gop

(Not recommended) Reduce arrays on spmd workers

gop is not recommended. Use spmdReduce instead. For more information, see Version History.

Syntax

B = gop(fcn,A)

B = gop(fcn,A,destination)

Description

B = gop(fcn,A) uses the function fcn to reduce the array A defined on each worker running an spmd block or communicating job.

For example, gop(@plus,A) returns the sum of the arrays A defined on each worker.

MATLAB uses the fcn function to reduce the array A defined on each worker by calling the function N-1 times.

• N is the value returned by numlabs in that spmd block or communicating job.

• A_j is the array A as defined on spmd worker j.

When you use parfor, parfeval, or parfevalOnAll to run code on a parallel pool, the workers are independent and do not communicate with each other. If you use gop on these workers, the result is the same as using gop on a client.

If numlabs is equal to 1, B is equal to A.

B = gop(fcn,A,destination) reduces the array A defined on each worker and stores the result on only one worker.

Examples

Find the Maximum of Worker Values in an spmd Block

This example shows how to use gop to calculate the maximum value for x across all workers.

Create a parallel pool with 4 workers.

parpool(4);

When you execute an spmd block after creating a parallel pool, by default all available workers in the pool will run the code inside the spmd block.

Run labindex on each worker in the spmd block, and store the result in a scalar A. Use gop and max to calculate the maximum value of A from each worker.

When you use gop, by default the result is stored on every worker. On the client, C is a Composite array. To get the result, index into the Composite array.

spmd
    A = labindex;
    C = gop(@max,A);
end
disp(C{1})

     4

Input Arguments

A — Input array

scalar | vector | matrix | multidimensional array | table | timetable | ...

Input array, specified as any MATLAB variable.

Example: magic(3)

fcn — Reduction function

function handle

Reduction function, specified as a function handle. The function must take two input arguments.

Example: fcn = @max

Data Types: function_handle

destination — Worker index

positive integer scalar

Index of destination worker, specified as a positive integer scalar. The value must be less than or equal to the value given by numlabs, the number of workers running the current spmd block or communicating job. When specified, the result of gop(fcn,A,destination) is only stored on one worker. On the worker with labindex equal to destination, C is the result of the operation. On all other workers, C is [].

Example: 1

Algorithms

When you use gop(fcn,A), fcn is used to combine the value of A from each worker.

To ensure that your spmd block or communicating job always produces the same results, specify fcn as an associative function.

Extended Capabilities

GPU Arrays

Accelerate code by running on a graphics processing unit (GPU) using Parallel Computing Toolbox™.

The gop function fully supports GPU arrays. To run the function on a GPU, specify the input data as a gpuArray. For more information, see Run MATLAB Functions on a GPU.

If x is a gpuArray, data transfer between multiple GPUs in a parallel pool uses fast peer-to-peer communication, including NVLink, if available.

Version History

Introduced before R2006a

R2022b: gop is not recommended

To indicate their intended use within spmd blocks, gop is renamed to spmdReduce. gop will continue to work but is no longer recommended. To update your code, replace any instance of gop with spmdReduce. There are no plans to remove gop.