Attempt to address issue #386. unique_code_id's have been removed and
replaced with a pointer to the "raw code" information. This pointer is
stored in the actual byte code (aligned, so the GC can trace it), so
that raw code (ie byte code, native code and inline assembler) is kept
only for as long as it is needed. In memory it's now like a tree: the
outer module's byte code points directly to its children's raw code. So
when the outer code gets freed, if there are no remaining functions that
need the raw code, then the children's code gets freed as well.
This is pretty much like CPython does it, except that CPython stores
indexes in the byte code rather than machine pointers. These indices
index the per-function constant table in order to find the relevant
code.
Based on the discussion in #433. mp_load_attr() is critical-path function,
so any extra check will slowdown any script. As supporting default val
required only for getattr() builtin, move correspending implementation
there (still as a separate function due to concerns of maintainability
of such almost-duplicated code instances).
Finishes addressing issue #424.
In the end this was a very neat refactor that now makes things a lot
more consistent across the py code base. It allowed some
simplifications in certain places, now that everything is a dict object.
Also converted builtins tables to dictionaries. This will be useful
when we need to turn builtins into a proper module.
It's not completely satisfactory, because a failed call to __getattr__
should not raise an exception.
__setattr__ could be implemented, but it would slow down all stores to a
user created object. Need to implement some caching system.
There was thinkos that either send_value or throw_value is specified, but
there were cases with both. Note that send_value is pushed onto generator's
stack - but that's probably only good, because if we throw exception into
gen, it should not ever use send_value, and that will be just extra "assert".
In this case, the exception is just re-thrown - the ideas is that object
doesn't handle this exception specially, so it will propagated per Python
semantics.
Adding this bytecode allows to remove 4 others related to
function/method calls with * and ** support. Will also help with
bytecodes that make functions/closures with default positional and
keyword args.
Pretty much everyone needs to include map.h, since it's such an integral
part of the Micro Python object implementation. Thus, the definitions
are now in obj.h instead. map.h is removed.
Mostly just a global search and replace. Except rt_is_true which
becomes mp_obj_is_true.
Still would like to tidy up some of the names, but this will do for now.
Rationale: setting up the stack (state for locals and exceptions) is
really part of the "code", it's the prelude of the function. For
example, native code adjusts the stack pointer on entry to the function.
Native code doesn't need to know n_state for any other reason. So
putting the state size in the bytecode prelude is sensible.
It reduced ROM usage on STM by about 30 bytes :) And makes it easier to
pass information about the bytecode between functions.
Originally, .methods was used for methods in a ROM class, and
locals_dict for methods in a user-created class. That distinction is
unnecessary, and we can use locals_dict for ROM classes now that we have
ROMable maps.
This removes an entry in the bloated mp_obj_type_t struct, saving a word
for each ROM object and each RAM object. ROM objects that have a
methods table (now a locals_dict) need an extra word in total (removed
the methods pointer (1 word), no longer need the sentinel (2 words), but
now need an mp_obj_dict_t wrapper (4 words)). But RAM objects save a
word because they never used the methods entry.
Overall the ROM usage is down by a few hundred bytes, and RAM usage is
down 1 word per user-defined type/class.
There is less code (no need to check 2 tables), and now consistent with
the way ROM modules have their tables initialised.
Efficiency is very close to equivaluent.
For this, needed to implement DELETE_NAME bytecode (because var bound
in except clause is automatically deleted at its end).
http://docs.python.org/3/reference/compound_stmts.html#except :
"When an exception has been assigned using as target, it is cleared at
the end of the except clause."
mp_module_obj_t can now be put in ROM.
Configuration of float type is now similar to longint: can now choose
none, float or double as the implementation.
math module has basic math functions. For STM port, these are not yet
implemented (they are just stub functions).
Each built-in exception is now a type, with base type BaseException.
C exceptions are created by passing a pointer to the exception type to
make an instance of. When raising an exception from the VM, an
instance is created automatically if an exception type is raised (as
opposed to an exception instance).
Exception matching (RT_BINARY_OP_EXCEPTION_MATCH) is now proper.
Handling of parse error changed to match new exceptions.
mp_const_type renamed to mp_type_type for consistency.
Linear table at the moment, to eventually be replaced with a hash table
generated by a preprocessor.
Dynamic table is retained so that builtins can be overridden.
sys.path is not initialized by rt_init(), that's left for platform-specific
startup code. (For example, bare metal port may have some hardcoded defaults,
and let user change sys.path directly; while port for OS with environment
feature can take path from environment). If it's not explicitly initialized,
modules will be imported only from a current directory.
TODO: Decide if we really need separate bytecode for creating functions
with default arguments - we would need same for closures, then there're
keywords arguments too. Having all combinations is a small exponential
explosion, likely we need just 2 cases - simplest (no defaults, no kw),
and full - defaults & kw.
__bool__() and __len__() are just the same as __neg__() or __invert__(),
and require efficient dispatching implementation (not requiring search/lookup).
type->unary_op() is just the right choice for this short of adding
standalone virtual method(s) to already big mp_obj_type_t structure.
We still have FAST_[0,1,2] byte codes, but they now just access the
fastn array (before they had special local variables). It's now
simpler, a bit faster, and uses a bit less stack space (on STM at least,
which is most important).
The only reason now to keep FAST_[0,1,2] byte codes is for compressed
byte code size.
LOAD_METHOD bug was: emitbc did not correctly calculate the amount of
stack usage for a LOAD_METHOD operation.
small int bug was: int was being used to pass small ints, when it should
have been machine_int_t.
Currently, MicroPython strings are mix between CPython byte and unicode
strings. So, conversion is null so far. This dummy implementation is
intended for compatibility with CPython (so, same code can run on both).
These are micropython.mem_total(), .mem_current(), .mem_peak(). These are 3
individual functions with simple scalar return value to make sure that
calls to these functions themselves have minimal (hopefully zero) impact on
memory allocation.
Change state layout in VM so the stack starts at state[0] and grows
upwards. Locals are at the top end of the state and number downwards.
This cleans up a lot of the interface connecting the VM to C: now all
functions that take an array of Micro Python objects are in order (ie no
longer in reverse).
Also clean up C API with keyword arguments (call_n and call_n_kw
replaced with single call method that takes keyword arguments). And now
make_new takes keyword arguments.
emitnative.c has not yet been changed to comply with the new order of
stack layout.
Implicit "micropython" module contains (at least) codegeneration decorators.
Make it explicit, so an app could have "import micropython". On MicroPython,
that will be no-op. On CPython, that will give a chance to have a module
with placeholder decorators.
We likely should make mp_obj_new_int() inline, and rely on its
encapsulated check rather than inline checks everywhere explicitly.
Also, parser for big small int values is still broken.
Qstr's are now split into a linked-list of qstr pools. This has 2
benefits: the first pool can be in ROM (huge benefit, since we no longer
use RAM for the core qstrs), and subsequent pools use m_new for the next
pool instead of m_renew (thus avoiding a huge single table for all the
qstrs).
Still would be better to use a hash table, but this scheme takes us part
of the way (eventually convert the pools to hash tables).
Also fixed bug with import.
Also improved the way the module code is referenced (not magic number 1
anymore).
A big change. Micro Python objects are allocated as individual structs
with the first element being a pointer to the type information (which
is itself an object). This scheme follows CPython. Much more flexible,
not necessarily slower, uses same heap memory, and can allocate objects
statically.
Also change name prefix, from py_ to mp_ (mp for Micro Python).
Paul Sokolovsky