Searching \ for '[OT] Byte Craft turns 30' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: massmind.org/techref/index.htm?key=byte+craft+turns
Search entire site for: 'Byte Craft turns 30'.

Exact match. Not showing close matches.
PICList Thread
'[OT] Byte Craft turns 30'
2009\05\22@091227 by Walter Banks

picon face
This is off topic. I won't do this again for another 20 years.

Thirty years today I walked out of the University of Waterloo
as a employee for the last time. These were the exponential
moments in the embedded system world. Early enough that
embedded systems and personal computers were just starting
to follow their own paths.

The story is an old one a start up company develops products
for a few years and finds that special niche that becomes
its signature. Byte Craft started developing micro-processor
based consumer, industrial and automotive products. Things
were different in those days, the first step in an embedded
systems development was to create tools to support the
processors used. We created tools and created products. We
created more tools and soon the tools were more important
to us than the products. Byte Craft became a software tool
company. We wrote a lot of assemblers for companies that
needed them.

Byte Craft developed MISTRAL in the mid 80's for the Motorola
6805 the first high level language that could compete with
well written hand coded assembler. MISTRAL evolved into C6805.
Proof that high level languages were effective on very small
processors, the first practical use of expert system technology
for optimization. This product set the standard for tight code
generation and innovative solutions found in our compiler
products.

Byte Craft has developed a lot of compiler innovation. We
developed industrial strength code generation, HLL support
for homogeneous and now heterogeneous multiprocessor
environments. The compiler innovation continues...

Less visible has been the work we have done on instruction
set design. We have been part of the design team on many of
the instructions sets that we support. Most recently working
instructions sets that are designed specifically for machine
generated code. (eTPU and eTPU2)

The Microchip MPC compiler was the fourth C compiler for
embedded systems that we wrote. It continues to be a
Byte Craft supported product with the recent implementation
of support for enhanced midrange instruction set to be
formally released later this year.

The business world is a full contact sport. It is important
to remember to give something back. Byte Craft created a
pleasant workplace for our employees. As we evolved into
a software tools company we put some our more interesting
technology into the public domain using the PIC list and
other methods to insure that some ideas would never
be owned. The various charge transfer sensor technologies
(touch switches and port based A/D converters) and
information based barcode technology are examples of this.

There are other ways to give back. Byte Craft represents
Canada at ISO SC22 on Working Group 14 (C standards).
Standards groups are a serious unpaid commitment of time
and resources. Here we co-authored IEC/ISO 18037
C standards for embedded systems that has brought the
unique requirements for embedded systems processors to
the language that is used to describe application solutions
to the tools we use. This work continues with C1X the
current work in progress.

Byte Craft has supplied thousands of copies of our tools
to individual students for their final year and masters
projects. Some of these projects are showcased on our
website.

There are a lot of outstanding people working in the embedded
systems world. I see creativity every day in my conversations
and emails with customers and friends.

For me I get up in the morning and look forward to the
day. I feel the thrill when project comes comes together
and the pride of looking at products knowing there is a
little bit of me in there.


Walter Banks
Byte Craft Limited
1 (519) 888-6911
http://www.bytecraft.com
spam_OUTwalterTakeThisOuTspambytecraft.com













2009\05\22@100411 by Bob Ammerman

picon face
Walter,

Congratulations!

My company is only 26 years old, I guess I'm just a young punk.

-- Bob Ammerman
RAm Systems

2009\05\22@104233 by Funny NYPD

picon face
Congratulations.
It seems you concentrated more on the Motorola (Freescale) MCU rather than PIC, right?

The PIC18 C compiler has been dominated by HI-tech PICC18 and Microchip C18 (now both owned by Microchip). The GNU based Microchip C30 took a lot place on relatively new dsPIC30/33, PIC24 too.

Now there are some opportunity on the PIC32 (everybody is new to this chip). Do you have any plan to expand your PIC C compiler?

Funny N.
Au Group Electronics, http://www.AuElectronics.com




________________________________
From: Walter Banks <.....walterKILLspamspam@spam@bytecraft.com>
To: piclistspamKILLspammit.edu
Sent: Friday, May 22, 2009 9:09:26 AM
Subject: [OT] Byte Craft turns 30

This is off topic. I won't do this again for another 20 years.

Thirty years today I walked out of the University of Waterloo
as a employee for the last time. These were the exponential
moments in the embedded system world. Early enough that
embedded systems and personal computers were just starting
to follow their own paths.

The story is an old one a start up company develops products
for a few years and finds that special niche that becomes
its signature. Byte Craft started developing micro-processor
based consumer, industrial and automotive products. Things
were different in those days, the first step in an embedded
systems development was to create tools to support the
processors used. We created tools and created products. We
created more tools and soon the tools were more important
to us than the products. Byte Craft became a software tool
company. We wrote a lot of assemblers for companies that
needed them.

Byte Craft developed MISTRAL in the mid 80's for the Motorola
6805 the first high level language that could compete with
well written hand coded assembler. MISTRAL evolved into C6805.
Proof that high level languages were effective on very small
processors, the first practical use of expert system technology
for optimization. This product set the standard for tight code
generation and innovative solutions found in our compiler
products.

Byte Craft has developed a lot of compiler innovation. We
developed industrial strength code generation, HLL support
for homogeneous and now heterogeneous multiprocessor
environments. The compiler innovation continues...

Less visible has been the work we have done on instruction
set design. We have been part of the design team on many of
the instructions sets that we support. Most recently working
instructions sets that are designed specifically for machine
generated code. (eTPU and eTPU2)

The Microchip MPC compiler was the fourth C compiler for
embedded systems that we wrote. It continues to be a
Byte Craft supported product with the recent implementation
of support for enhanced midrange instruction set to be
formally released later this year.

The business world is a full contact sport. It is important
to remember to give something back. Byte Craft created a
pleasant workplace for our employees. As we evolved into
a software tools company we put some our more interesting
technology into the public domain using the PIC list and
other methods to insure that some ideas would never
be owned. The various charge transfer sensor technologies
(touch switches and port based A/D converters) and
information based barcode technology are examples of this.

There are other ways to give back. Byte Craft represents
Canada at ISO SC22 on Working Group 14 (C standards).
Standards groups are a serious unpaid commitment of time
and resources. Here we co-authored IEC/ISO 18037
C standards for embedded systems that has brought the
unique requirements for embedded systems processors to
the language that is used to describe application solutions
to the tools we use. This work continues with C1X the
current work in progress.

Byte Craft has supplied thousands of copies of our tools
to individual students for their final year and masters
projects. Some of these projects are showcased on our
website.

There are a lot of outstanding people working in the embedded
systems world. I see creativity every day in my conversations
and emails with customers and friends.

For me I get up in the morning and look forward to the
day. I feel the thrill when project comes comes together
and the pride of looking at products knowing there is a
little bit of me in there.


Walter Banks
Byte Craft Limited
1 (519) 888-6911
http://www.bytecraft.com
.....walterKILLspamspam.....bytecraft.com













2009\05\22@111558 by Tamas Rudnai

face picon face
Great achievements, Walter, and Happy Birthday! :-)

Now that Byte Craft turns into a mature citizen is there any plan for a
marriage? :-) I mean nowadays we can see so many companies are acquired by
or acquire another company -- is there any plan on this?

Tamas


On Fri, May 22, 2009 at 2:09 PM, Walter Banks <EraseMEwalterspam_OUTspamTakeThisOuTbytecraft.com> wrote:

{Quote hidden}

> -

2009\05\22@111658 by Walter Banks

picon face
> It seems you concentrated more on the Motorola (Freescale) MCU rather than PIC, right?

Not really, we are mostly concentrated on embedded systems support in general.
As I noted we are still expanding inthe MPC C compiler family with new releases
schedualed for the comimg  month.

It have been quite a ride

Walter..



Funny NYPD wrote:

{Quote hidden}

> -

2009\05\22@113308 by Xiaofan Chen

face picon face
On Fri, May 22, 2009 at 10:42 PM, Funny NYPD <spamBeGonefunnynypdspamBeGonespamyahoo.com> wrote:
> Now there are some opportunity on the PIC32
> (everybody is new to this chip). Do you have any plan to
> expand your PIC C compiler?
>

PIC32 market is right now very small and C32 compiler itself
is actually free (gcc). It is just the library is not free. So I think
the market for another C compiler (other than based on GCC)
is really limited. The only two of the available PIC32 compiler
which are not based on GCC are from Green Hill and HiTech.
http://www.microchip.com/stellent/idcplg?IdcService=SS_GET_PAGE&nodeId=2602

Microchip has even posted the build script for C32 compiler
in Sourceforge.
http://sourceforge.net/projects/microchipopen/

And once OpenOCD for PIC32 is matured, you can use
cheaper Wiggler or USB JTAG tools for debugging.
http://www.microchip.com/forums/tm.aspx?m=346142

It is said you can use the free MIPS C/C++ compiler from
Code Sourcery (with newlib) for PIC32. You may just
need the header files from C32. The library will not be
as good as the highly optimized C32 library. Take note
Microchip does not distribute C32 library source codes,
even if you pay for the C32 compiler.

So I guess Byte Craft will have no interests in PIC32.

--
Xiaofan http://mcuee.blogspot.com

2009\05\22@113823 by Harold Hallikainen

face
flavicon
face
Walter,

THANKS for posting the brief history of Byte Craft! It's fun looking back.
I see that one of your early compilers was for the Motorola 6805. The 6805
was the first microcontroller I worked with. Prior to that, I had used the
MC6802, all in assembly, either using the assembler on a timeshare service
called The Source, or, later, using an assembler on my Cromemco CP/M
system. I also did 6805 code in assembly using the Cromemco. Fun stuff!
Debugging was largely staring at code. I also used the "crash and burn"
debugging technique, where my program would crash, I'd try to fix it, and
burn another eprom. For the 6802, I had written a monitor program that let
me look at memory, start execution at some point, and set breakpoints
(using the 6800 SWI instruction (0x3f)). When we hit an SWI instruction,
the monitor would dump the registers to my terminal (a Lear Siegler ADM-1)
and wait for further instructions.

Stuff is a little different today writing in C and using the Real ICE.

Harold

--
FCC Rules Updated Daily at http://www.hallikainen.com - Advertising
opportunities available!

2009\05\22@120126 by Funny NYPD

picon face
Cool. Good to know.

Funny N.
Au Group Electronics, http://www.AuElectronics.com




________________________________
From: Walter Banks <TakeThisOuTwalterEraseMEspamspam_OUTbytecraft.com>
To: Microcontroller discussion list - Public. <RemoveMEpiclistspamTakeThisOuTmit.edu>
Sent: Friday, May 22, 2009 11:13:56 AM
Subject: Re: [OT] Byte Craft turns 30

> It seems you concentrated more on the Motorola (Freescale) MCU rather than PIC, right?

Not really, we are mostly concentrated on embedded systems support in general.
As I noted we are still expanding inthe MPC C compiler family with new releases
schedualed for the comimg  month.

It have been quite a ride

Walter..



Funny NYPD wrote:

{Quote hidden}

> -

2009\05\22@144137 by Walter Banks

picon face
Tamas,

Byte Craft is a niche market company in a community of
many similar companies. We work quietly on some aspect
of the business that we have special skills in.  Locally we
meet at house parties still and have watched our families
grow up and leave home.

At mid day some of the local friends and family dropped
into Byte Craft and swapped war stories of projects past.

We have seen a lot of technology develop in the 30
years and been part of some of it.

On your real question, we don't have any plans to be
acquired or aquire other companies. It is these changing
times when many new opportunities will surface.

Regards

Walter Banks





Tamas Rudnai wrote:

{Quote hidden}

2009\05\22@145016 by Walter Banks

picon face



> It is said you can use the free MIPS C/C++ compiler from
> Code Sourcery (with newlib) for PIC32. You may just
> need the header files from C32. The library will not be
> as good as the highly optimized C32 library. Take note
> Microchip does not distribute C32 library source codes,
> even if you pay for the C32 compiler.
>
> So I guess Byte Craft will have no interests in PIC32.
>

It has not been hard to compete with free software. The
performance increase and support that is provided with
our commercial tools is worth the price. The economics
per seat of our tools is about the loaded cost of two man
days. The question becomes is the benefit that commercial
tools provides worth the cost compared to the cost of
maintaining and supporting free tools.

Thirty years later our customers think it is.

Regards,

--
Walter Banks
Byte Craft Limited
1 (519) 888-6911
http://www.bytecraft.com






2009\05\22@145648 by Walter Banks

picon face


Harold Hallikainen wrote:

{Quote hidden}

Lunch time discussion today was all the bad things that would happen
when a 2708 was plugged in backwards. We dug out the parts from
the ADAM project.  Including Serial number 2. A data logger
system with a cmos automotive 6805 whose code was written
in Mistral optimized for power consumption.

Remembered Gerry Wheeler the originator of the HCF (Halt and Catch
Fire) moniker for the 6800 op code DD and D9.

Lots of old memories.

Walter..


2009\05\22@195423 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 2:47 AM, Walter Banks <EraseMEwalterspamspamspamBeGonebytecraft.com> wrote:
{Quote hidden}

I am not so sure about the performance part in this paticular
case. Without significant efforts, it is hard to beat C32 (based
on GCC, optimized by MIPS and Microchip themself).

On the other hand, you can eary money to sell compiler based
on GCC (or C32 compiler in this case). Rowley Associate sells their
ARM C compiler. They provide their own C libraris, an IDE,
and support. Many companies sell GCC based suites and the
main thing is support. Yes I agree the support is an important
part for commercial projects.

--
Xiaofan http://mcuee.blogspot.com

2009\05\22@200719 by William \Chops\ Westfield

face picon face

> it is hard to beat C32 (based on GCC, optimized by MIPS and  
> Microchip themself).

A lot of very large ecompiler vendors think that there is a market for  
gcc competitors.  Intel, Greenspring, IAR...

It's not hard to make a compiler better than gcc for a particular cpu,  
just by rejecting some of the decisions that the gcc people make about  
cpu-independence and generality.

And it's absolutely trivial to provide better "support" than there is  
for gcc. (Shucks, there are whole companies that seem to profit merely  
from providing support for gcc.)  With gcc the compiler changes in  
gratuitously incompatible ways at the whim of some developer(s).  A  
for-profit compiler vendor is much more subject to feedback from  
actual paying customers!

BillW

2009\05\22@204451 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 8:07 AM, William "Chops" Westfield
<RemoveMEwestfwKILLspamspammac.com> wrote:
>
>> it is hard to beat C32 (based on GCC, optimized by MIPS and
>> Microchip themself).
>
> A lot of very large ecompiler vendors think that there is a market for
> gcc competitors.  Intel, Greenspring, IAR...

Not for smaller companies like Byte Craft in this case.

> It's not hard to make a compiler better than gcc for a particular cpu,
> just by rejecting some of the decisions that the gcc people make about
> cpu-independence and generality.

MIPS and Microchip have probably done that. C32 is based on GCC
but with a lot of patches.

> And it's absolutely trivial to provide better "support" than there is
> for gcc. (Shucks, there are whole companies that seem to profit merely
> from providing support for gcc.)  With gcc the compiler changes in
> gratuitously incompatible ways at the whim of some developer(s).  A
> for-profit compiler vendor is much more subject to feedback from
> actual paying customers!
>
That is true.




--
Xiaofan http://mcuee.blogspot.com

2009\05\22@204906 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 8:44 AM, Xiaofan Chen <xiaofancSTOPspamspamspam_OUTgmail.com> wrote:
> On Sat, May 23, 2009 at 8:07 AM, William "Chops" Westfield
>> It's not hard to make a compiler better than gcc for a particular cpu,
>> just by rejecting some of the decisions that the gcc people make about
>> cpu-independence and generality.
>
> MIPS and Microchip have probably done that. C32 is based on GCC
> but with a lot of patches.

By the way, the library is another big part. MIPS holds the IP so
dear that Microchip does not even ship the C library source codes
for C32.

So in the end, I guess it is very difficult to beat C32 in the performance
department. Not so sure about Green Hills. Maybe they can beat C32.

IAR probably accepts the defeat in PIC24/dsPIC market and think PIC32
market is too small to justify the efforts, so they are not coming out any
C compiler for PIC32.

--
Xiaofan http://mcuee.blogspot.com

2009\05\22@214612 by Walter Banks

picon face


Xiaofan Chen wrote:

{Quote hidden}

GCC is a generic compiler based on old technology. With the
right attention to detail and organization it is possible to
significantly beat its performance. Once you are out of the
framework of GCC with the freedom to design a code
generator to target a particular instructions set it becomes
little optimization gains that add up to significant differences.

Good compilers require a huge attention to detail and
an organized test environment.

The push to port generic solutions gave the advantage of
providing a quick tool set to meet a marking check box
requirement. Look at the either the new processors from
the EU or the successful processors including Microchip
and you will find that a significant number of the demanding
applications are implemented  using tools designed to support
that specific processor.

I have a self interest in this to be sure. I have worked on
many instruction set designs. If the instruction set needs
to conform to the narrow range of instruction sets that
are even reasonably supported by GCC that is a serious
limitation. My point is there isn't a GCC implementation
for many of Microchips 14bit parts for a reason. The special
purpose automotive processors that I have worked on that
have three interwoven execution threads running on each
instruction are unlikely to be able to implemented with
generic compiler tools.

Regards,

--
Walter Banks
Byte Craft Limited
http://www.bytecraft.com





2009\05\22@220447 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 9:47 AM, Walter Banks <spamBeGonewalterSTOPspamspamEraseMEbytecraft.com> wrote:

> I have a self interest in this to be sure. I have worked on
> many instruction set designs. If the instruction set needs
> to conform to the narrow range of instruction sets that
> are even reasonably supported by GCC that is a serious
> limitation.
>
> My point is there isn't a GCC implementation
> for many of Microchips 14bit parts for a reason.

Yes GCC needs a bit better instruction set than PIC12/16.
It might be possible to extend GCC to PIC18 but I think
there is no much gain there. But there are successful gcc
implementation for AVR and MSP430.

And you can not say AVR/MSP430 instruction sets
are not nice. They are optimized for C compiler, not
necessarily GCC. But by doing that, they give the
possibility to support GCC.

14bit PIC instruction sets are despised by many people
on the AVR and ARM side from what I read. ;-)
And there are many people who prefer to use asm
than any C compiler for these PICs.

> The special purpose automotive processors that I have worked
> on that have three interwoven execution threads running on each
> instruction are unlikely to be able to implemented with
> generic compiler tools.
>

That is probably a different thing. The specially processor
may need special compilers.

On the other hand, if we put another example of the
semiconductor industry. CMOS is said to be not suitable
for many things yet now CMOS is almost everywhere,
even in analog and RF.


--
Xiaofan http://mcuee.blogspot.com

2009\05\23@024949 by Wouter van Ooijen

face picon face
> It is said you can use the free MIPS C/C++ compiler from
> Code Sourcery (with newlib) for PIC32. You may just
> need the header files from C32. The library will not be
> as good as the highly optimized C32 library. Take note
> Microchip does not distribute C32 library source codes,
> even if you pay for the C32 compiler.

What is in these libraries that Microchip and/or MIPS hold so dearly?
Statrtup/initialization code? Math routines? something else?

--

Wouter van Ooijen

-- -------------------------------------------
Van Ooijen Technische Informatica: http://www.voti.nl
consultancy, development, PICmicro products
docent Hogeschool van Utrecht: http://www.voti.nl/hvu

2009\05\23@032012 by Walter Banks

picon face


Xiaofan Chen wrote:

{Quote hidden}

This basically makes my point. There are many effective
C compilers for small processors whose instructions sets
GCC can't handle very well. The GCC MSP430 is at best
a 30% solution (30% overhead) AVR probably more.

Put another way with a compiler change a developer might
get a 25% boost in performance and reduction in program
size. The advantages don't stop there. There is a corresponding
reduction in power requirements and emi as well.

These kinds of gains are significant to consumer products
application developer. For volume manufacturers it can mean
using a smaller or less expensive part.

Regards,

--
Walter Banks
Byte Craft Limited
http://www.bytecraft.com










2009\05\23@062948 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 2:49 PM, Wouter van Ooijen <EraseMEwouterspamEraseMEvoti.nl> wrote:
> What is in these libraries that Microchip and/or MIPS hold so dearly?
> Statrtup/initialization code? Math routines? something else?

The Statrtup/initialization code is there. Peripheral lib sources are there.
A small dsp library source is distributed but I am sure it is not the full
picture.

Other than that, everything is not distributed (the standard C library, the
math library, etc).

We asked in the forum before but got no answer. So I have to guess
it has to do with the fact you can build the full version compiler if
the library source codes are distributed. You know, the compiler
is free, the library is not.

Reference:
www.microchip.com/forums/tm.aspx?m=424127
www.microchip.com/forums/tm.aspx?m=368624
http://www.microchip.com/forums/tm.aspx?m=292995


--
Xiaofan http://mcuee.blogspot.com

2009\05\23@064805 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 6:29 PM, Xiaofan Chen <@spam@xiaofanc@spam@spamspam_OUTgmail.com> wrote:
{Quote hidden}

This the build script Microchip released.

http://microchipopen.svn.sourceforge.net/viewvc/microchipopen/ccompiler4pic32/buildscripts/trunk/build.sh?revision=32&view=markup

../src/configure  --target=pic32mx --program-prefix=mypic32-
--enable-languages=c --prefix=$WORKING_DIR/install-image
--libexecdir=$WORKING_DIR/install-image/pic32mx/bin
--disable-nls --disable-tui --disable-gdbtk --disable-shared
--enable-static --disable-threads --disable-bootstrap --with-dwarf2
--enable-multilib --enable-sim --without-headers
--with-lib-path=: --with-pkgversion="Unsupported community
build of C Compiler for PIC32 v1.05"
--with-bugurl="https://sourceforge.net/projects/microchipopen/"

If you have the library source (example: a proper patched version
of newlib, or MIPS/microchip's Proprietory C library), you
can change --without-headers to something like
--with-headers=$prefix/$target/include
and you will get a full version of C compilers with no
code size limit and full optimization.
http://www.microchip.com/forums/tm.aspx?m=292995

The case of C30 is different. In that case, Microchip
is using the standard GCC. But they add a proprietory
post optimizer (Procedure abstraction optimizer) called
pic30-pa.exe.

--
Xiaofan http://mcuee.blogspot.com

2009\05\23@070116 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 3:19 PM, Walter Banks <.....walterspam_OUTspambytecraft.com> wrote:
>> Yes GCC needs a bit better instruction set than PIC12/16.
>> It might be possible to extend GCC to PIC18 but I think
>> there is no much gain there. But there are successful gcc
>> implementation for AVR and MSP430.
>
> This basically makes my point. There are many effective
> C compilers for small processors whose instructions sets
> GCC can't handle very well. The GCC MSP430 is at best
> a 30% solution (30% overhead) AVR probably more.

I believe mspgcc and avrgcc developers do not quite agree. ;-)
It is said avrgcc is on par with even IAR AVR compiler.
Mspgcc is stuck in an rather old version of gcc (3.2.x).
But again new version gcc may not help these small
non-core-targets for gcc. But I am not using them so I have
no real first hand experiences.

> Put another way with a compiler change a developer might
> get a 25% boost in performance and reduction in program
> size. The advantages don't stop there. There is a corresponding
> reduction in power requirements and emi as well.
>
> These kinds of gains are significant to consumer products
> application developer. For volume manufacturers it can mean
> using a smaller or less expensive part.

This is only true for the low end 8-bit market. With Cortex M3,
M1 and M0 part out and low cost 16bit like PIC24, it will get
less and less critical to gain the 1% or even 10% code size.
You know, some PIC32 and PIC24 are cheaper than many
PIC18.


--
Xiaofan http://mcuee.blogspot.com

2009\05\23@092218 by Walter Banks

picon face


Xiaofan Chen wrote:

> > These kinds of gains are significant to consumer products
> > application developer. For volume manufacturers it can mean
> > using a smaller or less expensive part.
>
> This is only true for the low end 8-bit market. With Cortex M3,
> M1 and M0 part out and low cost 16bit like PIC24, it will get
> less and less critical to gain the 1% or even 10% code size.

What is the effect of 1% on a PID loop or digital filter?

As memory costs go to zero the benefits of optimization are
primarily on  execution speed.

I have actually been surprised that the Cortex M3 hasn't made
the impact that it potentially could have.

Regards,

--
Walter Banks
Byte Craft Limited
http://www.bytecraft.com



2009\05\23@093403 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 9:23 PM, Walter Banks <TakeThisOuTwalter.....spamTakeThisOuTbytecraft.com> wrote:

>> This is only true for the low end 8-bit market. With Cortex M3,
>> M1 and M0 part out and low cost 16bit like PIC24, it will get
>> less and less critical to gain the 1% or even 10% code size.
>
> What is the effect of 1% on a PID loop or digital filter?
>
> As memory costs go to zero the benefits of optimization are
> primarily on  execution speed.

Actually I was referring to code size. It is said actually
gcc is quite good in terms of execution speed.

> I have actually been surprised that the Cortex M3 hasn't made
> the impact that it potentially could have.

Because NXP/Atmel are in a bad shape financially? TI may make a
difference with the Luminary acqusition. On the other hand, I believe
ST has made significant gains with their STM32.

On the other hand, low end 4-bit and 8-bit MCUs will always
have a big market. Remember Russel is trying to find
sub US$0.30 MCUs. In those cut-throat competition
situation, every cent count and code size can be
important. And GCC is not for that market.

--
Xiaofan http://mcuee.blogspot.com

2009\05\23@100612 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 2:47 AM, Walter Banks <TakeThisOuTwalterKILLspamspamspambytecraft.com> wrote:
> It has not been hard to compete with free software. The
> performance increase and support that is provided with
> our commercial tools is worth the price. The economics
> per seat of our tools is about the loaded cost of two man
> days. The question becomes is the benefit that commercial
> tools provides worth the cost compared to the cost of
> maintaining and supporting free tools.
>
> Thirty years later our customers think it is.
>

And if I look at Byte Craft's product, they are all for small
MCUs which are not suitable for gcc.
http://www.bytecraft.com/Catalog

So I guess that it is actually hard to compete with
free or non-free C compilers after all for Byte Craft...

But you are right, there are small C compiler
vendors who are doing well for AVR market like
ImageCraft even though avr-gcc does get some
support from Atmel. And mspgcc may not be
good enough as well for many people so that
there are vendors who support MSP430 as well.

But the trend that big MCU vendors are also coming
out C compilers must be hard for 3rd parties. And sometimes
they just give these compiler for free to big customers.
Now that Microchip acquires HiTech and I believe
that this will be hard for 3rd parties as well. If HiTech
got rid of the annual fee, I think many SourceBoost,
CC5E/CC8E and and MPC customers may go
with that.

I know that competition and choices are good. But
it seems the life for 3rd party compiler vendors
are not good after all except the 8051/ARM market
where chip vendors do not offer C compilers in general.

--
Xiaofan http://mcuee.blogspot.com

2009\05\23@121715 by Xiaofan Chen

face picon face
On Sat, May 23, 2009 at 10:06 PM, Xiaofan Chen <.....xiaofancspamRemoveMEgmail.com> wrote:
> I know that competition and choices are good. But
> it seems the life for 3rd party compiler vendors
> are not good after all except the 8051/ARM market
> where chip vendors do not offer C compilers in general.
>

On the other hand, there will be good 3rd parties compiler
vendors who can produce good values for their target
customers. So in the end there will be a few dominate
players and many small vendors. After all, choices
are good. ;-)

But in the particular case of PIC32, I doubt that there
will be many players any time soon. I maintain that
PIC32 will be a niche market and Microchip can gain
some market share but not much.Therefore there will
not be many small vendors who are putting significant
resources to reinvent new compilers for PIC32. However,
the easiest way for them is to take gcc and then provide
other added values to the potential customers.

--
Xiaofan http://mcuee.blogspot.com

More... (looser matching)
- Last day of these posts
- In 2009 , 2010 only
- Today
- New search...