Lower power by using the Wait for Interrupt (WFI) instruction - #2685

@jepler Please take a look at this. I've only redone the nRF port so far and would like high level feedback before I do the others. The only remaining work for nRF is to get PulseIn working again which uses the monotonic clock for pulse timing.

I can take a closer look at this tomorrow morning, but it seems like good progress so far.

Thanks @jepler and @hierophect for the high level review. I've gotta propagate the changes to other ports and just wanted to make sure I was heading the right direction.

Thanks @dhalbert. sigh I may wait until next week to tackle M0 flakiness. I'm itching to work on ESP32-S2. Will see how I feel tomorrow.

@tannewt Yes, this can wait :)

@dhalbert I can't reproduce this flakiness with a Metro M0. Most of the time it works. Occasionally, I do see the boot hang very early but I think I can get the same behavior on master. I was testing it with an Arch Linux system and a Beagle sniffing the USB traffic.

I've merged in latest master which is what I tested with. Please try again and let me know if it's consistently bad for you still. I may need your help debugging it.

@tannewt It's still failing on Metro M0, somewhat more than half the time. I just press reset and see what happens. Attached is a Beagle trace from a failure.

metro-m0-fail.tdc.zip

I will not work on it more this evening.

I'm seeing a lot of assert errors in tinyusb with this patch when running on an nRF52833. For example:

(gdb) c
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
process_set_config (cfg_num=1 '\001', rhport=0 '\000') at ../../lib/tinyusb/src/device/usbd.c:731
731         TU_ASSERT( drv_id < USBD_CLASS_DRIVER_COUNT && drv_len >= sizeof(tusb_desc_interface_t) );
(gdb) c
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
process_control_request (p_request=0x2001ff20, rhport=0 '\000') at ../../lib/tinyusb/src/device/usbd.c:660
660         default: TU_BREAKPOINT(); return false;
(gdb)

(gdb) c
Continuing.

Program received signal SIGTRAP, Trace/breakpoint trap.
process_set_config (cfg_num=1 '\001', rhport=0 '\000') at ../../lib/tinyusb/src/device/usbd.c:731
731         TU_ASSERT( drv_id < USBD_CLASS_DRIVER_COUNT && drv_len >= sizeof(tusb_desc_interface_t) );
(gdb)

Also, the ASSERT that gets hit is probably due to the tinyusb process not being called often enough, causing the host to send a USB RESET, which hits this bug: hathach/tinyusb#179

@xobs Would you mind PRing the '833 board defs? I have a dev kit I can try it on here.

PR generated at #2828

Note that it requires a bootloader that's smaller (such as https://github.com/simmel-project/bootloader) in order to give more storage on the small internal flash.

@dhalbert This is ready for another look and test. Thanks!

Ok, @dhalbert. This commit should be a winner.

@jepler Ok for me to merge?

congrats on the big lift!