FPGARelated.com
Forums

Reconfiguration (on the fly) using SPARTAN 3A

Started by Unknown February 21, 2008
On Feb 21, 11:17 pm, austin <aus...@xilinx.com> wrote:
> Yes. For example, if you stopped the system clock, reconfigured, and > started the clock again, that would probably work just fine. > > It was decided (long ago) that some features were more important to some > customers, and since Spartan parts are concerned with: cost, cost, cost > (in order of importance), things like reconfigurability are reserved for > the Virtex line (where we are not so greatly constrained).
Austin, Just for clarification - for all S3 devices partial reconfiguration is either: A) guaranteed to work by stopping clocks, so to avoid glitches or B) not supported at all, so everyone just forget it Scenario "A" leaves for external reconfiguration, but rules out internal one.
puneetjamrani@gmail.com wrote:
> Hi again, > > Actually what i currently want to do is: > 1. divide the fpga into 2 parts > a) one part will have picoblaze controller. > b) other will have an up counter. > 2. Save a partial file for down counter (generated from difference > based partial reconfiguration option) in one of the ROMs. > 3. Run the up counter initially and then using picoblaze call the > partial file from the ROM to reconfigure the UP counter to a DOWN one. > Now the BIG question:Is it possible using Spartan 3a? Can multiboot > option be used for it or any other way possible? > Or should i go for Virtex??] > > Regards, > Parunoy.
Have you considered using both an up counter and a down counter and functionally selecting between the two of them?
I just reported on 3A,

I would have to go ask about 3, 3E.

Austin
On 22 Feb., 16:35, austin <aus...@xilinx.com> wrote:
> I just reported on 3A, > > I would have to go ask about 3, 3E. > > Austin
no need go asking. S3A is first ever Spartan with ICAP. ditto. whatever the supported features of Spartan3A-ICAP are, no other spartan family has anything related to self-reconfig at all, there is no ICAP. Antti
http://www.xilinx.com/support/documentation/user_guides/ug332.pdf

page 248:

Tells you exactly what is supported by ICAP (and the list does not
include partial reconfiguration).

If you want to reconfigure to a new, or different bitstream, ICAP may be
used to do that.

Austin
On 22 Feb., 17:45, austin <aus...@xilinx.com> wrote:
> http://www.xilinx.com/support/documentation/user_guides/ug332.pdf > > page 248: > > Tells you exactly what is supported by ICAP (and the list does not > include partial reconfiguration). > > If you want to reconfigure to a new, or different bitstream, ICAP may be > used to do that. > > Austin
well, page 248 says "same as slave-map"... via slavemap writing bitstream on frame-by-frame basis should be possible. if the same is possible via ICAP, i would say it is partial reconfiguration ?? maybe there are reasons why it would not work properly or have side issues glitches? but it should be possible to reconfigure some cells in S3a in runtime using ICAP? Antti
On Feb 21, 3:14=A0am, puneetjamr...@gmail.com wrote:
> Hi all, > > I want to know whether it is possible to program Spartan 3a for a run > time partial reconfiguration. > If yes are there any detailed demo programs available? > > Regards, > Parunoy
Here is a radically different idea: Why (partial) reconfiguration? Anything you can do with reconfiguration, you can also achieve by logic duplication. Reconfiguration is just more efficient and might give higher performance. The OP seems to aim for small changes, like changing up-count to down- count. For that, reconfiguration is gross overkill. I suggest looking at the traditional solution to such designs, namely logic duplication. An up/down counter is hardly bigger than an up-counter... Peter Alfke, Xilinx Applications
On 22 Feb., 18:08, Peter Alfke <pe...@xilinx.com> wrote:
> On Feb 21, 3:14 am, puneetjamr...@gmail.com wrote: > > > Hi all, > > > I want to know whether it is possible to program Spartan 3a for a run > > time partial reconfiguration. > > If yes are there any detailed demo programs available? > > > Regards, > > Parunoy > > Here is a radically different idea: > Why (partial) reconfiguration? > Anything you can do with reconfiguration, you can also achieve by > logic duplication. Reconfiguration is just more efficient and might > give higher performance. > The OP seems to aim for small changes, like changing up-count to down- > count. For that, reconfiguration is gross overkill. > I suggest looking at the traditional solution to such designs, namely > logic duplication. > An up/down counter is hardly bigger than an up-counter... > Peter Alfke, Xilinx Applications
sure, the OP was maybe barking under wrong tree. the topic itself is still interesting, if I have some soft-core say in right half of the die and if i want to change the PULLUP to PULLDOWN on the left side of the die I would assume this could be done by S3A ICAP ??? most MCU's can have pullup control in software, FPGA's have that control only via configuration, I could see S3A adds at least partially this possibilty or is there some thing that prevents such runtime reconfig? Antti
Possible,

Yes.

Supported,

No.

'Play' all you like, but it is unsupported.

If it does something useful for you, great.  But, as an unsupported
feature, we have no obligation to provide it (it might go away,
doubtful, but a possibility if we change something to fix something
else, if something needs fixing -- or if we cost reduce something to
reduce costs).

The pdf lists the supported features.  Reprogramming live is noticeably
absent.

Austin
On Feb 22, 9:16=A0am, Antti <Antti.Luk...@googlemail.com> wrote:
> On 22 Feb., 18:08, Peter Alfke <pe...@xilinx.com> wrote: > > > > > On Feb 21, 3:14 am, puneetjamr...@gmail.com wrote: > > > > Hi all, > > > > I want to know whether it is possible to program Spartan 3a for a run > > > time partial reconfiguration. > > > If yes are there any detailed demo programs available? > > > > Regards, > > > Parunoy > > > Here is a radically different idea: > > Why (partial) reconfiguration? > > Anything you can do with reconfiguration, you can also achieve by > > logic duplication. Reconfiguration is just more efficient and might > > give higher performance. > > The OP seems to aim for small changes, like changing up-count to down- > > count. For that, reconfiguration is gross =A0overkill. > > I suggest looking at the traditional solution to such designs, namely > > logic duplication. > > An up/down counter is hardly bigger than an up-counter... > > Peter Alfke, Xilinx Applications > > sure, the OP was maybe barking under wrong tree. > > the topic itself is still interesting, if I have some soft-core say in > right half of the die > and if i want to change the PULLUP to PULLDOWN on the left side of the > die > > I would assume this could be done by S3A ICAP ??? > > most MCU's can have pullup control in software, FPGA's have that > control > only via configuration, I could see S3A adds at least partially this > possibilty > or is there some thing that prevents such runtime reconfig? > > Antti
Antti, in partial reconfiguration (as in many things) the devil is in the details. Partial reconfiguration can inherently change only one frame at a time. In Virtex, the change is guaranteed to be glitchless (no erroneous codes), but in Spartan there will be glitches =3D erroneous codes. If you stop the clock, these glitches will probably (?) not cause any harm... If Xilinx were to support such a feature in Spartan, there would be so many "ifs and buts" that we would would perhaps make some designers happy, but definitely alienate others. So we decided not to support this feature in Spartan. Use Virtex if you need partial reconfiguration... Peter