columns. The underlying
51type of the *typedef* becomes a matrix type with the given dimensions and an
52element type of the former underlying type.
53
54If a declaration of a *typedef-name* has a ``matrix_type`` attribute, then all
55declaration of that *typedef-name* shall have a matrix_type attribute with the
56same element type, number of rows, and number of columns.
57
58Standard Conversions
59--------------------
60
61The standard conversions are extended as follows. Note that these conversions
62are intentionally not listed as satisfying the constraints for assignment,
63which is to say, they are only permitted as explicit casts, not as implicit
64conversions.
65
66A value of matrix type can be converted to another matrix type if the number of
67rows and columns are the same and the value's elements can be converted to the
68element type of the result type. The result is a matrix where each element is
69the converted corresponding element.
70
71A value of any real type (as in C2x 6.2.5p17) can be converted to a matrix type
72if it can be converted to the element type of the matrix. The result is a
73matrix where all elements are the converted original value.
74
75If the number of rows or columns differ between the original and resulting
76type, the program is ill-formed.
77
78
79Arithmetic Conversions
80----------------------
81
82The usual arithmetic conversions are extended as follows.
83
84Insert at the start:
85
86* If both operands are of matrix type, no arithmetic conversion is performed.
87* If one operand is of matrix type and the other operand is of a real type,
88  convert the real type operand to the matrix type
89  according to the standard conversion rules.
90
91Matrix Type Element Access Operator
92-----------------------------------
93
94An expression of the form ``E1 [E2] [E3]``, where ``E1`` has matrix type ``cv
95M``, is a matrix element access expression.  Let ``T`` be the element type
96of ``M``, and let ``R`` and ``C`` be the number of rows and columns in ``M``
97respectively.  The index expressions shall have integral or unscoped
98enumeration type and shall not be uses of the comma operator unless
99parenthesized.  The first index expression shall evaluate to a
100non-negative value less than ``R``, and the second index expression shall
101evaluate to a non-negative value less than ``C``, or else the expression has
102undefined behavior.  If ``E1`` is a prvalue, the result is a prvalue with type
103``T`` and is the value of the element at the given row and column in the matrix.
104Otherwise, the result is a glvalue with type ``cv T`` and with the same value
105category as ``E1`` which refers to the element at the given row and column in
106the matrix.
107
108Programs containing a single subscript expression into a matrix are ill-formed.
109
110**Note**: We considered providing an expression of the form
111``postfix-expression [expression]`` to access columns of a matrix. We think
112that such an expression would be problematic once both column and row major
113matrixes are supported: depending on the memory layout, either accessing columns
114or rows can be done efficiently, but not both. Instead, we propose to provide
115builtins to extract rows and columns from a matrix. This makes the operations
116more explicit.
117
118Matrix Type Binary Operators
119----------------------------
120
121Given two matrixes, the ``+`` and ``-`` operators perform element-wise addition
122and subtraction, while the ``*`` operator performs matrix multiplication.
123``+``, ``-``, ``*``, and ``/`` can also be used with a matrix and a scalar
124value, applying the operation to each element of the matrix.
125
126Earlier versions of this extension did not support division by a scalar.
127You can test for the availability of this feature with
128``__has_extension(matrix_types_scalar_division)``.
129
130For the expression ``M1 BIN_OP M2`` where
131
132* ``BIN_OP`` is one of ``+`` or ``-``, one of ``M1`` and ``M2`` is of matrix
133  type, and the other is of matrix type or real type; or
134* ``BIN_OP`` is ``*``, one of ``M1`` and ``M2`` is of matrix type, and the
135   other is of a real type; or
136* ``BIN_OP`` is ``/``, ``M1`` is of matrix type, and ``M2`` is of a real type:
137
138* The usual arithmetic conversions are applied to ``M1`` and ``M2``. [ Note: if ``M1`` or
139  ``M2`` are of a real type, they are broadcast to matrices here. — end note ]
140* ``M1`` and ``M2`` shall be of the same matrix type.
141* The result is equivalent to Res in the following where col is the number of
142  columns and row is the number of rows in the matrix type:
143
144.. code-block:: c++
145
146  decltype(M1) Res;
147  for (int C = 0; C < col; ++C)
148    for (int R = 0; R < row; ++R)
149      Res[R][C] = M1[R][C] BIN_OP M2[R][C];
150
151Given the expression ``M1 * M2`` where ``M1`` and ``M2`` are of matrix type:
152
153* The usual arithmetic conversions are applied to ``M1`` and ``M2``.
154* The type of ``M1`` shall have the same number of columns as the type of ``M2`` has
155  rows. The element types of ``M1`` and ``M2`` shall be the same type.
156* The resulting type, ``MTy``, is a matrix type with the common element type,
157  the number of rows of ``M1`` and the number of columns of ``M2``.
158* The result is equivalent to ``Res`` in the following where ``EltTy`` is the
159  element type of ``MTy``, ``col`` is the number of columns, ``row`` is the
160  number of rows in ``MTy`` and ``inner`` is the number of columns of ``M1``:
161
162.. code-block:: c++
163
164  MTy Res;
165  for (int C = 0; C < col; ++C) {
166    for (int R = 0; R < row; ++R) {
167      EltTy Elt = 0;
168      for (int K = 0; K < inner; ++K) {
169        Elt += M1[R][K] * M2[K][C];
170    }
171    Res[R][C] = Elt;
172  }
173
174All operations on matrix types match the behavior of the element type with
175respect to signed overflows.
176
177With respect to floating-point contraction, rounding and environment rules,
178operations on matrix types match the behavior of the elementwise operations
179in the corresponding expansions provided above.
180
181Operations on floating-point matrices have the same rounding and floating-point
182environment behavior as ordinary floating-point operations in the expression's
183context. For the purposes of floating-point contraction, all calculations done
184as part of a matrix operation are considered intermediate operations, and their
185results need not be rounded to the format of the element type until the final
186result in the containing expression. This is subject to the normal restrictions
187on contraction, such as ``#pragma STDC FP_CONTRACT``.
188
189For the ``+=``, ``-=`` and ``*=`` operators the semantics match their expanded
190variants.
191
192Matrix Type Builtin Operations
193------------------------------
194
195Each matrix type supports a collection of builtin expressions that look like
196function calls but do not form an overload set. Here they are described as
197function declarations with rules for how to construct the argument list types
198and return type and the library description elements from
199[library.description.structure.specifications]/3 in the C++ standard.
200
201Definitions:
202
203* *M*, *M1*, *M2*, *M3* - Matrix types
204* *T* - Element type
205* *row*, *col* - Row and column arguments respectively.
206
207
208``M2 __builtin_matrix_transpose(M1 matrix)``
209
210**Remarks**: The return type is a cv-unqualified matrix type that has the same
211element type as ``M1`` and has the same number of rows as ``M1`` has columns and
212the same number of columns as ``M1`` has rows.
213
214**Returns**: A matrix ``Res`` equivalent to the code below, where ``col`` refers to the
215number of columns of ``M``, and ``row`` to the number of rows of ``M``.
216
217**Effects**: Equivalent to:
218
219.. code-block:: c++
220
221  M Res;
222  for (int C = 0; C < col; ++C)
223    for (int R = 0; R < row; ++R)
224      Res[C][R] = matrix[R][C];
225
226
227``M __builtin_matrix_column_major_load(T *ptr, size_t row, size_t col, size_t columnStride)``
228
229**Mandates**: ``row`` and ``col`` shall be integral constants greater than 0.
230
231**Preconditions**: ``columnStride`` is greater than or equal to ``row``.
232
233**Remarks**: The return type is a cv-unqualified matrix type with an element
234type of the cv-unqualified version of ``T`` and a number of rows and columns equal
235to ``row`` and ``col`` respectively. The parameter ``columnStride`` is optional
236and if omitted ``row`` is used as ``columnStride``.
237
238**Returns**: A matrix ``Res`` equivalent to:
239
240.. code-block:: c++
241
242  M Res;
243  for (size_t C = 0; C < col; ++C) {
244    for (size_t R = 0; R < row; ++K)
245      Res[R][C] = ptr[R];
246    ptr += columnStride
247  }
248
249
250``void __builtin_matrix_column_major_store(M matrix, T *ptr, size_t columnStride)``
251
252**Preconditions**: ``columnStride`` is greater than or equal to the number of rows in ``M``.
253
254**Remarks**: The type ``T`` is the const-unqualified version of the matrix
255argument’s element type. The parameter ``columnStride`` is optional and if
256omitted, the number of rows of ``M`` is used as ``columnStride``.
257
258**Effects**: Equivalent to:
259
260.. code-block:: c++
261
262  for (size_t C = 0; C < columns in M; ++C) {
263    for (size_t R = 0; R < rows in M; ++K)
264      ptr[R] = matrix[R][C];
265    ptr += columnStride
266  }
267
268
269TODOs
270-----
271
272TODO: Does it make sense to allow M::element_type, M::rows, and M::columns
273where M is a matrix type? We don’t support this anywhere else, but it’s
274convenient. The alternative is using template deduction to extract this
275information. Also add spelling for C.
276
277Future Work: Initialization syntax.
278
279
280Decisions for the Implementation in Clang
281=========================================
282
283This section details decisions taken for the implementation in Clang and is not
284part of the draft specification.
285
286The elements of a  value of a matrix type are laid out in column-major order
287without padding.
288
289We propose to provide a Clang option to override this behavior and allow
290contraction of those operations (e.g. *-ffp-contract=matrix*).
291
292TODO: Specify how matrix values are passed to functions.
293