Main Content

bsxfun

Binary singleton expansion function for gpuArray

Description

example

Note

  • The function arrayfun offers improved functionality compared to bsxfun. arrayfun is recommended.

  • This function behaves similarly to the MATLAB® function bsxfun, except that the evaluation of the function happens on the GPU, not on the CPU. Any required data not already on the GPU is moved to GPU memory. The MATLAB function passed in for evaluation is compiled and then executed on the GPU. All output arguments are returned as gpuArray objects. You can retrieve gpuArray data using the gather function.

C = bsxfun(FUN,A,B) applies the element-by-element binary operation specified by FUN to arrays A and B, with singleton expansion enabled.

Examples

collapse all

Use bsxfun with a matrix to subtract the mean of each column from all elements in that column. Then normalize by the standard deviation of each column.

A = rand(4,'gpuArray');
B = bsxfun(@minus,A,mean(A));
C = bsxfun(@rdivide,B,std(B))

You can use bsxfun to evaluate a function for different combinations of inputs.

A = rand(4,'gpuArray');
B = bsxfun(@minus,A,mean(A));
C = bsxfun(@rdivide,B,std(B))
C =

   -1.2957   -1.1587   -0.8727    0.2132
   -0.2071    0.9960    0.3272   -1.2763
    0.4786    0.6523   -0.7228    1.1482
    1.0243   -0.4896    1.2684   -0.0851

Create a function handle that represents the function f(a,b) = 1 - ae-b. Use bsxfun to apply the function to vectors a and b. bsxfun uses singleton expansion to expand the vectors into matrices and evaluates the function with all permutations of the input variables.

a = gpuArray(1:7);
b = gpuArray(pi*[0 1/4 1/2 3/4 1 5/4 6/4 7/4 2]).';
fun = @(a,b) 1 - a.*exp(-b);
c = bsxfun(fun,a,b)
c =

         0   -1.0000   -2.0000   -3.0000   -4.0000   -5.0000   -6.0000
    0.5441    0.0881   -0.3678   -0.8238   -1.2797   -1.7356   -2.1916
    0.7921    0.5842    0.3764    0.1685   -0.0394   -0.2473   -0.4552
    0.9052    0.8104    0.7157    0.6209    0.5261    0.4313    0.3365
    0.9568    0.9136    0.8704    0.8271    0.7839    0.7407    0.6975
    0.9803    0.9606    0.9409    0.9212    0.9015    0.8818    0.8621
    0.9910    0.9820    0.9731    0.9641    0.9551    0.9461    0.9371
    0.9959    0.9918    0.9877    0.9836    0.9795    0.9754    0.9713
    0.9981    0.9963    0.9944    0.9925    0.9907    0.9888    0.9869

Input Arguments

collapse all

Function to apply to the elements of the input arrays, specified as a function handle. FUN must be a handle to a supported element-wise function, or an element-wise function written in the MATLAB language that uses supported functions and syntax. Fun must return scalar values. For each output argument, FUN must return values of the same class each time it is called.

FUN must be a handle to a function that is written in the MATLAB language. You cannot specify FUN as a handle to a MEX-function.

FUN can contain the following built-in MATLAB functions and operators.

abs
and
acos
acosh
acot
acoth
acsc
acsch
asec
asech
asin
asinh
atan
atan2
atanh
beta
betaln
bitand
bitcmp
bitget
bitor
bitset
bitshift
bitxor
cast
ceil
complex
conj
cos
cosh
cot
coth
csc
csch
double
eps
eq
erf
erfc
erfcinv
erfcx
erfinv
exp
expm1
false
fix
floor
gamma
gammaln
ge
gt
hypot
imag
Inf
int8
int16
int32
int64
intmax
intmin
isfinite
isinf
isnan
ldivide
le
log
log2
log10
log1p
logical
lt
max
min
minus
mod
NaN
ne
not
ones
or
pi
plus
pow2
power
rand
randi
randn
rdivide
real
reallog
realmax
realmin
realpow
realsqrt
rem
round
sec
sech
sign
sin
single
sinh
sqrt
tan
tanh
times
true
uint8
uint16
uint32
uint64
xor
zeros

+
-
.*
./
.\
.^
==
~=
<
<=
>
>=
&
|
~
&&
||

Scalar expansion versions of the following:

*
/
\
^
Branching instructions:
break
continue
else, elseif, if
for
return
switch, case, otherwise
while

Functions that create arrays (such as Inf, NaN, ones, rand, randi, randn, and zeros) do not support size specifications as input arguments. Instead, the size of the generated array is determined by the size of the input variables to your functions. Enough array elements are generated to satisfy the needs of your input or output variables. You can specify the data type using both class and "like" syntaxes. The following examples show supported syntaxes for array-creation functions:

a = rand;
b = ones();
c = zeros("like", x);
d = Inf("single");
e = randi([0 9], "uint32");

When you use rand, randi, and randn to generate random numbers within FUN, each element is generated from a different substream. For more information about generating random numbers on the GPU, see Random Number Streams on a GPU.

When you use switch, case, otherwise within FUN:

  • Case expressions support only numeric and logical values.

  • Using a cell array as the case expression to compare the switch expression against multiple values, for example, case {x1,y1}, is not supported.

Input arrays, specified as scalars, vectors, matrices, or multidimensional arrays. Inputs A and B must have compatible sizes. For more information, see Compatible Array Sizes for Basic Operations. Whenever a dimension of A or B is singleton (equal to one), bsxfun virtually replicates the array along that dimension to match the other array. In the case where a dimension of A or B is singleton, and the corresponding dimension in the other array is zero, bsxfun virtually diminishes the singleton dimension to zero.

At least one of the inputs must be a gpuArray. Each array that is stored on CPU memory is converted to a gpuArray before the function is evaluated. If you plan to make several calls to bsxfun with the same array, it is more efficient to convert that array to a gpuArray.

Data Types: single | double | int8 | int16 | int32 | int64 | uint8 | uint16 | uint32 | uint64 | logical
Complex Number Support: Yes

Output Arguments

collapse all

Output array, returned as a scalar, vector, matrix, or multidimensional array, depending on the sizes of A and B. C is returned as a gpuArray.

Tips

  • The first time you call bsxfun to run a particular function on the GPU, there is some overhead time to set up the function for GPU execution. Subsequent calls of bsxfun with the same function can run faster.

  • Nonsingleton dimensions of input arrays must match each other. In other words, the corresponding dimensions of arguments A, B, etc., must be equal to each other, or equal to one. Whenever a dimension of an input array is singleton (equal to 1), bsxfun uses singleton expansion. The array is replicated along the singleton dimension to match the largest of the other arrays in that dimension. When a dimension of an input array is singleton and the corresponding dimension in another argument array is zero, bsxfun virtually diminishes the singleton dimension to 0.

    Each dimension of the output array C is the same size as the largest of the input arrays in that dimension for nonzero size, or zero otherwise. The following code shows how dimensions of size 1 are scaled up or down to match the size of the corresponding dimension in other arguments.

    R1 = rand(2,5,4,'gpuArray');
    R2 = rand(2,1,4,3,'gpuArray');
    R = bsxfun(@plus,R1,R2);
    size(R)
    
      2     5     4     3
    R1 = rand(2,2,0,4,'gpuArray');
    R2 = rand(2,1,1,4,'gpuArray');
    R = bsxfun(@plus,R1,R2);
    size(R)
    
      2     2     0     4
    
  • Because the operations supported by bsxfun are strictly element-wise, and each computation of each element is performed independently of the others, certain restrictions are imposed:

    • Input and output arrays cannot change shape or size.

    • Functions such as rand do not support size specifications. Arrays of random numbers have independent streams for each element.

  • Like bsxfun in MATLAB, matrix exponential power, multiplication, and division (^, *, /, \) perform element-wise calculations only.

  • Operations that change the size or shape of the input or output arrays (cat, reshape, and so on), are not supported.

  • Read-only indexing (subsref) and access to variables of the parent (outer) function workspace from within nested functions is supported. You can index variables that exist in the function before the evaluation on the GPU. Assignment or subsasgn indexing of these variables from within the nested function is not supported. For an example of the supported usage, see Stencil Operations on a GPU

  • Anonymous functions do not have access to their parent function workspace.

  • Overloading the supported functions is not allowed.

  • The code cannot call scripts.

  • There is no ans variable to hold unassigned computation results. Make sure to explicitly assign to variables the results of all calculations.

  • The following language features are not supported: persistent or global variables, parfor, spmd, switch, and try/catch.

  • P-code files cannot contain a call to bsxfun with gpuArray data.

Version History

Introduced in R2012a