When Design Becomes Technology Specific ?

B

Bar Nash

Guest
Hi

When a design turns into technology specific ?

At the GATE LEVEL stage ?

At the NETLIST stage ?

Thanks
EC
 
On Mon, 6 Oct 2008 22:36:15 +0200, "Bar Nash"
<wavesoft@netvision.net.il> wrote:

Hi

When a design turns into technology specific ?

At the GATE LEVEL stage ?

At the NETLIST stage ?

Thanks
EC
Click to expand...
I am not sure about your terminology is above but here is what happens
(usually) during (ASIC) synthesis: The tool maps the RTL to its
internal representation; depending on the tool this representation may
involve primitive gates (flops, nand, nor etc) or it may also include
architectural constructs like adders (which can be changed from ripple
to cla etc. later). Then this netlist gets mapped to specific gates in
your library without concern to driving strength initially (again this
depends on the tool) at which point it becomes specific to your
technology library; at this stage there is also a timing optimization
before placement which can involve architectural selection (which type
of adders to use, remap muxes etc). Then these gates are placed and
sized/resized and timing optimized, buffered etc. After placement
clock tree is synthesized and inserted and then the design routed.

To answer your question more clearly the design becomes technology
specific after it gets converted to internal representation and before
first timing optimization is done. But even at that stage some tools
don't use the actual gates from your library but from an abstract one
generated from your library to enable them do resizing more optimally.
In today RTL to GDS flows, the exact point where your design becomes
technology specific is difficult to answer and probably irrelevant
too.

Muzaffer Kal
http://www.dspia.com
 
Thank you very much
Your answer made the subject clear .
EC

"Muzaffer Kal" <kal@dspia.com> ???
??????:6gmle4hbfd1cpcl0p583b7s82ejfh9grqa@4ax.com...
On Mon, 6 Oct 2008 22:36:15 +0200, "Bar Nash"
wavesoft@netvision.net.il> wrote:

Hi

When a design turns into technology specific ?

At the GATE LEVEL stage ?

At the NETLIST stage ?

Thanks
EC


I am not sure about your terminology is above but here is what happens
(usually) during (ASIC) synthesis: The tool maps the RTL to its
internal representation; depending on the tool this representation may
involve primitive gates (flops, nand, nor etc) or it may also include
architectural constructs like adders (which can be changed from ripple
to cla etc. later). Then this netlist gets mapped to specific gates in
your library without concern to driving strength initially (again this
depends on the tool) at which point it becomes specific to your
technology library; at this stage there is also a timing optimization
before placement which can involve architectural selection (which type
of adders to use, remap muxes etc). Then these gates are placed and
sized/resized and timing optimized, buffered etc. After placement
clock tree is synthesized and inserted and then the design routed.

To answer your question more clearly the design becomes technology
specific after it gets converted to internal representation and before
first timing optimization is done. But even at that stage some tools
don't use the actual gates from your library but from an abstract one
generated from your library to enable them do resizing more optimally.
In today RTL to GDS flows, the exact point where your design becomes
technology specific is difficult to answer and probably irrelevant
too.

Muzaffer Kal
http://www.dspia.com
Click to expand...
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top