Repository navigation
gpl: Incremental global placement using estimate_target_density - #4594
vvbandeira merged 4 commits into
Conversation
Signed-off-by: LucasYuki <lucasyuki@yahoo.com.br>
There was a problem hiding this comment.
Code Review
This pull request updates the OpenROAD subproject commit and modifies the resize.tcl script to estimate and apply target density during incremental global placement when arithmetic operator swapping is enabled. The feedback points out that estimate_target_density should be called after replace_arith_modules rather than before, because the latter modifies the netlist and changes the total cell area, which would lead to an inaccurate density target.
🔍 QoR checkMetrics reflect the PR merge build — i.e. what will land on the target branch. Commit 62 design(s) checked — 9 with regression(s), 0 without a comparable baseline. ❌ asap7/aes-block base — 1 failing metric(s)
❌ asap7/riscv32i-mock-sram base — 1 failing metric(s)
❌ gf180/jpeg base — 1 failing metric(s)
❌ ihp-sg13g2/jpeg base — 3 failing metric(s)
❌ nangate45/ibex base — 1 failing metric(s)
❌ nangate45/tinyRocket base — 2 failing metric(s)
❌ sky130hd/chameleon base — 4 failing metric(s)
❌ sky130hd/riscv32i base — 2 failing metric(s)
❌ sky130hs/riscv32i base — 2 failing metric(s)
|
| if { [env_var_equals SWAP_ARITH_OPERATORS 1] } { | ||
| # Enable sanity checker until replace_arith_modules becomes stable | ||
| set_debug_level ODB replace_design_check_sanity 1 | ||
| set density [estimate_target_density -overflow 0.1] |
There was a problem hiding this comment.
For incremental mode now we estimate the density before calling the incremental command. Isn't 0.1 the target overflow anyway?
There was a problem hiding this comment.
It is the default target overflow.
Do you want me to remove the -overflow argument from this call?
|
@LucasYuki conflict |
Signed-off-by: LucasYuki <lucasyuki@yahoo.com.br>
babdf6e
into
The-OpenROAD-Project:master
Uses estimate_target_density to pick an explicit density for the post-arithmetic-replacement incremental placement, and bumps the OpenROAD submodule for the filler-redistribution and overflow-recovery improvements it depends on.