guile/mt-gcbench.c

// This is adapted from a benchmark written by John Ellis and Pete Kovac
// of Post Communications.
// It was modified by Hans Boehm of Silicon Graphics.
// Translated to C++ 30 May 1997 by William D Clinger of Northeastern Univ.
// Translated to C 15 March 2000 by Hans Boehm, now at HP Labs.
//
//      This is no substitute for real applications.  No actual application
//      is likely to behave in exactly this way.  However, this benchmark was
//      designed to be more representative of real applications than other
//      Java GC benchmarks of which we are aware.
//      It attempts to model those properties of allocation requests that
//      are important to current GC techniques.
//      It is designed to be used either to obtain a single overall performance
//      number, or to give a more detailed estimate of how collector
//      performance varies with object lifetimes.  It prints the time
//      required to allocate and collect balanced binary trees of various
//      sizes.  Smaller trees result in shorter object lifetimes.  Each cycle
//      allocates roughly the same amount of memory.
//      Two data structures are kept around during the entire process, so
//      that the measured performance is representative of applications
//      that maintain some live in-memory data.  One of these is a tree
//      containing many pointers.  The other is a large array containing
//      double precision floating point numbers.  Both should be of comparable
//      size.
//
//      The results are only really meaningful together with a specification
//      of how much memory was used.  It is possible to trade memory for
//      better time performance.  This benchmark should be run in a 32 MB
//      heap, though we don't currently know how to enforce that uniformly.
//
//      Unlike the original Ellis and Kovac benchmark, we do not attempt
//      measure pause times.  This facility should eventually be added back
//      in.  There are several reasons for omitting it for now.  The original
//      implementation depended on assumptions about the thread scheduler
//      that don't hold uniformly.  The results really measure both the
//      scheduler and GC.  Pause time measurements tend to not fit well with
//      current benchmark suites.  As far as we know, none of the current
//      commercial Java implementations seriously attempt to minimize GC pause
//      times.

#include <errno.h>
#include <pthread.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/time.h>

#include "assert.h"
#include "gc-api.h"
#include "mt-gcbench-types.h"
#include "simple-roots-api.h"
#include "simple-allocator.h"

#define MAX_THREAD_COUNT 256

static const int long_lived_tree_depth = 16; // about 4Mb
static const int array_size = 500000; // about 4Mb
static const int min_tree_depth = 4;
static const int max_tree_depth = 16;

typedef HANDLE_TO(Node) NodeHandle;
typedef HANDLE_TO(DoubleArray) DoubleArrayHandle;

static Node* allocate_node(struct gc_mutator *mut) {
  // memset to 0 by the collector.
  return gc_allocate_with_kind(mut, ALLOC_KIND_NODE, sizeof (Node));
}

static DoubleArray* allocate_double_array(struct gc_mutator *mut,
                                                 size_t size) {
  // May be uninitialized.
  size_t bytes = sizeof(DoubleArray) + sizeof (double) * size;
  DoubleArray *ret =
    gc_allocate_pointerless_with_kind(mut, ALLOC_KIND_DOUBLE_ARRAY, bytes);
  ret->length = size;
  return ret;
}

static Hole* allocate_hole(struct gc_mutator *mut, size_t size) {
  size_t bytes = sizeof(Hole) + sizeof (uintptr_t) * size;
  Hole *ret = gc_allocate_with_kind(mut, ALLOC_KIND_HOLE, bytes);
  ret->length = size;
  return ret;
}

static unsigned long current_time(void) {
  struct timeval t = { 0 };
  gettimeofday(&t, NULL);
  return t.tv_sec * 1000 * 1000 + t.tv_usec;
}

static double elapsed_millis(unsigned long start) {
  return (current_time() - start) * 1e-3;
}

// Nodes used by a tree of a given size
static int tree_size(int i) {
  return ((1 << (i + 1)) - 1);
}

// Number of iterations to use for a given tree depth
static int compute_num_iters(int i) {
  return 2 * tree_size(max_tree_depth + 2) / tree_size(i);
}

// A power-law distribution.  Each integer was selected by starting at 0, taking
// a random number in [0,1), and then accepting the integer if the random number
// was less than 0.15, or trying again with the next integer otherwise.  Useful
// for modelling allocation sizes or number of garbage objects to allocate
// between live allocations.
static const uint8_t power_law_distribution[256] = {
  1, 15, 3, 12, 2, 8, 4, 0, 18, 7, 9, 8, 15, 2, 36, 5,
  1, 9, 6, 11, 9, 19, 2, 0, 0, 3, 9, 6, 3, 2, 1, 1,
  6, 1, 8, 4, 2, 0, 5, 3, 7, 0, 0, 3, 0, 4, 1, 7,
  1, 8, 2, 2, 2, 14, 0, 7, 8, 0, 2, 1, 4, 12, 7, 5,
  0, 3, 4, 13, 10, 2, 3, 7, 0, 8, 0, 23, 0, 16, 1, 1,
  6, 28, 1, 18, 0, 3, 6, 5, 8, 6, 14, 5, 2, 5, 0, 11,
  0, 18, 4, 16, 1, 4, 3, 13, 3, 23, 7, 4, 10, 5, 3, 13,
  0, 14, 5, 5, 2, 5, 0, 16, 2, 0, 1, 1, 0, 0, 4, 2,
  7, 7, 0, 5, 7, 2, 1, 24, 27, 3, 7, 1, 0, 8, 1, 4,
  0, 3, 0, 7, 7, 3, 9, 2, 9, 2, 5, 10, 1, 1, 12, 6,
  2, 9, 5, 0, 4, 6, 0, 7, 2, 1, 5, 4, 1, 0, 1, 15,
  4, 0, 15, 4, 0, 0, 32, 18, 2, 2, 1, 7, 8, 3, 11, 1,
  2, 7, 11, 1, 9, 1, 2, 6, 11, 17, 1, 2, 5, 1, 14, 3,
  6, 1, 1, 15, 3, 1, 0, 6, 10, 8, 1, 3, 2, 7, 0, 1,
  0, 11, 3, 3, 5, 8, 2, 0, 0, 7, 12, 2, 5, 20, 3, 7,
  4, 4, 5, 22, 1, 5, 2, 7, 15, 2, 4, 6, 11, 8, 12, 1
};

static size_t power_law(size_t *counter) {
  return power_law_distribution[(*counter)++ & 0xff];
}

struct thread {
  struct gc_mutator *mut;
  struct gc_mutator_roots roots;
  size_t counter;
};

static void allocate_garbage(struct thread *t) {
  size_t hole = power_law(&t->counter);
  if (hole) {
    allocate_hole(t->mut, hole);
  }
}

static void set_field(Node *obj, Node **field, Node *val) {
  gc_small_write_barrier(gc_ref_from_heap_object(obj),
                         gc_edge(field),
                         gc_ref_from_heap_object(val));
  *field = val;
}

// Build tree top down, assigning to older objects.
static void populate(struct thread *t, int depth, Node *node) {
  struct gc_mutator *mut = t->mut;
  if (depth <= 0)
    return;

  NodeHandle self = { node };
  PUSH_HANDLE(t, self);
  allocate_garbage(t);
  NodeHandle l = { allocate_node(mut) };
  PUSH_HANDLE(t, l);
  allocate_garbage(t);
  NodeHandle r = { allocate_node(mut) };
  PUSH_HANDLE(t, r);

  set_field(HANDLE_REF(self), &HANDLE_REF(self)->left, HANDLE_REF(l));
  set_field(HANDLE_REF(self), &HANDLE_REF(self)->right, HANDLE_REF(r));
  // i is 0 because the memory is zeroed.
  HANDLE_REF(self)->j = depth;

  populate(t, depth-1, HANDLE_REF(self)->left);
  populate(t, depth-1, HANDLE_REF(self)->right);

  POP_HANDLE(t);
  POP_HANDLE(t);
  POP_HANDLE(t);
}

// Build tree bottom-up
static Node* make_tree(struct thread *t, int depth) {
  struct gc_mutator *mut = t->mut;
  if (depth <= 0)
    return allocate_node(mut);

  NodeHandle left = { make_tree(t, depth-1) };
  PUSH_HANDLE(t, left);
  NodeHandle right = { make_tree(t, depth-1) };
  PUSH_HANDLE(t, right);

  allocate_garbage(t);
  Node *result = allocate_node(mut);
  result->left = HANDLE_REF(left);
  result->right = HANDLE_REF(right);
  // i is 0 because the memory is zeroed.
  result->j = depth;

  POP_HANDLE(t);
  POP_HANDLE(t);

  return result;
}

static void validate_tree(Node *tree, int depth) {
#ifndef NDEBUG
  ASSERT_EQ(tree->i, 0);
  ASSERT_EQ(tree->j, depth);
  if (depth == 0) {
    ASSERT(!tree->left);
    ASSERT(!tree->right);
  } else {
    ASSERT(tree->left);
    ASSERT(tree->right);
    validate_tree(tree->left, depth - 1);
    validate_tree(tree->right, depth - 1);
  }
#endif
}

static void time_construction(struct thread *t, int depth) {
  struct gc_mutator *mut = t->mut;
  int num_iters = compute_num_iters(depth);
  NodeHandle temp_tree = { NULL };
  PUSH_HANDLE(t, temp_tree);

  printf("Creating %d trees of depth %d\n", num_iters, depth);

  {
    unsigned long start = current_time();
    for (int i = 0; i < num_iters; ++i) {
      HANDLE_SET(temp_tree, allocate_node(mut));
      populate(t, depth, HANDLE_REF(temp_tree));
      validate_tree(HANDLE_REF(temp_tree), depth);
      HANDLE_SET(temp_tree, NULL);
    }
    printf("\tTop down construction took %.3f msec\n",
           elapsed_millis(start));
  }

  {
    long start = current_time();
    for (int i = 0; i < num_iters; ++i) {
      HANDLE_SET(temp_tree, make_tree(t, depth));
      validate_tree(HANDLE_REF(temp_tree), depth);
      HANDLE_SET(temp_tree, NULL);
    }
    printf("\tBottom up construction took %.3f msec\n",
           elapsed_millis(start));
  }

  POP_HANDLE(t);
}

struct call_with_gc_data {
  void* (*f)(struct thread *);
  struct gc_heap *heap;
};
static void* call_with_gc_inner(struct gc_stack_addr *addr, void *arg) {
  struct call_with_gc_data *data = arg;
  struct gc_mutator *mut = gc_init_for_thread(addr, data->heap);
  struct thread t = { mut, };
  gc_mutator_set_roots(mut, &t.roots);
  void *ret = data->f(&t);
  gc_finish_for_thread(mut);
  return ret;
}
static void* call_with_gc(void* (*f)(struct thread *),
                          struct gc_heap *heap) {
  struct call_with_gc_data data = { f, heap };
  return gc_call_with_stack_addr(call_with_gc_inner, &data);
}

static void* run_one_test(struct thread *t) {
  NodeHandle long_lived_tree = { NULL };
  NodeHandle temp_tree = { NULL };
  DoubleArrayHandle array = { NULL };

  PUSH_HANDLE(t, long_lived_tree);
  PUSH_HANDLE(t, temp_tree);
  PUSH_HANDLE(t, array);

  // Create a long lived object
  printf(" Creating a long-lived binary tree of depth %d\n",
         long_lived_tree_depth);
  HANDLE_SET(long_lived_tree, allocate_node(t->mut));
  populate(t, long_lived_tree_depth, HANDLE_REF(long_lived_tree));

  // Create long-lived array, filling half of it
  printf(" Creating a long-lived array of %d doubles\n", array_size);
  HANDLE_SET(array, allocate_double_array(t->mut, array_size));
  for (int i = 0; i < array_size/2; ++i) {
    HANDLE_REF(array)->values[i] = 1.0/i;
  }

  for (int d = min_tree_depth; d <= max_tree_depth; d += 2) {
    time_construction(t, d);
  }

  validate_tree(HANDLE_REF(long_lived_tree), long_lived_tree_depth);

  // Fake reference to LongLivedTree and array to keep them from being optimized
  // away.
  if (HANDLE_REF(long_lived_tree)->i != 0
      || HANDLE_REF(array)->values[1000] != 1.0/1000)
    fprintf(stderr, "Failed\n");

  POP_HANDLE(t);
  POP_HANDLE(t);
  POP_HANDLE(t);
  return NULL;
}

static void* run_one_test_in_thread(void *arg) {
  struct gc_heap *heap = arg;
  return call_with_gc(run_one_test, heap);
}

struct join_data { int status; pthread_t thread; };
static void *join_thread(void *data) {
  struct join_data *join_data = data;
  void *ret;
  join_data->status = pthread_join(join_data->thread, &ret);
  return ret;
}

int main(int argc, char *argv[]) {
  size_t sizeof_double_array = sizeof(size_t);
  size_t heap_max_live =
    tree_size(long_lived_tree_depth) * sizeof(Node) +
    tree_size(max_tree_depth) * sizeof(Node) +
    sizeof(DoubleArray) + sizeof(double) * array_size;
  if (argc != 4) {
    fprintf(stderr, "usage: %s MULTIPLIER NTHREADS PARALLELISM\n", argv[0]);
    return 1;
  }

  double multiplier = atof(argv[1]);
  size_t nthreads = atol(argv[2]);
  size_t parallelism = atol(argv[3]);

  if (!(0.1 < multiplier && multiplier < 100)) {
    fprintf(stderr, "Failed to parse heap multiplier '%s'\n", argv[1]);
    return 1;
  }
  if (nthreads < 1 || nthreads > MAX_THREAD_COUNT) {
    fprintf(stderr, "Expected integer between 1 and %d for thread count, got '%s'\n",
            (int)MAX_THREAD_COUNT, argv[2]);
    return 1;
  }
  if (parallelism < 1 || parallelism > MAX_THREAD_COUNT) {
    fprintf(stderr, "Expected integer between 1 and %d for parallelism, got '%s'\n",
            (int)MAX_THREAD_COUNT, argv[3]);
    return 1;
  }

  size_t heap_size = heap_max_live * multiplier * nthreads;

  struct gc_options *options = gc_allocate_options();
  gc_options_set_int(options, GC_OPTION_HEAP_SIZE_POLICY, GC_HEAP_SIZE_FIXED);
  gc_options_set_size(options, GC_OPTION_HEAP_SIZE, heap_size);
  gc_options_set_int(options, GC_OPTION_PARALLELISM, parallelism);

  struct gc_heap *heap;
  struct gc_mutator *mut;
  if (!gc_init(options, NULL, &heap, &mut)) {
    fprintf(stderr, "Failed to initialize GC with heap size %zu bytes\n",
            heap_size);
    return 1;
  }
  struct thread main_thread = { mut, };
  gc_mutator_set_roots(mut, &main_thread.roots);

  printf("Garbage Collector Test\n");
  printf(" Live storage will peak at %zd bytes.\n\n", heap_max_live);

  unsigned long start = current_time();
        
  pthread_t threads[MAX_THREAD_COUNT];
  // Run one of the threads in the main thread.
  for (size_t i = 1; i < nthreads; i++) {
    int status = pthread_create(&threads[i], NULL, run_one_test_in_thread, heap);
    if (status) {
      errno = status;
      perror("Failed to create thread");
      return 1;
    }
  }
  run_one_test(&main_thread);
  for (size_t i = 1; i < nthreads; i++) {
    struct join_data data = { 0, threads[i] };
    gc_call_without_gc(mut, join_thread, &data);
    if (data.status) {
      errno = data.status;
      perror("Failed to join thread");
      return 1;
    }
  }
  
  printf("Completed in %.3f msec\n", elapsed_millis(start));
  gc_print_stats(heap);
}