Compute Noop

Example compute shader that does … nothing but copy a value from one buffer into another.

import wgpu
from wgpu.utils.compute import compute_with_buffers  # Convenience function

shader_source = """

@group(0) @binding(0)
var<storage,read> data1: array<i32>;

@group(0) @binding(1)
var<storage,read_write> data2: array<i32>;

@compute
@workgroup_size(1)
fn main(@builtin(global_invocation_id) index: vec3<u32>) {
    let i: u32 = index.x;
    data2[i] = data1[i];
}
"""

# Create input data as a memoryview
n = 20
data = memoryview(bytearray(n * 4)).cast("i")
for i in range(n):
    data[i] = i

# The first arg is the input data, per binding
# The second arg is the output types, per binding
out = compute_with_buffers({0: data}, {1: (n, "i")}, shader_source, n=n)

# The result is a dict matching the output types
# Select data from buffer at binding 1
result = out[1].tolist()
print(result)
assert result == list(range(20))

# import numpy as np
#
# numpy_data = np.frombuffer(data, np.int32)
# out = compute_with_buffers({0: numpy_data}, {1: numpy_data.nbytes}, shader_source, n=n)
# result = np.frombuffer(out[1], dtype=np.int32)
# print(result.tolist())

Create device and shader object

device = wgpu.utils.get_default_device()

# Show all available adapters
adapters = wgpu.gpu.enumerate_adapters_sync()
for a in adapters:
    print(a.summary)

# You can select specific GPU from the available adapters
# adapter = None
# for a in adapters:
#     if "NVIDIA" in a.summary:
#         adapter = a
#         break
# assert adapter is not None
# device = adapter.request_device_sync()

cshader = device.create_shader_module(code=shader_source)

# Create buffer objects, input buffer is mapped.
buffer1 = device.create_buffer_with_data(data=data, usage=wgpu.BufferUsage.STORAGE)
buffer2 = device.create_buffer(
    size=data.nbytes, usage=wgpu.BufferUsage.STORAGE | wgpu.BufferUsage.COPY_SRC
)

# Setup layout and bindings
binding_layouts = [
    {
        "binding": 0,
        "visibility": wgpu.ShaderStage.COMPUTE,
        "buffer": {
            "type": wgpu.BufferBindingType.read_only_storage,
        },
    },
    {
        "binding": 1,
        "visibility": wgpu.ShaderStage.COMPUTE,
        "buffer": {
            "type": wgpu.BufferBindingType.storage,
        },
    },
]
bindings = [
    {
        "binding": 0,
        "resource": {"buffer": buffer1, "offset": 0, "size": buffer1.size},
    },
    {
        "binding": 1,
        "resource": {"buffer": buffer2, "offset": 0, "size": buffer2.size},
    },
]

# Put everything together
bind_group_layout = device.create_bind_group_layout(entries=binding_layouts)
pipeline_layout = device.create_pipeline_layout(bind_group_layouts=[bind_group_layout])
bind_group = device.create_bind_group(layout=bind_group_layout, entries=bindings)

# Create and run the pipeline
compute_pipeline = device.create_compute_pipeline(
    layout=pipeline_layout,
    compute={"module": cshader, "entry_point": "main"},
)
command_encoder = device.create_command_encoder()
compute_pass = command_encoder.begin_compute_pass()
compute_pass.set_pipeline(compute_pipeline)
compute_pass.set_bind_group(0, bind_group)
compute_pass.dispatch_workgroups(n, 1, 1)  # x y z
compute_pass.end()
device.queue.submit([command_encoder.finish()])

# Read result
# result = buffer2.read_data().cast("i")
out = device.queue.read_buffer(buffer2).cast("i")
result = out.tolist()
print(result)
assert result == list(range(20))

Interactive example

This uses Pyodide. If this does not work, your browser may not have sufficient support for wasm/pyodide/wgpu (check your browser dev console). Stdout (print statements) will also appear in the browser console.