Atmel's AVR Studio 5.x

[firewalker]

I downloaded today Atmel's AVR Studio 5. As it only supports Windows I installed a fresh copy in my system. I was amazed to see how "heavy" this version of AVR Studio is. The first thing I didn't like was all the .NET stuff it needs in order to run. I really couldn't pleasantly use it.

Is it my idea? Something with my system?

Your opinions about the new AVR Studio?

My system:

Code: [Select]

Windows Xp Sp3 pro.
Pentium 4 3.2 GHz (Hyperthreading) CPU.
2 Gbytes RAM.
250 Gbytes 7200 RMP sata drive.
GeForce 8400 GS GPU.

AVR Studio 5 requirements:

Code: [Select]

1.6GHz or faster processor
1 GB RAM for x86
5400 RPM hard disk drive
DirectX 9-capable video card that runs at 1024 x 768 or higher display resolution

[Bored@Work]

Your opinions about the new AVR Studio?

Downloaded it, was surprised about its size. Installed it, was surprised about installation quirks. Tried an old project, started to get annoyed. Tried to debug, debugging failed, started to get pissed. Deleted it, shook my head in disbelieve about that junk.

Later I read an article paid-for advertising in Electronic Design

http://electronicdesign.com/article/digital/Visual-Studio-based-IDE-Target-AVR-Microcontrollers.aspx

What a load of bull. But then I found an explanation how it could happen that Atmel so totally missed the boat. There is a spontaneous interview from eetimes with the AVR product marketing director about AVR Studio 5:

http://video.eetimes.com/playlist-video/latest-videos/636468326001/atmel-avr-studio-5-interview-with-haakon-skar/820373018001

Again a load of bull, but it occurred to me what went wrong with AVR Studio 5. Look at that kid. That manager is probably still wearing diapers. He'd probably never seen anything else then Word, PowerPoint and Outlook, and never worked with a tiny 8 Bit OS.

The project lacked a supervising adult.

[bilko]

Again a load of bull, but it occurred to me what went wrong with AVR Studio 5. Look at that kid. That manager is probably still wearing diapers. He'd probably never seen anything else then Word, PowerPoint and Outlook, and never worked with a tiny 8 Bit OS.

The project lacked a supervising adult.

Brilliant comment !!! and so true, had me in stitches.

[neoone]

there is version 5? wow, good thing I didn't notice that before reading your comments  I'm still pissed about switching from XP to win7 and if I added AVR Studio 5 it could be really bad heh. I see they are excited with merging 8 and 32bit version. Seriously - how many people use AVR32 ? It's almost dead platform due to ubiquitous ARM. And from I can see it no longer uses WinAVR as a C compiler which doesn't sound so good. The editor seams nicer though.

[alm]

AVR studio 5 is still in beta, I wouldn't use it for anything important (yet). It's now based on MS Visual Studio (hence the bloat). The compiler is still avr-gcc (WinAVR was just a convenient distribution of avr-gcc and related tools), but the name is different and it's now maintained by Atmel (I believe they hired/pay the original maintainer).

[neoone]

yes I know that WinAVR is based on GCC but you're saying tat they just renamed it ? Or you meant the GCC for AVR itself. WinAVR + AVR Studio is a good thing since there are always many interesting mods for some of tools included to WinAVR. But that's another topic.

[alm]

They basically renamed it (to AVR GCC toolchain I think) and bundled it with AVR Studio 5 (it used to be a separate download). You still get an IDE with a working C compiler. I'm sure there are some other changes, though: it's a newer version of gcc, for starters.

[Hypernova]

I would take the MS VS capabilities over 4.x, the original was pretty bad.

[SrS]

On AVRFreaks there’s a big topic about the new studio. Many people where complaining about VS and all the .net requirements because it will be limited to Windows only. But one of the Atmel employees said that the actual backend (including debugger) is cross platform.

Although VS 2010 is very heavy I think it’s a great IDE and you get it for free. For any real project in the near future they recommend to use v4.x anyway.

[madkiwi]

Again a load of bull, but it occurred to me what went wrong with AVR Studio 5. Look at that kid. That manager is probably still wearing diapers. He'd probably never seen anything else then Word, PowerPoint and Outlook, and never worked with a tiny 8 Bit OS.

The project lacked a supervising adult.

Brilliant comment !!! and so true, had me in stitches.

I'm really glad that I've watched the videos and read the thread - I was considering upgrading but not anymore

Time to learn the PIC and MPLAB!!

[firewalker]

So it isn't only me. First I thought "Ok it's because you don't use Windows". That .NET crap "triggered" me correctly.

The other thing I forgot to mention is the procedure in order to download it. Whats all those questions? It is just stupid. I resulted in "torrenting" the sucker.

A good text editor (vi, emacs, gedit) and gcc for me.

[apex]

I don't know what you don't like about it.
I can run it normally on every PC I'm using, like all the other Visual Studio Applications.
It is new, it has proceeded to go on in time and I like the new surface.
Debugging is much easier now, although not every processor is supported yet.
The old version didn't even have things like code highlighting beneath the c keywords or code completition.

If you want to use the 4.0 Version, you can do it!
No one stops you...

I really like the new version and already compiled two projects succesfully!

Maybe it's a question of time. I could imagine, if I had worked with the 3.0 and 4.0 Versions all day, then I wouldn't like the 5.0 Version...

apex

[Bored@Work]

I don't know what you don't like about it.

It is not about disliking it. It is about the thing not working and Atmel not being honest in my view.

I don't care much about the Visual Studio features. Frankly said, if you need stuff like IntelliSense, syntax highlighting and whatnot, to not lose track of what is going in your source code for a little microcontroller, then, well, you should probably consider another job. Visual Studio is just useless ballast, cracking a nut with a sledgehammer, for lack of (that manager kid?) having a clue. If you can't think, then VS won't think for you, and thinking is the most important task when coding.

What I do care about are the basics. Like, for example, the thing installing hassle-free. Or the thing starting up fast, and then, believe it or not, the thing working. Call it as much a Beta as you like. It isn't complete, and as such doesn't deserve Beta status. And Atmel conveniently likes to forget to mention that it is a Beta in interviews and the like.

I also care about a company being honest. The articlepress release I mentioned is not what I call honest.

Regarding AVR Toolchain and WinAvr, they are not the same, and it is not just a renaming. WinAvr was indeed build by an Atmel employ, but on his own time, without support from Atmel. But Atmel didn't take WinAvr for AVR Studio 5. Someone (that manager kid?) decided they need to do their own toolchain, and got a team within Atmel to start all over again, producing the AVR Toolchain. A bad case of not-invented-here syndrome.

Coming back to the I care about a company being honest thing,  Atmel added some proprietary extensions to the GCC in the AVR Toolchain. Last time I checked, people were still busy convincing Atmel that the GPL applies to them, too, and that they are required to release the source code. I don't call this Atmel behavior honest.

[firewalker]

I don't know what you don't like about it.

It is not about disliking it. It is about the thing not working and Atmel not being honest in my view.

I don't care much about the Visual Studio features. Frankly said, if you need stuff like IntelliSense, syntax highlighting and whatnot, to not lose track of what is going in your source code for a little microcontroller, then, well, you should probably consider another job. Visual Studio is just useless ballast, cracking a nut with a sledgehammer, for lack of (that manager kid?) having a clue. If you can't think, then VS won't think for you, and thinking is the most important task when coding.

What I do care about are the basics. Like, for example, the thing installing hassle-free. Or the thing starting up fast, and then, believe it or not, the thing working. Call it as much a Beta as you like. It isn't complete, and as such doesn't deserve Beta status. And Atmel conveniently likes to forget to mention that it is a Beta in interviews and the like.

I also care about a company being honest. The articlepress release I mentioned is not what I call honest.

Regarding AVR Toolchain and WinAvr, they are not the same, and it is not just a renaming. WinAvr was indeed build by an Atmel employ, but on his own time, without support from Atmel. But Atmel didn't take WinAvr for AVR Studio 5. Someone (that manager kid?) decided they need to do their own toolchain, and got a team within Atmel to start all over again, producing the AVR Toolchain. A bad case of not-invented-here syndrome.

Coming back to the I care about a company being honest thing,  Atmel added some proprietary extensions to the GCC in the AVR Toolchain. Last time I checked, people were still busy convincing Atmel that the GPL applies to them, too, and that they are required to release the source code. I don't call this Atmel behavior honest.

I couldn;t agree more.


A couple things more for Atmel about AVR Studio.

1. Cross platform
2. Cross platform
3. Cross platform

[apex]

Well, syntax highlighting and things like this support your process of thinking. If you have coloured an area, you can concentrate on it. Same point with commenting out whole segments by a simple click. It helps you doing more in shorter time.
It's is not the case that I can't program without it, but it makes it a simpler process for me so I can concentrate on the things that really matter, like the logic...

I like this software, as I like C# and other Visual Studio Applications, but working against legal policies is evil.
I understand what you're saying.

apex

[Bored@Work]

A couple things more for Atmel about AVR Studio.

1. Cross platform
2. Cross platform
3. Cross platform

From what I hear, many are upset because of the lack of cross-platform support. For some reason people expected more. But when did Atmel provide AVR cross-platform support in the past? Everything that is there on Linux, BSD and Mac was done by the community and systematically ignored by Atmel.

I would have been surprised if Atmel finally saw the light and started to support Linux, BSD or Mac. I would have some use for this, too. But, come on, they can't even get it right on Windows. Can you guess how they would screw up on Linux, BSD or Mac? I don't think it would be a pretty sight.

[alm]

As far as I know, AVR32 Studio runs on Linux and was based on Eclipse, but I never used AVR32 (I prefer ARM), so I don't know how well this worked / how bad they screwed up. This may be because the (now mature) AP7000 AVR32 series runs Linux, so it makes sense to use Linux as host platform. People probably assumed AVR Studio 5 would be based on AVR32 Studio, although part of it was likely wishful thinking.

[shadewind]

I don't care much about the Visual Studio features. Frankly said, if you need stuff like IntelliSense, syntax highlighting and whatnot, to not lose track of what is going in your source code for a little microcontroller, then, well, you should probably consider another job. Visual Studio is just useless ballast, cracking a nut with a sledgehammer, for lack of (that manager kid?) having a clue. If you can't think, then VS won't think for you, and thinking is the most important task when coding.

While I might agree when it comes to micro controllers where you usually have very few library functions to remember, code completion is very useful if your're developing applications which use larger libraries. If you're you've been doing lots of projects over the years and used lots of libraries, it's impossible to remember every signature for every function that you need. So you either look at the reference documentation or use code completion (a more general term for IntelliSense-like functionality). Guess which is the faster alternative?

I would have been surprised if Atmel finally saw the light and started to support Linux, BSD or Mac. I would have some use for this, too. But, come on, they can't even get it right on Windows. Can you guess how they would screw up on Linux, BSD or Mac? I don't think it would be a pretty sight.

As evidenced by the avr-libc commit messages, at least one Atmel employee is working actively on the toolchain most Linux and Mac people use:
http://lists.nongnu.org/archive/html/avr-libc-commit/2011-03/index.html

[Bored@Work]

Guess which is the faster alternative?

The thing is, it does not matter. The majority of time spent in a project is not about cranking out code, and typing feverish on the keyboard to produce as much LOC / hour. Even not in a software-only project. LOC / hour is not a productivity metric and especially not a quality metric.

As evidenced by the avr-libc commit messages, at least one Atmel employee is working actively on the toolchain most Linux and Mac people use

Yes, on his own time, not in the name of Atmel. This is the guy who did WinAvr, too. You know, the toolchain that got snubbed by Atmel, because they wanted to do their own AVR Toolchain, to have control over it.

[shadewind]

The thing is, it does not matter. The majority of time spent in a project is not about cranking out code, and typing feverish on the keyboard to produce as much LOC / hour. Even not in a software-only project. LOC / hour is not a productivity metric and especially not a quality metric.

Of course it matters. It might not matter a lot but it matters a little. That's why we're using editors that keep indentation across lines, for example. It might not matter a lot but it's a little thing that helps. Things like code completion and refactoring tools help make some tasks less boring which might not matter in the bigger picture but helps you concentrate on what really matters.

On the other hand IntelliSense is a really bad example when it comes to code completion simply because it isn't very good and often simply refuses to work properly. I much prefer both Eclipse and Xcode since they seem to have a much more accurate (or at least more useful) syntax tree representation.

[Bored@Work]

Of course it matters.

Before I continue, let me say that I consider LOCs in all its variations the worst metric for measuring performance, but since this discussion is about increasing the performance of cranking out code by using an IDE it is the metric that applies.

Now, look up the typical programmer performance figures as published since the '80 by various sources. Well, in fact, the first ones are from the '60. A typical figure is 1000 - 2000 LOC / programmer / year. Yes, per year! A pitty 2000 lines of code per year.

You'll see it does not matter a bit. I am not talking about getting an LED blink on an Arduino, but stuff on which your live and property depends. I.e. the stuff typically done by a microcontroller without anyone noticing its presence.

I know, this is the moment where the typical armchair hero code monkey comes forward and claims he can do that in half a day. Oh yes, can you do it with documentation, testing, and test documentation, code reviews, type approvals, endless meetings, the heaps of paperwork, parameter tuning, prove of stability, and all the overhead that is part of producing commercial, reliable embedded software for serious shit? You can't.

I have found that 1000 - 2000 figure to be true all over time, in multiple companies I worked for. And if you work in a typically disorganized company you can easily divide the LOC figures by ten.

A chisels and a stone plate is enough to get 2000 lines of code / year typed in. Properly indented, of course :-) This an IDE is a productivity enhancer is a myth. It it is part of the portray of the lonely programmer and his IDE sitting in his mom's basement and cranking out code to become filthy rich. The real world is different.