test/perf/msgpass: ubench a producer-consumer app

Approximate a message-passing application as a set of producers, a set of
consumers, and a set of proxies that do both.  We'll use this for some initial
insight for #634 but it seems worth
having in general.

src/test/perf/msgpass/msgpass.cc

@@ -0,0 +1,244 @@
+/**
+ * A simulation of a message-passing application workload for snmalloc.
+ *
+ * - N_PRODUCER producer threads allocate and queue spans of messages randomly,
+ * - to N_CONSUMER consumer threads, which dequeue messages and free() them.
+ *
+ * Optionally, N_PROXY threads act as both producers and consumers, forwarding
+ * received messages back to another queue rather than freeing them.
+ */
+
+#include "test/opt.h"
+#include "test/setup.h"
+#include "test/usage.h"
+#include "test/xoroshiro.h"
+
+constexpr static bool use_malloc = false;
+constexpr static bool be_chatty = false;
+
+#include <chrono>
+#include <iomanip>
+#include <iostream>
+#include <snmalloc/snmalloc.h>
+#include <stdarg.h>
+#include <thread>
+#include <vector>
+
+constexpr static size_t N_PRODUCER = 3;
+constexpr static size_t N_CONSUMER = 3;
+constexpr static size_t N_PROXY = 2;
+constexpr static size_t N_QUEUE = N_CONSUMER + N_PROXY;
+
+constexpr static size_t N_PRODUCER_BATCH = 256 * 1024;
+constexpr static size_t N_MAX_OUTSTANDING = 4 * 1024;
+
+using namespace snmalloc;
+
+void chatty(const char* p, ...)
+{
+  if constexpr (be_chatty)
+  {
+    va_list va;
+    va_start(va, p);
+    vfprintf(stderr, p, va);
+    va_end(va);
+  }
+}
+
+/*
+ * RemoteAllocators make for convenient MPSC queues, so we use those for sending
+ * "messages".  Each consumer or proxy has its own (source) queue.
+ */
+RemoteAllocator msgqueue[N_QUEUE];
+
+std::atomic<bool> producers_live;
+std::atomic<size_t> queue_gate;
+std::atomic<size_t> messages_outstanding;
+
+freelist::HeadPtr domesticate_nop(freelist::QueuePtr p)
+{
+  return freelist::HeadPtr::unsafe_from(p.unsafe_ptr());
+};
+
+void consumer(size_t qix)
+{
+  auto& a = ThreadAlloc::get();
+  auto& myq = msgqueue[qix];
+  size_t reap;
+
+  chatty("Cl %zu q is %p\n", qix, &myq);
+
+  do
+  {
+    reap = 0;
+
+    if (myq.can_dequeue(domesticate_nop))
+    {
+      myq.dequeue(
+        domesticate_nop,
+        domesticate_nop,
+        [qix, &a, &reap](freelist::HeadPtr o) {
+          UNUSED(qix);
+          auto p = o.as_void().unsafe_ptr();
+          chatty("Cl %zu free %p\n", qix, p);
+          (use_malloc ? free(p) : a.dealloc(p));
+          reap++;
+          return true;
+        });
+    }
+
+    messages_outstanding -= reap;
+
+    if (reap == 0)
+    {
+      std::this_thread::yield();
+    }
+    else
+    {
+      chatty("Cl %zu reap %zu\n", qix, reap);
+    }
+
+  } while (reap || producers_live || (queue_gate > N_CONSUMER));
+
+  chatty("Cl %zu fini\n", qix);
+}
+
+void proxy(size_t qix)
+{
+  auto& myq = msgqueue[qix];
+  bool any;
+
+  chatty("Px %zu q is %p\n", qix, &myq);
+
+  xoroshiro::p128r32 r(1234 + qix, qix);
+  do
+  {
+    any = false;
+
+    if (myq.can_dequeue(domesticate_nop))
+    {
+      myq.dequeue(
+        domesticate_nop, domesticate_nop, [qix, &r, &any](freelist::HeadPtr o) {
+          auto rcptqix = r.next() % qix;
+
+          chatty(
+            "Px %zu send %p to %zu\n", qix, o.as_void().unsafe_ptr(), rcptqix);
+
+          msgqueue[rcptqix].enqueue(o, o, domesticate_nop);
+          any = true;
+          return true;
+        });
+    }
+
+    std::this_thread::yield();
+  } while (any || producers_live || (queue_gate > qix + 1));
+
+  chatty("Px %zu fini\n", qix);
+  queue_gate--;
+}
+
+freelist::Builder<false> batch[N_PRODUCER];
+
+void producer(size_t pix)
+{
+  auto& a = ThreadAlloc::get();
+  static constexpr size_t msgsizes[] = {48, 64, 96, 128};
+  static constexpr size_t nmsgsizes = sizeof(msgsizes) / sizeof(msgsizes[0]);
+
+  xoroshiro::p128r32 r(5489 + pix, pix);
+  batch[pix].init(0, RemoteAllocator::key_global);
+
+  for (size_t batchix = N_PRODUCER_BATCH; batchix > 0; batchix--)
+  {
+    while (messages_outstanding >= N_MAX_OUTSTANDING)
+    {
+      std::this_thread::yield();
+    }
+
+    size_t nmsg = (r.next() & 15) + 1;
+    size_t msgsize = msgsizes[r.next() % nmsgsizes];
+
+    /* Allocate batch and form list */
+    for (size_t msgix = 0; msgix < nmsg; msgix++)
+    {
+      auto msg = (use_malloc ? malloc(msgsize) : a.alloc(msgsize));
+      chatty("Pd %zu make %p\n", pix, msg);
+
+      auto msgc = capptr::Alloc<void>::unsafe_from(msg)
+                    .template as_reinterpret<freelist::Object::T<>>();
+      batch[pix].add(msgc, RemoteAllocator::key_global);
+    }
+
+    /* Post to random queue */
+    auto [bfirst, blast] =
+      batch[pix].extract_segment(RemoteAllocator::key_global);
+    auto rcptqix = r.next() % N_QUEUE;
+    msgqueue[rcptqix].enqueue(bfirst, blast, domesticate_nop);
+    messages_outstanding += nmsg;
+
+    chatty("Pd %zu send %zu to %zu\n", pix, nmsg, rcptqix);
+
+    /* Occasionally yield the CPU */
+    if ((batchix & 0xF) == 1)
+      std::this_thread::yield();
+  }
+
+  chatty("Pd %zu fini\n", pix);
+}
+
+int main()
+{
+  std::thread producer_threads[N_PRODUCER];
+  std::thread queue_threads[N_QUEUE];
+
+  /*
+   * Force snmalloc to initialize itself, and especially
+   * RemoteAllocator::key_global, now, before we initialize our message queues.
+   * (Otherwise, we'll pick up the wrong key and pointers will fail to decode
+   * correctly and we will be sad for hours.)
+   */
+  auto& a = ThreadAlloc::get();
+  a.dealloc(a.alloc(32));
+
+  for (size_t i = 0; i < N_QUEUE; i++)
+  {
+    msgqueue[i].init();
+  }
+
+  producers_live = true;
+  queue_gate = N_QUEUE;
+  messages_outstanding = 0;
+
+  /* Spawn consumers */
+  for (size_t i = 0; i < N_CONSUMER; i++)
+  {
+    queue_threads[i] = std::thread(consumer, i);
+  }
+
+  /* Spawn proxies */
+  for (size_t i = N_CONSUMER; i < N_QUEUE; i++)
+  {
+    queue_threads[i] = std::thread(proxy, i);
+  }
+
+  /* Spawn producers */
+  for (size_t i = 0; i < N_PRODUCER; i++)
+  {
+    producer_threads[i] = std::thread(producer, i);
+  }
+
+  /* Wait for producers to finish */
+  for (size_t i = 0; i < N_PRODUCER; i++)
+  {
+    producer_threads[i].join();
+  }
+  producers_live = false;
+
+  /* Wait for proxies and consumers to finish */
+  for (size_t i = 0; i < N_QUEUE; i++)
+  {
+    queue_threads[N_QUEUE - 1 - i].join();
+  }
+
+  return 0;
+}