Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Page based heap size heuristics #50144

Merged
merged 13 commits into from
Jul 23, 2023
Merged

Page based heap size heuristics #50144

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
13 commits
Select commit Hold shift + click to select a range
5f36833
Implement new GC heuristics.
gbaraldi Jul 3, 2023
2f42cd5
Relax some of the atomics for sweeping
gbaraldi Jul 5, 2023
01d6e39
Apply review suggestions.
gbaraldi Jul 7, 2023
b033c0a
Add batching for counted functions
gbaraldi Jul 7, 2023
2ab7f95
Small fixup
gbaraldi Jul 7, 2023
7af549e
Batch other big allocations as well
gbaraldi Jul 7, 2023
4da775f
Readd testenv changes
gbaraldi Jul 13, 2023
b4bab03
Let the heap increase a bit if we are thrashing
gbaraldi Jul 14, 2023
2cbba34
Whitespace
gbaraldi Jul 14, 2023
df3fcb0
Merge branch 'master' into new-heuristics
gbaraldi Jul 17, 2023
e6a18f3
Experiment with not setting a max size by default at all
gbaraldi Jul 17, 2023
15b34a5
Add under pressure callback
gbaraldi Jul 19, 2023
9f3ca7c
Merge branch 'master' into new-heuristics
gbaraldi Jul 20, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
Apply review suggestions.
  • Loading branch information
@gbaraldi
gbaraldi committed Jul 7, 2023
commit 01d6e3910869c77a9d552fcdfb618fe6baec3b03
13 changes: 4 additions & 9 deletions src/gc-debug.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1229,19 +1229,14 @@ void _report_gc_finished(uint64_t pause, uint64_t freed, int full, int recollect
recollect ? "recollect" : ""
);

jl_safe_printf("Heap stats: bytes_mapped %.2f MB, bytes_allocd %.2f MB\nbytes_freed %.2f MB, bytes_mallocd %.1f, malloc_bytes_freed %.2f MB\npages_perm_allocd %zu, heap_size %.2f MB, heap_target %.2f MB, live_bytes %.2f MB\n",
jl_safe_printf("Heap stats: bytes_mapped %.2f MB, bytes_resident %.2f MB, heap_size %.2f MB, heap_target %.2f MB, live_bytes %.2f MB\n, Fragmentation %.3f",
jl_atomic_load_relaxed(&gc_heap_stats.bytes_mapped)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.bytes_allocd)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.bytes_freed)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.bytes_mallocd)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.malloc_bytes_freed)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.pages_perm_allocd),
jl_atomic_load_relaxed(&gc_heap_stats.bytes_resident)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.heap_size)/(double)(1<<20),
jl_atomic_load_relaxed(&gc_heap_stats.heap_target)/(double)(1<<20),
live_bytes/(double)(1<<20)
live_bytes/(double)(1<<20),
(double)live_bytes/(double)jl_atomic_load_relaxed(&gc_heap_stats.heap_size)
);
double bytes_mapped = (jl_atomic_load_relaxed(&gc_heap_stats.bytes_resident) + jl_atomic_load_relaxed(&gc_heap_stats.bytes_mallocd) - jl_atomic_load_relaxed(&gc_heap_stats.malloc_bytes_freed))/(double)(1<<20);
jl_safe_printf("Fragmentation %f, mapped_bytes %.2f MB\n", (double)live_bytes/(double)jl_atomic_load_relaxed(&gc_heap_stats.heap_size), bytes_mapped);
// Should fragmentation use bytes_resident instead of heap_size?
}

Expand Down
64 changes: 22 additions & 42 deletions src/gc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1010,7 +1010,6 @@ STATIC_INLINE jl_value_t *jl_gc_big_alloc_inner(jl_ptls_t ptls, size_t sz)
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + allocsz);
jl_atomic_store_relaxed(&ptls->gc_num.bigalloc,
jl_atomic_load_relaxed(&ptls->gc_num.bigalloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, allocsz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, allocsz);
#ifdef MEMDEBUG
memset(v, 0xee, allocsz);
Expand Down Expand Up @@ -1057,8 +1056,6 @@ static bigval_t **sweep_big_list(int sweep_full, bigval_t **pv) JL_NOTSAFEPOINT
if (nxt)
nxt->prev = pv;
gc_num.freed += v->sz&~3;
jl_atomic_store_relaxed(&gc_heap_stats.malloc_bytes_freed,
jl_atomic_load_relaxed(&gc_heap_stats.malloc_bytes_freed) + (v->sz&~3));
jl_atomic_store_relaxed(&gc_heap_stats.heap_size,
jl_atomic_load_relaxed(&gc_heap_stats.heap_size) - (v->sz&~3));
#ifdef MEMDEBUG
Expand Down Expand Up @@ -1120,7 +1117,6 @@ void jl_gc_count_allocd(size_t sz) JL_NOTSAFEPOINT
jl_ptls_t ptls = jl_current_task->ptls;
jl_atomic_store_relaxed(&ptls->gc_num.allocd,
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, sz);
}

Expand Down Expand Up @@ -1196,12 +1192,8 @@ static void jl_gc_free_array(jl_array_t *a) JL_NOTSAFEPOINT
jl_free_aligned(d);
else
free(d);
jl_atomic_store_relaxed(&gc_heap_stats.malloc_bytes_freed,
jl_atomic_load_relaxed(&gc_heap_stats.malloc_bytes_freed) + jl_array_nbytes(a));
jl_atomic_store_relaxed(&gc_heap_stats.heap_size,
jl_atomic_load_relaxed(&gc_heap_stats.heap_size) - jl_array_nbytes(a));
jl_atomic_fetch_add_relaxed(&gc_heap_stats.malloc_bytes_freed, jl_array_nbytes(a));
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, -jl_array_nbytes(a));
gc_num.freed += jl_array_nbytes(a);
gc_num.freecall++;
}
Expand Down Expand Up @@ -1273,7 +1265,6 @@ static NOINLINE jl_taggedvalue_t *gc_add_page(jl_gc_pool_t *p) JL_NOTSAFEPOINT
set_page_metadata(pg);
push_page_metadata_back(&ptls->page_metadata_allocd, pg);
jl_taggedvalue_t *fl = gc_reset_page(ptls, p, pg);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_allocd, GC_PAGE_SZ);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, GC_PAGE_SZ);
p->newpages = fl;
return fl;
Expand Down Expand Up @@ -1468,11 +1459,9 @@ static jl_taggedvalue_t **gc_sweep_page(jl_gc_pool_t *p, jl_gc_pagemeta_t **allo
}
else if (freed_lazily) {
push_page_metadata_back(lazily_freed, pg);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_freed, GC_PAGE_SZ);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, -GC_PAGE_SZ);
}
else {
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_freed, GC_PAGE_SZ);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, -GC_PAGE_SZ);
#ifdef _P64 // only enable concurrent sweeping on 64bit
if (jl_n_sweepthreads == 0) {
Expand Down Expand Up @@ -3279,29 +3268,30 @@ static int _jl_gc_collect(jl_ptls_t ptls, jl_gc_collection_t collection)
}

size_t heap_size = jl_atomic_load_relaxed(&gc_heap_stats.heap_size);
uint64_t alloc_diff = before_free_heap_size - old_heap_size;
uint64_t freed_diff = before_free_heap_size - heap_size;

double alloc_smooth_factor = 0.95;
double collec_smooth_factor = 0.5;
double tuning_factor = 0.03;
double alloc_mem = jl_gc_smooth(old_alloc_diff, alloc_diff, alloc_smooth_factor);
double alloc_time = jl_gc_smooth(old_mut_time, mutator_time, alloc_smooth_factor);
double gc_mem = jl_gc_smooth(old_freed_diff, freed_diff, collec_smooth_factor);
double gc_time = jl_gc_smooth(old_pause_time, pause, collec_smooth_factor);
old_alloc_diff = alloc_diff;
old_mut_time = mutator_time;
old_freed_diff = freed_diff;
old_pause_time = pause;
old_heap_size = heap_size;
double target_allocs = 0.0;
double min_interval = default_collect_interval;
double target_allocs;
if (alloc_mem != 0 && alloc_time != 0 && gc_mem != 0 && gc_time != 0) {
double alloc_rate = alloc_mem/alloc_time;
double gc_rate = gc_mem/gc_time;
target_allocs = sqrt(((double)heap_size/min_interval * alloc_rate)/(gc_rate * tuning_factor)); // work on multiples of min interval
if (collection == JL_GC_AUTO) {
uint64_t alloc_diff = before_free_heap_size - old_heap_size;
uint64_t freed_diff = before_free_heap_size - heap_size;
double alloc_smooth_factor = 0.95;
double collect_smooth_factor = 0.5;
double tuning_factor = 0.03;
double alloc_mem = jl_gc_smooth(old_alloc_diff, alloc_diff, alloc_smooth_factor);
double alloc_time = jl_gc_smooth(old_mut_time, mutator_time, alloc_smooth_factor);
double gc_mem = jl_gc_smooth(old_freed_diff, freed_diff, collect_smooth_factor);
double gc_time = jl_gc_smooth(old_pause_time, pause, collect_smooth_factor);
old_alloc_diff = alloc_diff;
old_mut_time = mutator_time;
old_freed_diff = freed_diff;
old_pause_time = pause;
old_heap_size = heap_size;
if (alloc_mem != 0 && alloc_time != 0 && gc_mem != 0 && gc_time != 0 ) {
double alloc_rate = alloc_mem/alloc_time;
double gc_rate = gc_mem/gc_time;
target_allocs = sqrt(((double)heap_size/min_interval * alloc_rate)/(gc_rate * tuning_factor)); // work on multiples of min interval
}
}
else
if (target_allocs == 0.0)
target_allocs = 2*sqrt((double)heap_size/min_interval);

uint64_t target_heap = (uint64_t)target_allocs*min_interval + heap_size;
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

this will be equivalent to sqrt(heap_sizealloc_rate/(gc_ratetuning_factor)) * sqrt(min_interval) + heap_size.
I am a bit confused - shouldnt this be independent of min_interval?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I used it as a scaling factor basically. It made it easier for me to reason about what taking the sqrt meant.

Expand Down Expand Up @@ -3611,7 +3601,6 @@ JL_DLLEXPORT void *jl_gc_counted_malloc(size_t sz)
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + sz);
jl_atomic_store_relaxed(&ptls->gc_num.malloc,
jl_atomic_load_relaxed(&ptls->gc_num.malloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, sz);
gbaraldi marked this conversation as resolved.
Show resolved Hide resolved
}
return malloc(sz);
Expand All @@ -3628,7 +3617,6 @@ JL_DLLEXPORT void *jl_gc_counted_calloc(size_t nm, size_t sz)
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + nm*sz);
jl_atomic_store_relaxed(&ptls->gc_num.malloc,
jl_atomic_load_relaxed(&ptls->gc_num.malloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, nm*sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, nm*sz);
}
return calloc(nm, sz);
Expand All @@ -3645,7 +3633,6 @@ JL_DLLEXPORT void jl_gc_counted_free_with_size(void *p, size_t sz)
jl_atomic_load_relaxed(&ptls->gc_num.freed) + sz);
jl_atomic_store_relaxed(&ptls->gc_num.freecall,
jl_atomic_load_relaxed(&ptls->gc_num.freecall) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.malloc_bytes_freed, sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, -sz);
}
}
Expand All @@ -3665,8 +3652,6 @@ JL_DLLEXPORT void *jl_gc_counted_realloc_with_old_size(void *p, size_t old, size
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + (sz - old));
jl_atomic_store_relaxed(&ptls->gc_num.realloc,
jl_atomic_load_relaxed(&ptls->gc_num.realloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.malloc_bytes_freed, old);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, sz-old);
}
return realloc(p, sz);
Expand Down Expand Up @@ -3742,7 +3727,6 @@ JL_DLLEXPORT void *jl_gc_managed_malloc(size_t sz)
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + allocsz);
jl_atomic_store_relaxed(&ptls->gc_num.malloc,
jl_atomic_load_relaxed(&ptls->gc_num.malloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, allocsz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, allocsz);
int last_errno = errno;
#ifdef _OS_WINDOWS_
Expand Down Expand Up @@ -3782,8 +3766,6 @@ static void *gc_managed_realloc_(jl_ptls_t ptls, void *d, size_t sz, size_t olds
jl_atomic_load_relaxed(&ptls->gc_num.allocd) + (allocsz - oldsz));
jl_atomic_store_relaxed(&ptls->gc_num.realloc,
jl_atomic_load_relaxed(&ptls->gc_num.realloc) + 1);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_mallocd, allocsz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.malloc_bytes_freed, oldsz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size, allocsz-oldsz);
int last_errno = errno;
#ifdef _OS_WINDOWS_
Expand Down Expand Up @@ -3873,7 +3855,6 @@ static void *gc_perm_alloc_large(size_t sz, int zero, unsigned align, unsigned o
#ifdef _OS_WINDOWS_
SetLastError(last_error);
#endif
jl_atomic_fetch_add_relaxed(&gc_heap_stats.bytes_allocd,sz);
jl_atomic_fetch_add_relaxed(&gc_heap_stats.heap_size,sz);
errno = last_errno;
jl_may_leak(base);
Expand Down Expand Up @@ -3918,7 +3899,6 @@ void *jl_gc_perm_alloc_nolock(size_t sz, int zero, unsigned align, unsigned offs
errno = last_errno;
if (__unlikely(pool == MAP_FAILED))
return NULL;
jl_atomic_fetch_add_relaxed(&gc_heap_stats.pages_perm_allocd, 1);
#endif
gc_perm_pool = (uintptr_t)pool;
gc_perm_end = gc_perm_pool + GC_PERM_POOL_SIZE;
Expand Down
5 changes: 0 additions & 5 deletions src/gc.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -261,11 +261,6 @@ typedef struct {
typedef struct {
_Atomic(size_t) bytes_mapped;
gbaraldi marked this conversation as resolved.
Show resolved Hide resolved
_Atomic(size_t) bytes_resident;
_Atomic(size_t) bytes_freed;
_Atomic(size_t) bytes_allocd;
_Atomic(size_t) bytes_mallocd;
_Atomic(size_t) malloc_bytes_freed;
_Atomic(size_t) pages_perm_allocd;
_Atomic(size_t) heap_size;
_Atomic(size_t) heap_target;
} gc_heapstatus_t;
gbaraldi marked this conversation as resolved.
Show resolved Hide resolved
Expand Down