7c1584aef1
When := is used in a comprehension the target variable is bound to the
parent scope, so it's either a global or a nonlocal. Prior to this commit
that was handled by simply using the parent scope's id_info for the
target variable. That's completely wrong because it uses the slot number
for the parent's Python stack to store the variable, rather than the slot
number for the comprehension. This will in most cases lead to incorrect
behaviour or memory faults.
This commit fixes the scoping of the target variable by explicitly
declaring it a global or nonlocal, depending on whether the parent is the
global scope or not. Then the id_info of the comprehension can be used to
access the target variable. This fixes a lot of cases of using := in a
comprehension.
Code size change for this commit:
bare-arm: +0 +0.000%
minimal x86: +0 +0.000%
unix x64: +152 +0.019% standard
stm32: +96 +0.024% PYBV10
cc3200: +96 +0.052%
esp8266: +196 +0.028% GENERIC
esp32: +156 +0.010% GENERIC[incl +8(data)]
mimxrt: +96 +0.027% TEENSY40
renesas-ra: +88 +0.014% RA6M2_EK
nrf: +88 +0.048% pca10040
rp2: +104 +0.020% PICO
samd: +88 +0.033% ADAFRUIT_ITSYBITSY_M4_EXPRESS
Fixes issue #10895.
Signed-off-by: Damien George <damien@micropython.org>
|
||
|---|---|---|
| .. | ||
| basics | ||
| cmdline | ||
| cpydiff | ||
| esp32 | ||
| extmod | ||
| feature_check | ||
| float | ||
| frozen | ||
| import | ||
| inlineasm | ||
| internal_bench | ||
| io | ||
| jni | ||
| micropython | ||
| misc | ||
| multi_bluetooth | ||
| multi_net | ||
| net_hosted | ||
| net_inet | ||
| perf_bench | ||
| pyb | ||
| qemu-arm | ||
| renesas-ra | ||
| stress | ||
| thread | ||
| unicode | ||
| unix | ||
| wipy | ||
| README.md | ||
| run-internalbench.py | ||
| run-multitests.py | ||
| run-natmodtests.py | ||
| run-perfbench.py | ||
| run-tests-exp.py | ||
| run-tests-exp.sh | ||
| run-tests.py | ||
README.md
MicroPython Test Suite
This directory contains tests for various functionality areas of MicroPython. To run all stable tests, run "run-tests.py" script in this directory.
Tests of capabilities not supported on all platforms should be written to check for the capability being present. If it is not, the test should merely output 'SKIP' followed by the line terminator, and call sys.exit() to raise SystemExit, instead of attempting to test the missing capability. The testing framework (run-tests.py in this directory, test_main.c in qemu_arm) recognizes this as a skipped test.
There are a few features for which this mechanism cannot be used to condition a test. The run-tests.py script uses small scripts in the feature_check directory to check whether each such feature is present, and skips the relevant tests if not.
Tests are generally verified by running the test both in MicroPython and in CPython and comparing the outputs. If the output differs the test fails and the outputs are saved in a .out and a .exp file respectively. For tests that cannot be run in CPython, for example because they use the machine module, a .exp file can be provided next to the test's .py file. A convenient way to generate that is to run the test, let it fail (because CPython cannot run it) and then copy the .out file (but not before checking it manually!)
When creating new tests, anything that relies on float support should go in the float/ subdirectory. Anything that relies on import x, where x is not a built-in module, should go in the import/ subdirectory.
perf_bench
The perf_bench directory contains some performance benchmarks that can be used
to benchmark different MicroPython firmwares or host ports.
The runner utility is run-perfbench,py. Execute ./run-perfbench.py --help
for a full list of command line options.
Benchmarking a target
To run tests on a firmware target using pyboard.py, run the command line like
this:
./run-perfbench.py -p -d /dev/ttyACM0 168 100
-pindicates running on a remote target via pyboard.py, not the host.-d PORTNAMEis the serial port,/dev/ttyACM0is the default if not provided.168is valueN, the approximate CPU frequency in MHz (in this case Pyboard V1.1 is 168MHz). It's possible to choose other values as well: lower values like10will run much the tests much quicker, higher values like1000will run much longer.100is valueM, the approximate heap size in kilobytes (can get this fromimport micropython; micropython.mem_info()or estimate it). It's possible to choose other values here too: lower values like10will run shorter/smaller tests, and higher values will run bigger tests. The maximum value ofMis limited by available heap, and the tests are written so the "recommended" value is approximately the upper limit.
Benchmarking the host
To benchmark the host build (unix/Windows), run like this:
./run-perfbench.py 2000 10000
The output of perfbench is a list of tests and times/scores, like this:
N=2000 M=10000 n_average=8
perf_bench/bm_chaos.py: SKIP
perf_bench/bm_fannkuch.py: 94550.38 2.9145 84.68 2.8499
perf_bench/bm_fft.py: 79920.38 10.0771 129269.74 8.8205
perf_bench/bm_float.py: 43844.62 17.8229 353219.64 17.7693
perf_bench/bm_hexiom.py: 32959.12 15.0243 775.77 14.8893
perf_bench/bm_nqueens.py: 40855.00 10.7297 247776.15 11.3647
perf_bench/bm_pidigits.py: 64547.75 2.5609 7751.36 2.5996
perf_bench/core_import_mpy_multi.py: 15433.38 14.2733 33065.45 14.2368
perf_bench/core_import_mpy_single.py: 263.00 11.3910 3858.35 12.9021
perf_bench/core_qstr.py: 4929.12 1.8434 8117.71 1.7921
perf_bench/core_yield_from.py: 16274.25 6.2584 12334.13 5.8125
perf_bench/misc_aes.py: 57425.25 5.5226 17888.60 5.7482
perf_bench/misc_mandel.py: 40809.25 8.2007 158107.00 9.8864
perf_bench/misc_pystone.py: 39821.75 6.4145 100867.62 6.5043
perf_bench/misc_raytrace.py: 36293.75 6.8501 26906.93 6.8402
perf_bench/viper_call0.py: 15573.00 14.9931 19644.99 13.1550
perf_bench/viper_call1a.py: 16725.75 9.8205 18099.96 9.2752
perf_bench/viper_call1b.py: 20752.62 8.3372 14565.60 9.0663
perf_bench/viper_call1c.py: 20849.88 5.8783 14444.80 6.6295
perf_bench/viper_call2a.py: 16156.25 11.2956 18818.59 11.7959
perf_bench/viper_call2b.py: 22047.38 8.9484 13725.73 9.6800
The numbers across each line are times and scores for the test:
- Runtime average (microseconds, lower is better)
- Runtime standard deviation as a percentage
- Score average (units depend on the benchmark, higher is better)
- Score standard deviation as a percentage
Comparing performance
Usually you want to know if something is faster or slower than a reference. To
do this, copy the output of each run-perfbench.py run to a text file.
This can be done multiple ways, but one way on Linux/macOS is with the tee
utility: ./run-perfbench.py -p 168 100 | tee pyb-run1.txt
Once you have two files with output from two different runs (maybe with
different code or configuration), compare the runtimes with ./run-perfbench.py -t pybv-run1.txt pybv-run2.txt or compare scores with ./run-perfbench.py -s pybv-run1.txt pybv-run2.txt:
> ./run-perfbench.py -s pyb-run1.txt pyb-run2.txt
diff of scores (higher is better)
N=168 M=100 pyb-run1.txt -> pyb-run2.txt diff diff% (error%)
bm_chaos.py 352.90 -> 352.63 : -0.27 = -0.077% (+/-0.00%)
bm_fannkuch.py 77.52 -> 77.45 : -0.07 = -0.090% (+/-0.01%)
bm_fft.py 2516.80 -> 2519.74 : +2.94 = +0.117% (+/-0.00%)
bm_float.py 5749.27 -> 5749.65 : +0.38 = +0.007% (+/-0.00%)
bm_hexiom.py 42.22 -> 42.30 : +0.08 = +0.189% (+/-0.00%)
bm_nqueens.py 4407.55 -> 4414.44 : +6.89 = +0.156% (+/-0.00%)
bm_pidigits.py 638.09 -> 632.14 : -5.95 = -0.932% (+/-0.25%)
core_import_mpy_multi.py 477.74 -> 477.57 : -0.17 = -0.036% (+/-0.00%)
core_import_mpy_single.py 58.74 -> 58.72 : -0.02 = -0.034% (+/-0.00%)
core_qstr.py 63.11 -> 63.11 : +0.00 = +0.000% (+/-0.01%)
core_yield_from.py 357.57 -> 357.57 : +0.00 = +0.000% (+/-0.00%)
misc_aes.py 397.27 -> 396.47 : -0.80 = -0.201% (+/-0.00%)
misc_mandel.py 3375.70 -> 3375.84 : +0.14 = +0.004% (+/-0.00%)
misc_pystone.py 2265.36 -> 2265.97 : +0.61 = +0.027% (+/-0.01%)
misc_raytrace.py 367.61 -> 368.15 : +0.54 = +0.147% (+/-0.01%)
viper_call0.py 605.92 -> 605.92 : +0.00 = +0.000% (+/-0.00%)
viper_call1a.py 576.78 -> 576.78 : +0.00 = +0.000% (+/-0.00%)
viper_call1b.py 452.45 -> 452.46 : +0.01 = +0.002% (+/-0.01%)
viper_call1c.py 457.39 -> 457.39 : +0.00 = +0.000% (+/-0.00%)
viper_call2a.py 561.37 -> 561.37 : +0.00 = +0.000% (+/-0.00%)
viper_call2b.py 389.49 -> 389.50 : +0.01 = +0.003% (+/-0.01%)
Note in particular the error percentages at the end of each line. If these are
high relative to the percentage difference then it indicates high variability in
the test runs, and the absolute difference value is unreliable. High error
percentages are particularly common on PC builds, where the host OS may
influence test run times. Increasing the N value may help average this out by
running each test longer.