Support swapping to uninstantiated modules #6616
Conversation
1) this requires OpenSTA changes to set top module name properly and return contents of unused modules 2) each unused module is stored in a separate child block of top block 3) swapMaster supports swapping to unused modules 4) some regression failures exist (dbSta/test has 3 failures) Signed-off-by: Cho Moon <[email protected]>
1) fixed regression failures 2) fixed connectivity bug due to wrong bus member bit 3) fixed missing mod nets and internal flat nets Signed-off-by: Cho Moon <[email protected]>
1) excluded black box cells from module swap candidates 2) this nedes OpenSTA update 3) fixed regression test failures Signed-off-by: Cho Moon <[email protected]>
Signed-off-by: Cho Moon <[email protected]>
1) linkNetwork change to defer module deletion is still outstanding 2) additional OpenSTA change is still needed Signed-off-by: Cho Moon <[email protected]>
Signed-off-by: Cho Moon <[email protected]>
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
Signed-off-by: Cho Moon <[email protected]>
Signed-off-by: Cho Moon <[email protected]>
Signed-off-by: Cho Moon <[email protected]>
|
clang-tidy review says "All clean, LGTM! 👍" |
1 similar comment
|
clang-tidy review says "All clean, LGTM! 👍" |
|
@andyfox-rushc, could you review the codes? |
There was a problem hiding this comment.
dbInst::makeUniqueInst.
Please check that dbInst::create really does return null if the named object exists (I don't believe it does). Perhaps another approach is to look in the hash on the dbModule (possibly getting the size of the hash and using that size as the id). See dbModule::findInst.
There was a problem hiding this comment.
Thank you for adding the printContent. it is very useful to have this exposed.
There was a problem hiding this comment.
Please add a comment to clarify the layout of the bus ports which is:
modbterm <-- busport type (the sentinel) has reference to size of bus port -- number and direction of elements
list of modbterms the actual bus content
So for a bus with N elements we see N + 1 modbterms. (The first one is the "bus port" sentinel).
There was a problem hiding this comment.
There is already a map for dbInsts. Perhaps this map is for something else (to help find undefined objects ?) As a dbInst is added it is added to the inst_tbl in the block. See the tables in the block (the tables are used by the hash functions). So it is unclear to me why there is a need for another hash.
_net_hash.setTable(_net_tbl);
_inst_hash.setTable(_inst_tbl);
_module_hash.setTable(_module_tbl);
_modinst_hash.setTable(_modinst_tbl);
The index for the has is the name. So this would appear at first blush to do what you want, but I might be wrong !
There was a problem hiding this comment.
This is a duplicate comment. Method makeUniqueInst dbInst::makeUniqueInst.
Please check that dbInst::create really does return null if the named object exists (I don't believe it does). Perhaps another approach is to look in the hash on the dbModule (possibly getting the size of the hash and using that size as the id). See dbModule::findInst.
There was a problem hiding this comment.
For the naming construction. Please can you add a "todo" item to get the hierarchy separator from the dbBlock (there is a field called hier_delimiter) in dbBlock.h. Likewise a todo item could be usefully added to state the array index delimiter on the dbBlock. I realize that I hardcoded those things ('/' and '[') myself and can now see it will proliferate. I think we need to get those characters from the same place (the dbBlock).
set top module name properly and
return contents of unused modules