Fastest way to run a software feedback loop?

Altera seems to keep theirs around for several years, anyway. Xilinx has been iffy with their Coolrunner series, IMO. In any case, there is a better chance the low-end, like CPLDs will be around longer than the high-end stuff. Other than some military parts, those things change faster than womens' fashion.

There are mixed signal FPGAs now. Haven't used one, so I don't know how good they are.

=A0I'll test

n and

r a bit

eeds to be

. =A0The

d a FET.

lta-sigma?

e for a

z 8051

ons of

has to

that'll

hat, e.g.

g from

r maybe

r,

nt

th

rous

of

the

like

many

for

nds

laser

ee

rea.

as been

chance

=A0Other

w good

But they are not cheap. Probably double the cost of a high end micro. BTW, I am running a 96MHz PIC32, roughly equivalent to a

=A0I'll test

n and

r a bit

eeds to be

. =A0The

d a FET.

lta-sigma?

e for a

z 8051

ons of

has to

that'll

hat, e.g.

g from

r maybe

r,

nt

th

rous

of

the

like

many

for

nds

laser

ee

ng

area.

has been

er chance

=A0Other

on.

how good

PIC32MX575F512H-80 overclocked to 96MHz

Which PIC # please? I would like to look at that.

Jamie

test

bit

to be

FET.

delta-sigma?

for a

of

to

that'll

e.g.

from

maybe

Twice the price of a high-end micro? I don't think so.

=A0I'll test

..

tion and

e or a bit

st.)

l needs to be

ter. =A0The

and a FET.

delta-sigma?

ague for a

e tons of

ely has to

But that'll

n that, e.g.

hing from

WM

t or maybe

can

ator,

-cent

nd

with

enerous

ch of

ith the

te like

or many

te for

with

Sounds

as laser

o see

long

t area.

x has been

ter chance

=A0Other

ion.

how good

Yes, you are right. Could be more, much more. The only pricing I can find online is a 600K Actel Fusion on Avnet, listed for $155.

Yes - I have been running an algorithm at 500kHz sampling on a 44MHz ADUC7021 for a while now (ARM32). It ran at 1MHz too but had no time for anything else. The STM32F2 replacing it runs at 120MHz (CortexM3), There is a 168MHz STM32F4 part (Cortex M4). All single chip microcontrollers.

How does that compare to execution speed of real C code? PIC32 is MIPS based. Since MIPS is not mainstream and ARM is I'd expect the compilers for the latter to be more efficient.

On a sunny day (Mon, 17 Oct 2011 08:38:08 GMT) it happened snipped-for-privacy@puntnl.niks (Nico Coesel) wrote in :

My gcc MIPS version is really OK.

Cray 1 was not mainstream either.

PIC32 has 5 stage pipeline. ARM Cortex M3 (ACM3) has 3. ACM3 is not bad, but PIC32 does not suck too badly either.

Compiled from pic32-gcc:

Around 3 millions C +=3D A * B per second at 96MHz running from flash, where A, B and C are 32 bits memory buffer.

People have it running 160MHz in the fridge (for cooling).

But they had loads of money to spend. AFAIK Microchip has made some proprietary additions to GCC for MIPS but the effort is nowhere near the effort that is being put into GCC for ARM.

A deeper pipeline isn't always better. Google for MIPS and 'one delay slot'. You'll learn two things: jumps and calls may be costly and never try to program assembly on a MIPS cpu.

I have been down the MIPS road and quickly discovered there are very few people you can ask for directions.

M, or

I don't plan on doing too much assembly, that's why we PICk on RISC in the first place.

There are too many forks (roads) for ARM. I/Os for STM is not necessary the same for NXP.

There are times for ARM, but should not rule out MIPS yet.

I am still looking into the PIC32 A2D. Perhaps we can get close to

Just be happy they use the same CPU core nowadays. In the old days each vendor had his own core complemented by a crappy toolchain. Ever chased a bug caused by an error in the compiler?

Using the peripherals is usually easy. Both ST and NXP have a large library on their websites which can serve as an example.

end

to a

in RA=3D

MIPS

=3DA0ACM3 is no=3D

Just so long as you do not assume that the examples are bug free.

?-)