32-Bit Microcontroller for New Project

"Everett M. Greene" <mojaveg@mojaveg.stkn.mdsg-pacwest.com> skrev i 
meddelandet news:20080910.79D70A8.741D@mojaveg.lsan.mdsg-pacwest.com...
> "Ulf Samuelsson" <ulf@a-t-m-e-l.com> writes:
>> >> AVR32 and AVR32 Studio
>> > AVR32 Studio is Eclipse, which is full-featured but for my money very
>> > slow.
>>
>> Saw some studies on this.
>> Eclipse uses a lot of RAM. Minimum 1 GB, possibly 2 GB  is needed.
>> If you have too little RAM you would obviously be swapping = SLOW.
>
> What kind of grotesque implementation is this?  Are they
> loading everything on the hard drive into memory just to
> have some ballast?

The above statement means that Eclipse will be slow on
a machine with only 512 MB of SDRAM, but it might
be acceptable with 1 GB memory.
A lot of that GB is used by other things like Windows.



-- 
Best Regards,
Ulf Samuelsson
This is intended to be my personal opinion which may,
or may not be shared by my employer Atmel Nordic AB 

Grant Edwards wrote:


> Even Win2K wasn't too bad.  XP was definitely a step down, and
> Vista is a leap off a cliff.
> 
> 
>>The Microsoft reputation was spoiled by the unreliable
>>MS-DOS/16 bit Windows operating systems and the Windows 9x
>>versions were not much better.
> 
> 
> Exactly.  None of the 16-bit stuff past about DOS 3.x was worth
> the cost of the floppies it came on.

I am tired of hearing the baseless complaints of linux zealots about 
DOS/Windows. DOS was pretty much bare bone hardware; there is nothing 
that could be good or bad about it: everything is in the hands of the 
programmer.  The Win16 API was illogical, unforgiving and difficult to 
program; however with enough care one could develop perfectly working 
applications. Win32 API is just fine; don't know why did they add .Net 
to it. NT->Win2k->WinXP is the path of small useful improvements in the 
core functionality. Vista is an attempt to do several things in the 
straight way without been burdened with legacy compatibility.  With 
bells, whistles and stupidity turned off, Vista works just fine.


Vladimir Vassilevsky
DSP and Mixed Signal Design Consultant
http://www.abvolt.com

Vladimir Vassilevsky wrote:
> 
> 
> Grant Edwards wrote:
> 
> 
>> Even Win2K wasn't too bad.  XP was definitely a step down, and
>> Vista is a leap off a cliff.
>>
>>
>>> The Microsoft reputation was spoiled by the unreliable
>>> MS-DOS/16 bit Windows operating systems and the Windows 9x
>>> versions were not much better.
>>
>>
>> Exactly.  None of the 16-bit stuff past about DOS 3.x was worth
>> the cost of the floppies it came on.
> 
> I am tired of hearing the baseless complaints of linux zealots about 
> DOS/Windows. DOS was pretty much bare bone hardware; there is nothing 
> that could be good or bad about it: everything is in the hands of the 
> programmer.  

DOS was mostly fine within its limits.  It didn't claim to do much, and 
it didn't do much.

> The Win16 API was illogical, unforgiving and difficult to 
> program; however with enough care one could develop perfectly working 

The Win32 API is also illogical - for any given feature, there are often 
dozens of ways to do it, all with slightly different bugs, corner cases, 
or undocumented features, all of which are different depending on the 
variety of the Win32 API (such as Win9x or WinNT).  But it's not the API 
that is the root cause of windows' main problems (such as security and 
stability).  These are more down to the design philosophy of Windows 
being basically a single-user, single-task, stand-alone system (like DOS 
was), with multi-user, multi-tasking and networking being addons, and 
with a consistent emphasis on ease-of-use and performance above all 
other considerations.

I have written applications that run on Win3.1 which have run 
continuously, in use all day every day, for years without so much as a 
reboot.  You couldn't possibly do that with Win95, and you'd be hard 
pushed to do so with newer windows versions - NT 3.51 was a peak for 
windows reliability.

 > applications. Win32 API is just fine; don't know why did they add .Net

They added .Net because Sun wouldn't let them control Java.  You can 
well argue that .Net and/or C# is technically better than Java (I don't 
know enough details to argue either way) - it certainly should be, 
having come after Java.  But it's existence is purely a matter of control.

> to it. NT->Win2k->WinXP is the path of small useful improvements in the 
> core functionality. Vista is an attempt to do several things in the 
> straight way without been burdened with legacy compatibility.  With 
> bells, whistles and stupidity turned off, Vista works just fine.
> 

NT -> W2K -> XP certainly have a number of small and useful 
improvements, but each generation got more bloated and buggier (these 
are somewhat related), and not all changes are for the better.  Roughly 
speaking, they made some easy tasks easier, but some harder or rarer 
tasks harder.  On balance, however, I prefer W2K to NT4, and XP 
marginally over W2K if I'm using the machine a lot (for simple tasks or 
more dedicated machines, W2K is simpler and easier).

My experience with Vista is not extensive, but from what I've seen, I 
just don't see the point.  When you turn off the bells and whistles 
until it looks and acts much like XP, it is much the same as XP apart 
from the memory and processor requirements.

sodaant@gmail.com wrote:
> CBFalconer <cbfalco...@yahoo.com> wrote:
>> soda...@gmail.com wrote:
>>
>>> Why do you Linux fanbois always turn everything into dig against
>>> Windows?
>>
>> Because, unlike Winders, it is reliable and consistent.  Much
>> better suited for programmers.  Winders is designed for the newbie
>> idiots and/or teen-age game players.  Most Winders users don't even
>> know how to get to a command line window.
> 
... snip ...
> 
> As a longtime Windows user, I find it to be reliable and consistent.
> I haven't seen a BSOD in over ten years, and that was due to a
> poorly written 3Com NIC driver. As for consistency, I'd give the nod
> to Windows. Linus comes in so many flavors and with so many
> interfaces (KDE, Gnome, Xfce, Icebox, etc.) that one can hardly call
> it consistent. The same can be said about Windows kernel internals,
> which seem more stable than, e.g., Linux with its scheduler-of-the-
> month, etc.

I suspect that you would say the same about Linux if you simply
stuck to one distribution.  The fundamental faults are generally
concerned with applications.  Windows does not have a good means of
mounting dangerous applications, and thus cannot easily isolate
applications from the OS.  It also costs considerably more, and it
is much harder to acquire source code.  :-)

-- 
 [mail]: Chuck F (cbfalconer at maineline dot net) 
 [page]: <http://cbfalconer.home.att.net>
            Try the download section.

On Sep 10, 2:05=A0pm, David Brown
<david.br...@hesbynett.removethisbit.no> wrote:

> stability). =A0These are more down to the design philosophy of Windows
> being basically a single-user, single-task, stand-alone system (like DOS
> was), with multi-user, multi-tasking and networking being addons, and
> with a consistent emphasis on ease-of-use and performance above all
> other considerations.

That may be true for the 16-bit versions of Windows (Win 3.1, 95, 98,
etc.), but Windows NT was designed from the ground up (by David
Cutler, the designer of VAX/VMS, among other OSes) to be multi-tasking
and with a modern (TCP/IP) networking stack.

On Sep 10, 3:45=A0pm, CBFalconer <cbfalco...@yahoo.com> wrote:
> I suspect that you would say the same about Linux if you simply
> stuck to one distribution. =A0The fundamental faults are generally
> concerned with applications. =A0Windows does not have a good means of
> mounting dangerous applications, and thus cannot easily isolate
> applications from the OS. =A0It also costs considerably more, and it
> is much harder to acquire source code. =A0:-)

What do you mean Windows can't easily isolate applications from the
OS? Sure it can--it uses essentially the same hardware paging and
memory protection mechanisms that Linux does. Newer versions support
data execution protection and other means to protect against malicious
programs.

On Tue, 09 Sep 2008 22:06:20 +0200, "Boudewijn Dijkstra"
<boudewijn@indes.com> wrote:

>Op Tue, 09 Sep 2008 21:22:26 +0200 schreef <sodaant@gmail.com>:
>> Why do you Linux fanbois always turn everything into dig against
>> Windows?
>
>Q: What is the number one reason to use Linux?
>A: Because it's not Windows.

Q: What are the alternatives of Linux? Is Windows the only one?
A: There are many members in the *nix family. Windows is not a member
     (despite of what BG has told: "NT is one kind of Unix") there  so
     it cannot  be a alternative for another Unix

>> If it all came down to a steel cage match pitting Dave Cutler against
>> Torvalds, my money would be on Cutler.
>
>Some people are better at cage fighting and some people are better at  
>designing operating systems.

Finally the question is whether (the not so well known) "Bill" (Joy)
and his followers were/are better in opsys'es than Linus and his
followers... (During the 14 years I have used Linux, it hasn't yet
becomen clear how on earth Linux is different from the Unices I
had used earlier and have used lately... Ok, Linux uses GNU tools
from the beginning, commercial Unices needded one to install them)

Can one build GCC, fit apps, etc. in a system which has only 32 MB
RAM, and uses :

1. Linux
2. one of those *BSD systems (NetBSD, OpenBSD,...)

I myself have not done much Linux vs BSD comparisons, only seen
that installing "current" Linuces into a old Pentium PC with 64 MB RAM
will not succeed very well but putting the current NetBSD, OpenBSD
and maybe even that "GUI'ish" FreeBSD distro should work ok...

The glibc usually used in Linuces seem to be a horrible monster, so
those "embedded Linuces" really don't use it but the uClibc... But
finding popular Linuces for PCs which use uClibc seems to be hard.
Producing static apps with glibc seems to result app sizes which are
10 or more times bigger than the same apps on a BSD-system or on
some old SVR3/4 PC-Unix system. A static "Hello World" on Linux
may be 16 kbytes on a BSD system but 400 kbytes on Linux...
How bit the installed shared libs in Linux/glibc, Linux/uClibc and in
a NetBSD system would be, could be an interesting comparison...

Let's borrow ones opinion from 10 years back :

----------------------- clip -------------------------------------
From: Ignatios Souvatzis ignatios@theory.cs.uni-bonn.de
Newsgroups: alt.humor.best-of-usenet
Subject: [comp.unix.advocacy] Any good comparison of
Linux/FreeBSD/NetBSD/OpenBSD/Hurd?
Followup-To: alt.humor.best-of-usenet.d
Date: 20 Jan 1998 09:17:03 -0700

BSD was designed by the government to be a really good OS, because
AT&T
can't program worth a darn.

Jordan Hubbard pointed out the differences between them, but it's
often
just a matter of style, a preference for taste:

Linux is Kaustkian socialist;
Hurd is Menshevik;
FreeBSD is Trotskyist;
OpenBSD is Leninist;
NetBSD is Maoist.

Hope this helps.
----------------------- clip -------------------------------------

sodaant@gmail.com wrote:
> On Sep 10, 2:05 pm, David Brown
> <david.br...@hesbynett.removethisbit.no> wrote:
> 
>> stability).  These are more down to the design philosophy of Windows
>> being basically a single-user, single-task, stand-alone system (like DOS
>> was), with multi-user, multi-tasking and networking being addons, and
>> with a consistent emphasis on ease-of-use and performance above all
>> other considerations.
> 
> That may be true for the 16-bit versions of Windows (Win 3.1, 95, 98,
> etc.), but Windows NT was designed from the ground up (by David
> Cutler, the designer of VAX/VMS, among other OSes) to be multi-tasking
> and with a modern (TCP/IP) networking stack.

I can see you've bought into the "Dave Cutler designed VMS, Dave Cutler 
designed NT, ergo NT is a type of VMS" myth.  Dave Cutler was heavily 
involved in the design of VMS, and no doubt knows a great deal about 
good operating system design.  But he did *not* have a free reign to 
make NT as some sort of ideal operating system - he had the job of 
leading the design team for *windows* NT.  Thus he had a lot of 
influence on the design of the NT kernel - and there is no doubt that it 
is far superior to that of non-NT windows - but the system was still 
Windows, based on Windows design philosophies.

So he put a lot of effort into the kernel, and made a fairly solid 
kernel (though not quite as originally envisioned - such is life in most 
software projects).  And on top of that, Microsoft wanted people to see 
Windows.

The kernel was not bad for multi-tasking, and had some reasonable 
security to start with - and then lost a lot of its elegance and 
reliability through things like integration with the Windows gui (in 
Cutler's vision, the gui was separate and replaceable, just like in 
*nix, since he understood it to be a major weak point).

Thus you have a kernel designed for multi-tasking, but a gui and the 
rest of the system designed around the idea of a single user working on 
a single task.  And each generation of NT (up to XP at least) has taken 
it further from Cutler's ideas as MS has moved user and gui features 
from the Win9x line into the NT line.

On Thu, 11 Sep 2008 19:43:11 +0200, David Brown
<david.brown@hesbynett.removethisbit.no> wrote:

>
>The kernel was not bad for multi-tasking, and had some reasonable 
>security to start with - and then lost a lot of its elegance and 
>reliability through things like integration with the Windows gui (in 
>Cutler's vision, the gui was separate and replaceable, just like in 
>*nix, since he understood it to be a major weak point).


In 32 bit console mode applications could be written, without using
any Win16 style GUI constructs. The GUI was more or less an add on in
WinNT 3.x, even the display driver was in user mode.

In NT 4, the display driver was moved into kernel mode. In the initial
released version, you could crash the whole system in some cases by
passing a null-pointer to a kernel mode display service :-). 

Anybody writing any kernel mode routines callable from the user mode,
the first thing is to validate the parameters that they were
accessible from the user mode, before doing anything else.

Paul

"Ulf Samuelsson" <ulf@a-t-m-e-l.com> writes:
> "Everett M. Greene" <mojaveg@mojaveg.stkn.mdsg-pacwest.com> skrev
> > "Ulf Samuelsson" <ulf@a-t-m-e-l.com> writes:

> >> >> AVR32 and AVR32 Studio
> >> > AVR32 Studio is Eclipse, which is full-featured but for my money very
> >> > slow.
> >>
> >> Saw some studies on this.
> >> Eclipse uses a lot of RAM. Minimum 1 GB, possibly 2 GB  is needed.
> >> If you have too little RAM you would obviously be swapping = SLOW.
> >
> > What kind of grotesque implementation is this?  Are they
> > loading everything on the hard drive into memory just to
> > have some ballast?
> 
> The above statement means that Eclipse will be slow on
> a machine with only 512 MB of SDRAM, but it might
> be acceptable with 1 GB memory.
> A lot of that GB is used by other things like Windows.

How much is used by Eclipse?  You seemed to be saying in the
earlier posting that Eclipse itself uses 1 Gbyte or more.