as Xilinx has dropped hard processor IP in the
latest families it
makes ACTEL the only FPGA vendor whos latest product family does have
hard processor IP.

Smart fusion includes Cortex-M3, and yes its available now, I did have
the kits in my hand at embedded

Antti

On Mar 5, 7:56=A0am, Antti
<antti.luk...@googlemail.com> wrote:
> as Xilinx has dropped hard processor IP in the latest families it
> makes ACTEL the only FPGA vendor whos latest product family does have
> hard processor IP.
>
> Smart fusion includes Cortex-M3, and yes its available now, I did have
> the kits in my hand at embedded
>
> Antti

 Did they mention a price for the A2F060 ?

The A2F200 hits that old 'all things to all users' conundrum : Price!
(and package..)

eg you can buy the 'right sized' uC for these sort of prices, and add
the 'right sized' ProASIC from another selection list ?

[STM32 Value Line
$0.85 : 10K : 16Kbyte 48-pin LQFP48
$1.44 : 10K : 64Kbyte 64-pin LQFP64
]

 Cypress has the PSoC3 in open samples, and supposedly the PSoC5 in
more stealth samples ?

-jg

On Mar 4, 11:56=A0am, Antti
<antti.luk...@googlemail.com> wrote:
> as Xilinx has dropped hard processor IP in the latest families it
> makes ACTEL the only FPGA vendor whos latest product family does have
> hard processor IP.

Putting a processor inside an FPGA has proven to us to be a bigger
PITA than it's worth.

Consider than instead of V4FX, you can use an S3AN and a standalone
PPC and you'll pay a whole lot less. Plus the various Freescale PPCs
have DDR memory and Ethernet and DMA controllers that don't suck, and
you're not stuck with crappy tools.

Embedding the processor in the FPGA is an interesting idea, but as
long as Brand X seems to think that the only people who do are the
types who want to run Linux on an FPGA, it's gonna suck for actual
embedded use.

-a

On Mar 5, 9:14=A0am, Andy Peters
<goo...@latke.net> wrote:
> Putting a processor inside an FPGA has proven to us to be a bigger PITA t=
han it's worth.
> Consider than instead of V4FX, you can use an S3AN and a standalone
> PPC and you'll pay a whole lot less. Plus the various Freescale PPCs
> have DDR memory and Ethernet and DMA controllers that don't suck, and
> you're not stuck with crappy tools.

Yes, you should always 'reality check' the alternatives.
It can make sense if the resource really is 'free', but SRAM FPGA's
still need external code memory, and when you look at those newest uC
prices, they are LESS than SRAM!!

We could use the A2F060, if the price and package are right. - but we
consider it more a uC + CPLD, than
a FPGA.

-jg

On Mar 4, 11:08=A0pm, -jg
<jim.granvi...@gmail.com> wrote:
> On Mar 5, 9:14=A0am, Andy Peters <goo...@latke.net> wrote:
>
> > Putting a processor inside an FPGA has proven to us to be a bigger PITA=
 than it's worth.
> > Consider than instead of V4FX, you can use an S3AN and a standalone
> > PPC and you'll pay a whole lot less. Plus the various Freescale PPCs
> > have DDR memory and Ethernet and DMA controllers that don't suck, and
> > you're not stuck with crappy tools.
>
> Yes, you should always 'reality check' the alternatives.
> It can make sense if the resource really is 'free', but SRAM FPGA's
> still need external code memory, and when you look at those newest uC
> prices, they are LESS than SRAM!!
>
> We could use the A2F060, if the price and package are right. - but we
> consider it more a uC + CPLD, than
> a FPGA.
>
> -jg

dont judge on the mouser pricing yet, the price issue will be key
issue
that is recognized by Actel also, specially A2F060 must come to low
price to make it an option to choose at all, at the mouser prices for
A2F200
i guess it would be very little applications where on would pay that
price

Antti

On Mar 4, 3:14 pm, Andy Peters
<goo...@latke.net> wrote:
> On Mar 4, 11:56 am, Antti <antti.luk...@googlemail.com> wrote:
>
> > as Xilinx has dropped hard processor IP in the latest families it
> > makes ACTEL the only FPGA vendor whos latest product family does have
> > hard processor IP.
>
> Putting a processor inside an FPGA has proven to us to be a bigger
> PITA than it's worth.

Who exactly is "us"?  Are you with Xilinx, Actel or someone else?


> Consider than instead of V4FX, you can use an S3AN and a standalone
> PPC and you'll pay a whole lot less. Plus the various Freescale PPCs
> have DDR memory and Ethernet and DMA controllers that don't suck, and
> you're not stuck with crappy tools.
>
> Embedding the processor in the FPGA is an interesting idea, but as
> long as Brand X seems to think that the only people who do are the
> types who want to run Linux on an FPGA, it's gonna suck for actual
> embedded use.

The big problem seems to be the problem of too many size combinations
to be practical.  Mixing FPGA, CPU, SRAM and Flash on one chip can't
be all things to all people.  Still, I think there are a few sweet
spots that can be profitable.  A smallish FPGA (~3000 to 5000 LUT4s)
combined with a CM3 in the typical MCU memory combinations is a
powerful device, especially if it is available with a decent dual
CODEC in a package on the smaller side (100 pins or less).  The cost
of low end FPGAs seem to be I/O count driven, so a 5000 LUT4 FPGA can
likely be combined with a CM3 (or maybe a CM0), etc, without
increasing the cost significantly.  I am paying $10 for a 3000 LUT4
FPGA in a 100 TQFP along with a $3 stereo CODEC.  If I could get an
MCU in that package with even just 64 kB of Flash it would allow a
very space constrained product to really have some capabilities... if
they could keep the cost down to $15 at qty 100.

Other than Cypress, no one else seems to see this as a viable market.
I'm still waiting to see if the Cypress PSOC5 is all its cracked up to
be.  The Actel parts are way too expensive from what I've heard and I
think the PSOC5 will be too rich as well.

Rick

On Mar 5, 10:44=A0pm, Antti
<antti.luk...@googlemail.com> wrote:
> recognized by Actel also, specially A2F060 must come to low
> price..

Hi Antti,

 Did you get everything finished for ESC ok >

 Speaking of low prices, have you seen price indicators yet on the
SiliconBlue iCE65L01 ?
(other than the 99c/1M pieces) - a 1K-10K price is
more useful :) - and availability timelines ?

-jg

On Mar 5, 11:51=A0am, rickman
<gnu...@gmail.com> wrote:
> On Mar 4, 3:14 pm, Andy Peters <goo...@latke.net> wrote:
>
> > On Mar 4, 11:56 am, Antti <antti.luk...@googlemail.com> wrote:
>
> > > as Xilinx has dropped hard processor IP in the latest families it
> > > makes ACTEL the only FPGA vendor whos latest product family does have
> > > hard processor IP.
>
> > Putting a processor inside an FPGA has proven to us to be a bigger
> > PITA than it's worth.
>
> Who exactly is "us"? =A0Are you with Xilinx, Actel or someone else?
>
> > Consider than instead of V4FX, you can use an S3AN and a standalone
> > PPC and you'll pay a whole lot less. Plus the various Freescale PPCs
> > have DDR memory and Ethernet and DMA controllers that don't suck, and
> > you're not stuck with crappy tools.
>
> > Embedding the processor in the FPGA is an interesting idea, but as
> > long as Brand X seems to think that the only people who do are the
> > types who want to run Linux on an FPGA, it's gonna suck for actual
> > embedded use.
>
> The big problem seems to be the problem of too many size combinations
> to be practical. =A0Mixing FPGA, CPU, SRAM and Flash on one chip can't
> be all things to all people. =A0Still, I think there are a few sweet
> spots that can be profitable. =A0A smallish FPGA (~3000 to 5000 LUT4s)
> combined with a CM3 in the typical MCU memory combinations is a
> powerful device, especially if it is available with a decent dual
> CODEC in a package on the smaller side (100 pins or less). =A0The cost
> of low end FPGAs seem to be I/O count driven, so a 5000 LUT4 FPGA can
> likely be combined with a CM3 (or maybe a CM0), etc, without
> increasing the cost significantly. =A0I am paying $10 for a 3000 LUT4
> FPGA in a 100 TQFP along with a $3 stereo CODEC. =A0If I could get an
> MCU in that package with even just 64 kB of Flash it would allow a
> very space constrained product to really have some capabilities... if
> they could keep the cost down to $15 at qty 100.
>
> Other than Cypress, no one else seems to see this as a viable market.
> I'm still waiting to see if the Cypress PSOC5 is all its cracked up to
> be. =A0The Actel parts are way too expensive from what I've heard and I
> think the PSOC5 will be too rich as well.
>
> Rick

The PSoC5 looks interesting, but the EMIF interface is pretty limited.
The external memory bus is restricted to a 32MHz clock, due to the pad
I/O apparently, and is hardwired to require two clock cycles per
access. This might be livable if the EMIF core had a ZBT or FIFO
option, but everything I've seen so far indicates that it only
supports traditional async and sync SRAM. The only saving grace is
that it can be 16-bits wide, giving you 32MB/s to an external host.

32MB/s is nothing to sneeze at for a lot of designs, but it would have
been nice to see them find a clever way to double that - even if it
meant DDR techniques.

I'm actually looking at putting together a little breadboard with a
PSoC5 and a Cyclone III to play around with, and see what kind of
sustained data rates can be achieved - once Cypress gets around to
sampling them and updates Creator with the EMIF core. I'll probably
emulate a sync SRAM in the FPGA, and memory map the I/O.

Of course, the real advantage of the PSoC's is the extremely flexible
analog. Strapping a 20-bit del-sig ADC to anything can be absurdly
expensive, and you are generally stuck with it once you put it down.
The PSoC, on the other hand, lets you trade speed for resolution, and
play all kinds of games with the clocks; after the board is done - and
generally create the "perfect" analog hardware for the task on-chip.
Even better, you can do it on the fly by poking the configuration
registers with software.

If they are even close to being as good as Cypress claims, I've got
lots of jobs for them.