End-to-end multi-replica case.

kernel/src/process.rs

@@ -51,7 +51,7 @@ pub(crate) type Eid = usize;
 pub(crate) const MAX_PROCESSES: usize = 12;
 
 /// How many registered "named" frames a process can have.
-pub(crate) const MAX_FRAMES_PER_PROCESS: usize = 1023;
+pub(crate) const MAX_FRAMES_PER_PROCESS: usize = 2047;
 
 /// How many writable sections a process can have (part of the ELF file).
 pub(crate) const MAX_WRITEABLE_SECTIONS_PER_PROCESS: usize = 4;
@@ -175,7 +175,7 @@ impl FrameManagement for ProcessFrames {
         // If `self.frames` is too big, the O(n) lookup in this fn might become
         // a problem. better to implement some reverse-map for PAddr -> FrameId
         // then.
-        static_assertions::const_assert!(MAX_FRAMES_PER_PROCESS < 1024);
+        static_assertions::const_assert!(MAX_FRAMES_PER_PROCESS < 2048);
 
         for (frame, ref mut refcnt) in self.frames.iter_mut() {
             if let Some(frame) = frame {

kernel/tests/s11_rackscale_benchmarks.rs

@@ -1087,12 +1087,49 @@ fn s11_rackscale_dynrep_userspace() {
     fn controller_match_fn(
         proc: &mut PtySession,
         output: &mut String,
-        _cores_per_client: usize,
-        _num_clients: usize,
-        _file_name: &str,
+        cores_per_client: usize,
+        num_clients: usize,
+        file_name: &str,
         _is_baseline: bool,
         _arg: Option<()>,
     ) -> Result<()> {
+        let expected_lines = num_clients * cores_per_client * 5;
+        for _i in 0..expected_lines {
+            let (prev, matched) =
+                proc.exp_regex(r#"init::dynrep: (.*),(\d+),(\d+),(\d+),(\d+)"#)?;
+            *output += prev.as_str();
+            *output += matched.as_str();
+
+            // Append parsed results to a CSV file
+            let write_headers = !Path::new(file_name).exists();
+            let mut csv_file = OpenOptions::new()
+                .append(true)
+                .create(true)
+                .open(file_name)
+                .expect("Can't open file");
+            if write_headers {
+                let row = "git_rev,nclients,cores_per_client,benchmark,machine_id,thread_id,time,operations\n";
+                let r = csv_file.write(row.as_bytes());
+                assert!(r.is_ok());
+            }
+
+            let parts: Vec<&str> = matched.split("init::dynrep: ").collect();
+            assert!(parts.len() >= 2);
+
+            let r = csv_file.write(format!("{},", env!("GIT_HASH")).as_bytes());
+            assert!(r.is_ok());
+
+            let r = csv_file.write(format!("{},", num_clients).as_bytes());
+            assert!(r.is_ok());
+
+            let r = csv_file.write(format!("{},", cores_per_client).as_bytes());
+            assert!(r.is_ok());
+
+            let r = csv_file.write(parts[1].as_bytes());
+            assert!(r.is_ok());
+            let r = csv_file.write("\n".as_bytes());
+            assert!(r.is_ok());
+        }
         *output += proc.exp_string("dynrep_test OK")?.as_str();
         Ok(())
     }

usr/init/src/dynrep/allocator.rs

@@ -1,60 +1,65 @@
 use core::slice::from_raw_parts_mut;
 use x86::bits64::paging::LARGE_PAGE_SIZE;
+use lazy_static::lazy_static;
+use core::sync::atomic::AtomicU8;
+use core::sync::atomic::Ordering;
 
+use log::info;
 use core::{
     alloc::{AllocError, Allocator, Layout},
     ptr::NonNull,
 };
-use log::info;
 
 pub const BASE: u64 = 0x0510_0000_0000;
-pub const MAX_FRAMES: u64 = 600;
+pub const MAX_FRAMES: u64 = 200;
+
+static NODE_ID: AtomicU8 = AtomicU8::new(1);
 
 #[derive(Clone, Copy)]
 pub struct MyAllocator;
 
-impl Default for MyAllocator {
-    fn default() -> Self {
+impl MyAllocator {
+    fn allocate_pages(node_id: u8) {
         let mut allocated = 0;
+        let node_offset = (node_id - 1) as u64 * LARGE_PAGE_SIZE as u64 * MAX_FRAMES;
         while allocated < MAX_FRAMES {
             // Allocate a large page of physical memory
             // Note that even if you allocate a base page, behind the scenes a large page is allocated
             // because DCM (and thus DiNOS) only allocates at large page granularity
             // 1 is the client machine id we want to allocate from
-            let (frame_id, paddr) = vibrio::syscalls::PhysicalMemory::allocate_large_page(1)
+            let (frame_id, paddr) = vibrio::syscalls::PhysicalMemory::allocate_large_page(node_id as usize)
                 .expect("Failed to get physical memory large page");
             info!("large frame id={:?}, paddr={:?}", frame_id, paddr);
 
             // Map allocated physical memory into user space so we can actually access it.
             unsafe {
-                vibrio::syscalls::VSpace::map_frame(
-                    frame_id,
-                    BASE + (allocated * LARGE_PAGE_SIZE as u64),
-                )
-                .expect("Failed to map base page");
+                vibrio::syscalls::VSpace::map_frame(frame_id, BASE + node_offset + (allocated * LARGE_PAGE_SIZE as u64)).expect("Failed to map base page");
             }
             allocated += 1;
         }
-        info!("# Allocated {} frames", allocated);
-        MyAllocator {}
+        info!("# Allocated {} frames on {}", allocated, node_id);
     }
 }
 
 unsafe impl Allocator for MyAllocator {
     fn allocate(&self, layout: Layout) -> Result<NonNull<[u8]>, AllocError> {
+        let node_id = NODE_ID.fetch_add(1, Ordering::SeqCst);
+        let node_offset = (node_id - 1) as u64 * LARGE_PAGE_SIZE as u64 * MAX_FRAMES;
+        MyAllocator::allocate_pages(node_id);
         info!("# Allocating {:?}", layout);
         if layout.size() > LARGE_PAGE_SIZE * MAX_FRAMES as usize {
             return Err(AllocError);
         }
 
-        let slice = unsafe { from_raw_parts_mut(BASE as *mut u8, layout.size()) };
-        // DO we need to zero the memory for mapping to work?
+        let slice = unsafe {from_raw_parts_mut((BASE + node_offset) as *mut u8, layout.size()) };
         Ok(NonNull::from(slice))
     }
 
     unsafe fn deallocate(&self, ptr: NonNull<u8>, layout: Layout) {
         info!("# Deallocating {:?}", layout);
-        vibrio::syscalls::VSpace::unmap(ptr.as_ptr() as u64, layout.size() as u64)
-            .expect("Failed to unmap base page");
+        /*for i in 0..MAX_FRAMES {
+            vibrio::syscalls::VSpace::unmap((BASE + (i * LARGE_PAGE_SIZE as u64)) as u64, LARGE_PAGE_SIZE as u64)
+                .expect("Failed to unmap base page");
+        }*/
     }
 }

usr/init/src/dynrep/mod.rs

@@ -8,14 +8,21 @@ use log::info;
 
 use hashbrown::{hash_map::DefaultHashBuilder, HashMap};
 
+use core::sync::atomic::AtomicUsize;
+use core::sync::atomic::Ordering;
+
 use lineup::tls2::{Environment, SchedulerControlBlock};
-use nr2::nr::{rwlock::RwLock, Dispatch, NodeReplicated};
+use nr2::nr::{Dispatch, NodeReplicated};
 use x86::bits64::paging::VAddr;
+use x86::random::rdrand64;
+use rawtime::Instant;
 
 mod allocator;
 use allocator::MyAllocator;
 
-pub const NUM_ENTRIES: u64 = 50_000_000;
+pub const NUM_ENTRIES: u64 = 500_000;
+
+static POOR_MANS_BARRIER: AtomicUsize = AtomicUsize::new(0);
 
 #[derive(Clone)]
 struct HashTable {
@@ -24,7 +31,7 @@ struct HashTable {
 
 impl Default for HashTable {
     fn default() -> Self {
-        let allocator = MyAllocator::default();
+        let allocator = MyAllocator{};
         let map = HashMap::<u64, u64, DefaultHashBuilder, MyAllocator>::with_capacity_in(
             NUM_ENTRIES as usize,
             allocator,
@@ -66,15 +73,44 @@ impl Dispatch for HashTable {
     }
 }
 
-unsafe extern "C" fn bencher_trampoline(_arg1: *mut u8) -> *mut u8 {
-    thread_routine();
-    ptr::null_mut()
+fn run_bench(mid: usize, core_id : usize, replica: Arc<NodeReplicated<HashTable>>) {
+    let ttkn = replica.register(mid - 1).unwrap();
+    let mut random_key :u64 = 0;
+    let batch_size = 64;
+    let duration = 5;
+
+    let mut iterations = 0;
+    while iterations <= duration {
+        let mut ops = 0;
+        let start = Instant::now();
+        while start.elapsed().as_secs() < 1 {
+        for i in 0..batch_size {
+                unsafe { rdrand64(&mut random_key) };
+                random_key = random_key % NUM_ENTRIES;
+                let _ = replica.execute(OpRd::Get(random_key), ttkn).unwrap();
+                ops += 1;
+        }
+        }
+        info!(
+            "dynhash,{},{},{},{}",mid,core_id, iterations, ops
+        );
+        iterations += 1;
+    }
 }
 
-fn thread_routine() {
+unsafe extern "C" fn bencher_trampoline(args: *mut u8) -> *mut u8 {
     let current_gtid = vibrio::syscalls::System::core_id().expect("Can't get core id");
     let mid = kpi::system::mid_from_gtid(current_gtid);
-    info!("I am thread {:?} and I am on node {:?}", current_gtid, mid);
+    let replica: Arc<NodeReplicated<HashTable>> = Arc::from_raw(args as *const NodeReplicated<HashTable>);
+
+    // Synchronize with all cores
+    POOR_MANS_BARRIER.fetch_sub(1, Ordering::Release);
+    while POOR_MANS_BARRIER.load(Ordering::Acquire) != 0 {
+        core::hint::spin_loop();
+    }
+
+    run_bench(mid, current_gtid, replica.clone());
+    ptr::null_mut()
 }
 
 pub fn userspace_dynrep_test() {
@@ -96,17 +132,11 @@ pub fn userspace_dynrep_test() {
 
     // Create data structure, with as many replicas as there are clients (assuming 1 numa node per client)
     // TODO: change # of replicas to nnodes
-    let replicas = NonZeroUsize::new(1).unwrap();
-    let nrht = NodeReplicated::<HashTable>::new(replicas, |_| 0).unwrap();
-
-    // TODO: populate data structure
-    let ttkn = nrht.register(0).unwrap();
-    nrht.execute(OpRd::Get(0), ttkn).unwrap();
+    let num_replicas = NonZeroUsize::new(nnodes).unwrap();
+    let replicas = Arc::new(NodeReplicated::<HashTable>::new(num_replicas, |_| 0).unwrap());
 
     let s = &vibrio::upcalls::PROCESS_SCHEDULER;
-
     let mut gtids = Vec::with_capacity(ncores);
-
     // We already have current core
     gtids.push(current_gtid);
 
@@ -137,13 +167,14 @@ pub fn userspace_dynrep_test() {
         32 * 4096,
         move |_| {
             let mut thandles = Vec::with_capacity(ncores);
+            POOR_MANS_BARRIER.store(ncores, Ordering::SeqCst);
 
             for core_index in 0..ncores {
                 thandles.push(
                     Environment::thread()
                         .spawn_on_core(
                             Some(bencher_trampoline),
-                            ncores as *mut u8,
+                            Arc::into_raw(replicas.clone()) as *const _ as *mut u8,
                             gtids[core_index],
                         )
                         .expect("Can't spawn bench thread?"),