nimcl

High level wrapper over OpenCL

Pure Nim score 15/100 · tests present · no docs generated

Summary

Latest Version	Unknown
License	Apache License 2.0
CI Status	Failing
Downloads	0
Last Indexed	2026-07-21 05:24

Installation

nimble install nimcl

choosenim install nimcl

git clone https://github.com/andreaferretti/nimcl

OS Compatibility

Platform	Linux	macOS	Windows	FreeBSD	OpenBSD	NetBSD	Android	iOS	WASM	Embedded
nimcl	✓	✓	✓	-	-	-	-	-	-	-

Source

Repository	https://github.com/andreaferretti/nimcl
Homepage	https://github.com/andreaferretti/nimcl
Registry Source	nimble_official

README

Nim OpenCL utilities

This is an attempt at a high level wrapper over OpenCL.

For now, things are added when needed and as such, they may not be perfectly coherent. Still, they should be enough to cover the simplest cases and get started.

Some API changes can also be expected as the library becomes more comprehensive.

Vector add example

The "hello, world!" of OpenCL:

const
  body = staticRead("vadd.cl")
  size = 1_000_000
var
  a = newSeq[float32](size)
  b = newSeq[float32](size)
  c = newSeq[float32](size)

for i in 0 .. a.high:
  a[i] = i.float32
  b[i] = (i * i).float32

let
  (device, context, queue) = singleDeviceDefaults()
  program = context.createAndBuild(body, device)
  add = program.createKernel("add_vector")
  gpuA = context.bufferLike(a)
  gpuB = context.bufferLike(b)
  gpuC = context.bufferLike(c)

add.args(gpuA, gpuB, gpuC, size.int32)

queue.write(a, gpuA)
queue.write(b, gpuB)
queue.run(add, size)
queue.read(c, gpuC)

echo c[1 .. 100]

# Clean up
release(queue)
release(add)
release(program)
release(gpuA)
release(gpuB)
release(gpuC)
release(context)

The kernel is just

__kernel void add_vector(__global float* a, __global float* b, __global float* c, int num_els) {
  int idx = get_global_id(0);
  if (idx < num_els) {
    c[idx] = a[idx] + b[idx];
  }
}