Circuit Parser

Graph Construction

• Put .v and .lib to OpenSTA to generate .sdf file, which contains cell delay and net delay, we can construct the circuit graph based on these connections

Documentation for DGL

• https://docs.dgl.ai/index.html

Data Format from Repo

• https://github.com/TimingPredict/Dataset

Edges (3 types)

cell_out

• fanin $\rightarrow$ fanout
• cell arcs.

Features:

• 2*4*(1+7+7)x: [E/L]* cell_{rise, fall}, {rise, fall}_transition {is_valid, xindex, yindex}

• 2*4*49x: [E/L]* cell_{rise, fall}, {rise, fall}_transition values

• 4x: cell delay annotations (EL/RF)

• Edge type = cell_out

  • fea_name = e_cell_delays, fea_scheme = Scheme(shape=(4,), dtype=torch.float32)
  • fea_name = ef, fea_scheme = Scheme(shape=(512,), dtype=torch.float32)

net_out

• net arc
• net driver $\rightarrow$ net sink
• 2x: relative position
• Edge type = net_out
  • fea_name = ef, fea_scheme = Scheme(shape=(2,), dtype=torch.float32)

net_in

• net arc
• net sink $\rightarrow$ net driver
• 2x: relative position
• Edge type = net_in
  • fea_name = ef, fea_scheme = Scheme(shape=(2,), dtype=torch.float32)

Node (1 type)

node

each node is pin Features:

• nf
  • 1x: is primary I/O pin (1) or not (0)
  • 1x: is fanin (0) or fanout (1)
  • 4x: relative to the top/left/right/bottom of die area
  • 4x: capacitance information (EL/RF) in cell library
• 4x: net delay annotations (EL/RF) for fanin pins
• 4x: arrival time annotations (EL/RF)
• 4x: slew annotations (EL/RF)
• 1x: is timing endpoint (i.e. has constraint) (1) or not (0)
• 4x: required arrival time annotations (EL/RF)

Node type = node

• fea_name = n_rats, fea_scheme = Scheme(shape=(4,), dtype=torch.float32)
• fea_name = n_net_delays, fea_scheme = Scheme(shape=(4,), dtype=torch.float32)
• fea_name = n_ats, fea_scheme = Scheme(shape=(4,), dtype=torch.float32)
• fea_name = n_slews, fea_scheme = Scheme(shape=(4,), dtype=torch.float32)
• fea_name = nf, fea_scheme = Scheme(shape=(10,), dtype=torch.float32)
• fea_name = n_is_timing_endpt, fea_scheme = Scheme(shape=(), dtype=torch.float32)