Hi-Tech Software bought by Microchip - no more other compilers

On Wed, 08 Jul 2009 17:48:03 +0200, "Boudewijn Dijkstra" <boudewijn@indes.com> wrote:

>Op Wed, 08 Jul 2009 17:17:59 +0200 schreef MC  
><for.address.look@www.ai.uga.edu.slash.mc>:
>> The Real Andy wrote:
>>> On Mon, 6 Jul 2009 22:01:49 +1000, "David L. Jones"
>>
>>> Its interesting this. Is microchip going to release the compiler free
>>> of charge?
>>
>> They release the "demo" version of the compiler free, without a time  
>> limit or (as I recall) any limits except that optimizations are turned  
>> off.
>
>That like a car dealer not allowing test drivers to go beyond first gear.   
>It's a great way to ensure that people continue to have no idea about the  
>quality of the product.
>
>> Having optimizations turned off is pretty bad -- the code frequently  
>> bank-switches to the memory bank it is already in, for instance.  So the  
>> un-optimized compiler is good for learning C and testing algorithms, but  
>> its output is rather bloated.
>
>As is any compiler's unoptimized output.

In the case of the PIC10/12/16 version it is not just bloated but horrific. 
Common simple expressions which should translate into single PIC instructions with minimal compiler
effort  ( e.g. var=0; or var++;) end up as 4 or 5.

On 8 July, 10:32, The Real Andy <thereala...@nospam.com> wrote:
> On Mon, 6 Jul 2009 22:01:49 +1000, "David L. Jones"
>
> <altz...@gmail.com> wrote:
> >nob...@nowhere.com wrote:
> >> Does anybody know what has happened to their old compilers e.g. H8/300
> >> or Z180?
>
> >They binned them, as to be expected.
>
> >> Is somebody buying the rights to these products?
>
> >Unlikely, as they would probably include common code that is part of the PIC
> >compiler which Microchip want to keep for themselves.
>
> >Dave.
>
> Its interesting this. Is microchip going to release the compiler free
> of charge? Atmel uses gcc, so it must be getting some good market
> share on that aspect alone. I recently did some work on the AP7000
> running linux, and whilst I hate linux its clear that Atmel has done a
> lot of work to make it all free. Gcc is pretty good, can microchip
> beat it?

Microchip makes the source code for their C30 and C32 compilers
available, they are based on gcc. They supply free versions of those
compilers as well, with some features disabled.

Leon

On Jul 9, 9:27=A0am, Mike Harrison <m...@whitewing.co.uk> wrote:
> In the case of the PIC10/12/16 version it is not just bloated but horrifi=
c.
> Common simple expressions which should translate into single PIC instruct=
ions with minimal compiler
> effort =A0( e.g. var=3D0; or var++;) end up as 4 or 5.

A Cynic might be drawn to observe, that is not merely 'non-optimised'
but rather deliberately de-optimised, or nobbled ! !  :(

-jg

On Wed, 08 Jul 2009 17:07:50 +0100, Nobody <nobody@nowhere.com> wrote:

>On Wed, 08 Jul 2009 19:32:25 +1000, The Real Andy wrote:
>
>>>> Is somebody buying the rights to these products?
>>>
>>>Unlikely, as they would probably include common code that is part of the PIC 
>>>compiler which Microchip want to keep for themselves.
>>>
>>>Dave.
>> 
>> Its interesting this. Is microchip going to release the compiler free
>> of charge? Atmel uses gcc, so it must be getting some good market
>> share on that aspect alone. I recently did some work on the AP7000
>> running linux, and whilst I hate linux its clear that Atmel has done a
>> lot of work to make it all free. Gcc is pretty good, can microchip
>> beat it?
>
>Microchip uses gcc for their higher-end products, but modifying gcc for
>an 8-bit target isn't realistic.
>
>I'm not expecting them to give away the Hi-Tech product for free.

(snip)

Why not?  That surely is the optimum business model.  Let's face it, their core
business is selling silicon.  If giving away the tools (which cost them
incrementally nothing per copy) gains/retains buyers and user base, they are in
front.  At the moment users are migrating away from Microchip in numbers,
contributed to by the tools situation.

On Thu, 09 Jul 2009 09:57:45 +0800, who where wrote:

>>I'm not expecting them to give away the Hi-Tech product for free.
> 
> (snip)
> 
> Why not?

Because they charge $500 for C18, and the Hi-Tech compiler is supposedly a
better product (why else would they have bought it?)

> That surely is the optimum business model.  Let's face it, their core
> business is selling silicon.  If giving away the tools (which cost them
> incrementally nothing per copy) gains/retains buyers and user base, they
> are in front.  At the moment users are migrating away from Microchip in
> numbers, contributed to by the tools situation.

I can't see the cost of tools being a significant factor for major
customers. Even if they charge nothing for the software, the time it
takes for an engineer to get up to speed on the platform will equate to a
few grand in salary.

OTOH, I wouldn't expect that Microchip earns a significant proportion of
their revenue from tools. I'm wondering if they do it to maintain some
modest barriers to entry, to give their more valuable customers some
advantage over the mom-&-pop outfits.

Op Wed, 08 Jul 2009 23:27:11 +0200 schreef Mike Harrison  
<mike@whitewing.co.uk>:
> On Wed, 08 Jul 2009 17:48:03 +0200, "Boudewijn Dijkstra"  
> <boudewijn@indes.com> wrote:
>> Op Wed, 08 Jul 2009 17:17:59 +0200 schreef MC
>> <for.address.look@www.ai.uga.edu.slash.mc>:
>>> The Real Andy wrote:
>>>> On Mon, 6 Jul 2009 22:01:49 +1000, "David L. Jones"
>>>
>>> Having optimizations turned off is pretty bad -- the code frequently
>>> bank-switches to the memory bank it is already in, for instance.  So  
>>> the un-optimized compiler is good for learning C and testing  
>>> algorithms, but its output is rather bloated.
>>
>> As is any compiler's unoptimized output.
>
> In the case of the PIC10/12/16 version it is not just bloated but  
> horrific.
> Common simple expressions which should translate into single PIC  
> instructions with minimal compiler
> effort  ( e.g. var=0; or var++;) end up as 4 or 5.

The first stage is often stateless.  Meaning that it doesn't care yet  
whether 'var' is located in a register.  It will load it into a register  
first, ensuring the right bank is selected, perform the operation, and  
store the value again.  I haven't seen it any other way with any RISC  
compiler I ever used.



-- 
Gemaakt met Opera's revolutionaire e-mailprogramma:  
http://www.opera.com/mail/

On Thu, 09 Jul 2009 04:55:11 +0100, Nobody <nobody@nowhere.com> wrote:

>On Thu, 09 Jul 2009 09:57:45 +0800, who where wrote:
>
>>>I'm not expecting them to give away the Hi-Tech product for free.
>> 
>> (snip)
>> 
>> Why not?
>
>Because they charge $500 for C18, and the Hi-Tech compiler is supposedly a
>better product (why else would they have bought it?)
>
>> That surely is the optimum business model.  Let's face it, their core
>> business is selling silicon.  If giving away the tools (which cost them
>> incrementally nothing per copy) gains/retains buyers and user base, they
>> are in front.  At the moment users are migrating away from Microchip in
>> numbers, contributed to by the tools situation.
>
>I can't see the cost of tools being a significant factor for major
>customers. Even if they charge nothing for the software, the time it
>takes for an engineer to get up to speed on the platform will equate to a
>few grand in salary.
>
>OTOH, I wouldn't expect that Microchip earns a significant proportion of
>their revenue from tools. I'm wondering if they do it to maintain some
>modest barriers to entry, to give their more valuable customers some
>advantage over the mom-&-pop outfits.

Microchip have always supported all sizes of customer, and this is probably a significant reason for
their success. ISTR reading a while ago that no one customer represents more than 5% of their
business.

Unlike other MCUs that were designed for high-level language friendliness (ARM, AVR etc.), the
Microchip parts absolutely need a compiler that is heavily tweaked to suit the architecture to
produce reasonable code, so they are more dependent on commercial/customised compilers than their
competition, who can get by with GCC and get reasonable performance. 

Their current policy of supplying a free, highly inefficient compiler, as opposed to something like
a fully optimising but code-size limited one is a really good way to put off potential new customers
when they evaluate it, get poor results and dismiss their product in favour of the competition. 
   

On Jul 9, 5:55 am, Nobody <nob...@nowhere.com> wrote:

> I can't see the cost of tools being a significant factor for major
> customers. Even if they charge nothing for the software, the time it
> takes for an engineer to get up to speed on the platform will equate to a
> few grand in salary.

Sure, but cost of tools is a serious issue for hobbyists and small
companies. Experience acquired as a result of hobby projects could
later influence some big purchasing decisions.

On 9 July, 02:57, who where <no...@home.net> wrote:
> On Wed, 08 Jul 2009 17:07:50 +0100, Nobody <nob...@nowhere.com> wrote:
> >On Wed, 08 Jul 2009 19:32:25 +1000, The Real Andy wrote:
>
> >>>> Is somebody buying the rights to these products?
>
> >>>Unlikely, as they would probably include common code that is part of t=
he PIC
> >>>compiler which Microchip want to keep for themselves.
>
> >>>Dave.
>
> >> Its interesting this. Is microchip going to release the compiler free
> >> of charge? Atmel uses gcc, so it must be getting some good market
> >> share on that aspect alone. I recently did some work on the AP7000
> >> running linux, and whilst I hate linux its clear that Atmel has done a
> >> lot of work to make it all free. Gcc is pretty good, can microchip
> >> beat it?
>
> >Microchip uses gcc for their higher-end products, but modifying gcc for
> >an 8-bit target isn't realistic.
>
> >I'm not expecting them to give away the Hi-Tech product for free.
>
> (snip)
>
> Why not? =A0That surely is the optimum business model. =A0Let's face it, =
their core
> business is selling silicon. =A0If giving away the tools (which cost them
> incrementally nothing per copy) gains/retains buyers and user base, they =
are in
> front. =A0At the moment users are migrating away from Microchip in number=
s,
> contributed to by the tools situation.

Evidence?

Leon

On Thu, 09 Jul 2009 04:55:11 +0100, Nobody <nobody@nowhere.com> wrote:

>I can't see the cost of tools being a significant factor for major
>customers.

Not to existing and short term new customers. But gaining new
customers in the long term is important too.

>Even if they charge nothing for the software, the time it
>takes for an engineer to get up to speed on the platform will equate to a
>few grand in salary.

Exactly. By giving away good tools, they will attract young hobbyists
and students who don't have a lot of money. Years later, some of these
people will inevitably end up in jobs in large companies.

Imagine a situation where the new employee tells his boss: "I can do
it. If we use an Atmel chip, I'll have it done by Friday, but if we
use a Microchip chip, I'll need a few weeks to learn.".
-- 
RoRo