nodepp-gpu: A GPGPU Library for Nodepp

nodepp-gpu is a proof-of-concept library that brings the power of General-Purpose GPU (GPGPU) computing to C++ and Nodepp, inspired by the popular JavaScript library gpu.js. It allows you to write C++ functions that are executed on the GPU, leveraging the parallel processing capabilities of modern graphics hardware.

🔗: Nodepp GPU: Accelerating C++ with GPU and Nodepp

Dependencies

include(FetchContent)

FetchContent_Declare(
	nodepp
	GIT_REPOSITORY   https://github.com/NodeppOfficial/nodepp
	GIT_TAG          origin/main
	GIT_PROGRESS     ON
)
FetchContent_MakeAvailable(nodepp)

FetchContent_Declare(
	nodepp-raylib
	GIT_REPOSITORY   https://github.com/NodeppOfficial/nodepp-raylib
	GIT_TAG          origin/main
	GIT_PROGRESS     ON
)
FetchContent_MakeAvailable(nodepp-raylib)

FetchContent_Declare(
	nodepp-gpu
	GIT_REPOSITORY   https://github.com/NodeppOfficial/nodepp-gpu
	GIT_TAG          origin/main
	GIT_PROGRESS     ON
)
FetchContent_MakeAvailable(nodepp-gpu)

#[...]

target_link_libraries( #[...]
	PUBLIC nodepp nodepp-raylib nodepp-gpu #[...]
)

Features

• GPU Kernel Execution: Write your computational logic in GLSL (OpenGL Shading Language) and have it executed on the GPU.
• Seamless Integration with Nodepp: Uses Nodepp's core data structures and concepts for a familiar developer experience.
• Type-Safe Data Handling: Provides a matrix_t class to represent and manage data, including support for various vector types (vec2, vec3, vec4, sampler2D, etc.).
• Flexible Input/Output: Easily set input variables and textures for your GPU kernel and retrieve the computed result as a matrix_t object.
• Raylib Integration: Built on top of the Raylib library for its OpenGL context management and shader capabilities.
• Image Processing: Load images from memory, file paths, or existing textures and use them as input for your GPU computations.

Simple Example: Matrix Addition

#include <nodepp/nodepp.h>
#include <gpu/gpu.h>

using namespace nodepp;

void onMain() { 

    if( !gpu::start_machine() ) 
      { throw except_t("Failed to start GPU machine"); }

    gpu::gpu_t gpu ( GPU_KERNEL(

        vec2 idx = uv / vec2( 2, 2 );

        float color_a = texture( image_a, idx ).x;
        float color_b = texture( image_b, idx ).x;
        float color_c = color_a * color_b;
        
        return vec4( vec3( color_c ), 1. );

    ));

    gpu::matrix_t matrix_a( 2, 2, ptr_t<float>({
        10., 10.,
        10., 10.,
    }) );

    gpu::matrix_t matrix_b( 2, 2, ptr_t<float>({
        .1, .2,
        .4, .3,
    }) );

    gpu.set_output(2, 2, gpu::OUT_DOUBLE4);
    gpu.set_input (matrix_a, "image_a");
    gpu.set_input (matrix_b, "image_b");

    for( auto x: gpu().data() ) 
       { console::log(x); }

    gpu::stop_machine();

}

This example demonstrates how to set up the inputs, execute the kernel, and retrieve the result. The gpu_t::operator() handles the entire process of compiling the shader, rendering to a texture, and converting the texture back into a matrix_t object.

Build & Usage

    🪟: time g++ -o main main.cpp -L./lib -I./include -lraylib -lssl -lcrypto -lws2_32 ; ./main.exe
    🐧: time g++ -o main main.cpp -L./lib -I./include -lraylib -lssl -lcrypto ; ./main

License

Nodepp-gpu is distributed under the MIT License. See the LICENSE file for more details.