Re: Digest for electricvlsi@googlegroups.com - 2 updates in 1 topic
4 views
Skip to first unread message
Alexandre Rusev
Nov 29, 2020, 4:56:41 AM11/29/20
to electr...@googlegroups.com
Ok.
Before having good understanding of scaling algos in Electric I am not ready to code scaling rule infering appropriately ;)
(straight forward approaches to fix it don't work for me)
Postponing my efforts on this direction in favour of using XML description
But it would be convenient to use advantage of placing geometries using GUI, than add scaling rules in textual form.
Adam had to implement a special code generator due to generate complex geometries for Sky130..
Would you like to include his code to the main repository as example tech-generator?
Alex
On Sun, Nov 29, 2020 at 9:30 AM <electr...@googlegroups.com> wrote:
Alexandre Rusev <cybe...@gmail.com>: Nov 28 07:11PM +0300
Converted MOCMOS tech to library, then removed scalable and NPN transistors
and and encountered the same issue for *all* resistors:
Number of cut-s supposed to be different for different size-of-resistor
examples due to infer scaling rules for amount of vias later.
But current implementation of associateExamples(...) verifies that number
of instances of "similar layer" is the same in all examples.
This equality is intentionally NOT right for cuts in resistor examples ,
and possibly also not right for vias in case of capacitors
What is right approach to fix it?
Cell node-Hi-Res-Poly2-Resistor{lay}, node artwork:Filled-Box[art@23]: *Layer
Poly-Cut found 2 times in main example, 6 in others*
Failed to convert the library to a technology
private boolean associateExamples(List<Example> neList, Cell np)
{
// if there is only one example, no association
if (neList.size() <= 1) return false;
// associate each example "ne" with the original in "neList"
Example firstEx = neList.get(0);
for(int n=1; n<neList.size(); n++)
{
Example ne = neList.get(n);
// clear associations for every sample "ns" in the example "ne"
for(Sample ns : ne.samples)
ns.assoc = null;
// associate every sample "ns" in the example "ne"
for(Sample ns : ne.samples)
{
if (ns.assoc != null) continue;
// cannot have center in other examples
if (ns.layer == Generic.tech().cellCenterNode)
{
error.markError(ns.node, np, "Grab point should only be in main example");
return true;
}
// count number of similar layers in original example "neList"
int total = 0;
Sample nsFound = null;
for(Sample nsList : firstEx.samples)
{
if (nsList.layer != ns.layer) continue;
total++;
nsFound = nsList;
}
// no similar layer found in the original: error
if (total == 0)
{
error.markError(ns.node, np, "Layer " + Info.getSampleName(ns.layer) + "
not found in main example");
return true;
}
// just one in the original: simple association
if (total == 1)
{
ns.assoc = nsFound;
continue;
}
// if it is a port, associate by port name
if (ns.layer == Generic.tech().portNode)
{
String name = Info.getPortName(ns.node);
if (name == null)
{
error.markError(ns.node, np, "Port does not have a name");
return true;
}
// search the original for that port
boolean found = false;
for(Sample nsList : firstEx.samples)
{
if (nsList.layer == Generic.tech().portNode)
{
String otherName = Info.getPortName(nsList.node);
if (otherName == null)
{
error.markError(nsList.node, np, "Port does not have a name");
return true;
}
if (!name.equalsIgnoreCase(otherName)) continue;
ns.assoc = nsList;
found = true;
break;
}
}
if (!found)
{
error.markError(null, np, "Could not find port " + name + " in all
examples");
return true;
}
continue;
}
// count the number of this layer in example "ne"
int i = 0;
for(Sample nsList : ne.samples)
{
if (nsList.layer == ns.layer) i++;
}
// if number of similar layers differs: error
if (total != i)
{
*error.markError(ns.node, np, "Layer " + Info.getSampleName(ns.layer) + "
found " + total + " times in main example, " + i + " in others");*
return true;
}
[image: image.png]
Steven Rubin <str...@staticfreesoft.com>: Nov 28 08:51AM -0800
You need to understand that the "Technology Editor" (one of three ways
to define a technology) is limited because it uses graphical examples to
describe the shape rules. It does not understand examples where the
number of cuts grows, and I believe it uses other methods to determine
this phenomenon. But in the long run, when you need specialized
technology features, the Technology Editor is not able to help anymore.
The "Technology Creation Wizard" is also limited. But the third way to
edit a technology is to hand edit the XML, and that is how you can get
those unusual rules included.
-Steven Rubin
On 11/28/2020 8:11 AM, Alexandre Rusev wrote:
You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page.
To unsubscribe from this group and stop receiving emails from it send an email to electricvlsi...@googlegroups.com.
Reply all
Reply to author
Forward
0 new messages