1 USING: alien byte-arrays help.markup help.syntax math math.blas.vectors sequences strings ;
4 ARTICLE: "math.blas-summary" "Basic Linear Algebra Subroutines (BLAS) interface"
5 "Factor provides an interface to high-performance vector and matrix math routines available in implementations of the BLAS math library. A set of specialized types are provided for handling packed, unboxed vector data:"
6 { $subsections "math.blas-types" }
7 "Scalar-vector and vector-vector operations are available in the " { $vocab-link "math.blas.vectors" } " vocabulary:"
8 { $subsections "math.blas.vectors" }
9 "Vector-matrix and matrix-matrix operations are available in the " { $vocab-link "math.blas.matrices" } " vocabulary:"
10 { $subsections "math.blas.matrices" }
11 "The low-level BLAS Fortran interface can be accessed directly through the " { $vocab-link "math.blas.ffi" } " vocabulary. The BLAS interface can be configured to use different underlying BLAS implementations:"
12 { $subsections "math.blas.config" } ;
14 ARTICLE: "math.blas-types" "BLAS interface types"
15 "BLAS vectors come in single- and double-precision, real and complex flavors:"
19 complex-float-blas-vector
20 complex-double-blas-vector
22 "These vector types all follow the " { $link sequence } " protocol. In addition, there are corresponding types for matrix data:"
26 complex-float-blas-matrix
27 complex-double-blas-matrix
29 "There are BOA constructors for all vector and matrix types, which provide the most flexibility in specifying memory layout:"
33 <complex-float-blas-vector>
34 <complex-double-blas-vector>
37 <complex-float-blas-matrix>
38 <complex-double-blas-matrix>
40 "For the simple case of creating a dense, zero-filled vector or matrix, simple empty object constructors are provided:"
45 "BLAS vectors and matrices can also be constructed from other Factor sequences:"
49 >complex-float-blas-vector
50 >complex-double-blas-vector
53 >complex-float-blas-matrix
54 >complex-double-blas-matrix
57 ARTICLE: "math.blas.matrices" "BLAS interface matrix operations"
58 "Transposing and slicing matrices:"
65 "Matrix-vector products:"
72 "Vector outer products:"
90 "Scalar-matrix products:"
106 ABOUT: "math.blas.matrices"
108 HELP: blas-matrix-base
109 { $class-description "The base class for all BLAS matrix types. Objects of this type should not be created directly; instead, instantiate one of the typed subclasses:"
111 { { $link float-blas-matrix } }
112 { { $link double-blas-matrix } }
113 { { $link complex-float-blas-matrix } }
114 { { $link complex-double-blas-matrix } }
116 "All of these subclasses share the same tuple layout:"
118 { { $snippet "underlying" } " contains an alien pointer referencing or byte-array containing a packed, column-major array of float, double, float complex, or double complex values;" }
119 { { $snippet "ld" } " indicates the distance, in elements, between matrix columns;" }
120 { { $snippet "rows" } " and " { $snippet "cols" } " indicate the number of significant rows and columns in the matrix;" }
121 { "and " { $snippet "transpose" } ", if set to a true value, indicates that the matrix should be treated as transposed relative to its in-memory representation." }
124 { blas-vector-base blas-matrix-base } related-words
126 HELP: float-blas-matrix
127 { $class-description "A matrix of single-precision floating-point values. For details on the tuple layout, see " { $link blas-matrix-base } "." } ;
128 HELP: double-blas-matrix
129 { $class-description "A matrix of double-precision floating-point values. For details on the tuple layout, see " { $link blas-matrix-base } "." } ;
130 HELP: complex-float-blas-matrix
131 { $class-description "A matrix of single-precision floating-point complex values. Complex values are stored in memory as two consecutive float values, real part then imaginary part. For details on the tuple layout, see " { $link blas-matrix-base } "." } ;
132 HELP: complex-double-blas-matrix
133 { $class-description "A matrix of double-precision floating-point complex values. Complex values are stored in memory as two consecutive float values, real part then imaginary part. For details on the tuple layout, see " { $link blas-matrix-base } "." } ;
136 float-blas-matrix double-blas-matrix complex-float-blas-matrix complex-double-blas-matrix
137 float-blas-vector double-blas-vector complex-float-blas-vector complex-double-blas-vector
141 { $values { "matrix" blas-matrix-base } { "width" integer } }
142 { $description "Returns the number of columns in " { $snippet "matrix" } "." } ;
145 { $values { "matrix" blas-matrix-base } { "height" integer } }
146 { $description "Returns the number of rows in " { $snippet "matrix" } "." } ;
148 { Mwidth Mheight } related-words
151 { $values { "alpha" number } { "A" blas-matrix-base } { "x" blas-vector-base } { "beta" number } { "y" blas-vector-base } { "y=alpha*A.x+b*y" blas-vector-base } }
152 { $description "Calculate the matrix-vector product " { $snippet "αAx + βy" } ", and overwrite the current contents of " { $snippet "y" } " with the result. The width of " { $snippet "A" } " must match the length of " { $snippet "x" } ", and the height must match the length of " { $snippet "y" } ". Corresponds to the xGEMV routines in BLAS." }
153 { $side-effects "y" } ;
156 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "A" blas-matrix-base } { "A=alpha*x(*)y+A" blas-matrix-base } }
157 { $description "Calculate the outer product " { $snippet "αx⊗y + A" } " and overwrite the current contents of A with the result. The width of " { $snippet "A" } " must match the length of " { $snippet "y" } ", and its height must match the length of " { $snippet "x" } ". Corresponds to the xGER and xGERU routines in BLAS." }
158 { $side-effects "A" } ;
161 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "A" blas-matrix-base } { "A=alpha*x(*)yconj+A" blas-matrix-base } }
162 { $description "Calculate the conjugate outer product " { $snippet "αx⊗y̅ + A" } " and overwrite the current contents of A with the result. The width of " { $snippet "A" } " must match the length of " { $snippet "y" } ", and its height must match the length of " { $snippet "x" } ". Corresponds to the xGERC routines in BLAS." }
163 { $side-effects "A" } ;
166 { $values { "alpha" number } { "A" blas-matrix-base } { "B" blas-matrix-base } { "beta" number } { "C" blas-matrix-base } { "C=alpha*A.B+beta*C" blas-matrix-base } }
167 { $description "Calculate the matrix product " { $snippet "αAB + βC" } " and overwrite the current contents of C with the result. The width of " { $snippet "A" } " and the height of " { $snippet "B" } " must match, as must the heights of " { $snippet "A" } " and " { $snippet "C" } ", and the widths of " { $snippet "B" } " and " { $snippet "C" } ". Corresponds to the xGEMM routines in BLAS." }
168 { $side-effects "C" } ;
171 { $values { "rows" integer } { "cols" integer } { "exemplar" blas-vector-base blas-matrix-base } { "matrix" blas-matrix-base } }
172 { $description "Create a matrix of all zeros with the given dimensions and the same element type as " { $snippet "exemplar" } "." } ;
174 { <zero-vector> <empty-vector> <empty-matrix> } related-words
177 { $values { "alpha" number } { "A" blas-matrix-base } { "x" blas-vector-base } { "beta" number } { "y" blas-vector-base } { "alpha*A.x+b*y" blas-vector-base } }
178 { $description "Calculate the matrix-vector product " { $snippet "αAx + βy" } " and return a freshly allocated vector containing the result. The width of " { $snippet "A" } " must match the length of " { $snippet "x" } ", and the height must match the length of " { $snippet "y" } ". The returned vector will have the same length as " { $snippet "y" } ". Corresponds to the xGEMV routines in BLAS." } ;
181 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "A" blas-matrix-base } { "alpha*x(*)y+A" blas-matrix-base } }
182 { $description "Calculate the outer product " { $snippet "αx⊗y + A" } " and return a freshly allocated matrix containing the result. The width of " { $snippet "A" } " must match the length of " { $snippet "y" } ", and its height must match the length of " { $snippet "x" } ". The returned matrix will have the same dimensions as " { $snippet "A" } ". Corresponds to the xGER and xGERU routines in BLAS." } ;
185 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "A" blas-matrix-base } { "alpha*x(*)yconj+A" blas-matrix-base } }
186 { $description "Calculate the conjugate outer product " { $snippet "αx⊗y̅ + A" } " and return a freshly allocated matrix containing the result. The width of " { $snippet "A" } " must match the length of " { $snippet "y" } ", and its height must match the length of " { $snippet "x" } ". The returned matrix will have the same dimensions as " { $snippet "A" } ". Corresponds to the xGERC routines in BLAS." } ;
189 { $values { "alpha" number } { "A" blas-matrix-base } { "B" blas-matrix-base } { "beta" number } { "C" blas-matrix-base } { "alpha*A.B+beta*C" blas-matrix-base } }
190 { $description "Calculate the matrix product " { $snippet "αAB + βC" } " and overwrite the current contents of C with the result. The width of " { $snippet "A" } " and the height of " { $snippet "B" } " must match, as must the heights of " { $snippet "A" } " and " { $snippet "C" } ", and the widths of " { $snippet "B" } " and " { $snippet "C" } ". Corresponds to the xGEMM routines in BLAS." } ;
193 { $values { "alpha" number } { "A" blas-matrix-base } { "x" blas-vector-base } { "alpha*A.x" blas-vector-base } }
194 { $description "Calculate the matrix-vector product " { $snippet "αAx" } " and return a freshly allocated vector containing the result. The width of " { $snippet "A" } " must match the length of " { $snippet "x" } ". The length of the returned vector will match the height of " { $snippet "A" } ". Corresponds to the xGEMV routines in BLAS." } ;
197 { $values { "A" blas-matrix-base } { "x" blas-vector-base } { "A.x" blas-vector-base } }
198 { $description "Calculate the matrix-vector product " { $snippet "Ax" } " and return a freshly allocated vector containing the result. The width of " { $snippet "A" } " must match the length of " { $snippet "x" } ". The length of the returned vector will match the height of " { $snippet "A" } ". Corresponds to the xGEMV routines in BLAS." } ;
200 { n*M.V+n*V! n*M.V+n*V n*M.V M.V } related-words
203 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "alpha*x(*)y" blas-matrix-base } }
204 { $description "Calculate the outer product " { $snippet "αx⊗y" } " and return a freshly allocated matrix containing the result. The returned matrix's height will match the length of " { $snippet "x" } ", and its width will match the length of " { $snippet "y" } ". Corresponds to the xGER and xGERU routines in BLAS." } ;
207 { $values { "alpha" number } { "x" blas-vector-base } { "y" blas-vector-base } { "alpha*x(*)yconj" blas-matrix-base } }
208 { $description "Calculate the outer product " { $snippet "αx⊗y̅" } " and return a freshly allocated matrix containing the result. The returned matrix's height will match the length of " { $snippet "x" } ", and its width will match the length of " { $snippet "y" } ". Corresponds to the xGERC routines in BLAS." } ;
211 { $values { "x" blas-vector-base } { "y" blas-vector-base } { "x(*)y" blas-matrix-base } }
212 { $description "Calculate the outer product " { $snippet "x⊗y" } " and return a freshly allocated matrix containing the result. The returned matrix's height will match the length of " { $snippet "x" } ", and its width will match the length of " { $snippet "y" } ". Corresponds to the xGER and xGERU routines in BLAS." } ;
215 { $values { "x" blas-vector-base } { "y" blas-vector-base } { "x(*)yconj" blas-matrix-base } }
216 { $description "Calculate the conjugate outer product " { $snippet "x⊗y̅" } " and return a freshly allocated matrix containing the result. The returned matrix's height will match the length of " { $snippet "x" } ", and its width will match the length of " { $snippet "y" } ". Corresponds to the xGERC routines in BLAS." } ;
218 { n*V(*)V+M! n*V(*)Vconj+M! n*V(*)V+M n*V(*)Vconj+M n*V(*)V n*V(*)Vconj V(*) V(*)conj V. V.conj } related-words
221 { $values { "alpha" number } { "A" blas-matrix-base } { "B" blas-matrix-base } { "alpha*A.B" blas-matrix-base } }
222 { $description "Calculate the matrix product " { $snippet "αAB" } " and return a freshly allocated matrix containing the result. The width of " { $snippet "A" } " and the height of " { $snippet "B" } " must match. The returned matrix's height will be the same as " { $snippet "A" } "'s, and its width will match " { $snippet "B" } "'s. Corresponds to the xGEMM routines in BLAS." } ;
225 { $values { "A" blas-matrix-base } { "B" blas-matrix-base } { "A.B" blas-matrix-base } }
226 { $description "Calculate the matrix product " { $snippet "AB" } " and return a freshly allocated matrix containing the result. The width of " { $snippet "A" } " and the height of " { $snippet "B" } " must match. The returned matrix's height will be the same as " { $snippet "A" } "'s, and its width will match " { $snippet "B" } "'s. Corresponds to the xGEMM routines in BLAS." } ;
228 { n*M.M+n*M! n*M.M+n*M n*M.M M. } related-words
231 { $values { "matrix" blas-matrix-base } { "row" integer } { "col" integer } { "height" integer } { "width" integer } { "sub" blas-matrix-base } }
232 { $description "Select a rectangular submatrix of " { $snippet "matrix" } " with the given dimensions. The returned submatrix will share the parent matrix's storage." } ;
235 { $values { "A" blas-matrix-base } { "rows" sequence } }
236 { $description "Return a sequence of BLAS vectors representing the rows of " { $snippet "matrix" } ". Each vector will share the parent matrix's storage." } ;
239 { $values { "A" blas-matrix-base } { "cols" sequence } }
240 { $description "Return a sequence of BLAS vectors representing the columns of " { $snippet "matrix" } ". Each vector will share the parent matrix's storage." } ;
243 { $values { "n" number } { "A" blas-matrix-base } { "A=n*A" blas-matrix-base } }
244 { $description "Calculate the scalar-matrix product " { $snippet "nA" } " and overwrite the current contents of A with the result." }
245 { $side-effects "A" } ;
248 { $values { "n" number } { "A" blas-matrix-base } { "n*A" blas-matrix-base } }
249 { $description "Calculate the scalar-matrix product " { $snippet "nA" } " and return a freshly allocated matrix with the same dimensions as " { $snippet "A" } " containing the result." } ;
252 { $values { "A" blas-matrix-base } { "n" number } { "A*n" blas-matrix-base } }
253 { $description "Calculate the scalar-matrix product " { $snippet "nA" } " and return a freshly allocated matrix with the same dimensions as " { $snippet "A" } " containing the result." } ;
256 { $values { "A" blas-matrix-base } { "n" number } { "A/n" blas-matrix-base } }
257 { $description "Calculate the scalar-matrix product " { $snippet "(1/n)A" } " and return a freshly allocated matrix with the same dimensions as " { $snippet "A" } " containing the result." } ;
259 { n*M! n*M M*n M/n } related-words
262 { $values { "matrix" blas-matrix-base } { "matrix^T" blas-matrix-base } }
263 { $description "Returns the transpose of " { $snippet "matrix" } ". The returned matrix shares storage with the original matrix." } ;
266 { $values { "v" blas-vector-base blas-matrix-base } { "type" string } }
267 { $description "Return the C type of the elements in the given BLAS vector or matrix." } ;
270 { $values { "length" "The length of the new vector" } { "exemplar" blas-vector-base blas-matrix-base } { "vector" blas-vector-base } }
271 { $description "Return a vector of zeros with the given " { $snippet "length" } " and the same element type as " { $snippet "v" } "." } ;
274 { $syntax """smatrix{
280 { $description "Construct a literal " { $link float-blas-matrix } ". Note that although BLAS matrices are stored in column-major order, the literal is specified in row-major order." } ;
283 { $syntax """dmatrix{
289 { $description "Construct a literal " { $link double-blas-matrix } ". Note that although BLAS matrices are stored in column-major order, the literal is specified in row-major order." } ;
292 { $syntax """cmatrix{
294 { 0.0 C{ 0.0 1.0 } 0.0 2.0 }
296 { 0.0 0.0 0.0 C{ 0.0 -1.0 } }
298 { $description "Construct a literal " { $link complex-float-blas-matrix } ". Note that although BLAS matrices are stored in column-major order, the literal is specified in row-major order." } ;
301 { $syntax """zmatrix{
303 { 0.0 C{ 0.0 1.0 } 0.0 2.0 }
305 { 0.0 0.0 0.0 C{ 0.0 -1.0 } }
307 { $description "Construct a literal " { $link complex-double-blas-matrix } ". Note that although BLAS matrices are stored in column-major order, the literal is specified in row-major order." } ;
310 POSTPONE: smatrix{ POSTPONE: dmatrix{
311 POSTPONE: cmatrix{ POSTPONE: zmatrix{