cudaz: CUDA GPU Library for Zig

const std = @import("std");
const Cuda = @import("cudaz");

const increment_kernel =
    \\extern "C" __global__ void increment(float *out)
    \\{
    \\    int i = blockIdx.x * blockDim.x + threadIdx.x;
    \\    out[i] = out[i] + 1;
    \\}
;

pub fn main() !void {
    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
    defer _ = gpa.deinit();
    const allocator = gpa.allocator();

    // Initialize GPU 0
    const device = try Cuda.Device.default();
    defer device.deinit();

    // Copy array to GPU
    const data = [_]f32{ 1.2, 2.8, 0.123 };
    const cu_slice = try device.htodCopy(f32, &data);
    defer cu_slice.free();

    // Compile and run kernel
    const ptx = try Cuda.Compile.cudaText(increment_kernel, .{}, allocator);
    defer allocator.free(ptx);
    const module = try Cuda.Device.loadPtxText(ptx);
    const function = try module.getFunc("increment");
    try function.run(
        .{&cu_slice.device_ptr},
        Cuda.LaunchConfig{ .block_dim = .{ 3, 1, 1 }, .grid_dim = .{ 1, 1, 1 }, .shared_mem_bytes = 0 },
    );

    // Retrieve results
    var result = try Cuda.Device.syncReclaim(f32, allocator, cu_slice);
    defer result.deinit(allocator);
    // result.items == [2.2, 3.8, 1.123]
}