I'm trying to extract a spice netlist from the car...
# openram
m
I'm trying to extract a spice netlist from the caravel gds which includes
sram_1rw1r_32_256_8_sky130
. This is openRAM, right? I'm getting unexpected results in the
pk_control_logic_r
extraction. @User says that it looks like some cells might be smashed and the lower labels are clashing. Here's a small sample of the extracted
mag
file. Who do I contact about this?
Copy code
node "D" 0 0 2791 2953 mvpdif 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
node "S" 0 0 3159 2703 mvpdif 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
node "D" 0 0 3159 2703 mvpdif 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
node "gnd" 0 0 3109 5580 li 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
node "Q" 0 0 1716 51 ndif 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
equiv "Q" "dout_0"
equiv "Q" "D"
equiv "Q" "D"
equiv "Q" "Z"
node "ON" 0 0 619 102 ndif 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
node "csb" 0 0 143 531 p 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
equiv "csb" "D"
equiv "csb" "D"
equiv "csb" "D"
m
This seems like a Caraval issue. We have a lot of nodes in the transistors called D, S, and G but given hierarchy this should be fine... Also, the SRAM should be read as a read only GDS.
m
@Matthew Guthaus Thanks for the quick response. Does the SRAM pass
netgen
LVS or do you use another tool for verification? I was thinking that the subcells of
pk_control_logic_r
may have been inadvertently flattened. I'm trying to read in the GDS and extract. Is that what you mean by "read as a read only GDS."
m
It passes LVS with Calibre.
You should read in the GDS with the read only option so that Magic doesn't convert to its internal structures (eg vias). This could mess up some layers that Magic auto generates like implants and npc.
t
@Mitch Bailey (@Matthew Guthaus): Reading GDS with the "readonly" option is only important if you are going to either save the resulting layout or re-write the GDS. If read as an abstract view, then it will extract as a black-box entry. If not read as an abstract view, then it will be extracted from the database contents, which are not correct GDS but are good enough to do extraction with.