Hello,
let me introduce a little thing, and I swear I have no shares in
this company.
Microsemi (now owner of Actel) is launching a new generation of
Flash-based FPGA with more performance, more capacity, integrating
a well-featured ARM core (M3@166MHz) and all the possible I/Os
that sophisticated embedded systems dream for. It's called "SmartFusion
2"
and it drops analog I/Os from the earlier SmartFusion (still available
and targeted at slightly different markets).
Even as I knew the previous generations, it took a while to understand
what this new beast can be used for. It tries to catch up with Altera
and Xilinx
in density and performance, and blows the competition out of their
socks
on other aspects. Read their docs very carefully.
One missing interface is to video/display. There are embedded SERDES
but they don't seem rated for HDMI (it's a speed issue, I've been told,
HDMI is slower than their SERDES... but they should look into it).
Now look at it this way : here is a chip that basicly only needs a few
(LP)DDR2/DDR3 chips and it's almost a full featured
LOW POWER, CONFIGURABLE SOC. (as in : not a current-hungry
beast like some X&A references as found on MilkyMist)
Considering that the HDMI/DVI/whatever interface can be solved soon
(like, maybe mid-2013 when the production has ramped up),
it's an amazing platform for small volumes, with security
features that are beyond my expectations, as they enable
secure yet open and transparent configuration delivery.
More on this later.
What's the point, why bug the OGML ?
Well, the FPGA fabric has between 11 and 240 Math blocks,
10 and 236 18Kb SRAM blocks, 11 and 240 1Kb SRAM blocks...
Are you seeing where it can go ?
* With a scalable shader and the OGD1s programmable scanline
generator, it's possible to add cool 2D and 3D features
to the very moderately-powered M3 core.
* Skip the M3 and you get a PCIe-compliant graphic card.
* Use the XGMII (up to 10Gbps) and you can plug the board
to your computer to have a remote X display.
So to me, the SmartFusion2 is a viable, very interesting target
for the OpenGraphics soft shader. It can get it to work quite
fast, in a very decent timeframe (i expect decent devkits
to be available by mid-2013) and a lot of flexibility for
the end applications : as a main accelerator for the M3,
as a coproc over the SERDES...
Furthermore : the devkits will be already designed, soldered,
distributed. It won't be too hard to get some recognition
(eventually endorsement) from Microsemi if the result is great
(Actel has Gaisler's SPARC cores too).
What do you think ? Isn't it a good, cost-effective, fast
and open target for a first iteration of the OpenGraphics shader ?
yg
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
