Enable protomatter on RP2040 builds #4186

jepler · 2021-02-11T22:38:40Z

Testing performed: ran on a 32x32 display

Other testing performed:

created a 128x64 display, 3 bits resolution (didn't drive anything)
checked across soft resets
checked recreating display

Also found a race condition between timer_disable and redraw, which would happen if I debugger-paused inside common_hal_rgbmatrix_timer_disable or put a delay or print inside it. That's what pausing inside reconstruct
fixes.

So that the "right timer" can be chosen, timer_allocate now gets the self pointer. It's guaranteed at this point that the pin information is accurate,so you can e.g., find a PWM unit related to the pins themselves. This required touching each port to add the parameter even though it's unused everywhere but raspberrypi.

Here's my testing code, with a 32x32 matrix:

import rgbmatrix
import board
import framebufferio
import displayio

displayio.release_displays()

MATRIX = rgbmatrix.RGBMatrix(
    width=32, height=32, bit_depth=3, tile=1,
    rgb_pins=[board.GP0,
              board.GP1,
              board.GP2,
              board.GP3,
              board.GP4,
              board.GP5],
    addr_pins=[board.GP9,
               board.GP10,
               board.GP11,
               board.GP12],
    clock_pin=board.GP6, latch_pin=board.GP7,
    output_enable_pin=board.GP8)

DISPLAY = framebufferio.FramebufferDisplay(MATRIX, auto_refresh=True,
                                           rotation=0)

Also found a race condition between timer_disable and redraw, which would happen if I debugger-paused inside common_hal_rgbmatrix_timer_disable or put a delay or print inside it. That's what pausing inside reconstruct fixes. So that the "right timer" can be chosen, `timer_allocate` now gets the `self` pointer. It's guaranteed at this point that the pin information is accurate, so you can e.g., find a PWM unit related to the pins themselves. This required touching each port to add the parameter even though it's unused everywhere but raspberrypi.

These changes are tested along with adafruit/circuitpython#4186 and work on a 32x32 matrix.

tannewt

I don't see any issues. Thanks! (Not approving because it's a draft.)n

jepler · 2021-02-12T19:56:11Z

Thanks! We should probably wait for the pull request up at protomatter to be accepted first, but I marked this ready for review anyhow.

These changes are tested along with adafruit/circuitpython#4186 and work on a 32x32 matrix.

jepler · 2021-02-12T23:15:11Z

I was confused about the "red X" failed CI notes, but those are for the Protomatter CI running in my fork, on a commit that refers to this PR. Click over to the "checks" tab to see a green board.

ladyada · 2021-02-14T19:16:35Z

ok so... technically yes this does work.
but it seems like when it tries to get into REPL-on-display mode it crashes, and if thats during a file save sometimes so hard that the USB filesys gets corrupted and i have to nuke it to get it back.

jepler · 2021-02-14T19:36:48Z

Uh oh, it worked better than that for me. I'll test further on Monday. I know I tested being in the REPL, soft restarting, but mmmmaybe not writing to CIRCUITPY.

jepler · 2021-02-26T16:29:59Z

I tested by loading an adapted version of the fruit demo from https://github.com/adafruit/Adafruit_Learning_System_Guides/blob/master/CircuitPython_RGBMatrix/fruit.py and causing heavy CIRCUITPY filesystem activity by running cat emoji.bmp{,,,} > bigmoji.bin; sync but didn't reproduce a crash.

I also re-tested entering and using the REPL and didn't encounter problems.

While writing to CIRCUITPY, the display will temporarily glitch. I believe this is because code can only execute from SRAM during flash erase (and write?) operations. Since protomatter (like most of the rest of circuitpython) resides in the flash it can't be executed during this time.

Anyhow, as a merge conflict had arisen I've merged main into this PR and with any luck there will be fresh artifacts shortly.

jepler · 2021-02-26T17:13:21Z

We figured out why @ladyada and I were having ahem varying levels of success. Limor was testing the prototype Feather, while I was testing the pico. Feather has a status LED.

At the time everything goes south, GDB says...

#0  0x1006926c in _PM_swapbuffer_maybe ()
#1  0x10050986 in common_hal_rgbmatrix_rgbmatrix_refresh ()
#2  0x1004606e in rgbmatrix_rgbmatrix_swapbuffers ()
#3  0x1004e5b4 in _refresh_display ()
#4  0x1004eb04 in framebufferio_framebufferdisplay_background ()
#5  0x1004d5ec in displayio_background ()
#6  0x10027aaa in supervisor_background_tasks ()
#7  0x10026780 in background_callback_run_all ()
#8  0x10027b46 in supervisor_run_background_tasks_if_tick ()
#9  0x100325f8 in common_hal_rp2pio_statemachine_write ()
#10 0x1003ef9c in common_hal_neopixel_write ()
#11 0x10026ed6 in rgb_led_status_init ()
#12 0x1003ec24 in reset_all_pins ()
#13 0x1002652e in reset_port ()
#14 0x1002623a in main ()

I can now investigate this properly.

jepler · 2021-02-26T22:42:19Z

I got another hung backtrace, again using the feather uf2 but not obviously related to neopixel:

#0  0x10069088 in _PM_row_handler (core=0x20000994 <displays+20>)
    at ../../lib/protomatter/src/core.c:534
#1  0x10069238 in _PM_PWM_ISR () at ../../lib/protomatter/src/arch/rp2040.h:148
#2  <signal handler called>
#3  0x10068698 in _PM_init (core=core@entry=0x20000994 <displays+20>, 
    bitWidth=bitWidth@entry=64, bitDepth=bitDepth@entry=4 '\004', 
    rgbCount=<optimized out>, rgbCount@entry=1 '\001', 
    rgbList=rgbList@entry=0x20000a0c <displays+140> "\006\005\t\v\n\f", 
    addrCount=<optimized out>, 
    addrList=addrList@entry=0x20000a2a <displays+170> "\031\030\035\034", 
    clockPin=13 '\r', latchPin=latchPin@entry=1 '\001', 
    oePin=oePin@entry=0 '\000', doubleBuffer=doubleBuffer@entry=true, 
    tile=-1 '\377', timer=0x10003) at ../../lib/protomatter/src/core.c:119
#4  0x1005071a in common_hal_rgbmatrix_rgbmatrix_reconstruct (
    self=self@entry=0x20000980 <displays>, framebuffer=framebuffer@entry=0x0)
    at ../../shared-module/rgbmatrix/RGBMatrix.c:97
#5  0x1002fad6 in supervisor_display_move_memory ()
    at ../../supervisor/shared/display.c:144
#6  0x10039e2e in supervisor_move_memory ()
    at ../../supervisor/shared/memory.c:316
#7  0x1002621a in cleanup_after_vm (heap=0x200033b8 <allocations+8>)
    at ../../main.c:233
#8  run_repl () at ../../main.c:528
#9  main () at ../../main.c:593

jepler · 2021-03-01T21:38:22Z

@ladyada I re-tested the slot machine demo from Adafruit_Learning_System_Guides/CircuitPython_RGBMatrix using the Feather RP2040 build and my Pico pinout on a Pico, but with R and B pin swapped to trigger the bug I was able to reproduce.

During the testing process, I ten times hit ctrl-c, ctrl-d to re-load the program, and caused filesystem activity with cat emoji.bmp{,,,} > large.bin; sync; rm large.bin; sync also 10 times with 5-second sleeps between. (of course, large.bin is only about 115kB big)

I did not observe any hard-faults or lockups, only the expected screen flickering while writing to CIRCUITPY. I did the same test with your Feather pinout, and while I couldn't see the display I also didn't have any lockups or hard-faults.

jepler · 2021-03-01T21:39:44Z

See adafruit/Adafruit_Protomatter#36 for the PWM slice number fix, already merged in yay

dhalbert · 2021-03-02T22:16:41Z

@jepler is this all fixed and ready to merge? Did @ladyada test?

ladyada · 2021-03-03T00:31:43Z

i did test, i had a wierd hardfault once but couldnt repro it, and i think it had more to do with USB so this is ok to merge!

ladyada

tested with feather + matrix featherwing!

jepler mentioned this pull request Feb 11, 2021

Changes needed to let the code work on CircuitPython & RP2040 adafruit/Adafruit_Protomatter#33

Merged

jepler force-pushed the update-protomatter-rp2 branch from 763f7ac to ff1942c Compare February 12, 2021 14:28

jepler added a commit to jepler/Adafruit_Protomatter that referenced this pull request Feb 12, 2021

Changes needed to let the code work on CircuitPython & RP2040

78cde80

These changes are tested along with adafruit/circuitpython#4186 and work on a 32x32 matrix.

tannewt self-requested a review February 12, 2021 19:13

tannewt reviewed Feb 12, 2021

View reviewed changes

jepler marked this pull request as ready for review February 12, 2021 19:55

jepler added a commit to jepler/Adafruit_Protomatter that referenced this pull request Feb 12, 2021

Changes needed to let the code work on CircuitPython & RP2040

c3cccab

These changes are tested along with adafruit/circuitpython#4186 and work on a 32x32 matrix.

jepler requested a review from tannewt February 12, 2021 22:07

Merge remote-tracking branch 'origin/main' into update-protomatter-rp2

1d1ff5f

jepler mentioned this pull request Feb 26, 2021

rp2040: Need to enable/disable the PWM_IRQ_WRAP at these times adafruit/Adafruit_Protomatter#35

Closed

jepler added 2 commits March 1, 2021 15:13

protomatter: get an rp2 fix by updating

407a8c2

Merge remote-tracking branch 'origin/main' into update-protomatter-rp2

238484e

ladyada approved these changes Mar 3, 2021

View reviewed changes

jepler merged commit efc2667 into adafruit:main Mar 3, 2021

jepler deleted the update-protomatter-rp2 branch November 3, 2021 21:11

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Enable protomatter on RP2040 builds #4186

Enable protomatter on RP2040 builds #4186

Uh oh!

jepler commented Feb 11, 2021 •

edited

Loading

Uh oh!

tannewt left a comment

Uh oh!

jepler commented Feb 12, 2021

Uh oh!

jepler commented Feb 12, 2021

Uh oh!

ladyada commented Feb 14, 2021

Uh oh!

jepler commented Feb 14, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Mar 1, 2021

Uh oh!

jepler commented Mar 1, 2021

Uh oh!

dhalbert commented Mar 2, 2021

Uh oh!

ladyada commented Mar 3, 2021

Uh oh!

ladyada left a comment

Uh oh!

Uh oh!

Enable protomatter on RP2040 builds #4186

Enable protomatter on RP2040 builds #4186

Uh oh!

Conversation

jepler commented Feb 11, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

tannewt left a comment

Choose a reason for hiding this comment

Uh oh!

jepler commented Feb 12, 2021

Uh oh!

jepler commented Feb 12, 2021

Uh oh!

ladyada commented Feb 14, 2021

Uh oh!

jepler commented Feb 14, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Feb 26, 2021

Uh oh!

jepler commented Mar 1, 2021

Uh oh!

jepler commented Mar 1, 2021

Uh oh!

dhalbert commented Mar 2, 2021

Uh oh!

ladyada commented Mar 3, 2021

Uh oh!

ladyada left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

jepler commented Feb 11, 2021 •

edited

Loading