Db network connect #6727
Db network connect #6727
Conversation
Signed-off-by: andyfox-rushc <andy@rushc.com>
Signed-off-by: andyfox-rushc <andy@rushc.com>
Signed-off-by: andyfox-rushc <andy@rushc.com>
|
clang-tidy review says "All clean, LGTM! 👍" |
…ring swapMaster Signed-off-by: andyfox-rushc <andy@rushc.com>
|
clang-tidy review says "All clean, LGTM! 👍" |
Signed-off-by: andyfox-rushc <andy@rushc.com>
|
clang-tidy review says "All clean, LGTM! 👍" |
|
I would recommend printing the offending net name(s) in the errors if practical. |
|
Cho @cmoon has been notified of the change to swapMaster and I respectfully suggest that if necessary any further error handling be added in a separate pull request. I can imagine a situation in which it is total ok to kill any existing modNet on a new iterm during swapMaster (eg when swapping in something that was not properly disconnected). My problem was the implied disconnection of the hierarchical net when connecting to a flat net (connect(dbNet) was implicitly killing both the hierarchical net and the dbNet if present). So I fixed that part (made connect(dbNet) only kill an existent dbNet) and then made the modNet removal explicit in the swapMaster code. But I defer to Cho on this issue in swapMaster. |
Signed-off-by: andyfox-rushc <andy@rushc.com>
|
clang-tidy review says "All clean, LGTM! 👍" |
Signed-off-by: andyfox-rushc <andy@rushc.com>
|
clang-tidy review says "All clean, LGTM! 👍" |
|
|
||
| */ | ||
|
|
||
| void dbNetwork::connectPin(Pin* pin, Net* flat_net, Net* hier_net) |
There was a problem hiding this comment.
If both flat_net and hier_net are supplied then check that hier_net belongs to flat_net
There was a problem hiding this comment.
No, it really does not. The point of this api is just to allow a pin to be hooked to a hier net and a flat net simultaneously.
Also you comment evidences lack of knowledge of the architecture used: hier nets do not "belong" to flat nets. I think that is the root of you mis-understanding. Hier nets simply decorate the flat network (eg at the edge of hierarchical module boundaries to connect to the flat instances).
There was a problem hiding this comment.
To help you learn a little more of the architecture used, here is some documentation on the hierarchy representation. Note that we never assume that there is a hierarchical net for each flat net.
NotesOnHierarchicalRepresentation.pdf
This draft pull request includes modifications to the dbNetwork connect apis.
Specifically the new api call:
void dbNetwork::connectPin(Pin* pin, Net* flat_net, Net* hier_net)
This one allows a pin to be simultaneously hooked to a flat net and a hierarchical net.
The intent is to simplify the coding so a pin is not first checked for a hierarchical net, disconnected and then hooked to new flat and hierarchical nets.
Some other clean up performed (eg explicit removal of flat net and hier net and some further journalling/undo support).