We are running a survey to help us improve the experience for all of our members. If you see the survey appear, please take the time to tell us about your experience if you can.
Hello!
I've been using Keil IDE for 8051 since years and it always worked fine. Now I need to work with ARM processor. Since I'm new to it, I wanted to start with the examples. I got the MCB2300 board v4.0 with uLink 2 and, of course, the appropriate Keil ARM IDE with RV.
1. In all the documentation it tells about the JTAG jumper to be set for uLink - but there is no JTAG jumper
2. When I open the example project "blinky", the target is set to LPC2378, though the MCB2300 has LPC2368. The abstract.txt says, there is a target called "mcb2300 flash", but it's actually named "lpc2378 flash".
3. In the target settings, it already selected ULINK debugger, which is set to 1MHz speed, though the online step-by-step guide says it should be set to 200kHz
4. Compiling the project works fine, but going to debug reports error "memory mismatch, address: 0x0, value: 0x0, expected: 0x18" - this looks like there was no code. Is it required to create a hex file and if so, why is this not activated in the target options? I thought these are ready-to-run projects.
5. After activating "Create hex file", it still shows the error, but enters debug mode and the disassembly windows shows more than zeros
6. The code is running instantly, at least the JTAG interface says so. But there is no LED bar or character display on the LCD. The little beeper produces low, high frequency noise. Looking into the code (as far as I understand the ARM assembler) tells me, it is not correct. Resetting the target to 0 and going stepwise lets the debugger jump to address 0xfffffff3 after a few steps. So the debugger does not really work. If I let the the code run after reset and stop it, it halts at 0x0000000C, going in a loop to 0x00000014. Not really a working code...
7. Opening example projects with abstract.txt in it gives error "abstract.txt contains a invalid path". Huh? Since when can a text file contain a path?
8. At least, example project EasyWeb compiles and uploads to the MCU. But then it says "Can't stop the ARM device - check JTAG cable". Hilarious! There is a hardware debug tool with a reset line and it can't stop the controller! Even I manually reset the whole board by unplugging the USB cable and inserting it again, the flash uploader gives the same error. Sure, it is possible that the cable is broken, but board and uLink are quite new and the cable has only beed unplugged a several times.
Now I wonder how this can be possible. I mean, it can't be the way to read hundreds of pages and books just to be able to setup Keil! Why must every tiny piece of software have thousands of settings? An ARM controller might be more complex than a 8051, but from the IDE it is still the same. Setup, project, compilation, debug.
This is a hardware/software package that costs some hundred or even some thousand Euros. I simply expect it to work! I mean, what else can occur if not even small example projects work?
(By the way, selecting a different country still shows the american flag on the preview)
Geez, this is not about Keil. I wouldn't complain if this were all free software and we only had to buy the hardware. But this is not free software. It has high prices and should the hell do its job. I can read x times through the manuals without any sense, if the damn tool chain does not do as expected. This is annoying. We don't have time to waste with problems like these. Time is money.
Do you think I enjoy seeing some tasks of the latest RTX locking up after 2-3 days of uninterrupted work? this is what you get with complex software (sometimes), and the developers at Keil are as susceptible as you and me to mistakes. be patient...
I did not have these problems when starting with Keil/ARM. The only problem I had, was that I had to wait a number of months to get simulation support for the LPC23xx chips. And after that, I had to wait further for a fix for correctly reading back fast-GPIO state using the JTAG interface.
I can't say I think the tools are non-intuitive. I got most if it working without the manuals. I often like to start out directly, and as step 2 - when I have some general knowledge about a product - pick up the manuals to see if there are features or limitations mentioned, that are not obviously observed by just using the tools.
Reading the manuals first often just requires a second pass through them because I'm not able to correlate some descriptions with the actual use until I have tried the product or already read the manuals once before.
A big problem when writing a manual is to write it in an order so the user has picked up the required background knowledge from earlier chapters. An example is a part of the manual that talked a lot about the use of a wizard. It took some time to realize that I could switch view for the startup file - either seeing the assembler source, or an input form. I'm so used to wizards being invoked as toolbar buttons or menu choices that I didn't think about looking for alternating views for open windows.
Maybe take a couple of seconds and look at the "The Insider's Guide To The NXP LPC2300/2400 Based Microcontrollers" from HITEX it helped me a lot when starting with ARM development.
But other than that getting from your description of how the board runs when uploaded with compiled application, I would guess that the target uC is not correctly selected/setup in startup file. LPC2378 instead of LPC2368 - where LPC2368 doesn't have external memory controller and if selected in startup file the board will not start properly.
If it was only that... I don't know if Keil has released different versions of the MCB2300 with different LPC models, but ours has the 2368. Also, the example projects are more or less simple demonstration software that doesn't use external memory management etc. But, like I wrote above, I took a project from uTasker that is available as hex file for LPC23XX, ergo not specifically for the 2368, and this runs fine. Taking the code project as it is, opening it in Keil, setting the processor type in config.h and compiling it - does not run correctly (the tasks don't seem to be initialised). Debugging the code annoys me with error requesters like I could only set two breakpoints when debugging in Flash and unlimited when debugging in RAM. So I deactivated "Dowload to flash", assuming this would result in RAM debugging, but after two breakpoints I couldn't set another one. Is this because of the debugger or is JTAG not capable of doing more? I often cursed the Hitex DProbe we use for the 8051, because the Hitop software is also a PITA, but compared to this it is heaven. The Keil debugger is not able to perform true step-by-step code processing.
Geez, this is not about Keil.
You don't say. Seriously: you don't say anything to that effect.
So far everything you've written in this thread, from the Summary entry to the text itself, is quite clearly about Keil.
tool chain does not do as expected.
Expectations are a business involving two parties. One of them usually is a human being, i.e. you. It doesn't have to be the other end that's at fault if they fail to come true. Did it occur to you that maybe your expectations might be based on bad assumptions?
In other, words: you need to cool down, man. Get a hug from someone, grab a beer, or whatever usually helps.
Debugging the code annoys me with error requesters like I could only set two breakpoints when debugging in Flash and unlimited when debugging in RAM. So I deactivated "Dowload to flash", assuming this would result in RAM debugging, but after two breakpoints I couldn't set another one. Is this because of the debugger or is JTAG not capable of doing more? I often cursed the Hitex DProbe we use for the 8051, because the Hitop software is also a PITA, but compared to this it is heaven. The Keil debugger is not able to perform true step-by-step code processing.
Err, what?!?!?!?! how is this related to Keil , exactly?!?! Hardware breakpoints are implemented using an EmbeddedICE logic point to detect an instruction fetch from the appropriate address. This works in all cases, even if the program being debugged modifies itself as it executes, or if the code is in ROM. However, it completely ties up one of the two available EmbeddedICE logic point units.
Gute Idee. Leider steht dem der Zeitdruck, Termine und Auftraege entgegen. Und das in einer Zeit, in der die Wirtschaftslage alles andere als rosig ist. Da kann man sich solche Spielchen einfach nicht leisten. Man steigt auch nicht mal eben so auf ein anderes Entwicklungssystem um, zumal man nicht sicher sein kann, dass dieses besser ist. Falsche Vorstellungen? Vielleicht. Wie gesagt, bis dahin, was wir uns vom ARM vorstellen sind wir ja noch gar nicht gekommen. Es hapert ja schon bei dem, was eigentlich laufen muesste.
I realize that you are under time pressure, but that is a problem on your side, not relevant to the Keil tools.
Yes, Keil has multiple versions of the development board. You could have it with 100-pin or 144-pin 23xx chips - if your board has a 100-pin chip, you will notice the second set of pads on the outside of the chip.
What happens if you do not try to build any code, but just download the precompiled examples?
About hardware breakpoints - when running in flash, you are limited by the support in the ARM chip. The JTAG interface is just an interface - the big part of the debugging is inside the ARM chip. This is a difference from a real emulator.
When debugging in RAM, then the JTAG interface can be used to replace instructions, allowing software-generated breakpoints.
The above isn't unique for Keil development with the LPC23xx chips. If you switched to Atmel AVR chips, you would find similar behaviour, but with the optional extra that the Atmel AVR tools would allow programming of software breakpoints in the flash too. Helps getting an extra breakpoint, but also helps greatly if the goal is to wear down the flash sectors. In the case of the Atmel AVR chips, each flash sector is tiny. The LPC23xx flash sectors are big. Big means big wear, since there will be way less wear leveling.
Not downloading to flash will just stop the download cycle - the debugger will assume that you already have the firmware downloaded. Either since a previous run, or that you have used FlashMagic or similar to program the chip.
The demo applications has - as already mentioned - more than one target defined. When you switch the build target to RAM, you will get the project settings optimized for RAM debugging. The memory regions specified in the project will not mention the flash - instead some of the RAM will be specified for use for code. The RAM target will also specify a debugger script, that will set the PC to the start of the RAM copy, instead of trying to start program execution from the reset vector. This step is needed, since the RAM target will not download any code into flash - the reset vector will have undefined contents.
In this case, you have lost a lot of time because you wanted to save time. Instead of doing things one step at a time, you went five steps ahead. Then you got lost, and angry about the tools. Having already taken 5 steps, you are now in a no-mans-land, where you do not know which of the five steps that where wrong.
It is always advantageous to take a bit of time to familiarize yourself with new tools before expecting to be productive with them. The play time will pay off in the end. It's the early shortcuts that will cost the most money.
Thanks for telling me. Now I see that the controller is the reason. How could we imagine to have a fully functional development system which saves time and effort with just switching to a controller with a JTAG interface? You read everywhere that this is called "integrated debugging". This is not debugging, this is a joke. Anyway, the Keil debugger is a part of the IDE, it controls the interface (uLink). It should at least be able to perform step-by-step code execution, no matter what controller I have. If required, it should be supported by hardware to circumvent the limitations of the controller. In german, the word "ARM" means "poor" and that's what this chip is...
The demo applications has - as already mentioned - more than one target defined. When you switch the build target to RAM, you will get the project settings optimized for RAM debugging. I don't know what demo apps you are referring to, but most of the demo apps have only one target (either 'MCB2300' or 'MCB2300 board'), while "blinky" has a "LPC2378 Flash" target. This is the mess I was talking about. These are all demo apps in the /boards/keil/mcb2300 folder and the all have different target names. If I open the "blinky" demo from that folder with "LPC2378 Flash" target and change the controller type to LPC2368, compile it and want upload it, the flash uploader tells me "Could not load flash programming algorithms" (LPC_2000_IAP2_512.flx). Now you tell me this is all my fault?? The target dialogue has no option for which memory segment the code is compiled. The only difference is in the debugger dialogue, where you can select "Download to flash". But even if it was as you said and the demo apps would have two targets, then both targets should work and the compiler should do accordingly.
The different boards have their examples in different directories.
But when you open a Keil example for a specific board, then the project file contains one build target for running in flash, and one build target for debugging directly in RAM.
There is most definitely input boxes to specify the memory range to use for code and the memory range to use for RAM. Each target in the project file has its own set of all settings.
No, Per, they don't do! There is mostly only one target and in some cases also the simulator as second target. But they are not distinguished by memory type. Also, a target has settings. In the target settings dialogue I can not selected whether the code shall be compiled for RAM or ROM. I can only set the debugger to not use flash (ROM), so it must be using RAM. The input boxes you refer to are to define and tell the compiler what memory areas the specific controller has. I don't think they are used to define the target memory type, because if both, RAM and ROM, are enabled, there must be a priority of one (in this case ROM) and so the compiler should compile the code for ROM. If "Download to flash" is activated, the code should work now. If not, then it would run from RAM and the debugger/flashloader would tell me. This would explain one of the items I told in my first post. But how will I know? These demo apps are actually supposed to work as is.
I'm running a bit older version of the tools - I don't much like upgrading when there is no need.
Seems like you are right - Keil has been a bit lazy.
For MCB2300 board: Blinky,RTX_Blinky,RTX_Traffic seems to have Simulator + Flash
For MCB2100 board: Bliny,BlinkyCAN,BlinkyIRQ has Simulator + Flash + RAM RTX_Blinky: Simulator + Flash
For MCB2400 board: Blinky has Simulator + Flash + Ext SDRAM + Ext NOR Flash
Seems like Keil has been slowing down a bit. I can't see any real reason for the big lack of RAM targets for newer boards. It isn't completely obvious how to set up one, without having an example to look at.
=>If I open the "blinky" demo from that folder with "LPC2378 Flash" target and change the controller type to LPC2368, compile it and want upload it, the flash uploader tells me "Could not load flash programming algorithms" (LPC_2000_IAP2_512.flx). Now you tell me this is all my fault??<=
(I don't know much about all these.)
Hi Maik,
There is one setting to tell KEIL compiler what the MCU you are using. And there is another setting to tell the JTAG Tool what the MCU you are using.
Maybe you can check this:
Options for Target -> Debug and Utilities -> there is a button "Setting". If you haven't done this, click that button, and configure it.
Adding New Algorithms
www.keil.com/.../ulinkme_su_newalgorithms.htm
www.keil.com/.../ulink2_su_newalgorithms.htm
www.keil.com/.../ulink_su_newalgorithms.htm