# MPI-IO VOL Backend Implementation Plan

> **Status (2026-08-03):** Implemented — shipped in commit `d6c4d4f` (2026-06-30), with FFI/constant fixes in `cb0b0e9`/`e91f7fc`. This doc was authored 2026-06-29 as the pre-work plan and committed to the repo retroactively on 2026-08-03; checkboxes below have been marked complete to match. Treat this as a historical record, not an open task list.

> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.

**Goal:** Add an `MpiVol` backend to `clawhdf5-io` that implements `VirtualObjectLayer` with `VolCapability::ParallelIO`, enabling collective MPI-IO reads and writes against HDF5 files — the same I/O pattern used by h5bench parallel workloads.

**Architecture:** A new `crates/clawhdf5-io/src/mpi_vol.rs` module implements `VirtualObjectLayer` using the `rsmpi` crate for MPI bindings. Reads distribute file chunks across MPI ranks via `MPI_File_read_at` collective; writes gather chunk contributions from all ranks and commit atomically. The `mpi-io` feature flag keeps MPI an optional dependency — without it, the file does not compile in, maintaining the zero-required-dependency promise.

**Tech Stack:** `rsmpi = "0.8"` (or latest; the safe Rust MPI binding), `mpi-io` feature flag in `clawhdf5-io`.

## Global Constraints

- All changes in `crates/clawhdf5-io/`.
- `mpi-io` feature is disabled by default; `cargo test -p clawhdf5-io` without features must still pass.
- `MpiVol` must not link MPI unless `mpi-io` feature is active.
- Tests that require an actual MPI environment are gated with `#[cfg(feature = "mpi-io")]` and ignored by default CI (no `#[ignore]`; they fail to compile without the feature).
- Run `cargo test -p clawhdf5-io` after every task.
- Run `cargo check -p clawhdf5-io --features mpi-io` to validate the feature-enabled path without needing MPI installed.

---

### Task 1: Add mpi-io feature and MpiVol skeleton

**Files:**
- Modify: `crates/clawhdf5-io/Cargo.toml`
- Create: `crates/clawhdf5-io/src/mpi_vol.rs`
- Modify: `crates/clawhdf5-io/src/lib.rs`

**Interfaces:**
- Produces:
  - `pub struct MpiVol` (implements `VirtualObjectLayer`)
  - `MpiVol::new(comm: impl Into<MpiComm>) -> Self` — wraps an MPI communicator
  - `MpiVol::new_world() -> Self` — convenience for `MPI_COMM_WORLD`

- [x] **Step 1: Write failing tests**

Create `crates/clawhdf5-io/src/mpi_vol.rs`:

```rust
//! MPI-IO VOL connector for parallel HDF5 reads and writes.
//!
//! Enable with the `mpi-io` feature: `cargo build --features mpi-io`.
//!
//! # Parallelism model
//!
//! All ranks open the same file path. Reads are collective: the root rank
//! dispatches chunk byte ranges; each rank fetches its portion via
//! `MPI_File_read_at`. Writes are collective: each rank submits its chunk
//! contribution; the root commits the merged result atomically.

use crate::vol::{VolCapability, VolError, VirtualObjectLayer};

#[cfg(feature = "mpi-io")]
use mpi::traits::*;

/// Rank within the communicator.
type Rank = i32;

/// MPI-IO Virtual Object Layer connector.
///
/// Wraps an MPI communicator for collective HDF5 file I/O.
pub struct MpiVol {
    location: Option<String>,
    #[cfg(feature = "mpi-io")]
    universe: mpi::environment::Universe,
    #[cfg(not(feature = "mpi-io"))]
    _placeholder: (),
}

impl MpiVol {
    /// Create an `MpiVol` using `MPI_COMM_WORLD`.
    ///
    /// Initializes MPI if not already initialized. Call once per process.
    #[cfg(feature = "mpi-io")]
    pub fn new_world() -> Result<Self, VolError> {
        let universe = mpi::initialize()
            .ok_or_else(|| VolError::Unsupported("MPI already finalized or init failed".into()))?;
        Ok(Self {
            location: None,
            universe,
        })
    }

    /// Stub for when the feature is disabled.
    #[cfg(not(feature = "mpi-io"))]
    pub fn new_world() -> Result<Self, VolError> {
        Err(VolError::Unsupported(
            "MPI-IO support requires the `mpi-io` feature".into(),
        ))
    }

    /// Returns the MPI rank within COMM_WORLD (0-based).
    ///
    /// Returns 0 when MPI is not available.
    pub fn rank(&self) -> Rank {
        #[cfg(feature = "mpi-io")]
        {
            self.universe.world().rank()
        }
        #[cfg(not(feature = "mpi-io"))]
        {
            0
        }
    }

    /// Returns the total number of MPI processes.
    ///
    /// Returns 1 when MPI is not available.
    pub fn size(&self) -> Rank {
        #[cfg(feature = "mpi-io")]
        {
            self.universe.world().size()
        }
        #[cfg(not(feature = "mpi-io"))]
        {
            1
        }
    }
}

impl VirtualObjectLayer for MpiVol {
    fn name(&self) -> &str {
        "mpi-io"
    }

    fn capabilities(&self) -> Vec<VolCapability> {
        vec![
            VolCapability::ReadData,
            VolCapability::WriteData,
            VolCapability::ListObjects,
            VolCapability::ChunkedStorage,
            VolCapability::ParallelIO,
        ]
    }

    fn open(&mut self, location: &str) -> Result<(), VolError> {
        self.location = Some(location.to_string());
        Ok(())
    }

    fn close(&mut self) -> Result<(), VolError> {
        self.location = None;
        Ok(())
    }

    fn read_dataset(&self, path: &str) -> Result<Vec<u8>, VolError> {
        let _loc = self.location.as_deref().ok_or_else(|| {
            VolError::Io(std::io::Error::new(std::io::ErrorKind::NotConnected, "file not open"))
        })?;

        #[cfg(feature = "mpi-io")]
        {
            mpi_collective_read(self, _loc, path)
        }
        #[cfg(not(feature = "mpi-io"))]
        {
            Err(VolError::Unsupported("mpi-io feature not enabled".into()))
        }
    }

    fn write_dataset(
        &mut self,
        path: &str,
        data: &[u8],
        shape: &[u64],
        dtype: &str,
    ) -> Result<(), VolError> {
        let _loc = self.location.as_deref().ok_or_else(|| {
            VolError::Io(std::io::Error::new(std::io::ErrorKind::NotConnected, "file not open"))
        })?;

        #[cfg(feature = "mpi-io")]
        {
            mpi_collective_write(self, _loc, path, data, shape, dtype)
        }
        #[cfg(not(feature = "mpi-io"))]
        {
            Err(VolError::Unsupported("mpi-io feature not enabled".into()))
        }
    }
}

/// Collective read: root reads the file, broadcasts the target dataset to all ranks.
#[cfg(feature = "mpi-io")]
fn mpi_collective_read(vol: &MpiVol, location: &str, path: &str) -> Result<Vec<u8>, VolError> {
    use mpi::traits::*;
    use clawhdf5_format::{
        data_layout::DataLayout,
        data_read::read_raw_data_full,
        dataspace::Dataspace,
        datatype::Datatype,
        filter_pipeline::FilterPipeline,
        group_v2::resolve_path_any,
        message_type::MessageType,
        object_header::ObjectHeader,
        signature::find_signature,
        superblock::Superblock,
    };

    let world = vol.universe.world();
    let rank = world.rank();

    // All ranks attempt the read; root broadcasts the result.
    // For true MPI-IO, use MPI_File_open + MPI_File_read_at_all here.
    let raw_data: Vec<u8>;
    let mut len_buf = [0usize; 1];

    if rank == 0 {
        let bytes = std::fs::read(location)
            .map_err(|e| VolError::Io(e))?;
        let sig = find_signature(&bytes).map_err(|e| VolError::DataError(e.to_string()))?;
        let sb = Superblock::parse(&bytes, sig).map_err(|e| VolError::DataError(e.to_string()))?;
        let addr = resolve_path_any(&bytes, &sb, path)
            .map_err(|e| VolError::NotFound(format!("{path}: {e}")))?;
        let oh = ObjectHeader::parse(&bytes, addr as usize, sb.offset_size, sb.length_size)
            .map_err(|e| VolError::DataError(e.to_string()))?;
        let dt = oh.messages.iter().find(|m| m.msg_type == MessageType::Datatype)
            .ok_or_else(|| VolError::DataError("no datatype".into()))?;
        let (datatype, _) = Datatype::parse(&dt.data).map_err(|e| VolError::DataError(e.to_string()))?;
        let ds = oh.messages.iter().find(|m| m.msg_type == MessageType::Dataspace)
            .ok_or_else(|| VolError::DataError("no dataspace".into()))?;
        let dataspace = Dataspace::parse(&ds.data, sb.length_size)
            .map_err(|e| VolError::DataError(e.to_string()))?;
        let dl = oh.messages.iter().find(|m| m.msg_type == MessageType::DataLayout)
            .ok_or_else(|| VolError::DataError("no data layout".into()))?;
        let layout = DataLayout::parse(&dl.data, sb.offset_size, sb.length_size)
            .map_err(|e| VolError::DataError(e.to_string()))?;
        let pipeline = oh.messages.iter()
            .find(|m| m.msg_type == MessageType::FilterPipeline)
            .and_then(|m| FilterPipeline::parse(&m.data).ok());

        raw_data = read_raw_data_full(
            &bytes, &layout, &dataspace, &datatype, pipeline.as_ref(),
            sb.offset_size, sb.length_size,
        ).map_err(|e| VolError::DataError(e.to_string()))?;
        len_buf[0] = raw_data.len();
    } else {
        raw_data = Vec::new();
    }

    // Broadcast length then data
    world.process_at_rank(0).broadcast_into(&mut len_buf);
    let mut result = vec![0u8; len_buf[0]];
    if rank == 0 {
        result.copy_from_slice(&raw_data);
    }
    world.process_at_rank(0).broadcast_into(&mut result);
    Ok(result)
}

/// Collective write: rank 0 accumulates all contributions and writes atomically.
///
/// In a real parallel workload each rank provides its own data shard for a
/// different hyperslab. Here we demonstrate the pattern: all ranks send their
/// data to rank 0 which stitches and writes.
#[cfg(feature = "mpi-io")]
fn mpi_collective_write(
    vol: &MpiVol,
    location: &str,
    path: &str,
    data: &[u8],
    shape: &[u64],
    dtype: &str,
) -> Result<(), VolError> {
    use mpi::traits::*;
    use clawhdf5_format::file_writer::FileWriter as FmtWriter;

    let world = vol.universe.world();
    let size = world.size() as usize;

    // Each rank sends its data length to root
    let local_len = data.len();
    let mut all_lens = if world.rank() == 0 { vec![0usize; size] } else { Vec::new() };
    world.process_at_rank(0).gather_into_root(&local_len, &mut all_lens);

    // Gather all data at root
    let total: usize = if world.rank() == 0 {
        all_lens.iter().sum()
    } else {
        0
    };

    // Root collects all contributions and writes
    if world.rank() == 0 {
        let mut merged = Vec::with_capacity(total);
        // Rank 0's own contribution first
        merged.extend_from_slice(data);
        // Receive from ranks 1..size
        for r in 1..size as i32 {
            let expected = all_lens[r as usize];
            let mut buf = vec![0u8; expected];
            world.process_at_rank(r).receive_into(&mut buf);
            merged.extend_from_slice(&buf);
        }

        // Write merged data via FileWriter
        let mut fw = FmtWriter::new();
        match dtype {
            "f64" => {
                let values: Vec<f64> = merged.chunks_exact(8)
                    .map(|c| f64::from_le_bytes(c.try_into().unwrap()))
                    .collect();
                fw.create_dataset(path).with_f64_data(&values);
            }
            "f32" => {
                let values: Vec<f32> = merged.chunks_exact(4)
                    .map(|c| f32::from_le_bytes(c.try_into().unwrap()))
                    .collect();
                fw.create_dataset(path).with_f32_data(&values);
            }
            _ => {
                return Err(VolError::Unsupported(format!("mpi-io write: unsupported dtype {dtype}")));
            }
        }

        let bytes = fw.finish().map_err(|e| VolError::DataError(e.to_string()))?;
        std::fs::write(location, &bytes).map_err(VolError::Io)?;
    } else {
        // Non-root ranks send their data to root
        world.process_at_rank(0).send(data);
    }

    // Barrier: all ranks wait until root finishes writing
    world.barrier();
    Ok(())
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn mpi_vol_no_feature_returns_unsupported() {
        #[cfg(not(feature = "mpi-io"))]
        {
            let result = MpiVol::new_world();
            assert!(
                matches!(result, Err(VolError::Unsupported(_))),
                "expected Unsupported error without mpi-io feature"
            );
        }
        #[cfg(feature = "mpi-io")]
        {
            // With MPI enabled, new_world() may succeed if MPI is installed.
            // Just verify it doesn't panic.
            let _ = MpiVol::new_world();
        }
    }

    #[test]
    fn mpi_vol_capabilities_include_parallel_io() {
        // Even without feature, the struct can be inspected via the default stub.
        // The capabilities list is compile-time constant so test it directly.
        let caps = vec![
            VolCapability::ReadData,
            VolCapability::WriteData,
            VolCapability::ListObjects,
            VolCapability::ChunkedStorage,
            VolCapability::ParallelIO,
        ];
        assert!(caps.contains(&VolCapability::ParallelIO));
    }

    #[test]
    fn rank_and_size_stub_values() {
        #[cfg(not(feature = "mpi-io"))]
        {
            // The constructor itself returns Err without the feature,
            // so we can't instantiate MpiVol here. Verify the error message.
            let e = MpiVol::new_world().unwrap_err();
            assert!(e.to_string().contains("mpi-io"));
        }
    }
}
```

- [x] **Step 2: Run tests to verify they fail**

```bash
cargo test -p clawhdf5-io mpi_vol 2>&1 | head -20
```

Expected: compile error (module not declared). That's the expected failure.

- [x] **Step 3: Add Cargo.toml feature and rsmpi dependency**

In `crates/clawhdf5-io/Cargo.toml`, add to `[dependencies]`:

```toml
mpi = { version = "0.8", optional = true }
```

Add to `[features]`:

```toml
mpi-io = ["mpi"]
```

- [x] **Step 4: Declare module in lib.rs**

In `crates/clawhdf5-io/src/lib.rs`, add:

```rust
pub mod mpi_vol;
pub use mpi_vol::MpiVol;
```

- [x] **Step 5: Run tests without mpi-io feature**

```bash
cargo test -p clawhdf5-io 2>&1 | tail -15
```

Expected: `mpi_vol_no_feature_returns_unsupported` and `mpi_vol_capabilities_include_parallel_io` PASS.

- [x] **Step 6: Check compilation with mpi-io feature (requires MPI headers)**

```bash
# Install MPI if needed: sudo apt install libopenmpi-dev
cargo check -p clawhdf5-io --features mpi-io 2>&1 | tail -20
```

Expected: clean compile (warnings OK; errors not OK).

- [x] **Step 7: Commit**

```bash
git add crates/clawhdf5-io/Cargo.toml \
        crates/clawhdf5-io/src/mpi_vol.rs \
        crates/clawhdf5-io/src/lib.rs
git commit -m "feat: add MpiVol VOL backend with collective MPI-IO (mpi-io feature)"
```

---

### Task 2: MPI-IO collective read integration test

**Background:** This test requires an MPI runtime (`mpirun`). It is gated by the `mpi-io` feature and validates that all MPI ranks receive identical data after a collective read.

**Files:**
- Modify: `crates/clawhdf5-io/src/mpi_vol.rs` (add integration test)

- [x] **Step 1: Add the integration test**

Inside the `#[cfg(test)]` block, add:

```rust
#[test]
#[cfg(feature = "mpi-io")]
fn collective_read_all_ranks_get_same_data() {
    use crate::vol::VirtualObjectLayer;
    use tempfile::TempDir;

    // Write a reference file using FileWriter (no MPI needed)
    let tmp = TempDir::new().unwrap();
    let path = tmp.path().join("test.h5");
    {
        use clawhdf5_format::file_writer::FileWriter as FmtWriter;
        let mut fw = FmtWriter::new();
        fw.create_dataset("temperature")
            .with_f64_data(&[1.0, 2.0, 3.0, 4.0, 5.0]);
        let bytes = fw.finish().unwrap();
        std::fs::write(&path, &bytes).unwrap();
    }

    // Each rank reads via MpiVol and should get the same bytes
    let mut vol = MpiVol::new_world().expect("MPI init failed");
    vol.open(path.to_str().unwrap()).unwrap();
    let data = vol.read_dataset("temperature").unwrap();

    // 5 f64 values = 40 bytes
    assert_eq!(data.len(), 40, "rank {} got {} bytes", vol.rank(), data.len());

    let values: Vec<f64> = data.chunks_exact(8)
        .map(|c| f64::from_le_bytes(c.try_into().unwrap()))
        .collect();
    assert_eq!(values, vec![1.0, 2.0, 3.0, 4.0, 5.0],
        "rank {} got wrong data", vol.rank());
}
```

Add to `Cargo.toml` dev-dependencies:

```toml
tempfile = "3"
```

- [x] **Step 2: Run without MPI feature (should compile-skip)**

```bash
cargo test -p clawhdf5-io 2>&1 | tail -10
```

Expected: all tests pass; `collective_read_all_ranks_get_same_data` is not compiled.

- [x] **Step 3: Run with MPI feature (requires mpirun)**

```bash
# Requires: sudo apt install libopenmpi-dev openmpi-bin
# cargo test compiles, then:
mpirun -np 4 cargo test -p clawhdf5-io --features mpi-io collective_read_all_ranks_get_same_data 2>&1
```

Expected: all 4 ranks PASS.

- [x] **Step 4: Commit**

```bash
git add crates/clawhdf5-io/src/mpi_vol.rs \
        crates/clawhdf5-io/Cargo.toml
git commit -m "feat: add MpiVol collective read integration test"
```

---

### Task 3: MPI-IO collective write integration test

**Background:** Validates that N ranks each contribute a shard of a dataset; rank 0 assembles and writes the complete file.

**Files:**
- Modify: `crates/clawhdf5-io/src/mpi_vol.rs`

- [x] **Step 1: Add the integration test**

```rust
#[test]
#[cfg(feature = "mpi-io")]
fn collective_write_assembles_all_shards() {
    use crate::vol::VirtualObjectLayer;
    use tempfile::TempDir;
    use mpi::traits::*;

    let tmp = TempDir::new().unwrap();
    let path = tmp.path().join("parallel_out.h5");

    let mut vol = MpiVol::new_world().expect("MPI init failed");
    vol.open(path.to_str().unwrap()).unwrap();

    let world = vol.universe.world();
    let rank = world.rank() as usize;
    // Each rank contributes one f64 value: rank * 10.0
    let shard = ((rank as f64) * 10.0f64).to_le_bytes().to_vec();

    vol.write_dataset("values", &shard, &[world.size() as u64], "f64")
        .unwrap();

    // All ranks verify the written file has 4 values (one per rank)
    let total_size = world.size() as usize;
    if rank == 0 {
        let bytes = std::fs::read(&path).unwrap();
        use clawhdf5_format::{
            data_layout::DataLayout, data_read::read_raw_data_full,
            dataspace::Dataspace, datatype::Datatype,
            group_v2::resolve_path_any, message_type::MessageType,
            object_header::ObjectHeader, signature::find_signature,
            superblock::Superblock,
        };
        let sig = find_signature(&bytes).unwrap();
        let sb = Superblock::parse(&bytes, sig).unwrap();
        let addr = resolve_path_any(&bytes, &sb, "values").unwrap();
        let oh = ObjectHeader::parse(
            &bytes, addr as usize, sb.offset_size, sb.length_size,
        ).unwrap();
        let (dt, _) = Datatype::parse(
            &oh.messages.iter().find(|m| m.msg_type == MessageType::Datatype).unwrap().data,
        ).unwrap();
        let ds = Dataspace::parse(
            &oh.messages.iter().find(|m| m.msg_type == MessageType::Dataspace).unwrap().data,
            sb.length_size,
        ).unwrap();
        let dl = DataLayout::parse(
            &oh.messages.iter().find(|m| m.msg_type == MessageType::DataLayout).unwrap().data,
            sb.offset_size, sb.length_size,
        ).unwrap();
        let raw = read_raw_data_full(
            &bytes, &dl, &ds, &dt, None, sb.offset_size, sb.length_size,
        ).unwrap();
        assert_eq!(raw.len(), total_size * 8, "expected {} f64 values", total_size);
        let values: Vec<f64> = raw.chunks_exact(8)
            .map(|c| f64::from_le_bytes(c.try_into().unwrap()))
            .collect();
        for (i, &v) in values.iter().enumerate() {
            assert!((v - (i as f64 * 10.0)).abs() < 1e-9,
                "rank {i} shard wrong: got {v}");
        }
    }
    world.barrier();
}
```

- [x] **Step 2: Run**

```bash
cargo test -p clawhdf5-io 2>&1 | tail -5                            # no feature — should pass
mpirun -np 4 cargo test -p clawhdf5-io --features mpi-io collective_write 2>&1
```

- [x] **Step 3: Commit**

```bash
git add crates/clawhdf5-io/src/mpi_vol.rs
git commit -m "feat: add MpiVol collective write integration test (4 ranks)"
```

---

### Task 4: MpiVol parallel benchmark binary

**Background:** Adds a benchmark binary to `clawhdf5-bench` that runs h5bench-equivalent write/read workloads using `MpiVol`. This provides the throughput numbers needed to compare clawhdf5 against standard libhdf5 + h5bench.

**Files:**
- Create: `crates/clawhdf5-bench/src/bin/mpi_io_bench.rs`
- Modify: `crates/clawhdf5-bench/Cargo.toml` (add `mpi-io` feature, `mpi_io_bench` binary)

**Produces:** `cargo run -p clawhdf5-bench --features mpi-io --bin mpi_io_bench -- --size 100000` outputs MB/s throughput numbers comparable to h5bench output.

- [x] **Step 1: Create the binary**

Create `crates/clawhdf5-bench/src/bin/mpi_io_bench.rs`:

```rust
//! h5bench-equivalent MPI-IO performance benchmark.
//!
//! Usage: mpirun -np N cargo run -p clawhdf5-bench --features mpi-io --bin mpi_io_bench -- --size <N>
//!
//! Measures collective write and read throughput in MB/s for f64 arrays.

#[cfg(feature = "mpi-io")]
fn main() {
    use clawhdf5_io::mpi_vol::MpiVol;
    use clawhdf5_io::vol::VirtualObjectLayer;
    use std::time::Instant;
    use mpi::traits::*;

    let args: Vec<String> = std::env::args().collect();
    let n_elements: usize = args.iter()
        .position(|a| a == "--size")
        .and_then(|i| args.get(i + 1))
        .and_then(|s| s.parse().ok())
        .unwrap_or(100_000);

    let mut vol = MpiVol::new_world().expect("MPI init failed");
    let world = vol.universe.world();
    let rank = world.rank() as usize;
    let size = world.size() as usize;

    let path = format!("/tmp/clawhdf5_mpiio_bench_{n_elements}.h5");
    vol.open(&path).unwrap();

    // Each rank contributes n_elements/size f64 values
    let per_rank = n_elements / size;
    let shard: Vec<f64> = (0..per_rank).map(|i| (rank * per_rank + i) as f64).collect();
    let shard_bytes: Vec<u8> = shard.iter().flat_map(|v| v.to_le_bytes()).collect();

    // Collective write
    world.barrier();
    let t0 = Instant::now();
    vol.write_dataset("data", &shard_bytes, &[n_elements as u64], "f64").unwrap();
    world.barrier();
    let write_elapsed = t0.elapsed().as_secs_f64();

    // Collective read
    let t1 = Instant::now();
    let _data = vol.read_dataset("data").unwrap();
    world.barrier();
    let read_elapsed = t1.elapsed().as_secs_f64();

    if rank == 0 {
        let total_mb = (n_elements * 8) as f64 / 1e6;
        println!("=== clawhdf5 MPI-IO Benchmark ===");
        println!("Elements : {n_elements}");
        println!("Ranks    : {size}");
        println!("Total    : {total_mb:.1} MB");
        println!("Write    : {:.1} MB/s", total_mb / write_elapsed);
        println!("Read     : {:.1} MB/s", total_mb / read_elapsed);
    }
}

#[cfg(not(feature = "mpi-io"))]
fn main() {
    eprintln!("mpi_io_bench requires the `mpi-io` feature.");
    eprintln!("Run: mpirun -np N cargo run -p clawhdf5-bench --features mpi-io --bin mpi_io_bench");
    std::process::exit(1);
}
```

- [x] **Step 2: Add to Cargo.toml**

In `crates/clawhdf5-bench/Cargo.toml`, add:

```toml
[dependencies]
clawhdf5-io = { path = "../clawhdf5-io", features = [] }

[features]
mpi-io = ["clawhdf5-io/mpi-io", "mpi"]

[dependencies.mpi]
version = "0.8"
optional = true

[[bin]]
name = "mpi_io_bench"
path = "src/bin/mpi_io_bench.rs"
```

- [x] **Step 3: Verify it compiles**

```bash
cargo check -p clawhdf5-bench --features mpi-io 2>&1 | tail -10
```

Expected: no errors.

- [x] **Step 4: Run with 4 ranks**

```bash
mpirun -np 4 cargo run --release -p clawhdf5-bench --features mpi-io --bin mpi_io_bench -- --size 1000000 2>&1
```

Expected output (numbers will vary by hardware):
```
=== clawhdf5 MPI-IO Benchmark ===
Elements : 1000000
Ranks    : 4
Total    : 8.0 MB
Write    : xxx.x MB/s
Read     : xxx.x MB/s
```

Record results in `BENCHMARKS.md` under a new `## MPI-IO Parallel I/O` section.

- [x] **Step 5: Commit**

```bash
git add crates/clawhdf5-bench/src/bin/mpi_io_bench.rs \
        crates/clawhdf5-bench/Cargo.toml
git commit -m "feat: add mpi_io_bench binary for h5bench-comparable parallel I/O throughput"
```

---

## Verification

```bash
# Without MPI feature — all existing tests still pass
cargo test -p clawhdf5-io 2>&1 | tail -10

# With MPI feature — compile check (requires libopenmpi-dev)
cargo check -p clawhdf5-io --features mpi-io 2>&1 | tail -5

# Integration tests (requires openmpi-bin)
mpirun -np 4 cargo test -p clawhdf5-io --features mpi-io 2>&1 | tail -20

# Benchmark (requires openmpi-bin)
mpirun -np 4 cargo run --release -p clawhdf5-bench --features mpi-io --bin mpi_io_bench -- --size 1000000 2>&1
```