MCU 'mcu' shutdown: Rescheduled timer in the past - Host Buffer 100%

General Discussion
20

Thanks. What are the benchmarks if just the cache is enabled and no other changes?

Just 42 ticks per stepper is troublesome as that is only 105ns for time between steps - which is very close to the minimum supported by the trinamic drivers (~103ns). It may be necessary to force Klipper to implement explicit stepper timing checks on the stm32h7 (which will reduce the total steps per second but avoid lost steps between mcu and driver).

As indicated by Jake, various Klipper code assumes a 32-bit counter will not overflow within 10s, which would not be true with a 480Mhz mcu.

-Kevin

EDIT: if you want to test with explicit Klipper step timing (slower but potentially safer) it is the following change (rerun make menuconfig, make, and make flash to apply):

--- a/src/stm32/Kconfig
+++ b/src/stm32/Kconfig
@@ -13,7 +13,7 @@ config STM32_SELECT
     select HAVE_GPIO_BITBANGING if !MACH_STM32F031
     select HAVE_STRICT_TIMING
     select HAVE_CHIPID
-    select HAVE_STEPPER_BOTH_EDGE
+    select HAVE_STEPPER_BOTH_EDGE if !MACH_STM32H7
     select SERIAL_BOOTLOADER_SIDECHANNEL
 
 config BOARD_DIRECTORY
1 Like