Another reduction of -48 bytes can be had if the fine calculation
step is skipped. The worst difference compared to the old reference
code with my calibration values in the 0° to 60° was 2°C,
and the difference at 25°C is 1°C.
The final size decrease for non-full builds like Trinket M0 is 268
bytes.
Perform most arithmetic with scaled integer values.
For my calibration values
```
const uint32_t NVMCTRL_TEMP_LOG[]={0xfc05511e, 0xcc7ac0f7};
```
the maximum difference between the old and new calculation is 0.50°C.
The difference is smallest (0.13°) at 25.87°C in the old scale.
This reduces mcu_processor_get_temperature from 568 bytes to 348 bytes
(-220 bytes)
Originally, black_bindings found each contiguous "//|" block and sent
it to black independently. This was slower than it needed to be.
Instead, swap the comment prefix: when running black, take off
"//|" prefixes and put "##|" prefixes on all un-prefixed lines.
Then, after black is run, do the opposite operation
This more than doubles the overall speed of "pre-commit run --all",
from 3m20s to 55s CPU time on my local machine (32.5s to under 10s
"elapsed" time)
It also causes a small amount of churn in the bindings, because
black now sees enough context to know whether one 'def' follows another
or ends the 'def's in a 'class'. In the latter case, it adds an extra
newline, which becomes a "//|" line.
I'm less sure why a trailing comma was omitted before down in
rp2pio/StateMachine.c but let's roll with it.
- define CIRCUITPY_BUILD_EXTENSIONS to predefined values
- set CIRCUITPY_BUILD_EXTENSIONS in port and board config
- reuse the support matrix "get_settings_from_makefile" to get it
- move the existing port and board specific values
- remove the C3 specific board values because it's not the default
- update build_release_files.py to use get_settings_from_makefile
Everything should be using the keypad module instead.
Note: there are several boards that still had gamepadshift enabled. I
did not contact their authors to make sure they already switched to
keypad in their code and documentation. We should probably wait with
merging this for their go ahead.
.. and enable it on atmel-samd and raspberrypi. On trinket_m0 this saves
96 net bytes of flash. There are 216 bytes actually saved by reducing
the flash storage size of the property descriptors, but added code in
several paths takes back over half of the 'raw savings'.
By organizing the "get-only" and "get-set" (but no delete) properties
each in a different section, we can represent then more efficiently.
Testing performed: that a get-only property can still be gotten but
can't be set or deleted; that a get-set property can sill be gotten or
set but can't be deleted. Tested on pygamer.
Because this requires linker file support, I only enabled it on two of
the ports.
I instrumented RTC_Handler and determined that on SAMD51 it was possible
for the interrupt to be delivered well before the actual overflow of the
RTC COUNT register (e.g., a value as small as 0xffff_fffd could be seen
at the time of overflow)
Rather than depending on the overflow interrupt coming in at the same time
as COUNT overflows (exactly), rely only on observed values of COUNT in
_get_count, overflowing when it wraps around from a high value to a low
one.
With this change, PLUS a second change so that it is possible to warp
the RTC counter close to an overflow and test in 20ms instead of 3 days,
there was no problem detected over 20000+ overflows. Before, a substantial
fraction (much greater than 10%) of overflows failed.
Fixes#5985
Change to common-hal/rtc/RTC.c for time warping (plus make rtc_old_count non-static):
```patch
void common_hal_rtc_set_calibration(int calibration) {
+
+ common_hal_mcu_disable_interrupts();
+
+ RTC->MODE0.COUNT.reg = 0xffffff00;
+ rtc_old_count = 0;
+ do {
+ while ((RTC->MODE0.SYNCBUSY.reg & (RTC_MODE0_SYNCBUSY_COUNTSYNC | RTC_MODE0_SYNCBUSY_COUNT)) != 0) { }
+ }
+ while(RTC->MODE0.COUNT.reg < 0xffffff00);
+ common_hal_mcu_enable_interrupts();
+
+ mp_printf(&mp_plat_print, "Warping RTC in calibration setter count=%08x rtc_old_count=%08x\n", RTC->MODE0.COUNT.reg, rtc_old_count);
```
Test program:
```python
import time
from rtc import RTC
i = 0
while True:
RTC().calibration = 1 # Warps to ~16ms before overflow, with patch to RTC code
t0 = time.monotonic_ns()
et = t0 + 20_000_000 # 20ms
while (t1 := time.monotonic_ns()) < et: pass
i += 1
print(f"{i:6d}: duration {t1-t0}")
if t1-t0 > 200_000_000: break
print()
```
To me, it made more sense to track which boards go together in a cluster;
when reviewing a request to actually use a duplicate vid/pid, you want
to know what board(s) it is aliasing.
I also revamped the detection of non-USB boards so that a board .mk file
that couldn't be parsed by the code here would raise a problem instead
of just being skipped for the purposes of checking.
There were some lines with comments on the end, and some variation in
capitalization of the IDs. These are all normalized and a (sometimes
unfriendly!) error printed when it's incorrect.
Before this, here were some ways to trick the duplicate vid/pid checker:
```
USB_PID = 0XABCD
USB_PID = 0xAbCd
USB_PID = 0xABCD # harmless comment?
```
None of these things were ever done on purpose.
Removes:
* AUTORESET_DELAY_MS which never did anything but was introduced
somehow.
* CIRCUITPY_BOOT_BUTTON in all but one ESP board because they all have
them. There is a default based on the strapping pins.
* BOARD_USER_SAFE_MODE_ACTION because it was all the same for boards
with boot buttons. Now the safe mode code manages the message.
- Add reset for autoreload. De-request ticks.
- Separate state a little more in autoreload.c
- Rename some routines.
- Remove redundant settings of CIRCUITPY_AUTORELOAD_DELAY_MS.
