Actel vs. Xilinx and Altera

Hi,

It's been a couple of years since I've been a heavy FPGA user, but it appears
that I'll now be getting back into them.  As of a few years back, I was using
Xilinx Virtex IIe parts and was quite happy with them... I kept up with what
Altera was doing as well, and while it always seemed to me that for DSP
applications Xilinx tended to have the edge, in many ways Xilnx and Altera
were the Coke and Pepsi of FPGAs -- both were good, solid products where
either could have gotten the job done in the vast majority of applications.

Where I am now there's been some historical use of the Actel 54SX parts,
something I've never used.  However, I do recall that -- as of a few years
ago -- the deal with Actel was always that the parts were antifuse-based, so
while you _might_ be able to gain something in speed, you gave up a lot in the
way of being able to issue field upgrades, bug fixes, etc.  However, I now see
that Actel has their ProASIC line of parts so they can perhaps compete
somewhat closer to Xilinx and Altera than previously.  Could anyone summarize
how the ProASIC parts stack up to the contemporary Xilinx and Altera parts?
(E.g., Xilinx Virtex II or 4, Altera Stratix II.)  In particular I'm
interested in:

-- DSP usage.  Things seemed to get a lot easier when Xilinx starrted
introducing fixed DSP blocks (e.g., multiply-accumulate blocks) within the
FPGA fabric.
-- Embedded processor usage.  I never used them, but Xilinx and Altera's
embeeded "soft cores" (microblaze and NIOS) both seemed pretty neat, and
Xilinx was offering ARM hard cores if you really wanted "big iron."
-- Debugging support.  Xilinx had some "soft probe" thing that would let you
poke around the internal nets of the FPGA as it was running, and I believe
Altera had something like this even before Xilinx.
-- Tool support.  I used to use Synplify for VHDL synthesis, which worked
quite well.  I tried Xilinx's built-in synthesis tool, and given the price
(vs. Synplify), it was really pretty good as well.

How does Actel performs in these area?  I realize they're very general
questions, but I'm trying to get a feeling for how viable ProASICs are for
something like a software defined radio (i.e., plenty of "real" DSP, desire
for some "supervisory" soft core CPU, etc.) vs. just going with what I know
would work -- Xilinx or Altera.

Thanks,
---Joel Kolstad

Joel Kolstad wrote:
> -- Embedded processor usage.  I never used them, but Xilinx and Altera's
> embeeded "soft cores" (microblaze and NIOS) both seemed pretty neat, and
> Xilinx was offering ARM hard cores if you really wanted "big iron."

Actually, Altera was the one with the ARM9 core that came out in their
Excalibur family, this wasn't really a "big iron" processor and the family
is no longer actively pushed.

Xilinx has the PowerPC405 hard core which spans 3 families, Virtex-II Pro,
Virtex-II Pro X and Virtex-4 FX.  With a 705 DMIPS this would be considered
"big iron" for FPGA offerings.

> -- Debugging support.  Xilinx had some "soft probe" thing that would let you
> poke around the internal nets of the FPGA as it was running, and I believe
> Altera had something like this even before Xilinx.
> 

There are several levels of the debugging support from Xilinx.  The
"soft probe" function allows you to use FPGA Editor to route any net to
an output pin.  This has been around for 10+ years.  There is also the
ChipScope Pro cores and software that allow you to insert logic analyzers
(ILA), processor bus analyzers (IBA), and virtual I/O (VIO) cores into your
design for debug and control.  And through a joint project with Agilent
you can use an external logic analyzer with internal cores (ATC2) to provide
easy control and deep trace analysis.

Ed

Hi Joel,

As suggested before I would simply write some time/area critical blocks of 
your design, then synthesize/P&R for the three different families and see 
what you get. obviously you need to study all 3 architectures since you 
might need to do some low level stuff. If you don't have an all vendors 
synthesis tool then get an evaluation copy of Precision/Synplicity which 
should sort you out for at least 30 days :-) You might also want to speak to 
an FAE from all three companies (ask the Actel guy when the ProASIC3+ARM 
core will be available :-) to see what support is like. For debugging I 
would suggest you have a look at Temento's Dialite which is an FPGA 
independent JTAG debugger on steroids :-)

Regards,
Hans.
www.ht-lab.com

"Joel Kolstad" <JKolstad71HatesSpam@yahoo.com> wrote in message 
news:11cp5ptpk3jg1c2@corp.supernews.com...
> Hi,
>
> It's been a couple of years since I've been a heavy FPGA user, but it 
> appears
> that I'll now be getting back into them.  As of a few years back, I was 
> using
> Xilinx Virtex IIe parts and was quite happy with them... I kept up with 
> what
> Altera was doing as well, and while it always seemed to me that for DSP
> applications Xilinx tended to have the edge, in many ways Xilnx and Altera
> were the Coke and Pepsi of FPGAs -- both were good, solid products where
> either could have gotten the job done in the vast majority of 
> applications.
>
> Where I am now there's been some historical use of the Actel 54SX parts,
> something I've never used.  However, I do recall that -- as of a few years
> ago -- the deal with Actel was always that the parts were antifuse-based, 
> so
> while you _might_ be able to gain something in speed, you gave up a lot in 
> the
> way of being able to issue field upgrades, bug fixes, etc.  However, I now 
> see
> that Actel has their ProASIC line of parts so they can perhaps compete
> somewhat closer to Xilinx and Altera than previously.  Could anyone 
> summarize
> how the ProASIC parts stack up to the contemporary Xilinx and Altera 
> parts?
> (E.g., Xilinx Virtex II or 4, Altera Stratix II.)  In particular I'm
> interested in:
>
> -- DSP usage.  Things seemed to get a lot easier when Xilinx starrted
> introducing fixed DSP blocks (e.g., multiply-accumulate blocks) within the
> FPGA fabric.
> -- Embedded processor usage.  I never used them, but Xilinx and Altera's
> embeeded "soft cores" (microblaze and NIOS) both seemed pretty neat, and
> Xilinx was offering ARM hard cores if you really wanted "big iron."
> -- Debugging support.  Xilinx had some "soft probe" thing that would let 
> you
> poke around the internal nets of the FPGA as it was running, and I believe
> Altera had something like this even before Xilinx.
> -- Tool support.  I used to use Synplify for VHDL synthesis, which worked
> quite well.  I tried Xilinx's built-in synthesis tool, and given the price
> (vs. Synplify), it was really pretty good as well.
>
> How does Actel performs in these area?  I realize they're very general
> questions, but I'm trying to get a feeling for how viable ProASICs are for
> something like a software defined radio (i.e., plenty of "real" DSP, 
> desire
> for some "supervisory" soft core CPU, etc.) vs. just going with what I 
> know
> would work -- Xilinx or Altera.
>
> Thanks,
> ---Joel Kolstad
>
> 

Thanks for the info, Hans.  BTW, I like your web site, but I'd like it even
more if you didn't "trap" other web sites within your own frames. :-)

---Joel