ADSP 21262 - GPL toolchain

Jim Thomas <jthomas@bittware.com> wrote in news:11ou4tm1eomttb6
@corp.supernews.com:

> giangiammy wrote:
>> hi all,
>> 
>> I have googled some time to look for GPl development tool
>> of ADSP 21262, with no luck: any one have informations
>> about such tools?
>> Does exist something?
>> 
> 
> ADI used to provide a GPL tool chain, but switched to closed source in 
> the late 90's.  If memory serves, the last processor ADI supported with 
> gcc was the 21062.  The compiler was called g21k, so there's a term for 
> you to feed to Google.
> 
> The 21160 has a lot of features the 2106x family lacks (SIMD in 
> particular, and the 2126x stuff looks a lot like the 2116x).  It is a 
> NON-trivial task to add 2116x support to g21k.  I have the source tree 
> hanging around on an old not-currently-installed-in-a-PC hard drive, so 
> if you want it, send me an email.  It might take me a couple of days to 
> unearth it though.
> 
> Maybe Al Clark knows something more up-to-date than me.


Not much to add, except that for all practical purposes, 2116x, 2126x and 
2136x code is all compatible from a language point of view. These DSPs 
all support SIMD. 

We tend to use serial flash for bootloading these days. You would need to 
create an appropriate loader file program that certainly did not exist in 
the old g21k.


-- 
Al Clark
Danville Signal Processing, Inc.
--------------------------------------------------------------------
Purveyors of Fine DSP Hardware and other Cool Stuff
Available at http://www.danvillesignal.com

Donald wrote:
 > giangiammy wrote:
 >>
 >> thank you very much for the disponibility, but I found the g21k: the
 >
 >                               ^^^^^^^^^^^^^
 >                                 WHAT !!??
 >

To which giangiammy replied:
> sorry for my english, I adapted from italian
> I just wanted thank Jim for offering his help
> (and hope that this is grammatically correct :-)
> bye
> giammy
> 

OK you have me intrigued. I have a hobbyist interest in linguistics.
[ My 9th grade {1st yr secondary school} English teacher had her masters 
degree in Indo-European. I've not been same since ;]

~50 yrs ago I had 3 years of Latin and 3 years of French.
I can't figure out the specific word you wanted.

What was the Italian word you were thinking of?
What is its Latin root (if any)?

My intention is to dive into WEB on an etymological search of discovery.

On Thu, 01 Dec 2005 13:40:00 -0600, Richard Owlett
<rowlett@atlascomm.net> wrote:
>
>Donald wrote:
> > giangiammy wrote:
> >>
> >> thank you very much for the disponibility, but I found the g21k: the
> >
> >                               ^^^^^^^^^^^^^
> >                                 WHAT !!??

http://www.google.com/translate_t?text=available&langpair=en%7Cit&hl=en&ie=UTF8

<abariska@student.ethz.ch> wrote in message 
news:1133456869.280780.11830@g49g2000cwa.googlegroups.com...
> giangiammy wrote:
>> > if you want it, send me an email.  It might take me a couple of days to
>> > unearth it though.
>>
>> thank you very much for the disponibility, but I found the g21k: the
>> problem
>> is exactly its missing support for SIMD :-(
>
> If it is only that, then that is not really a problem. The command to
> activate SIMD is
>
> "bit set mode1 PEYEN;"   = 0x140200200000
>
> and to disable SIMD you need
>
> "bit clr mode1 PEYEN"  = 0x142200200000
>
> Include the following macros into your code:
>
> #define ENABLE_SIMD INSTR1: px =  0x140200200000; pm(INSTR1+2) = px;
> nop;
> #define DISABLE_SIMD INSTR2: px =  0x142200200000; pm(INSTR2+2) = px;
> nop;
>
> Now whenever you want to enable or disable SIMD in your code, call the
> respective macro instead.

Self-modifying code.  Cool! 

Hi,

i was thinking:
Grazie per la tua disponibilit=E0

I do not know the latin root, if I find something
in my dictionary i'll let you know

bye
giammy

abariska@student.ethz.ch wrote:
> I just thought of three things: First, the above construct doesn't work
> (the multiple use of INSTR1 and INSTR2 will generate errors), second
> enabling and disabling SIMD can be achieved much simpler (define PEYEN
> and you're done), and thirdly, adding SIMD support will be the least of
> your problems. The major problem is how you can fool the linker into
> generating a memory image of the 212xx instead of the 2106x. There
> might be a chance with the "nomemcheck" switch of the linker ...
> 

You're right.  I wouldn't fret too much over the lack of compiler 
support for SIMD.  Even with VisualDSP, when I want SIMD, I resort to 
assembly.  The bigger issue is the memory map and linker.  I fiddled 
with g21k back in the day to try to get it to support the 21065. 
There's an architecture file parser (g21k used .ach files intead of the 
.ldr files used by VisualDSP) that is absolute spaghetti (there's some 
Italian for you!)

Most of the time I spent on it was in trying to clean up the existing 
code rather than adding 21065 support, and then I ran out of enthusiasm.

It would probably be well worth the time to track down the Beastrider 
offerings.

-- 
Jim Thomas            Principal Applications Engineer  Bittware, Inc
jthomas@bittware.com  http://www.bittware.com    (603) 226-0404 x536
When you know how things work, the world is one big sandbox.  - Avins