use std::path::{Path, PathBuf};
use std::process::Command;

fn main() {
    println!("cargo:rerun-if-changed=build.rs");

    // Only compile CUDA kernels when the cuda feature is enabled.
    if std::env::var("CARGO_FEATURE_CUDA").is_err() {
        return;
    }

    let out_dir = PathBuf::from(std::env::var("OUT_DIR").unwrap());
    let kernel_dir = Path::new("src");

    // Detect GPU architecture from env or default to sm_80 (matches crates/ml/build.rs).
    println!("cargo:rerun-if-env-changed=CUDA_COMPUTE_CAP");
    let cuda_compute_cap = std::env::var("CUDA_COMPUTE_CAP").unwrap_or_else(|_| "80".to_string());
    let arch = format!("sm_{cuda_compute_cap}");

    // Check if nvcc is available -- gracefully skip if not (non-CUDA builds).
    let nvcc_path = find_nvcc();
    let nvcc = match nvcc_path {
        Some(p) => p,
        None => {
            eprintln!("  warning: nvcc not found, skipping CUDA kernel precompilation");
            eprintln!("           Install CUDA toolkit or set CUDA_HOME for GPU builds");
            return;
        }
    };

    let kernel_name = "ig_kernels.cu";
    if !compile_kernel(&nvcc, kernel_dir, kernel_name, &arch, &out_dir) {
        panic!("nvcc failed to compile {kernel_name}");
    }
    eprintln!("  Precompiled 1/1 CUDA kernels ({arch}) — f32");
}

/// Compile a single .cu kernel file to a .cubin via nvcc.
fn compile_kernel(
    nvcc: &Path,
    kernel_dir: &Path,
    kernel_name: &str,
    arch: &str,
    out_dir: &Path,
) -> bool {
    let kernel_path = kernel_dir.join(kernel_name);
    let cubin_name = kernel_name.replace(".cu", ".cubin");
    let cubin_path = out_dir.join(&cubin_name);

    println!("cargo:rerun-if-changed={}", kernel_path.display());

    // Compile with nvcc.
    let status = Command::new(nvcc)
        .args([
            "-cubin",
            &format!("-arch={arch}"),
            "-O3",
            "--ftz=true",
            "--fmad=true",
            "--prec-div=true",
            "--prec-sqrt=true",
            "-o", cubin_path.to_str().unwrap(),
            kernel_path.to_str().unwrap(),
        ])
        .status();

    match status {
        Ok(s) if s.success() => {
            eprintln!("  Compiled {kernel_name} -> {cubin_name} ({arch})");
            true
        }
        Ok(s) => {
            eprintln!("  FAILED: {kernel_name} (exit={})", s.code().unwrap_or(-1));
            false
        }
        Err(e) => {
            eprintln!("  FAILED: {kernel_name} (nvcc error: {e})");
            false
        }
    }
}

/// Find nvcc: prefer $CUDA_HOME/bin/nvcc, then check common paths / $PATH.
fn find_nvcc() -> Option<PathBuf> {
    if let Ok(home) = std::env::var("CUDA_HOME") {
        let nvcc = PathBuf::from(home).join("bin/nvcc");
        if nvcc.exists() {
            return Some(nvcc);
        }
    }

    for path in &["/usr/local/cuda/bin/nvcc", "/usr/bin/nvcc"] {
        let p = PathBuf::from(path);
        if p.exists() {
            return Some(p);
        }
    }

    match Command::new("nvcc").arg("--version").output() {
        Ok(output) if output.status.success() => Some(PathBuf::from("nvcc")),
        _ => None,
    }
}