You can now do:
X = const(123)
Y = const(456 + X)
and the compiler will replace X and Y with their values.
See discussion in issue #266 and issue #573.
This is primarily intended to provide testing of Thumb-specific code within
Travis CI as well as if anyone else want to run it locally. As discussed in
purposes. This is currently agains an emulated Cortex-M3 core, however in
the near future it can extended to support M0, M0+ as well M4 (work in
progress exists in sushihangover/qemu).
It's probably true that most of the code base can be covered running uPy
natively on a POSIX system, however we do have the tiny bit of assembly
code. There may exist bugs related to endianness and type aliases, let
alone potential standard library or compiler bugs or even
architecture-specific optimisations.
This could also incorporate lwIP (or other TCP/IP stack) integration as well
as SDIO+FATFS drivers.
The solution to inline the test cases was chose due to simplicity. It could
alternatively be implemented in a number of different way (see #515), but
this looked the simplest.
Inclusion of tinytest was just to avoid writing boilerplate code for
counting failed tests and other utility functions. Currently only a few
functions are used, however this could be extended. Checking in the code
instead of using submodule was a personal preference, but if people do want
the pain of submodules, this can provided. This particular framework is
also pretty good if one desires to run unit test on target. The approach
with scripts being inlined is probably not quite suited for the size of
memory an MCU has, but the tinytest itself should be good, if lower-level C
code is to be unit tested.
... and we have not that bad mapping type after all - lookup time is ~ the
same as in one-attr instance. My namedtuple implementation on the other
hand degrades awfully.
So, need to rework it. First observation is that named tuple fields are
accessed as attributes, so all names are interned at the program start.
Then, really should store field array as qstr[], and do quick 32/64 bit
scan thru it.
In case of empty non-blocking read()/write(), both return None. read()
cannot return 0, as that means EOF, so returns another value, and then
write() just follows. This is still pretty unexpected, and typical
"if not len:" check would treat this as EOF. Well, non-blocking files
require special handling!
This also kind of makes it depending on POSIX, but well, anything else
should emulate POSIX anyway ;-).
Need to have a policy as to how far we go adding keyword support to
built ins. It's nice to have, and gets better CPython compatibility,
but hurts the micro nature of uPy.
Addresses issue #577.
There are 2 locations in parser, and 1 in compiler, where memory
allocation is not precise. In the parser it's the rule stack and result
stack, in the compiler it's the array for the identifiers in the current
scope. All other mallocs are exact (ie they don't allocate more than is
needed).
This patch adds tuning options (MP_ALLOC_*) to mpconfig.h for these 3
inexact allocations.
The inexact allocations in the parser should actually be close to
logarithmic: you need an exponentially larger script (absent pathological
cases) to use up more room on the rule and result stacks. As such, the
default allocation policy for these is now to start with a modest sized
stack, but grow only in small increments.
For the identifier arrays in the compiler, these now start out quite
small (4 entries, since most functions don't have that many ids), and
grow incrementally by 6 (since if you have more ids than 4, you probably
have quite a few more, but it wouldn't be exponentially more).
Partially addresses issue #560.
Motivation is optimizing handling of various constructs as well as
understanding which constructs are more efficient in MicroPython.
More info: http://forum.micropython.org/viewtopic.php?f=3&t=77
Results are wildly unexpected. For example, "optimization" of range
iteration into while loop makes it twice as slow. Generally, the more
bytecodes, the slower the code.
