rp2/main.c: Set the default clock frequency at boot. #13718

robert-hh · 2024-02-21T08:52:42Z

As a side effect, the peripheral clock will be set to 48Mhz and both UART and I2C baud rates will not be affected any more by CPU speed changes using machine.freq().

With the change the UART baud rate range is 50 to 3_000_000. That is a somewhat breaking change since baud rates beyond 3MBaud are not supported any more.

Addresses the discussion #13679.

dpgeorge · 2024-02-26T00:51:53Z

ports/rp2/main.c

@@ -223,6 +227,12 @@ int main(int argc, char **argv) {
        soft_timer_deinit();
        gc_sweep_all();
        mp_deinit();
+        // Reset the MCU frequency
+        set_sys_clock_khz(125000, false);


I'm not sure it should reset the frequency on soft reset. Soft reset is about resetting all the Python state, whereas it's best to leave a lot of the hardware state as-is. Eg the WiFi is still connected. Eg on stm32 the frequency is not reset on soft reset.

The intention here is to have the device at the same state as when entering the for (;;) { loop. If code e.g. in boot.py and main.py changes the frequency, it will still do it.

The other discussion which is open in the forum is, whether the peripheral clock should be set to 48MHz. Or should it better follow the CPU freq change. That can be implemented as well by changing the code for machine.freq(). My preference was for the more conservative and less surprising state, but others missed the high speed options.

Gadgetoid · 2024-02-27T09:54:00Z

Could this possibly also include some accommodation for #8208

I'd very much like a board config option to change that default, please!

github-actions · 2024-06-01T18:19:35Z

Code size report:

   bare-arm:    +0 +0.000% 
minimal x86:    +0 +0.000% 
   unix x64:    +0 +0.000% standard
      stm32:    +0 +0.000% PYBV10
     mimxrt:    +0 +0.000% TEENSY40
        rp2:  +136 +0.015% RPI_PICO_W
       samd:    +0 +0.000% ADAFRUIT_ITSYBITSY_M4_EXPRESS

robert-hh · 2024-06-01T18:30:54Z

@dpgeorge I dropped resetting the clock at soft reboot. Only at cold boot the clock it set to have the side effect of changing the peripheral frequency to 48MHz.

dpgeorge · 2024-07-03T14:25:07Z

ports/rp2/modmachine.c

+    //                 0,
+    //                 CLOCKS_CLK_PERI_CTRL_AUXSRC_VALUE_CLK_SYS,
+    //                 SYS_CLK_KHZ * KHZ,
+    //                 SYS_CLK_KHZ * KHZ);


I'm not sure what this commented-out code is aiming to do. When set_sys_clock_khz() is called, does that update the PERI clock or not?

The whole change is about the inconsistency, that when machine.freq() is called to change the frequency, then PERI clock is set to 48 MHz as result of calling set_sys_clock_khz(). The change I suggested would set PERI clock always to 48MHz. Then, changing the CPU frequency would not affect e.g. the UART or PWM settings, at the penalty of limiting the peripherals baud rates.

This commented code is about the alternative suggestion to set PERI clock always to the CPU clock. Then there would be changes to the baud rates of the peripherals when changing the CPU clock, but at known rates. I added that code here just to recall it,

OK, I understand more.

So without this PR (ie current master), what is the value of PERI at start up, is it the system clock, or system clock divided by some factor?

Note that machine.freq() takes multiple arguments to specify the frequency of the buses (see eg stm32 port). So maybe it's a good idea to allow the user to specify the PERI clock as the second (optional) argument to machine.freq()? Something like:

machine.freq(sys, peri=48000000)

At the moment the PERI freq at boot is the system clock 125Mhz and it will be set to 48Mhz when calling machine.freq(n).

Note that machine.freq() takes multiple arguments to specify the frequency of the buses (see eg stm32 port).

Actually, the stm32 port lost this feature with commit 7d39db2. machine.freq() takes just 1 argument.

I changed machine.freq() to have a second boolean type agrument. If set to False or omitted, the peripheral frequency will be 48MHz, which is the default value now at boot. If set to True, the peripheral frequency is the same as the MCU. The clock section for the peripheral clock has no divider. The documentation is wrong. So the peripheral clock can only be one of the other clock sources, like USB clock or SYS clock.
The API could also be changed to accept a frequency instead of a boolean value, with 48_000_000 and the value of the first argument as the permitted values. But then, an error check & message would have to be added.
As a required side change, machine.freq() now accepts up to 4 arguments. I made a separate PR for that 1 char change as well.

The API could also be changed to accept a frequency instead of a boolean value, with 48_000_000 and the value of the first argument as the permitted values. But then, an error check & message would have to be added.

It looks like this is how you ended up implementing it (ie not as a boolean). I think what you did -- specifying a freq -- is the right way to do it. That way is more general, and more consistent with stm32.

But do you think it's better to make the peri frequency a keyword-only argument? That forces the user to be explicit and makes it even more future proof/general. Eg machine.freq(125000000, peri=48000000).

But do you think it's better to make the peri frequency a keyword-only argument? That forces the user to be explicit and makes it even more future proof/general. Eg machine.freq(125000000, peri=48000000).

The Python API is is extmod/modmachine.c. Making peri=xxx a keyword only argument would either require to change that or to move the API back to the rp2 port.

The Python API is is extmod/modmachine.c. Making peri=xxx a keyword only argument would either require to change that

Yes, if we want to make this a keyword-only argument, we'd need to change the API in extmod/modmachine.c. We can do that if we think it's the best option. All other ports can remain as they are (stm32 accepting 4 positional args, all other ports accepting 1 positional arg).

I do not know if it the best option, and it's a larger change. The STM32 port does not have the keywords.

OK, let's leave it as a positional argument.

codecov · 2024-07-06T11:05:56Z

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.43%. Comparing base (55b2720) to head (4b3e419).

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #13718   +/-   ##
=======================================
  Coverage   98.43%   98.43%           
=======================================
  Files         161      161           
  Lines       21281    21281           
=======================================
  Hits        20948    20948           
  Misses        333      333

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

dpgeorge · 2024-07-08T13:30:47Z

ports/rp2/modmachine.c

+        mp_obj_new_int(mp_hal_get_cpu_freq()),
+        mp_obj_new_int(clock_get_hz(clk_peri)),
+    };
+    return mp_obj_new_tuple(MP_ARRAY_SIZE(tuple), tuple);


Returning a 2-tuple here is a big breaking change... not sure how to handle this. My suggestion for now is to just retain the old behaviour here (ie just return the CPU freq), and we can work out how to query the peri frequency in the future (if anyone ever needs it).

(I know stm32 returns a tuple, but a lot of code will be written already for rp2 that expects this to return a single integer.)

Keeping the old behavior is easy. That's why it's a separate commit.

Thanks for removing the additional commit (and returning just a single int).

As a side effect, the peripheral clock will be set to 48Mhz and both UART and I2C will not be affected by CPu speed changed using `machine.freq()`. With the change the UART baud rate range is 50 to 3_000_000. Signed-off-by: robert-hh <robert@hammelrath.com>

By default, the peripheral clock for UART and SPI is set to 48 MHz and will not be affected by the MCU clock change. This can be changed by a second argument to `machine.freq(freq, peripheral_freq)`. The second argument must be either 48 MHz or identical with the first argument. Note that UART and SPI baud rates may have to be re-configured after changing the MCU clock. Signed-off-by: robert-hh <robert@hammelrath.com>

dpgeorge · 2024-08-20T02:36:22Z

Could this possibly also include some accommodation for #8208

I'd very much like a board config option to change that default, please!

@Gadgetoid that was not done in this PR (a config option for default CPU freq). That should be pretty easy to do, but someone needs to do it 😄

robert-hh · 2024-08-20T07:05:43Z

Thank you.

dpgeorge reviewed Feb 26, 2024

View reviewed changes

dpgeorge added the port-rp2 label Feb 26, 2024

robert-hh force-pushed the rp2_clock_init branch 2 times, most recently from e5a8965 to af7470a Compare March 7, 2024 15:09

robert-hh force-pushed the rp2_clock_init branch from af7470a to 9e6cd27 Compare June 1, 2024 18:13

robert-hh changed the title ~~rp2/main.c: Set the default clock frequency at boot and soft reset.~~ rp2/main.c: Set the default clock frequency at boot. Jun 5, 2024

dpgeorge reviewed Jul 3, 2024

View reviewed changes

robert-hh force-pushed the rp2_clock_init branch 2 times, most recently from f35c4a9 to c8e4511 Compare July 6, 2024 11:03

robert-hh force-pushed the rp2_clock_init branch 2 times, most recently from 9b2a6a9 to 7c53d70 Compare July 8, 2024 11:35

dpgeorge reviewed Jul 8, 2024

View reviewed changes

robert-hh force-pushed the rp2_clock_init branch from 7c53d70 to 4956a08 Compare July 8, 2024 19:31

lucblender mentioned this pull request Jul 15, 2024

Speed up display lucblender/lx-euclid-001#4

Closed

robert-hh force-pushed the rp2_clock_init branch from 4956a08 to 4b3e419 Compare August 7, 2024 18:21

robert-hh added 2 commits August 20, 2024 12:26

dpgeorge force-pushed the rp2_clock_init branch from 4b3e419 to 7270b87 Compare August 20, 2024 02:27

dpgeorge merged commit 7270b87 into micropython:master Aug 20, 2024
9 checks passed

robert-hh deleted the rp2_clock_init branch August 20, 2024 07:21

projectgus mentioned this pull request Sep 26, 2024

rp2: Fix dropped UART REPL bytes on soft reset. #15914

Merged

MilhouseVH mentioned this pull request Jan 20, 2025

W5500-EVB-PICO - Unexplained TCP & UDP Transfer Speed Delays and Variations...Very Slow #15645

Open

Uh oh!

rp2/main.c: Set the default clock frequency at boot. #13718

rp2/main.c: Set the default clock frequency at boot. #13718

Uh oh!

Conversation

robert-hh commented Feb 21, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Gadgetoid commented Feb 27, 2024

Uh oh!

github-actions bot commented Jun 1, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

robert-hh commented Jun 1, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

robert-hh Jul 3, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Jul 6, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

dpgeorge commented Aug 20, 2024

Uh oh!

robert-hh commented Aug 20, 2024

Uh oh!

Uh oh!

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

github-actions bot commented Jun 1, 2024 •

edited

Loading

robert-hh Jul 3, 2024 •

edited

Loading

codecov bot commented Jul 6, 2024 •

edited

Loading