Magma Blast Synthesis Question

[nanuji]

gain based synthesis

Hi ,

I am new to Magma Blast. I need some basic fundamental information on it.
While we do logic synthesis with other tools like DC, we generally over cosntrain our design by x% (as per guided by architect...nomally 20%) because the wire load assumptions and other stuff.

Now my question is , do we need to over constrain our RTL while doing synthesis with Magma Blast? I have read magma articles and it says that magma uses fixe time methodology, and as per this fix time methodology it assumes the fix timing (i.e. cosntrains given by user) and vary the size of the cell.

so from above I understand is that, even if we over constrain the design with magma, it will put size accordingly, but while we do p&r with magma, where we give our actuall constrains (i.e. targeted freq), it will resize all the cells as per new actual cosntrains...

Please let me know your views on the same...

 

[bjbhargav]

magma blast

Hi buddy,

In Magma You will have a gain (Virtually a Size) for each design cell untill real placement is done.

Magma has an additional step called library preparation, in that magma decides gain for each cell available in library from tau(T) value derived from minimum size inverter.
likewise,
Cell Gain
------- ----------
BUF1X 3
BUF2X 5
etc...,

At the time of mapping (synthesis) tool can pick any cell from library having similar foorprint based on gain requirement in design (decided on basis of fanout load).

So, No need to overconstraint your design at synthesis stage,
Feel free to tell me if you want to learn fix timing methodology (a very nice topic to learn in Magma)

bj

 

[nanuji]

synthesis based on cell footprint

Bhargav,
You anaswer does not answer my qesution completely... so plese tyr to answer in details from pratcial perspective.... as per industry standards...
What ever you explained is the theory of Magma. I know this.
I am asking about the pratical thing. I'm looking for standard industry pratice.
One more thing, regarding your statement "At the time of mapping (synthesis) tool can pick any cell from library having similar foorprint based on gain requirement in design (decided on basis of fanout load)". As per fix time methodoloy... at the time of synthesis it will alwasy implement with SUPER cells only. It wont look for suitable footprint and sutff like that... it just uses sper cell ...

Regarding the learning the fix time methodoloy... I'm eager to know more... if you have any documents, ppts pdfs ... that you can share... please share it on the boards... so that evey one can take advanctage of it... or you can send me to my personal email account as well.

One more queston...
what about specifying the PVT conditions at synthesis time... is there any command in magma like DC have "set_operating_condition ".. ? please let me know if you have some sort of list or guide that gives DC equivalent magma commands...

 

[denmos]

fix time magma

Regrading your questions, I think you should read carefully about Gain based synthesis documents ( that come from Magma binary and application notes). For Super-cell model, Magma Blast modeled that virtual cell base on cell footprint (with the same functional and voltage threshold cells Magma will create a super cell on behaves). The wire-load model is not longer support in Magma. The Gain-based have more advantages vs wire-load synthesis.

To specifying PVT condition in Magma, please use 'rule condition create/default' and 'config condition type' command. Please refer the manpages for more detail. I believe the DC commands have the equivalent Magma command in the same name, you can use the super-help command 'help find' to list the DC equivalent command on Magma

, denmos