Commit Graph

96 Commits

Author SHA1 Message Date
Paul Sokolovsky 93e353e384 py/gc: Implement GC running by allocation threshold.
Currently, MicroPython runs GC when it could not allocate a block of memory,
which happens when heap is exhausted. However, that policy can't work well
with "inifinity" heaps, e.g. backed by a virtual memory - there will be a
lot of swap thrashing long before VM will be exhausted. Instead, in such
cases "allocation threshold" policy is used: a GC is run after some number of
allocations have been made. Details vary, for example, number or total amount
of allocations can be used, threshold may be self-adjusting based on GC
outcome, etc.

This change implements a simple variant of such policy for MicroPython. Amount
of allocated memory so far is used for threshold, to make it useful to typical
finite-size, and small, heaps as used with MicroPython ports. And such GC policy
is indeed useful for such types of heaps too, as it allows to better control
fragmentation. For example, if a threshold is set to half size of heap, then
for an application which usually makes big number of small allocations, that
will (try to) keep half of heap memory in a nice defragmented state for an
occasional large allocation.

For an application which doesn't exhibit such behavior, there won't be any
visible effects, except for GC running more frequently, which however may
affect performance. To address this, the GC threshold is configurable, and
by default is off so far. It's configured with gc.threshold(amount_in_bytes)
call (can be queries without an argument).
2016-07-21 00:37:30 +03:00
Paul Sokolovsky 749cbaca7f py/gc: Calculate (and report) maximum contiguous free block size.
Just as maximum allocated block size, it's reported in allocation units
(not bytes).
2016-07-01 00:09:55 +03:00
Paul Sokolovsky 6a6e0b7e05 py/gc: Be sure to count last allocated block at heap end in stats.
Previously, if there was chain of allocated blocks ending with the last
block of heap, it wasn't included in number of 1/2-block or max block
size stats.
2016-06-30 12:56:21 +03:00
Damien George a1c93a62b1 py: Don't use gc or qstr mutex when the GIL is enabled.
There is no need since the GIL already makes gc and qstr operations
atomic.
2016-06-28 11:28:50 +01:00
Damien George 3653f5144a py/gc: Fix GC+thread bug where ptr gets lost because it's not computed.
GC_EXIT() can cause a pending thread (waiting on the mutex) to be
scheduled right away.  This other thread may trigger a garbage
collection.  If the pointer to the newly-allocated block (allocated by
the original thread) is not computed before the switch (so it's just left
as a block number) then the block will be wrongly reclaimed.

This patch makes sure the pointer is computed before allowing any thread
switch to occur.
2016-06-28 11:28:49 +01:00
Damien George e33806aaff py/gc: Fix 2 cases of concurrent access to ATB and FTB. 2016-06-28 11:28:49 +01:00
Damien George c93d9caa8b py/gc: Make memory manager and garbage collector thread safe.
By using a single, global mutex, all memory-related functions (alloc,
free, realloc, collect, etc) are made thread safe.  This means that only
one thread can be in such a function at any one time.
2016-06-28 11:28:49 +01:00
Damien George 330165a2cc py: Add MP_STATE_THREAD to hold state specific to a given thread. 2016-06-28 11:09:31 +01:00
Paul Sokolovsky 68a7a92cec py/gc: gc_dump_alloc_table(): Dump heap offset instead of actual address.
Address printed was truncated anyway and in general confusing to outsider.
A line which dumps it is still left in the source, commented, for peculiar
cases when it may be needed (e.g. when running under debugger).
2016-05-13 00:16:38 +03:00
Paul Sokolovsky 9a8751b006 gc: gc_dump_alloc_table(): Use '=' char for tail blocks.
'=' is pretty natural character for tail, and gives less dense picture
where it's easier to see what object types are actually there.
2016-05-13 00:16:38 +03:00
Paul Sokolovsky bc04dc277e py/gc: Make (byte)array type dumping conditional on these types being enabled. 2016-05-11 19:21:53 +03:00
Paul Sokolovsky 3d7f3f00e0 py/gc: gc_dump_alloc_table(): Show byte/str and (byte)array objects.
These are typical consumers of large chunks of memory, so it's useful to
see at least their number (how much memory isn't clearly shown, as the data
for these objects is allocated elsewhere).
2016-05-11 19:00:15 +03:00
Paul Sokolovsky 3ea03a1188 py/gc: Improve mark/sweep debug output.
Previously, mark operation weren't logged at all, while it's quite useful
to see cascade of marks in case of over-marking (and in other cases too).
Previously, sweep was logged for each block of object in memory, but that
doesn't make much sense and just lead to longer output, harder to parse
by a human. Instead, log sweep only once per object. This is similar to
other memory manager operations, e.g. an object is allocated, then freed.
Or object is allocated, then marked, otherwise swept (one log entry per
operation, with the same memory address in each case).
2015-12-27 20:40:36 +02:00
Damien George acaccb37ec py/gc: When printing info, use %u instead of UINT_FMT for size_t args.
Ideally we'd use %zu for size_t args, but that's unlikely to be supported
by all runtimes, and we would then need to implement it in mp_printf.
So simplest and most portable option is to use %u and cast the argument
to uint(=unsigned int).

Note: reason for the change is that UINT_FMT can be %llu (size suitable
for mp_uint_t) which is wider than size_t and prints incorrect results.
2015-12-18 12:52:45 +00:00
Damien George d977d268e8 py/gc: Use size_t instead of mp_uint_t to count things related to heap.
size_t is the correct type to use to count things related to the size of
the address space.  Using size_t (instead of mp_uint_t) is important for
the efficiency of ports that configure mp_uint_t to larger than the
machine word size.
2015-12-16 20:09:11 -05:00
Damien George f7782f8082 py/gc: For finaliser, interpret a pointer into the heap as concrete obj. 2015-12-16 19:45:42 -05:00
Damien George 969e4bbe6a py/gc: Scan GC blocks as an array of pointers, not an array of objects.
The GC should search for pointers within the heap.  This patch makes a
difference when an object is larger than a pointer (eg 64-bit NaN
boxing).
2015-12-16 19:41:37 -05:00
Paul Sokolovsky 75feece208 py/gc: Make GC block size be configurable. 2015-12-03 01:40:52 +02:00
Damien George 999cedb90f py: Wrap all obj-ptr conversions in MP_OBJ_TO_PTR/MP_OBJ_FROM_PTR.
This allows the mp_obj_t type to be configured to something other than a
pointer-sized primitive type.

This patch also includes additional changes to allow the code to compile
when sizeof(mp_uint_t) != sizeof(void*), such as using size_t instead of
mp_uint_t, and various casts.
2015-11-29 14:25:35 +00:00
Damien George 94fe6e523d py/gc: Move away from using mp_uint_t, instead use uintptr_t and size_t.
The GC works with concrete pointers and so the types should reflect this.
2015-11-29 14:25:04 +00:00
Dave Hylands 7f3c0d1ea8 py: Clear finalizer flag when calling gc_free.
Currently, the only place that clears the bit is in gc_collect.
So if a block with a finalizer is allocated, and subsequently
freed, and then the block is reallocated with no finalizer then
the bit remains set.

This could also be fixed by having gc_alloc clear the bit, but
I'm pretty sure that free is called way less than alloc, so doing
it in free is more efficient.
2015-11-07 14:26:11 +00:00
Damien George 3a2171e406 py: Eliminate some cases which trigger unused parameter warnings. 2015-09-04 16:53:46 +01:00
Damien George ade9a05236 py: Improve allocation policy of qstr data.
Previous to this patch all interned strings lived in their own malloc'd
chunk.  On average this wastes N/2 bytes per interned string, where N is
the number-of-bytes for a quanta of the memory allocator (16 bytes on 32
bit archs).

With this patch interned strings are concatenated into the same malloc'd
chunk when possible.  Such chunks are enlarged inplace when possible,
and shrunk to fit when a new chunk is needed.

RAM savings with this patch are highly varied, but should always show an
improvement (unless only 3 or 4 strings are interned).  New version
typically uses about 70% of previous memory for the qstr data, and can
lead to savings of around 10% of total memory footprint of a running
script.

Costs about 120 bytes code size on Thumb2 archs (depends on how many
calls to gc_realloc are made).
2015-07-14 22:56:32 +01:00
Damien George e72cda99fd py: Convert occurrences of non-debug printf to mp_printf. 2015-04-16 14:30:16 +00:00
Damien George 12ab9eda8d py: Make heap printing compatible with 16-bit word size. 2015-04-03 14:11:13 +01:00
Damien George e1e359ff59 py: Put mp_sys_path, mp_sys_argv and gc_collected in mp_state_ctx_t.
Without mp_sys_path and mp_sys_argv in the root pointer section of the
state, their memory was being incorrectly collected by GC.
2015-02-07 17:24:10 +00:00
Damien George abc1959e2c py, unix, lib: Allow to compile with -Wold-style-definition. 2015-01-12 22:34:38 +00:00
Damien George ec21405821 py: Add (commented out) code to gc_dump_alloc_table for qstr info. 2015-01-11 14:37:06 +00:00
stijn afd6c8e1d2 Remove obsolete bss-related code/build features
GC for unix/windows builds doesn't make use of the bss section anymore,
so we do not need the (sometimes complicated) build features and code related to it
2015-01-08 15:29:44 +01:00
Damien George b4b10fd350 py: Put all global state together in state structures.
This patch consolidates all global variables in py/ core into one place,
in a global structure.  Root pointers are all located together to make
GC tracing easier and more efficient.
2015-01-07 20:33:00 +00:00
Damien George fd40a9c38e py: Make GC's STACK_SIZE definition a proper MICROPY_ config variable. 2015-01-01 22:04:46 +00:00
Damien George 51dfcb4bb7 py: Move to guarded includes, everywhere in py/ core.
Addresses issue #1022.
2015-01-01 20:32:09 +00:00
Damien George 7860c2a68a py: Fix some macros defines; cleanup some includes. 2014-11-05 21:16:41 +00:00
Damien George 109c1de015 py: Make gc.enable/disable just control auto-GC; alloc is still allowed.
gc.enable/disable are now the same as CPython: they just control whether
automatic garbage collection is enabled or not.  If disabled, you can
still allocate heap memory, and initiate a manual collection.
2014-10-31 21:30:46 +00:00
Damien George 0b13f3e026 py: Improve memory usage debugging; better GC AT dumping.
In unix port, mem_info(1) now prints pretty GC alloc table.
2014-10-24 23:12:25 +01:00
Damien George e7bb0443cd py: Properly free string parse-node; add assertion to gc_free. 2014-10-23 14:13:05 +01:00
Damien George 37378f8a9d py: Clean up edge cases of malloc/realloc/free. 2014-10-23 12:02:00 +01:00
Damien George c30595eb1b py: Add more debug printing code in gc_dump_alloc_table. 2014-10-17 14:12:57 +00:00
Damien George 37ada236b3 py: Take gc_pool_start out of bss section, to reclaim 1st block of heap. 2014-10-16 21:50:39 +01:00
Damien George 9b0b373e5e py: Fix GC realloc issue, where memory chunks were never shrunk.
Previously, a realloc to a smaller memory chunk size would not free the
unused blocks in the tail of the chunk.
2014-10-15 18:24:47 +00:00
Damien George 516b09efc3 py, gc: Further reduce heap fragmentation with new, faster gc alloc.
The heap allocation is now exactly as it was before the "faster gc
alloc" patch, but it's still nearly as fast.  It is fixed by being
careful to always update the "last free block" pointer whenever the heap
changes (eg free or realloc).

Tested on all tests by enabling EXTENSIVE_HEAP_PROFILING in py/gc.c:
old and new allocator have exactly the same behaviour, just the new one
is much faster.
2014-08-28 23:06:38 +01:00
Damien George b796e3d848 py: Reduce fragmentation of GC heap.
Recent speed up of GC allocation made the GC have a fragmented heap.
This patch restores "original fragmentation behaviour" whilst still
retaining relatively fast allocation.  This patch works because there is
always going to be a single block allocated now and then, which advances
the gc_last_free_atb_index pointer often enough so that the whole heap
doesn't need scanning.

Should address issue #836.
2014-08-28 10:18:40 +01:00
Damien George d5e7f6e37e py: Speed up GC allocation.
This simple patch gives a very significant speed up for memory allocation
with the GC.

Eg, on PYBv1.0:
tests/basics/dict_del.py: 3.55 seconds -> 1.19 seconds
tests/misc/rge_sm.py:     15.3 seconds -> 2.48 seconds
2014-08-22 18:17:02 +01:00
Damien George a1d3ee376c py: Fix bug where GC finaliser table was not completely zeroed out.
This was a nasty bug to track down.  It only had consequences when the
heap size was just the right size to expose the rounding error in the
calculation of the finaliser table size.  And, a script had to allocate
a small (1 or 2 cell) object at the very end of the heap.  And, this
object must not have a finaliser.  And, the initial state of the heap
must have been all bits set to 1.  All these conspire on the pyboard,
but only if your run the script fresh (so unused memory is all 1's),
and if your script allocates a lot of small objects (eg 2-char strings
that are not interned).
2014-08-08 12:33:49 +01:00
Damien George 40f3c02682 Rename machine_(u)int_t to mp_(u)int_t.
See discussion in issue #50.
2014-07-03 13:25:24 +01:00
Dave Hylands 2fe841d2fa Try not to cause a MemoryError when raising an exception during nterrupt handling.
Step 1 fixes #732
2014-06-30 22:49:21 -07:00
Paul Sokolovsky 59c675a64c py: Include mpconfig.h before all other includes.
It defines types used by all other headers.

Fixes #691.
2014-06-21 22:43:22 +03:00
stijn 9acb5e4cf0 gc: Turn off debugging info again 2014-06-18 12:29:03 +02:00
stijn def10cecd1 gc: Keep debug statements at beginning of scope where possible 2014-06-18 10:20:41 +02:00
stijn bbcea3f62b gc: More verbose debugging
Add more DEBUG_printf statements to trace gc behaviour
2014-06-16 12:43:35 +02:00