> I grinned to myself while reading that rah-rah press release.  "Synopsys, a
> world leader in blah blah blah..." had acquired Virtio HW/SW co-design for
> an undisclosed price.  (I later heard it was a $15 million cash deal.)
>
>     - from http://www.deepchip.com/gadfly/gad051806.html


After I published that Industry Gadfly about Synopsys acquiring Virtio, I got
emails from two EDA entrepreneurs whose opinions I respect.  (I like hearing
from these guys because they can often give me the unofficial, unapproved
story-behind-the-story that explains some odd EDA actions.  Of course, these
types ALWAYS want to be anonymous if I quote them, but that's the price of
getting the truth vs. the bullshit press release version of the story.)

The first EDA entrepreneur wrote:

  "I have no inside knowledge, but my understanding is that Synopsys is
   an investor in Virtio -- so their purchase could just be their way of
   maximizing the value of the future money needed from Synopsys as an
   investor to implement Virtio's business plan.  Presumably Virtio needed
   another round, and went to the existing investors looking for dough.

   Synopsys decides to buy out the other investors, bringing the company
   inside, eliminate the redundant sales & marketing costs of Virtio, and
   invest future Synopsys dollars only in engineering & support costs.

   The other shareholders presumably are happy with a sure $15 million
   rather than run the risk Virtio won't fly.  Synopsys pays $15 million
   (presumably some to themselves as an investor and some to folks who
   are now Synopsys employees and likely have earn-outs) to get to see
   100% of their future investment go to where they want it to go, and
   remove the uncertainty that Virtio will do anything other than
   Synopsys wants them to do.

The second EDA entrepreneur wrote:

  "Great history!  But I think SNPS will have to go a fourth round.

   There's a reason why the V-guys (Vast, Virtutech, Virtio) and their
   instruction set simulator forefathers don't sell -- it's lack of
   performance, stupid!  If you code in C/C++ (where 95% of the embedded
   programmers do), you are accustomed to writing software on a computer
   running 2,000+ MIPS (laptop performance).  It's brutal to go back to
   a 10 to 50 MIPs environment that an ISS environment gives you.  It is
   the negative 100X performance that makes traditional ISS simulators
   uninteresting to embedded software developers.  Since there is no
   value, there is no one in the system houses demanding hardware models
   to make HW/SW co-design market ultimately happen.

   So how could Synopsys make such an obvious mistake?  My take is that
   they first talked to semiconductor company GMs who said their chip
   revenue ramps were hindered by their customer's inability to write
   software earlier, and faster.  If Synopsys could give their customers
   a software model of the chip in the proper simulation/IDE, they could
   get started writing software earlier.  So Synopsys correctly
   identified the problem to be solved.

   Where Synopsys erred was talking to the wrong end users.  I suspect
   that they talked to chip designers in these semiconductor companies,
   who talk to their own internal reference board designers.  The
   reference board designers might also write software drivers.  In most
   cases they may in fact be doing some assembly coding, because chip
   timing might be that critical.   But assembly language programmers at
   best are 5% of all embedded programmers, and they are not the guys
   at Synopsys's customers's customer.

   What Synopsys should have done was to talk to any one of the 4,000
   C-programmers in the Nokia software development ecosystem that write
   software to the TI OMAP platform (one HW model Virtio was able to get
   done).  They would have told Synopsys "It's all about simulator and
   model performance.  We don't write in assembly code.  We write in C."
   Nokia won't use that Virtio assembly language OMAP model because its
   a 5 MIPS dog.

   Anyway, as long as EDA marketing departments think embedded software
   guys only write hardware drivers,  EDA companies will keep buying the
   assembly code HW/SW companies that cater to less than 5% of the world's
   600,000 embedded software programmers.  There's only one company I am
   aware that offers a HW/SW tool with 2000+ MIP performance for the 95%
   crowd: Paravirtual.  It was originally called Accenia.  We looked at
   them for a round.  Their simulator was originally called DoubleWide
   Studio.  Today the money in this industry is in the models, not the
   simulator.  Most of these ISS deals are $50K for the simulator and
   $300K in design services for PhDs to build the models it needs.  It's
   these missing models that stunts all the ISS players."

Mind you, this is only backdoor gossip about the Virtio acquisition.  For the
official story, you should *only* trust the Synopsys press release.  :)

P.S. As a side note, I'm curious about how this impacts CoWare's business now.