Movatterモバイル変換

[0]ホーム
URL:

Skip to content

Navigation Menu

Sign in
Appearance settings

Search code, repositories, users, issues, pull requests...

Provide feedback

We read every piece of feedback, and take your input very seriously.

Saved searches

Use saved searches to filter your results more quickly

 Sign up
Appearance settings

Add print(float) to save space (VS double).#2036

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to ourterms of service andprivacy statement. We’ll occasionally send you account related emails.

Already on GitHub?Sign in to your account

Merged
fpistm merged 1 commit intostm32duino:mainfromhonnet:main
Jun 27, 2023

Conversation

honnet
Copy link
Contributor

Summary

This PR implements the following methods in Print.h and Print.cpp:

  • Feature 1: Print::print(float) - - - only double was available
  • Feature 2: Print::println(float) - - - only double was available
  • Feature 2: Print::printFloat(float) - - - only double was available => maybe it should be called printDouble(double)?

Motivation

Some of the new MCUs have a limited memory (ex: 32KB for STM32C011D6) and don't have floating point calculation unit, so a simple accelerometer driver (example) can take more than 100% of the flash if we don't optimize it a bit (the compilation fails at first).

This PR saves up to 10K by avoiding the compiler to use ARM libs for double calculation subroutines such as__aeabi_dsub,__aeabi_dadd,__aeabi_dmul, etc.

Validation

The following command allows listing the heaviest functions after compilation:
arm-none-eabi-nm --print-size --size-sort --radix=d -l file.elf | tail -n 3

134242369 00001416 T __aeabi_dmul134240049 00001724 T __aeabi_dadd134243785 00001796 T __aeabi_dsub

The problem can also be observed in the assembly code [with C code + comments using debug enabled (-g) if there's enough space]:
arm-none-eabi-objdump -S file.elf

Note
The printFloat(float) function repeats a lot of what printFloat(double) does and it should probably be factorized.
I expect that this PR could be rejected for this reason at least, let me know if you have any suggestion of how to implement it (using templates?).

@fpistmfpistm self-requested a reviewJune 7, 2023 13:05
@fpistm
Copy link
Member

HI@honnet
Thanks for this PR.
do you evaluate theSerial.printf("%f"); (enable Newlib + prinrtf float support in the menu) or the dtostrf()?

@honnet
Copy link
ContributorAuthor

honnet commentedJun 7, 2023•
edited
Loading

Thanks for the feedback!

Sadly, with the simple accelerometer example, enabling Newlib with prinrtf takes too much space (=> linker error).

...but it says how much memory is missing so I got the following values:

  • newlib + float + Serial.print() : 42K
  • newlib + float + Serial.printf(): 41K
  • dtostrf: 30K

As a comparison, with my PR, the space used is 27K (which is already tight for the 32K available flash).

Edit - other example:

For theC011 orC031 (32KB max), the following code will not compile without my PR:
https://github.com/adafruit/Adafruit_MMC56x3/tree/main/examples/compass

@fpistmfpistm added the enhancementNew feature or request labelJun 12, 2023
@fpistm
Copy link
Member

Hi@honnet
I've updated your PR with template to avoid code duplication.
I guess this enhancement could be shared with ArduinoCore-API in order to be aligned:
https://github.com/arduino/ArduinoCore-API/blob/master/api/Print.h

honnet reacted with hooray emoji

@fpistmfpistm added this to the2.6.0 milestoneJun 26, 2023
Copy link
Member

@fpistmfpistm left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others.Learn more.

LGTM

@honnet
Copy link
ContributorAuthor

Great!
Thanks a lot for the improvement!

@honnethonnet closed thisJun 26, 2023
@fpistmfpistm reopened thisJun 26, 2023
@fpistm
Copy link
Member

It is not yet merged 😉

@honnet
Copy link
ContributorAuthor

My bad, sorry!

@fpistm
Copy link
Member

@honnet seems you rebase and removed all changes....

Signed-off-by: Cedric Honnet <honnet@telecom-paristech.org>Co-Authored-By: Frederic Pillon <frederic.pillon@st.com>
@honnet
Copy link
ContributorAuthor

honnet commentedJun 26, 2023•
edited
Loading

Damn, I didn't pay attention and I thought I integrated your modifications!

I found a hash in my terminal history, cherry-picked it and pushed it:
main...honnet:Arduino_Core_STM32:main

Does it make the PR usable now, or should I do it again?

@fpistm
Copy link
Member

I will restore it tomorrow. Do nothing until I fix the issue.

honnet reacted with thumbs up emoji

@fpistmfpistm reopened thisJun 26, 2023
@fpistmfpistm merged commitc6bc5b2 intostm32duino:mainJun 27, 2023
@honnet
Copy link
ContributorAuthor

I just tested a c/cpp compiler option to forbid doubles:-fsingle-precision-constant:
https://github.com/stm32duino/Arduino_Core_STM32/blob/main/platform.txt#L35-L37

It's even more systematic and future proof against libraries that hide implicit doubles (in constants for example).

...but I'm not sure how to make it user friendly:

  1. in the Optimize options?
    https://github.com/stm32duino/Arduino_Core_STM32/blob/main/boards.txt#L11490

  2. in the C Runtime Library options?
    https://github.com/stm32duino/Arduino_Core_STM32/blob/main/boards.txt#L12404

  3. something else? nowhere? (maybe it's just a bad idea)

Is it worth exploring something clean? (any advice would be appreciated)

@fpistm
Copy link
Member

You can usebuild_opt.h feature, it was added to support such kind of request:
https://github.com/stm32duino/Arduino_Core_STM32/wiki/Customize-build-options-using-build_opt.h

honnet reacted with thumbs up emoji

@honnet
Copy link
ContributorAuthor

Perfect!
Thanks again so much!!

Sign up for freeto join this conversation on GitHub. Already have an account?Sign in to comment
Reviewers

@fpistmfpistmfpistm approved these changes

Assignees
No one assigned
Labels
enhancementNew feature or request
Projects
None yet
Milestone
2.6.0
Development

Successfully merging this pull request may close these issues.

2 participants
@honnet@fpistm
[8]ページ先頭
©2009-2025 Movatter.jp