Fcalva/cyclone
1
0
Fork 0
mirror of https://github.com/justinethier/cyclone.git synced 2025-05-29 06:55:06 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions
cyclone/gc.c
Justin Ethier 72eda149c5 Use minor GC code to move GC coop result to heap 
This allows the code to move any object to the heap, instead of the previous hack. There may be more overhead calling the full minor GC code, however, most objects should already have been moved and by definition this would only happen after a function has blocked for a length of time. So, both should minimize the additional overhead.
2016-01-04 22:39:37 -05:00

1316 lines
42 KiB
C
Raw Blame History

/**
* Cyclone Scheme
* Copyright (c) 2015-2016, Justin Ethier
* All rights reserved.
*
* Heap garbage collector used by the Cyclone runtime for major collections.
*
* Tracing GC algorithm is based on the one from "Implementing an on-the-fly
* garbage collector for Java", by Domani et al.
*
* The heap implementation (alloc / sweep, etc) is based on code from Chibi Scheme.
*
* Note there is also a minor GC (in runtime.c) that collects objects allocated
* on the stack, based on "Cheney on the MTA" (but without the copying collector).
*/
#include <ck_array.h>
#include <ck_pr.h>
#include "cyclone/types.h"
#include <time.h>
////////////////////
// Global variables
// Note: will need to use atomics and/or locking to access any
// variables shared between threads
static int gc_color_mark = 1; // Black, is swapped during GC
static int gc_color_clear = 3; // White, is swapped during GC
// unfortunately this had to be split up; const colors are located in types.h
static int gc_status_col = STATUS_SYNC1;
static int gc_stage = STAGE_RESTING;
// Does not need sync, only used by collector thread
static void **mark_stack = NULL;
static int mark_stack_len = 0;
static int mark_stack_i = 0;
// Lock to protect the heap from concurrent modifications
static pthread_mutex_t heap_lock;
// Cached heap statistics
// Note this assumes a single overall heap "chain". Code would need to
// be modified to support multiple independent heaps
static int cached_heap_free_size = 0;
static int cached_heap_total_size = 0;
// Data for each individual mutator thread
ck_array_t Cyc_mutators, old_mutators;
static pthread_mutex_t mutators_lock;
static void my_free(void *p, size_t m, bool d)
{
free(p);
return;
}
static void *my_malloc(size_t b)
{
return malloc(b);
}
static void *my_realloc(void *r, size_t a, size_t b, bool d)
{
return realloc(r, b);
}
static struct ck_malloc my_allocator = {
.malloc = my_malloc,
.free = my_free,
.realloc = my_realloc
};
/////////////
// Functions
// Perform one-time initialization before mutators can be executed
void gc_initialize()
{
if (ck_array_init(&Cyc_mutators, CK_ARRAY_MODE_SPMC, &my_allocator, 10) == 0){
fprintf(stderr, "Unable to initialize mutator array\n");
exit(1);
}
if (ck_array_init(&old_mutators, CK_ARRAY_MODE_SPMC, &my_allocator, 10) == 0){
fprintf(stderr, "Unable to initialize mutator array\n");
exit(1);
}
// Initialize collector's mark stack
mark_stack_len = 128;
mark_stack = vpbuffer_realloc(mark_stack, &(mark_stack_len));
// Here is as good a place as any to do this...
if (pthread_mutex_init(&(heap_lock), NULL) != 0) {
fprintf(stderr, "Unable to initialize heap_lock mutex\n");
exit(1);
}
if (pthread_mutex_init(&(mutators_lock), NULL) != 0) {
fprintf(stderr, "Unable to initialize mutators_lock mutex\n");
exit(1);
}
}
// Add data for a new mutator
void gc_add_mutator(gc_thread_data *thd)
{
pthread_mutex_lock(&mutators_lock);
if (ck_array_put_unique(&Cyc_mutators, (void *)thd) < 0) {
fprintf(stderr, "Unable to allocate memory for a new thread, exiting\n");
exit(1);
}
ck_array_commit(&Cyc_mutators);
pthread_mutex_unlock(&mutators_lock);
}
// Remove selected mutator from the mutator list.
// This is done for terminated threads. Note data is queued to be
// freed, to prevent accidentally freeing it while the collector
// thread is potentially accessing it.
void gc_remove_mutator(gc_thread_data *thd)
{
pthread_mutex_lock(&mutators_lock);
if (!ck_array_remove(&Cyc_mutators, (void *)thd)) {
fprintf(stderr, "Unable to remove thread data, exiting\n");
exit(1);
}
ck_array_commit(&Cyc_mutators);
// Place on list of old mutators to cleanup
if (ck_array_put_unique(&old_mutators, (void *)thd) < 0) {
fprintf(stderr, "Unable to add thread data to GC list, existing\n");
exit(1);
}
ck_array_commit(&old_mutators);
pthread_mutex_unlock(&mutators_lock);
}
void gc_free_old_thread_data()
{
ck_array_iterator_t iterator;
gc_thread_data *m;
int freed = 0;
pthread_mutex_lock(&mutators_lock);
CK_ARRAY_FOREACH(&old_mutators, &iterator, &m){
//printf("JAE DEBUG - freeing old thread data...");
gc_thread_data_free(m);
if (!ck_array_remove(&old_mutators, (void *)m)) {
fprintf(stderr, "Error removing old mutator data\n");
exit(1);
}
freed = 1;
//printf(" done\n");
}
if (freed) {
ck_array_commit(&old_mutators);
//printf("commited old mutator data deletions\n");
}
pthread_mutex_unlock(&mutators_lock);
}
gc_heap *gc_heap_create(size_t size, size_t max_size, size_t chunk_size)
{
gc_free_list *free, *next;
gc_heap *h;
// TODO: mmap?
h = malloc(gc_heap_pad_size(size));
if (!h) return NULL;
h->size = size;
//h->free_size = size;
cached_heap_total_size += size;
cached_heap_free_size += size;
h->chunk_size = chunk_size;
h->max_size = max_size;
h->data = (char *) gc_heap_align(sizeof(h->data) + (unsigned int)&(h->data));
h->next = NULL;
free = h->free_list = (gc_free_list *)h->data;
next = (gc_free_list *)(((char *) free) + gc_heap_align(gc_free_chunk_size));
free->size = 0; // First one is just a dummy record
free->next = next;
next->size = size - gc_heap_align(gc_free_chunk_size);
next->next = NULL;
#if GC_DEBUG_PRINTFS
fprintf(stderr, "DEBUG h->data addr: %p\n", &(h->data));
fprintf(stderr, "DEBUG h->data addr: %p\n", h->data);
fprintf(stderr, ("heap: %p-%p data: %p-%p size: %d\n"),
h, ((char*)h)+gc_heap_pad_size(size), h->data, h->data + size, size);
fprintf(stderr, ("first: %p end: %p\n"),
(object)gc_heap_first_block(h), (object)gc_heap_end(h));
fprintf(stderr, ("free1: %p-%p free2: %p-%p\n"),
free, ((char*)free)+free->size, next, ((char*)next)+next->size);
#endif
return h;
}
// Copy given object into given heap object
char *gc_copy_obj(object dest, char *obj, gc_thread_data *thd)
{
// NOTE: no additional type checking because this is called from gc_move
// which already does that
switch(type_of(obj)){
case cons_tag: {
list hp = dest;
hp->hdr.mark = thd->gc_alloc_color;
type_of(hp) = cons_tag;
car(hp) = car(obj);
cdr(hp) = cdr(obj);
return (char *)hp;
}
case macro_tag: {
macro_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = macro_tag;
hp->fn = ((macro) obj)->fn;
hp->num_args = ((macro) obj)->num_args;
return (char *)hp;
}
case closure0_tag: {
closure0_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closure0_tag;
hp->fn = ((closure0) obj)->fn;
hp->num_args = ((closure0) obj)->num_args;
return (char *)hp;
}
case closure1_tag: {
closure1_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closure1_tag;
hp->fn = ((closure1) obj)->fn;
hp->num_args = ((closure1) obj)->num_args;
hp->elt1 = ((closure1) obj)->elt1;
return (char *)hp;
}
case closure2_tag: {
closure2_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closure2_tag;
hp->fn = ((closure2) obj)->fn;
hp->num_args = ((closure2) obj)->num_args;
hp->elt1 = ((closure2) obj)->elt1;
hp->elt2 = ((closure2) obj)->elt2;
return (char *)hp;
}
case closure3_tag: {
closure3_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closure3_tag;
hp->fn = ((closure3) obj)->fn;
hp->num_args = ((closure3) obj)->num_args;
hp->elt1 = ((closure3) obj)->elt1;
hp->elt2 = ((closure3) obj)->elt2;
hp->elt3 = ((closure3) obj)->elt3;
return (char *)hp;
}
case closure4_tag: {
closure4_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closure4_tag;
hp->fn = ((closure4) obj)->fn;
hp->num_args = ((closure4) obj)->num_args;
hp->elt1 = ((closure4) obj)->elt1;
hp->elt2 = ((closure4) obj)->elt2;
hp->elt3 = ((closure4) obj)->elt3;
hp->elt4 = ((closure4) obj)->elt4;
return (char *)hp;
}
case closureN_tag: {
int i;
closureN_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = closureN_tag;
hp->fn = ((closureN) obj)->fn;
hp->num_args = ((closureN) obj)->num_args;
hp->num_elt = ((closureN) obj)-> num_elt;
hp->elts = (object *)(((char *)hp) + sizeof(closureN_type));
for (i = 0; i < hp->num_elt; i++) {
hp->elts[i] = ((closureN) obj)->elts[i];
}
return (char *)hp;
}
case vector_tag: {
int i;
vector_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = vector_tag;
hp->num_elt = ((vector) obj)-> num_elt;
hp->elts = (object *)(((char *)hp) + sizeof(vector_type));
for (i = 0; i < hp->num_elt; i++) {
hp->elts[i] = ((vector) obj)->elts[i];
}
return (char *)hp;
}
case string_tag: {
char *s;
string_type *hp = dest;
s = ((char *)hp) + sizeof(string_type);
memcpy(s, string_str(obj), string_len(obj) + 1);
mark(hp) = thd->gc_alloc_color;
type_of(hp) = string_tag;
string_len(hp) = string_len(obj);
string_str(hp) = s;
return (char *)hp;
}
case integer_tag: {
integer_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = integer_tag;
hp->value = ((integer_type *) obj)->value;
return (char *)hp;
}
case double_tag: {
double_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = double_tag;
hp->value = ((double_type *) obj)->value;
return (char *)hp;
}
case port_tag: {
port_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = port_tag;
hp->fp = ((port_type *) obj)->fp;
hp->mode = ((port_type *) obj)->mode;
return (char *)hp;
}
case cvar_tag: {
cvar_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = cvar_tag;
hp->pvar = ((cvar_type *) obj)->pvar;
return (char *)hp;
}
case mutex_tag: {
mutex_type *hp = dest;
mark(hp) = thd->gc_alloc_color;
type_of(hp) = mutex_tag;
// NOTE: don't copy mutex itself, caller will do that (this is a special case)
return (char *)hp;
}
case forward_tag:
return (char *)forward(obj);
case eof_tag:
case primitive_tag:
case boolean_tag:
case symbol_tag:
break;
default:
fprintf(stderr, "gc_copy_obj: bad tag obj=%p obj.tag=%ld\n",(object) obj, type_of(obj));
exit(1);
}
return (char *)obj;
}
int gc_grow_heap(gc_heap *h, size_t size, size_t chunk_size)
{
size_t cur_size, new_size;
gc_heap *h_last, *h_new;
pthread_mutex_lock(&heap_lock);
h_last = gc_heap_last(h);
cur_size = h_last->size;
// JAE - For now, just add a new page
new_size = cur_size; //gc_heap_align(((cur_size > size) ? cur_size : size) * 2);
h_new = gc_heap_create(new_size, h_last->max_size, chunk_size);
h_last->next = h_new;
pthread_mutex_unlock(&heap_lock);
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - grew heap\n");
#endif
return (h_new != NULL);
}
void *gc_try_alloc(gc_heap *h, size_t size, char *obj, gc_thread_data *thd)
{
gc_free_list *f1, *f2, *f3;
pthread_mutex_lock(&heap_lock);
for (; h; h = h->next) { // All heaps
// TODO: chunk size (ignoring for now)
for (f1 = h->free_list, f2 = f1->next; f2; f1 = f2, f2 = f2->next) { // all free in this heap
if (f2->size >= size) { // Big enough for request
// TODO: take whole chunk or divide up f2 (using f3)?
if (f2->size >= (size + gc_heap_align(1) /* min obj size */)) {
f3 = (gc_free_list *) (((char *)f2) + size);
f3->size = f2->size - size;
f3->next = f2->next;
f1->next = f3;
} else { /* Take the whole chunk */
f1->next = f2->next;
}
// Copy object into heap now to avoid any uninitialized memory issues
gc_copy_obj(f2, obj, thd);
//h->free_size -= gc_allocated_bytes(obj, NULL, NULL);
cached_heap_free_size -= gc_allocated_bytes(obj, NULL, NULL);
pthread_mutex_unlock(&heap_lock);
return f2;
}
}
}
pthread_mutex_unlock(&heap_lock);
return NULL;
}
void *gc_alloc(gc_heap *h, size_t size, char *obj, gc_thread_data *thd, int *heap_grown)
{
void *result = NULL;
size_t max_freed = 0, sum_freed = 0, total_size;
// TODO: check return value, if null (could not alloc) then
// run a collection and check how much free space there is. if less
// the allowed ratio, try growing heap.
// then try realloc. if cannot alloc now, then throw out of memory error
size = gc_heap_align(size);
result = gc_try_alloc(h, size, obj, thd);
if (!result) {
// A vanilla mark&sweep collector would collect now, but unfortunately
// we can't do that because we have to go through multiple stages, some
// of which are asynchronous. So... no choice but to grow the heap.
gc_grow_heap(h, size, 0);
*heap_grown = 1;
result = gc_try_alloc(h, size, obj, thd);
if (!result) {
fprintf(stderr, "out of memory error allocating %d bytes\n", size);
exit(1); // could throw error, but OOM is a major issue, so...
}
}
#if GC_DEBUG_TRACE
fprintf(stderr, "alloc %p size = %d, obj=%p, tag=%ld, mark=%d\n", result, size, obj, type_of(obj), mark(((object)result)));
// Debug check, should no longer be necessary
//if (is_value_type(result)) {
// printf("Invalid allocated address - is a value type %p\n", result);
//}
#endif
return result;
}
size_t gc_allocated_bytes(object obj, gc_free_list *q, gc_free_list *r)
{
tag_type t;
#if GC_SAFETY_CHECKS
if (is_value_type(obj)) {
fprintf(stderr,
"gc_allocated_bytes - passed value type %p q=[%p, %d] r=[%p, %d]\n",
obj, q, q->size, r, r->size);
exit(1);
}
#endif
t = type_of(obj);
if (t == cons_tag) return gc_heap_align(sizeof(cons_type));
if (t == macro_tag) return gc_heap_align(sizeof(macro_type));
if (t == closure0_tag) return gc_heap_align(sizeof(closure0_type));
if (t == closure1_tag) return gc_heap_align(sizeof(closure1_type));
if (t == closure2_tag) return gc_heap_align(sizeof(closure2_type));
if (t == closure3_tag) return gc_heap_align(sizeof(closure3_type));
if (t == closure4_tag) return gc_heap_align(sizeof(closure4_type));
if (t == closureN_tag){
return gc_heap_align(sizeof(closureN_type) + sizeof(object) * ((closureN_type *)obj)->num_elt);
}
if (t == vector_tag){
return gc_heap_align(sizeof(vector_type) + sizeof(object) * ((vector_type *)obj)->num_elt);
}
if (t == string_tag){
return gc_heap_align(sizeof(string_type) + string_len(obj) + 1);
}
if (t == integer_tag) return gc_heap_align(sizeof(integer_type));
if (t == double_tag) return gc_heap_align(sizeof(double_type));
if (t == port_tag) return gc_heap_align(sizeof(port_type));
if (t == cvar_tag) return gc_heap_align(sizeof(cvar_type));
if (t == mutex_tag) return gc_heap_align(sizeof(mutex_type));
fprintf(stderr, "gc_allocated_bytes: unexpected object %p of type %ld\n", obj, t);
exit(1);
return 0;
}
gc_heap *gc_heap_last(gc_heap *h)
{
while (h->next)
h = h->next;
return h;
}
size_t gc_heap_total_size(gc_heap *h)
{
size_t total_size = 0;
pthread_mutex_lock(&heap_lock);
while(h) {
total_size += h->size;
h = h->next;
}
pthread_mutex_unlock(&heap_lock);
return total_size;
}
//size_t gc_heap_total_free_size(gc_heap *h)
//{
// size_t total_size = 0;
// pthread_mutex_lock(&heap_lock);
// while(h) {
// total_size += h->free_size;
// h = h->next;
// }
// pthread_mutex_unlock(&heap_lock);
// return total_size;
//}
size_t gc_sweep(gc_heap *h, size_t *sum_freed_ptr)
{
size_t freed, max_freed=0, heap_freed = 0, sum_freed=0, size;
object p, end;
gc_free_list *q, *r, *s;
//
// Lock the heap to prevent issues with allocations during sweep
// It sucks to have to use a coarse-grained lock like this, but let's
// be safe and prevent threading issues right now. Once the new GC
// works we can go back and try to speed things up (if possible)
// by using more fine-grained locking. Can also profile to see
// how much time is even spent sweeping
//
pthread_mutex_lock(&heap_lock);
for (; h; h = h->next) { // All heaps
#if GC_DEBUG_TRACE
fprintf(stderr, "sweep heap %p, size = %d\n", h, h->size);
#endif
p = gc_heap_first_block(h);
q = h->free_list;
end = gc_heap_end(h);
while (p < end) {
// find preceding/succeeding free list pointers for p
for (r = q->next; r && ((char *)r < (char *)p); q=r, r=r->next);
if ((char *)r == (char *)p) { // this is a free block, skip it
p = (object) (((char *)p) + r->size);
#if GC_DEBUG_TRACE
fprintf(stderr, "skip free block %p size = %d\n", p, r->size);
#endif
continue;
}
size = gc_heap_align(gc_allocated_bytes(p, q, r));
#if GC_SAFETY_CHECKS
if (!is_object_type(p)) {
fprintf(stderr, "sweep: invalid object at %p", p);
exit(1);
}
if ((char *)q + q->size > (char *)p) {
fprintf(stderr, "bad size at %p < %p + %u", p, q, q->size);
exit(1);
}
if (r && ((char *)p) + size > (char *)r) {
fprintf(stderr, "sweep: bad size at %p + %d > %p", p, size, r);
exit(1);
}
#endif
if (mark(p) == gc_color_clear) {
#if GC_DEBUG_VERBOSE
fprintf(stderr, "sweep is freeing unmarked obj: %p with tag %ld\n", p, type_of(p));
#endif
mark(p) = gc_color_blue; // Needed?
if (type_of(p) == mutex_tag) {
#if GC_DEBUG_VERBOSE
fprintf(stderr, "pthread_mutex_destroy from sweep\n");
#endif
if (pthread_mutex_destroy(&(((mutex)p)->lock)) != 0) {
fprintf(stderr, "Error destroying mutex\n");
exit(1);
}
}
// free p
heap_freed += size;
if (((((char *)q) + q->size) == (char *)p) && (q != h->free_list)) {
/* merge q with p */
if (r && r->size && ((((char *)p)+size) == (char *)r)) {
// ... and with r
q->next = r->next;
freed = q->size + size + r->size;
p = (object) (((char *)p) + size + r->size);
} else {
freed = q->size + size;
p = (object) (((char *)p) + size);
}
q->size = freed;
} else {
s = (gc_free_list *)p;
if (r && r->size && ((((char *)p) + size) == (char *)r)) {
// merge p with r
s->size = size + r->size;
s->next = r->next;
q->next = s;
freed = size + r->size;
} else {
s->size = size;
s->next = r;
q->next = s;
freed = size;
}
p = (object) (((char *)p) + freed);
}
if (freed > max_freed)
max_freed = freed;
} else {
//#if GC_DEBUG_VERBOSE
// fprintf(stderr, "sweep: object is marked %p\n", p);
//#endif
p = (object)(((char *)p) + size);
}
}
//h->free_size += heap_freed;
cached_heap_free_size += heap_freed;
sum_freed += heap_freed;
heap_freed = 0;
}
pthread_mutex_unlock(&heap_lock);
if (sum_freed_ptr) *sum_freed_ptr = sum_freed;
return max_freed;
}
void gc_thr_grow_move_buffer(gc_thread_data *d)
{
if (!d) return;
if (d->moveBufLen == 0) { // Special case
d->moveBufLen = 128;
d->moveBuf = NULL;
} else {
d->moveBufLen *= 2;
}
d->moveBuf = realloc(d->moveBuf, d->moveBufLen * sizeof(void *));
#if GC_DEBUG_TRACE
fprintf(stderr, "grew moveBuffer, len = %d\n", d->moveBufLen);
#endif
}
void gc_thr_add_to_move_buffer(gc_thread_data *d, int *alloci, object obj)
{
if (*alloci == d->moveBufLen) {
gc_thr_grow_move_buffer(d);
}
d->moveBuf[*alloci] = obj;
(*alloci)++;
}
// Generic buffer functions
void **vpbuffer_realloc(void **buf, int *len)
{
return realloc(buf, (*len) * sizeof(void *));
}
void **vpbuffer_add(void **buf, int *len, int i, void *obj)
{
if (i == *len) {
*len *= 2;
buf = vpbuffer_realloc(buf, len);
}
buf[i] = obj;
return buf;
}
void vpbuffer_free(void **buf)
{
free(buf);
}
// END heap definitions
// Tri-color GC section
/////////////////////////////////////////////
// GC functions called by the Mutator threads
/**
* Determine if object lives on the thread's stack
*/
int gc_is_stack_obj(gc_thread_data *thd, object obj)
{
char tmp;
object low_limit = &tmp;
object high_limit = thd->stack_start;
return (check_overflow(low_limit, obj) &&
check_overflow(obj, high_limit));
}
/**
* Write barrier for updates to heap-allocated objects
* The key for this barrier is to identify stack objects that contain
* heap references, so they can be marked to avoid collection.
*/
void gc_mut_update(gc_thread_data *thd, object old_obj, object value)
{
int status = ck_pr_load_int(&gc_status_col),
stage = ck_pr_load_int(&gc_stage);
if (ck_pr_load_int(&(thd->gc_status)) != STATUS_ASYNC) {
pthread_mutex_lock(&(thd->lock));
gc_mark_gray(thd, old_obj);
if (gc_is_stack_obj(thd, value)) {
// Set object to be marked after moved to heap by next GC.
// This avoids having to recursively examine the stack now,
// which we have to do anyway during minor GC.
grayed(value) = 1;
} else {
// Value is on the heap, mark gray right now
gc_mark_gray(thd, value);
}
pthread_mutex_unlock(&(thd->lock));
} else if (stage == STAGE_TRACING) {
//fprintf(stderr, "DEBUG - GC async tracing marking heap obj %p ", old_obj);
//Cyc_display(old_obj, stderr);
//fprintf(stderr, "\n");
pthread_mutex_lock(&(thd->lock));
gc_mark_gray(thd, old_obj);
pthread_mutex_unlock(&(thd->lock));
#if GC_DEBUG_VERBOSE
if (is_object_type(old_obj) && mark(old_obj) == gc_color_clear) {
fprintf(stderr, "added to mark buffer (trace) from write barrier %p:mark %d:", old_obj, mark(old_obj));
Cyc_display(old_obj, stderr);
fprintf(stderr, "\n");
}
#endif
}
}
void gc_mut_cooperate(gc_thread_data *thd, int buf_len)
{
int i, status_c, status_m;
#if GC_DEBUG_VERBOSE
int debug_print = 0;
#endif
// Handle any pending marks from write barrier
pthread_mutex_lock(&(thd->lock));
thd->last_write += thd->pending_writes;
thd->pending_writes = 0;
pthread_mutex_unlock(&(thd->lock));
// I think below is thread safe, but this code is tricky.
// Worst case should be that some work is done twice if there is
// a race condition
//
// TODO: should use an atomic comparison here
status_c = ck_pr_load_int(&gc_status_col);
status_m = ck_pr_load_int(&(thd->gc_status));
if (status_m != status_c) {
ck_pr_cas_int(&(thd->gc_status), status_m, status_c);
if (status_m == STATUS_ASYNC) {
// Async is done, so clean up old mark data from the last collection
pthread_mutex_lock(&(thd->lock));
thd->last_write = 0;
thd->last_read = 0;
thd->pending_writes = 0;
pthread_mutex_unlock(&(thd->lock));
}
else if (status_m == STATUS_SYNC2) {
#if GC_DEBUG_VERBOSE
debug_print = 1;
#endif
// Mark thread "roots":
// Begin my marking current continuation, which may have already
// been on the heap prior to latest minor GC
pthread_mutex_lock(&(thd->lock));
gc_mark_gray(thd, thd->gc_cont);
for (i = 0; i < thd->gc_num_args; i++) {
gc_mark_gray(thd, thd->gc_args[i]);
}
// Also, mark everything the collector moved to the heap
for (i = 0; i < buf_len; i++) {
gc_mark_gray(thd, thd->moveBuf[i]);
}
pthread_mutex_unlock(&(thd->lock));
thd->gc_alloc_color = ck_pr_load_int(&gc_color_mark);
}
}
#if GC_DEBUG_VERBOSE
if (debug_print) {
fprintf(stderr, "coop mark gc_cont %p\n", thd->gc_cont);
for (i = 0; i < thd->gc_num_args; i++) {
fprintf(stderr, "coop mark gc_args[%d] %p\n", i, thd->gc_args[i]);
}
for (i = 0; i < buf_len; i++) {
fprintf(stderr, "coop mark from move buf %i %p\n", i, thd->moveBuf[i]);
}
}
#endif
// Initiate collection cycle if free space is too low.
// Threshold is intentially low because we have to go through an
// entire handshake/trace/sweep cycle, ideally without growing heap.
if (ck_pr_load_int(&gc_stage) == STAGE_RESTING &&
(cached_heap_free_size < (cached_heap_total_size * 0.50))){
#if GC_DEBUG_TRACE
fprintf(stdout, "Less than 50%% of the heap is free, initiating collector\n");
#endif
ck_pr_cas_int(&gc_stage, STAGE_RESTING, STAGE_CLEAR_OR_MARKING);
}
}
/////////////////////////////////////////////
// Collector functions
/**
* Mark the given object gray if it is on the heap.
* Note marking is done implicitly by placing it in a buffer,
* to avoid repeated re-scanning.
*
* This function must be executed once the thread lock has been acquired.
*/
void gc_mark_gray(gc_thread_data *thd, object obj)
{
// From what I can tell, no other thread would be modifying
// either object type or mark. Both should be stable once the object is placed
// into the heap, with the collector being the only thread that changes marks.
if (is_object_type(obj) && mark(obj) == gc_color_clear) { // TODO: sync??
// Place marked object in a buffer to avoid repeated scans of the heap.
// TODO:
// Note that ideally this should be a lock-free data structure to make the
// algorithm more efficient. So this code (and the corresponding collector
// trace code) should be converted at some point.
thd->mark_buffer = vpbuffer_add(thd->mark_buffer,
&(thd->mark_buffer_len),
thd->last_write,
obj);
(thd->last_write)++; // Already locked, just do it...
}
}
void gc_mark_gray2(gc_thread_data *thd, object obj)
{
if (is_object_type(obj) && mark(obj) == gc_color_clear) {
thd->mark_buffer = vpbuffer_add(thd->mark_buffer,
&(thd->mark_buffer_len),
(thd->last_write + thd->pending_writes),
obj);
thd->pending_writes++;
}
}
void gc_collector_trace()
{
ck_array_iterator_t iterator;
gc_thread_data *m;
int clean = 0;
while (!clean) {
clean = 1;
CK_ARRAY_FOREACH(&Cyc_mutators, &iterator, &m){
// TODO: ideally, want to use a lock-free data structure to prevent
// having to use a mutex here. see corresponding code in gc_mark_gray
pthread_mutex_lock(&(m->lock));
while (m->last_read < m->last_write) {
clean = 0;
#if GC_DEBUG_VERBOSE
fprintf(stderr, "gc_mark_black mark buffer %p, last_read = %d last_write = %d\n",
(m->mark_buffer)[m->last_read],
m->last_read, m->last_write);
#endif
gc_mark_black((m->mark_buffer)[m->last_read]);
gc_empty_collector_stack();
(m->last_read)++; // Inc here to prevent off-by-one error
}
pthread_mutex_unlock(&(m->lock));
// Try checking the condition once more after giving the
// mutator a chance to respond, to prevent exiting early.
// This is experimental, not sure if it is necessary
if (clean) {
pthread_mutex_lock(&(m->lock));
if (m->last_read < m->last_write) {
fprintf(stderr, "JAE DEBUG - might have exited trace early\n");
clean = 0;
}
else if (m->pending_writes) {
clean = 0;
}
pthread_mutex_unlock(&(m->lock));
}
}
}
}
// TODO: seriously consider changing the mark() macro to color(),
// and sync up the header variable. that would make all of this code
// bit clearer...
void gc_mark_black(object obj)
{
// TODO: is sync required to get colors? probably not on the collector
// thread (at least) since colors are only changed once during the clear
// phase and before the first handshake.
int markColor = ck_pr_load_int(&gc_color_mark);
if (is_object_type(obj) && mark(obj) != markColor) {
// Gray any child objects
// Note we probably should use some form of atomics/synchronization
// for cons and vector types, as these pointers could change.
switch(type_of(obj)) {
case cons_tag: {
gc_collector_mark_gray(obj, car(obj));
gc_collector_mark_gray(obj, cdr(obj));
break;
}
case closure1_tag:
gc_collector_mark_gray(obj, ((closure1) obj)->elt1);
break;
case closure2_tag:
gc_collector_mark_gray(obj, ((closure2) obj)->elt1);
gc_collector_mark_gray(obj, ((closure2) obj)->elt2);
case closure3_tag:
gc_collector_mark_gray(obj, ((closure3) obj)->elt1);
gc_collector_mark_gray(obj, ((closure3) obj)->elt2);
gc_collector_mark_gray(obj, ((closure3) obj)->elt3);
case closure4_tag:
gc_collector_mark_gray(obj, ((closure4) obj)->elt1);
gc_collector_mark_gray(obj, ((closure4) obj)->elt2);
gc_collector_mark_gray(obj, ((closure4) obj)->elt3);
gc_collector_mark_gray(obj, ((closure4) obj)->elt4);
break;
case closureN_tag: {
int i, n = ((closureN) obj)->num_elt;
for (i = 0; i < n; i++) {
gc_collector_mark_gray(obj, ((closureN) obj)->elts[i]);
}
break;
}
case vector_tag: {
int i, n = ((vector) obj)->num_elt;
for (i = 0; i < n; i++) {
gc_collector_mark_gray(obj, ((vector) obj)->elts[i]);
}
break;
}
case cvar_tag: {
cvar_type *c = (cvar_type *)obj;
object pvar = *(c->pvar);
if (pvar) {
gc_collector_mark_gray(obj, pvar);
}
break;
}
default:
break;
}
if (mark(obj) != gc_color_red) {
// Only blacken objects on the heap
mark(obj) = markColor;
}
#if GC_DEBUG_VERBOSE
if (mark(obj) != gc_color_red) {
fprintf(stderr, "marked %p %d\n", obj, markColor);
} else {
fprintf(stderr, "not marking stack obj %p %d\n", obj, markColor);
}
#endif
}
}
void gc_collector_mark_gray(object parent, object obj)
{
// "Color" objects gray by adding them to the mark stack for further processing.
//
// Note that stack objects are always colored red during creation, so
// they should never be added to the mark stack. Which would be bad because it
// could lead to stack corruption.
if (is_object_type(obj) && mark(obj) == gc_color_clear) {
mark_stack = vpbuffer_add(mark_stack, &mark_stack_len, mark_stack_i++, obj);
#if GC_DEBUG_VERBOSE
fprintf(stderr, "mark gray parent = %p (%ld) obj = %p\n", parent, type_of(parent), obj);
#endif
}
}
void gc_empty_collector_stack()
{
// Mark stack is only used by the collector thread, so no sync needed
while (mark_stack_i > 0) { // not empty
mark_stack_i--;
//#if GC_DEBUG_VERBOSE
// fprintf(stderr, "gc_mark_black mark stack %p \n",
// mark_stack[mark_stack_i]);
//#endif
gc_mark_black(mark_stack[mark_stack_i]);
}
}
void gc_handshake(gc_status_type s)
{
gc_post_handshake(s);
gc_wait_handshake();
}
void gc_post_handshake(gc_status_type s)
{
int status = ck_pr_load_int(&gc_status_col);
while (!ck_pr_cas_int(&gc_status_col, status, s)){}
}
void gc_wait_handshake()
{
ck_array_iterator_t iterator;
gc_thread_data *m;
int statusm, statusc, thread_status, i, buf_len;
struct timespec tim;
tim.tv_sec = 0;
tim.tv_nsec = 1000000; // 1 millisecond
CK_ARRAY_FOREACH(&Cyc_mutators, &iterator, &m) {
while (1) {
// TODO: use an atomic comparison
statusc = ck_pr_load_int(&gc_status_col);
statusm = ck_pr_load_int(&(m->gc_status));
if (statusc == statusm) {
// Handshake succeeded, check next mutator
break;
}
thread_status = ck_pr_load_int((int *)&(m->thread_state));
if (thread_status == CYC_THREAD_STATE_BLOCKED ||
thread_status == CYC_THREAD_STATE_BLOCKED_COOPERATING) {
if (statusm == STATUS_ASYNC) { // Prev state
ck_pr_cas_int(&(m->gc_status), statusm, statusc);
// Async is done, so clean up old mark data from the last collection
pthread_mutex_lock(&(m->lock));
m->last_write = 0;
m->last_read = 0;
m->pending_writes = 0;
pthread_mutex_unlock(&(m->lock));
}else if (statusm == STATUS_SYNC1) {
ck_pr_cas_int(&(m->gc_status), statusm, statusc);
} else if (statusm == STATUS_SYNC2) {
//printf("DEBUG - is mutator still blocked?\n");
// Check again, if thread is still blocked we need to cooperate
if (ck_pr_cas_int((int *)&(m->thread_state),
CYC_THREAD_STATE_BLOCKED,
CYC_THREAD_STATE_BLOCKED_COOPERATING)
||
ck_pr_cas_int((int *)&(m->thread_state),
CYC_THREAD_STATE_BLOCKED_COOPERATING,
CYC_THREAD_STATE_BLOCKED_COOPERATING)
) {
//printf("DEBUG - update mutator GC status\n");
ck_pr_cas_int(&(m->gc_status), statusm, statusc);
pthread_mutex_lock(&(m->lock));
//printf("DEBUG - collector is cooperating for blocked mutator\n");
buf_len = gc_minor(m, m->stack_limit, m->stack_start, m->gc_cont, NULL, 0);
// Mark thread "roots", based on code from mutator's cooperator
gc_mark_gray(m, m->gc_cont);
//for (i = 0; i < m->gc_num_args; i++) {
// gc_mark_gray(m, m->gc_args[i]);
//}
// Also, mark everything the collector moved to the heap
for (i = 0; i < buf_len; i++) {
gc_mark_gray(m, m->moveBuf[i]);
}
m->gc_alloc_color = ck_pr_load_int(&gc_color_mark);
pthread_mutex_unlock(&(m->lock));
}
}
} else if (thread_status == CYC_THREAD_STATE_TERMINATED) {
// Thread is no longer running
break;
}
// At least for now, just give up quantum and come back to
// this quickly to test again. This probably could be more
// efficient.
nanosleep(&tim, NULL);
}
}
}
/////////////////////////////////////////////
// GC Collection cycle
void debug_dump_globals();
// Main collector function
void gc_collector()
{
int old_clear, old_mark;
size_t freed = 0, max_freed = 0, total_size, total_free;
#if GC_DEBUG_TRACE
time_t sweep_start = time(NULL);
#endif
//clear :
ck_pr_cas_int(&gc_stage, STAGE_RESTING, STAGE_CLEAR_OR_MARKING);
// exchange values of markColor and clearColor
old_clear = ck_pr_load_int(&gc_color_clear);
old_mark = ck_pr_load_int(&gc_color_mark);
while(!ck_pr_cas_int(&gc_color_clear, old_clear, old_mark)){}
while(!ck_pr_cas_int(&gc_color_mark, old_mark, old_clear)){}
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - swap clear %d / mark %d\n", gc_color_clear, gc_color_mark);
#endif
gc_handshake(STATUS_SYNC1);
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - after handshake sync 1\n");
#endif
//mark :
gc_handshake(STATUS_SYNC2);
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - after handshake sync 2\n");
#endif
ck_pr_cas_int(&gc_stage, STAGE_CLEAR_OR_MARKING, STAGE_TRACING);
gc_post_handshake(STATUS_ASYNC);
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - after post_handshake async\n");
#endif
gc_mark_globals();
gc_wait_handshake();
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - after wait_handshake async\n");
#endif
//trace :
gc_collector_trace();
#if GC_DEBUG_TRACE
fprintf(stderr, "DEBUG - after trace\n");
//debug_dump_globals();
#endif
ck_pr_cas_int(&gc_stage, STAGE_TRACING, STAGE_SWEEPING);
//
//sweep :
max_freed = gc_sweep(gc_get_heap(), &freed);
total_size = cached_heap_total_size; //gc_heap_total_size(gc_get_heap());
total_free = cached_heap_free_size; //gc_heap_total_free_size(gc_get_heap());
if (total_free < (total_size * 0.10)) {
#if GC_DEBUG_TRACE
fprintf(stdout, "Less than 10%% of the heap is free, growing it\n",
total_free, total_size);
#endif
gc_grow_heap(gc_get_heap(), 0, 0);
}
#if GC_DEBUG_TRACE
fprintf(stderr, "sweep done, total_size = %d, total_free = %d, freed = %d, max_freed = %d, elapsed = %ld\n",
total_size, total_free,
freed, max_freed, time(NULL) - sweep_start);
#endif
#if GC_DEBUG_TRACE
fprintf(stderr, "cleaning up any old thread data\n");
#endif
gc_free_old_thread_data();
// Idle the GC thread
ck_pr_cas_int(&gc_stage, STAGE_SWEEPING, STAGE_RESTING);
}
void *collector_main(void *arg)
{
int stage;
struct timespec tim;
tim.tv_sec = 0;
//JAE TODO: this is still not good enough, seems memory grows still grows fast with this.
//alternatively, may want to consider shrinking the heap if possible after a collection, if it is
//sparse enough (would be difficult to do without relocations, though
tim.tv_nsec = 100 * NANOSECONDS_PER_MILLISECOND;
while (1) {
stage = ck_pr_load_int(&gc_stage);
if (stage != STAGE_RESTING) {
gc_collector();
}
nanosleep(&tim, NULL);
}
return NULL;
}
static pthread_t collector_thread;
void gc_start_collector()
{
if (pthread_create(&collector_thread, NULL, collector_main, &collector_thread)) {
fprintf(stderr, "Error creating collector thread\n");
exit(1);
}
}
/////////////////////////////////////////////
// END tri-color marking section
/////////////////////////////////////////////
// Initialize runtime data structures for a thread.
// Must be called on the target thread itself during startup,
// to verify stack limits are setup correctly.
void gc_thread_data_init(gc_thread_data *thd, int mut_num, char *stack_base, long stack_size)
{
char stack_ref;
thd->stack_start = stack_base;
#if STACK_GROWS_DOWNWARD
thd->stack_limit = stack_base - stack_size;
#else
thd->stack_limit = stack_base + stack_size;
#endif
if (check_overflow(stack_base, &stack_ref)){
fprintf(stderr,
"Error: recompile with STACK_GROWS_DOWNWARD set to %d\n",
(1 - STACK_GROWS_DOWNWARD));
exit(1);
}
thd->stack_traces = calloc(MAX_STACK_TRACES, sizeof(char *));
thd->stack_trace_idx = 0;
thd->stack_prev_frame = NULL;
thd->mutations = NULL;
// thd->thread = NULL;
thd->thread_state = CYC_THREAD_STATE_NEW;
//thd->mutator_num = mut_num;
thd->jmp_start = malloc(sizeof(jmp_buf));
thd->gc_args = malloc(sizeof(object) * NUM_GC_ANS);
thd->gc_num_args = 0;
thd->moveBufLen = 0;
gc_thr_grow_move_buffer(thd);
thd->gc_alloc_color = ck_pr_load_int(&gc_color_clear);
thd->gc_status = ck_pr_load_int(&gc_status_col);
thd->last_write = 0;
thd->last_read = 0;
thd->mark_buffer_len = 128;
thd->mark_buffer = vpbuffer_realloc(thd->mark_buffer, &(thd->mark_buffer_len));
if (pthread_mutex_init(&(thd->lock), NULL) != 0) {
fprintf(stderr, "Unable to initialize thread mutex\n");
exit(1);
}
}
void gc_thread_data_free(gc_thread_data *thd)
{
if (thd) {
if (pthread_mutex_destroy(&thd->lock) != 0) {
// TODO: can only destroy the lock if it is unlocked. need to make sure we
// can guarantee that is the case prior to making this call
// On the other hand, can we just use sleep and a loop to retry??
fprintf(stderr, "Thread mutex is locked, unable to free\n");
exit(1);
}
if (thd->jmp_start) free(thd->jmp_start);
if (thd->gc_args) free(thd->gc_args);
if (thd->moveBuf) free(thd->moveBuf);
if (thd->mark_buffer) free(thd->mark_buffer);
if (thd->stack_traces) free(thd->stack_traces);
if (thd->mutations) {
clear_mutations(thd);
}
free(thd);
}
}
void gc_mutator_thread_blocked(gc_thread_data *thd, object cont)
{
if(!ck_pr_cas_int((int *)&(thd->thread_state),
CYC_THREAD_STATE_RUNNABLE,
CYC_THREAD_STATE_BLOCKED)){
fprintf(stderr, "Unable to change thread from runnable to blocked. status = %d\n", thd->thread_state);
exit(1);
}
thd->gc_cont = cont;
thd->gc_num_args = 0; // Will be set later, after collection
}
void gc_mutator_thread_runnable(gc_thread_data *thd, object result)
{
// Transition from blocked back to runnable using CAS.
// If we are unable to transition back, assume collector
// has cooperated on behalf of this mutator thread.
if (!ck_pr_cas_int((int *)&(thd->thread_state),
CYC_THREAD_STATE_BLOCKED,
CYC_THREAD_STATE_RUNNABLE)){
//printf("DEBUG - Collector cooperated, wait for it to finish. status is %d\n", thd->thread_state);
// wait for the collector to finish
pthread_mutex_lock(&(thd->lock));
pthread_mutex_unlock(&(thd->lock));
// update thread status
while(!ck_pr_cas_int((int *)&(thd->thread_state),
CYC_THREAD_STATE_BLOCKED_COOPERATING,
CYC_THREAD_STATE_RUNNABLE)){}
// Setup value to send to continuation
thd->gc_args[0] = result;
thd->gc_num_args = 1;
// Move any remaining stack objects (should only be the result?) to heap
gc_minor(thd, thd->stack_limit, thd->stack_start, thd->gc_cont, thd->gc_args, thd->gc_num_args);
//printf("DEBUG - Call into gc_cont after collector coop\n");
// Whoa.
longjmp(*(thd->jmp_start), 1);
} else {
// Collector didn't do anything; make a normal continuation call
(((closure)(thd->gc_cont))->fn)(thd, 1, thd->gc_cont, result);
}
}
//// Unit testing:
//int main(int argc, char **argv) {
// int a = 1, b = 2, c = 3, i;
// void **buf = NULL;
// int size = 1;
//
// buf = vpbuffer_realloc(buf, &size);
// printf("buf = %p, size = %d\n", buf, size);
// buf = vpbuffer_add(buf, &size, 0, &a);
// printf("buf = %p, size = %d\n", buf, size);
// buf = vpbuffer_add(buf, &size, 1, &b);
// printf("buf = %p, size = %d\n", buf, size);
// buf = vpbuffer_add(buf, &size, 2, &c);
// printf("buf = %p, size = %d\n", buf, size);
// buf = vpbuffer_add(buf, &size, 3, &a);
// printf("buf = %p, size = %d\n", buf, size);
// buf = vpbuffer_add(buf, &size, 4, &b);
// printf("buf = %p, size = %d\n", buf, size);
// for (i = 5; i < 20; i++) {
// buf = vpbuffer_add(buf, &size, i, &c);
// }
//
// for (i = 0; i < 20; i++){
// printf("%d\n", *((int *) buf[i]));
// }
// vpbuffer_free(buf);
// printf("buf = %p, size = %d\n", buf, size);
// return 0;
//}
//
Reference in a new issue View git blame Copy permalink