gpuArray — Move data to the GPU as gpuArray values in MATLAB and RunMat.

gpuArray(X) moves data to the active GPU and returns a gpuArray handle for accelerated execution. RunMat mode additionally supports explicit dtype, size, and 'like' construction extensions.

Syntax

G = gpuArray(X)
G = gpuArray(X, dim, ...)
G = gpuArray(X, dtype)
G = gpuArray(X, "like", prototype)
G = gpuArray(X, dim, ..., option, ...)

Inputs

Returns

Errors

How gpuArray works

• Accepts numeric tensors, complex tensors, logical arrays, booleans, character vectors, and existing gpuArray handles. Double, single, and all eight integer classes retain native real or supported complex storage during transfer.
• In runmat compatibility mode, optional leading size arguments reshape the uploaded value while preserving element count.
• In runmat compatibility mode, class strings convert real data before upload and complex inputs can select double or single precision.
• In runmat compatibility mode, 'like', prototype infers dtype and logical state from the prototype.
• "gpuArray" strings are accepted as no-ops so call-sites that forward arguments from constructors such as zeros(..., 'gpuArray') remain compatible.
• Inputs that are already gpuArray handles pass through by default. When a class change is requested, RunMat gathers the data, performs the conversion, and uploads a fresh buffer without invalidating the original handle.
• When no acceleration provider is registered, the builtin raises gpuArray: no acceleration provider registered.

Does RunMat run gpuArray on the GPU?

gpuArray itself runs on the CPU. Host values enter one provider transfer contract carrying their native element type, shape, and real or interleaved-complex layout. gather uses the matching native download contract to reconstruct the same host class without widening single or integer values. A requested dtype conversion creates a new buffer and leaves the source gpuArray valid; providers that cannot represent the requested native transfer return an informative gpuArray: error.

GPU memory and residency

RunMat’s auto-offload planner transparently moves and keeps tensors on the GPU when it predicts a benefit. MATLAB-compatible scripts use gpuArray(X) for explicit upload; dtype, size, and 'like' construction forms are RunMat-only extensions.

Examples

Moving a matrix to the GPU for elementwise work

A = [1 2 3; 4 5 6];
G = gpuArray(A);
out = gather(sin(G))

Expected output:

out =
  2×3

    0.8415    0.9093    0.1411
   -0.7568   -0.9589   -0.2794

Uploading a scalar with dtype conversion

pi_single = gpuArray(pi, 'single');
isa(pi_single, 'gpuArray');
class(gather(pi_single))

Expected output:

ans =
  logical
     1

ans =
  single

Converting host data to a logical gpuArray

mask = gpuArray([0 2 -5 0], 'logical');
gather(mask)

Expected output:

ans =
  1×4 logical array

   0   1   1   0

Matching an existing prototype with 'like'

template = gpuArray(true(2, 2));
values = gpuArray([10 20 30 40], [2 2], 'like', template);
isequal(gather(values), logical([10 20; 30 40]))

Expected output:

ans =
  logical
     1

Reshaping during upload

flat = 1:6;
G = gpuArray(flat, 2, 3);
size(G)

Expected output:

ans =
     2     3

Calling gpuArray on an existing gpuArray handle

G = gpuArray([1 2 3]);
H = gpuArray(G, 'double');
isequal(G, H)

Expected output:

ans =
  logical
     1

Using gpuArray with coding agents

Open a RunMat example with live inputs, then ask the agent to explain how gpuArray changes the result.

Related Acceleration functions

Open-source implementation

Unlike proprietary runtimes, every RunMat function is open-source. Read exactly how gpuArray is executed, line by line, in Rust.

• View the source for gpuArray in Rust on GitHub
• Learn how the RunMat runtime works
• Found a bug? Open an issue with a minimal reproduction.

About RunMat

RunMat is an open-source runtime that executes MATLAB-syntax code blazing on any GPU. It is licensed under the Apache 2.0 license.

• RunMat automatically optimizes your math for GPU execution on Apple, Nvidia, and AMD hardware. No code changes needed. Simulations that took hours now take minutes.
• Start running code in seconds. RunMat runs in the browser, on the desktop, or from the CLI. No license server, no IT ticket.

Getting started · Benchmarks · Pricing