I programmed an STM32 for the first time 2 days ago. This really gave me a so much deeper insight into how these controllers work. Honestly loving how simple stupid it really is. I am also once again surprised and also not surprised that we can just use C to generate the startup assembly only using a few compiler flags.
So much really interesting stuff. I feel like this 30 minute video is equivalent to weeks of reading the manuals without fully understanding them. Really well presented. Wow.
Great video! Two small notes: most recent controllers will actually run vendor code from a ROM before jumping into user code - you sadly can't debug that in many cases since the debugger gets enabled after that... What is done there? E.g. loading calibration values that must be loaded in all circumstances. And the other: even without linking the C standard library you may have to provide memset/memcpy/... since gcc will (at some optimization levels) replace loops with those functions - independent of command line flags
No shade, makes total sense - as I said its a great video as it is Just thought it might be interesting to some - surprised me a bit in the beginning but makes sense that there might be some other code hidden...
Excellent, very straightforward explanation that completely cleaned up my confusions. I'm trying to bringup my own programming language on RP2040, that was super helpful to me. Although I have a few questions about bootloaders: * how and where do they fit into the whole picture * why someone would need one * how to deal (or not) with them from the point of view of writing everything from scratch. Basically how to write a barebones C program that will work under a bootloader.
It depends :) But I think it makes sense to think of a bootloader as another small standalone program that runs before your main program. So, just like you need startup code to make the MCU ready to run the main program, you need startup code before the bootloader program. A bootloader is optional, but it's often used to separate the update mechanism from the main program, and to allow the main program to be upgraded without a dedicated debugger. If your bootloader is simple and perhaps not written in C, it may not require as much startup code though. In practice it may be like this: startup code 1->bootloader->startup code 2->main program.
@@artfulbytes Thanks for the reply, that makes sense. But what I still do not fully understand is if the bootloader has to live somewhere in the flash, it will basically have to take the space of the interrupt table, so the main program cannot provide ISRs directly via a table. Is that correct? So the bootloader has to provide the interrupt stubs, and then before loading the main program the bootloader has to do some kind of dynamic linking to wire up main program's ISRs, or am I compeletely wrong here?
@@stainlessCode STM32 allows you to change the address of the interrupt vector table (IVT) by writing to a register (VTOR). And yes bootloader would live in flash and it can have its own IVT and startup code. You would then place your main program including its IVT somewhere else in flash memory. Then just before the bootloader jumps to the main program it can write to VTOR to change the address to the IVT of the main program. Moving the IVT is not possible on all MCUs and in those cases you may have to partly share the IVT or solve it some other clever way.
Thanks for making this video. I had saved Memfault and EmbeddedArtistry's articles in my "read someday" list, but they are not light reading. This video makes these concepts much more enjoyable for me to follow along to (an engineer transitioning from mechatronics to embedded systems).
this is gold! 🙏 thank you for laying it out, can't tell how much time I spent digging through internet trying to make sense out of linker scripts and how they are related to the startup scripts
I was super excited to see a new video from you, and it did not disappoint! I always thought it was interesting that C-only startup code was a selling point of ARM Cortex M, but most vendors still used assembly. Very interesting to see how it's done. Keep up the awesome videos!
It’s the most useful video i’ve ever seen. My life will never be the same. Awesome, please, do more like this. Maybe compile smallest linux kernel for stm32?🤓
@@diegoporras7769 The sp increases from 0x200017f8 to 0x20001800 after stepping over the first instruction (ldr and not mov). I haven't checked, but I think it may be a debug artifact, i.e. register view not refreshing properly until stepping over the first instruction.
This video is so much gold. It gives so much insight of how things work in a microcontroller. Thank you so much. Keep that going! PS: Are you considering linking agains picolibc?
I think I've linked with picolibc in a few zephyr projects without thinking about it :) Not really considered the differences between picolibc vs newlib.
Compiler/toolchain may not support it yet, compatibility with existing code (e.g. stm32 HAL), potential issues/bugs. So unless you really need the latest C features, probably safer to stick with something more established. With that said, I have not tested it extensively.
Nice explanation, excelent job and mega respect, outstanding knowledge ! But i have feeling like with ORM vs SQL, so much boilerplate around C code, that assembly looks clearer to me ;) Yes, i know, C will be partialy portable between simillar architecures/boards, but still i have no clear winner.
Hi, have you ever used the C2000/C6000 MCUs from Texas Instruments? There isn't much information about them on this site and since their architecture differs from that of ARM, I thought I'd be interesting to create a series around them. Although, I would think it'd be for a very niche audience.
@@artfulbytes No, I meant actually in the code. Maybe not in this case, but sometimes I see that the debug trace doesnt progress sequentially, especially in low level code (goes back and forth)
@@Jonathan-ru9zl Normally the program counter just increments, but there are branch/jump instructions such as bcc, which can jump to a specific address. For example, these instructions are used when repeating a code block such as in a loop, or when jumping to a function.
Can you do a long series like you did the past but with STM32 microcontrollers, I really wanted to write code on it from scratch, and understand all the concept MCU-related also, but I can’t find any suitable resources than yours. Thank you for bringing this such a nice educational videos!
The value of sp increases from 0x200017f8 to 0x20001800 on running the first two assembly code lines. If estack was 0x20001800, then why was sp initialized by HW to 0x200017f8?
The sp increases from 0x200017f8 to 0x20001800 after stepping over the first instruction. I haven't checked, but I guess it may be a debug artifact, i.e. register view not refreshing properly until stepping over the first instruction.
@artfulbytes that's possible - may be due to debug interface Ldr command doesn't change the value of sp so it does not make sense that it changes after ldr.
I have stm32 bule phill it comes under arm Cortex m3 architecture I but doesn't reach main it goes half fault . I don't where the mistake is where any problem in openocd
kinda similar to writing an OS from scratch and writing a bootloader I passed that stage so I could understand what's going on in here, of course this is more specific to the MCU archtiecture and startup code needs to be configured according to the manufacturer and needs lots of datasheet checking about MCU specifically STM32 for this video In the end, it's a great vid. And I wonder your background like Electronics or CS?
yo pls do a vid about the stm32f407 i got one as a gift and i'm stuck with it i wanna use do something and i need to work around the roll,pitch & yaw it has builtin osciloscope and idk how to work with it. thx btw for your content
In practice, there may be some overlap, but roughly speaking, a bootloader is a small program that runs before the main application, while startup code is code that runs before any program, including the bootloader.
go more stm32? And if it possible, I think parallel compare with AVR8 like atmega8 most be very understandable and opens up understanding. Yes it must be hard to make video, but u can make most greate course and try sell them
It is possible to expect vedios on weekend . We are already waited for longtime so plese make Quickly and discuss various concepts as soon as possible. thanks artful bytes❤
I appreciate the lesson as always. However this is way too dense. I, and perhaps others, may be out my depth especially as I don't know what 'gcc expects...' really entails.
Please never leave us again 🙏🏾
ditto
yeaaassss
miss you!!!!!!
I programmed an STM32 for the first time 2 days ago. This really gave me a so much deeper insight into how these controllers work.
Honestly loving how simple stupid it really is.
I am also once again surprised and also not surprised that we can just use C to generate the startup assembly only using a few compiler flags.
So much really interesting stuff. I feel like this 30 minute video is equivalent to weeks of reading the manuals without fully understanding them. Really well presented. Wow.
This is the type of content, I want to pay the internet bill for. Thanks a lot bro❤
Very clearly, thanks man. I also devevop the bare metal embedded at the MCU company. This is very valuable lesson.
Every video is a banger. I love these deep dives into the "basics" we often forget or never learn properly
Great video! Two small notes: most recent controllers will actually run vendor code from a ROM before jumping into user code - you sadly can't debug that in many cases since the debugger gets enabled after that... What is done there? E.g. loading calibration values that must be loaded in all circumstances.
And the other: even without linking the C standard library you may have to provide memset/memcpy/... since gcc will (at some optimization levels) replace loops with those functions - independent of command line flags
Thanks, appreciate the input! Lots of details to keep track of and I try to get away with brushing over many of them :)
No shade, makes total sense - as I said its a great video as it is
Just thought it might be interesting to some - surprised me a bit in the beginning but makes sense that there might be some other code hidden...
You’re knowledgeable of the gcc tool chain is insane!!! Keep up the great videos!!
Literally I was thinking about this channel yesterday.
He's back boys!
Haha same here. I actively searched for his channel yesterday and noted it was long time ago he uploaded. Quite scary
Excellent, very straightforward explanation that completely cleaned up my confusions. I'm trying to bringup my own programming language on RP2040, that was super helpful to me.
Although I have a few questions about bootloaders:
* how and where do they fit into the whole picture
* why someone would need one
* how to deal (or not) with them from the point of view of writing everything from scratch. Basically how to write a barebones C program that will work under a bootloader.
It depends :) But I think it makes sense to think of a bootloader as another small standalone program that runs before your main program. So, just like you need startup code to make the MCU ready to run the main program, you need startup code before the bootloader program. A bootloader is optional, but it's often used to separate the update mechanism from the main program, and to allow the main program to be upgraded without a dedicated debugger. If your bootloader is simple and perhaps not written in C, it may not require as much startup code though. In practice it may be like this: startup code 1->bootloader->startup code 2->main program.
@@artfulbytes Thanks for the reply, that makes sense. But what I still do not fully understand is if the bootloader has to live somewhere in the flash, it will basically have to take the space of the interrupt table, so the main program cannot provide ISRs directly via a table. Is that correct?
So the bootloader has to provide the interrupt stubs, and then before loading the main program the bootloader has to do some kind of dynamic linking to wire up main program's ISRs, or am I compeletely wrong here?
@@stainlessCode STM32 allows you to change the address of the interrupt vector table (IVT) by writing to a register (VTOR). And yes bootloader would live in flash and it can have its own IVT and startup code. You would then place your main program including its IVT somewhere else in flash memory. Then just before the bootloader jumps to the main program it can write to VTOR to change the address to the IVT of the main program. Moving the IVT is not possible on all MCUs and in those cases you may have to partly share the IVT or solve it some other clever way.
@@artfulbytes I see, that makes sense, thank you for the reply!
Welcome back
This time, teach us electronics and programming from scratch.
C and ASM
Thanks for making this video. I had saved Memfault and EmbeddedArtistry's articles in my "read someday" list, but they are not light reading. This video makes these concepts much more enjoyable for me to follow along to (an engineer transitioning from mechatronics to embedded systems).
👍
Nice to see this channel uploading again! Hope to see more content on low-level stuffs!
this is gold! 🙏 thank you for laying it out, can't tell how much time I spent digging through internet trying to make sense out of linker scripts and how they are related to the startup scripts
Super informative! I do not work with embedded systems, but I found it fascinating to see how the start up process works here. Thank you!
welcome back for detailed teaching
I was super excited to see a new video from you, and it did not disappoint! I always thought it was interesting that C-only startup code was a selling point of ARM Cortex M, but most vendors still used assembly. Very interesting to see how it's done. Keep up the awesome videos!
Buddy, your videos are a goldmine! Thanks for sharing.
It’s the most useful video i’ve ever seen. My life will never be the same. Awesome, please, do more like this. Maybe compile smallest linux kernel for stm32?🤓
So glad to see you're posting.
Your videos are the best on microcontrollers and embedded systems!
6:01 I think the SP is actually changing after the mov sp, r0
Amazing video so far, watching till the end.
From …17F8 to …1800
@@diegoporras7769 The sp increases from 0x200017f8 to 0x20001800 after stepping over the first instruction (ldr and not mov). I haven't checked, but I think it may be a debug artifact, i.e. register view not refreshing properly until stepping over the first instruction.
So glad to see you back 🎉
Welcome back, keep uploading :)
Thankyou for coming back please keep carry on quality informative contents for Us
You're back! I really enjoy watching your videos very informative and full of valuable information. Keep up the great work!
Ohhhh, finally a video where they explain this! Thanks!!
Happy yo see you back again! Thank you
This video is so much gold. It gives so much insight of how things work in a microcontroller. Thank you so much. Keep that going! PS: Are you considering linking agains picolibc?
I think I've linked with picolibc in a few zephyr projects without thinking about it :) Not really considered the differences between picolibc vs newlib.
This is great if you want to play with things like Ben Eater's 6502 project and want to add C code
👍
Teach us embedded systems you are so good in this please share your knowledge with us🙏🏻
Deep. Learned so much!
9:38 Just curious what reason is there not to use -std=c23 every time?
Compiler/toolchain may not support it yet, compatibility with existing code (e.g. stm32 HAL), potential issues/bugs. So unless you really need the latest C features, probably safer to stick with something more established. With that said, I have not tested it extensively.
@ interesting…
Good to be back 😊
Thank you for the amazing content 🎉
Amazing video! Great work!! I learned a lot!
Welcome back man!!
I am waiting for your video so long
Nice explanation, excelent job and mega respect, outstanding knowledge !
But i have feeling like with ORM vs SQL, so much boilerplate around C code, that assembly looks clearer to me ;)
Yes, i know, C will be partialy portable between simillar architecures/boards, but still i have no clear winner.
I was a bit more verbose in my C code than necessary here as well
long time no see !!!!
Really nice content!
Thanks for the video!
Incredible video
welcom back
Welcome back 🎉
He's back!!
Thanks for the video.
Welcome back!
Amazing video!
Hi, have you ever used the C2000/C6000 MCUs from Texas Instruments? There isn't much information about them on this site and since their architecture differs from that of ARM, I thought I'd be interesting to create a series around them. Although, I would think it'd be for a very niche audience.
I haven't, yes quite niche :)
Hi!
In 7:14, doesn't debug trace usually move forward in a program?
Is it not moving forward? You mean in the terminal below right? I may have cut the video a bit weird in certain places so they don't always line up.
@@artfulbytes No, I meant actually in the code. Maybe not in this case, but sometimes I see that the debug trace doesnt progress sequentially, especially in low level code (goes back and forth)
@@Jonathan-ru9zl Normally the program counter just increments, but there are branch/jump instructions such as bcc, which can jump to a specific address. For example, these instructions are used when repeating a code block such as in a loop, or when jumping to a function.
Welcome back
Roberto, good to see you
@27:15 i'm not sure but call to _start() should be befor call to main() ?
You would call _start instead of main(). The function _start includes a call to main.
@artfulbytes i see, now it make sense 😀
Can you do a long series like you did the past but with STM32 microcontrollers, I really wanted to write code on it from scratch, and understand all the concept MCU-related also, but I can’t find any suitable resources than yours. Thank you for bringing this such a nice educational videos!
the goat return
The value of sp increases from 0x200017f8 to 0x20001800 on running the first two assembly code lines.
If estack was 0x20001800, then why was sp initialized by HW to 0x200017f8?
The sp increases from 0x200017f8 to 0x20001800 after stepping over the first instruction. I haven't checked, but I guess it may be a debug artifact, i.e. register view not refreshing properly until stepping over the first instruction.
@artfulbytes that's possible - may be due to debug interface
Ldr command doesn't change the value of sp so it does not make sense that it changes after ldr.
Awesome!
Eu não entendi nada, mas entendi tudo.
ótimo trabalho amigo.
It would be great if we move from bare metal to any RTOS and do some stuff there as well 😊.
bare metal programming , is field iam looking to breaking into , i thought you were going to take out the bios chip set and flash it with program
Microcontrollers typically don't have any bios chipset.
@artfulbytes its kinda interesting ,maybe i should look at the architecture including flash memory and how its stores bootloader
To long since the last video Niklas! Please give us more content!
I have stm32 bule phill it comes under arm Cortex m3 architecture I but doesn't reach main it goes half fault . I don't where the mistake is where any problem in openocd
You can step from the beginning like I did in the video to identify the instruction that causes the hard fault.
kinda similar to writing an OS from scratch and writing a bootloader I passed that stage so I could understand what's going on in here, of course this is more specific to the MCU archtiecture and startup code needs to be configured according to the manufacturer and needs lots of datasheet checking about MCU specifically STM32 for this video
In the end, it's a great vid.
And I wonder your background like Electronics or CS?
Yes some similarities indeed. Thanks. I studied CS, but do quite a bit of hardware design professionally atm.
Can you make video about make files for stm32.
What Desktop Environment so you use?
ubuntu linux
@ But the default from Ubuntu? It doesn't look like that
@@m-electronics5977 i3 window manager
I wish I could've seen this two years ago 😅
I just love you!
how to set up these environment
Install ubuntu linux and vs code.
Can you make vedio makefile for this project.
I explain Makefiles in my video series. A different microcontroller/toolchain, but the Makefile structure would be the same.
just wow. very cool. 👍 will it run PHP tho 🤣
yo pls do a vid about the stm32f407 i got one as a gift and i'm stuck with it i wanna use do something and i need to work around the roll,pitch & yaw it has builtin osciloscope and idk how to work with it. thx btw for your content
Can you make a mini C course
I think there already is quite good content out there on that already.
Could this startup code can technically be called bootloader?
In practice, there may be some overlap, but roughly speaking, a bootloader is a small program that runs before the main application, while startup code is code that runs before any program, including the bootloader.
go more stm32? And if it possible, I think parallel compare with AVR8 like atmega8 most be very understandable and opens up understanding. Yes it must be hard to make video, but u can make most greate course and try sell them
We'll see
It is possible to expect vedios on weekend . We are already waited for longtime so plese make Quickly and discuss various concepts as soon as possible. thanks artful bytes❤
No specific day. Quality videos take time :)
Ommmggg you're baaack!
nice
Please make more videos
0:02 a lot of prayer
😀
The content is great. However this is not an action movie, no need for fast cuts. This is a bit heavy topic, it'd be better to slow down a bit.
good
🙏
Microntoller.
Bruh
Windows 😒
?
?
?
I appreciate the lesson as always. However this is way too dense. I, and perhaps others, may be out my depth especially as I don't know what 'gcc expects...' really entails.
This probably isn’t for people with little to no knowledge. Might need watch other videos for that
I assume some prior knowledge to be able to follow along in this video.