c = gpuArrayFunctions()
gpuArrayFunctions()
| Parameter | Description |
|---|---|
| c | a column cell of character vectors: the sorted names of the functions overloaded for gpuArray. |
c = gpuArrayFunctions() returns the list of functions and operators that run on a gpuArray. Calling gpuArrayFunctions() without an output argument displays the list.
The list is derived from the operations actually registered by the gpu_engine module (both the device builtins and the method overloads), so it always reflects the current build. It groups, among others:
Operators: plus, minus, times, rdivide, ldivide, power, mtimes, uminus, transpose, ctranspose, eq, ne, lt, le, gt, ge, and, or, xor, not.
Element-wise math: abs, sqrt, exp, expm1, log, log1p, log2, log10, pow2, sin, cos, tan, sec, csc, cot, asin, acos, atan, asec, acsc, acot, atan2, hypot, sinh, cosh, tanh, sech, csch, coth, asinh, acosh, atanh, deg2rad, rad2deg, floor, ceil, round, fix, sign, mod, rem.
Bitwise (integer): bitand, bitor, bitxor, bitshift.
Reductions and scans: sum, prod, mean, var, std, any, all, cumsum, cumprod, cummax, cummin, sort, find, dot, norm.
Linear algebra: mtimes and, on the device where possible, dot and the vector norms; the factorizations (mldivide, inv, lu, qr, eig, svd), the matrix functions (expm, logm, sqrtm) and fft/ifft fall back to the host.
Complex: real, imag, conj, angle, isreal (floor/ceil/round/fix/sign and the NaN/Inf predicates also accept complex).
Shape: reshape, transpose, ctranspose, flip, fliplr, flipud, rot90, circshift, repmat, squeeze, diff, triu, tril, diag, kron, trace, cross, horzcat, vertcat, cat.
Order statistics and set functions: median, mode, prctile, histcounts, unique, ismember, sortrows, issorted, nonzeros, gradient (evaluated on the host, re-wrapped as a gpuArray).
Predicates and sizing: isnan, isinf, isfinite, isreal, isempty, isscalar, isvector, ismatrix, isrow, iscolumn, isnumeric, isfloat, isinteger, islogical, length, nnz, underlyingType, isUnderlyingType.
Constructors: zeros, ones, eye, nan, inf, true, false, rand, randn (through the 'gpuArray' / 'like' forms), and linspace / colon from a gpuArray endpoint.
Data is moved with gpuArray and gather; use arrayfun to run a fused element-wise expression on the device.
c = gpuArrayFunctions();
numel(c)
gpuArrayFunctions()
| Version | Description |
|---|---|
| 2.0.0 | initial version |