optuna: prune to simplest interesting case

multiobjective driven by synthesis parameters is the minimum interesting
case, prune everything else and reduce size of python code and
dependencies in terms of what a reader has to understand.

Fancy dependencies to visualize, if they are easy to use, such as generating
interactive 3d interactive HTML for pareto fron is fine.

Signed-off-by: Øyvind Harboe <oyvind.harboe@zylin.com>

Author: oharboe

optuna/BUILD

@@ -13,6 +13,7 @@ orfs_flow(
     name = "mock-cpu",
     abstract_stage = "cts",
     arguments = {
+        "VERILOG_TOP_PARAMS": "NUM_CORES 2 PIPELINE_DEPTH 4 WORK_PER_STAGE 8",
         "CORE_ASPECT_RATIO": "0.5",
         "CORE_UTILIZATION": "50",  # Higher utilization
         "PLACE_DENSITY": "0.20",
@@ -44,19 +45,10 @@ orfs_flow(
     verilog_files = [":mock-cpu.sv"],
 )
 
-# Extract timing information from CTS
-orfs_run(
-    name = "mock-cpu_timing",
-    src = ":mock-cpu_cts",
-    outs = ["mock-cpu_timing.txt"],
-    arguments = {"OUTFILE": "$(location mock-cpu_timing.txt)"},
-    script = ":extract_timing.tcl",
-)
-
 # Extract PPA (Performance, Power, Area) metrics from CTS
 orfs_run(
     name = "mock-cpu_ppa",
-    src = ":mock-cpu_cts",
+    src = ":mock-cpu_synth",
     outs = ["mock-cpu_ppa.txt"],
     arguments = {"OUTFILE": "$(location mock-cpu_ppa.txt)"},
     script = ":extract_ppa.tcl",
@@ -73,5 +65,8 @@ py_binary(
         "@bazel-orfs-pip//matplotlib",
         "@bazel-orfs-pip//numpy",
         "@bazel-orfs-pip//optuna",
+        "@bazel-orfs-pip//pandas",
+        "@bazel-orfs-pip//plotly",
+        "@bazel-orfs-pip//kaleido",
     ],
 )

optuna/README.md

@@ -23,20 +23,17 @@ Side note: ORFS has a large number of small designs for testing purposes, wherea
 - Supports constrained optimization (must meet timing) and multi-objective optimization (area + power simultaneously)
 - Optuna shines in simplicity and good results for limited effort. No need install or configure a server for remote execution.
 
-## Screenshots
+## 3D HTML intearctvie pareto front snapshot
 
-![](screenshots/optuna-plot.webp)
+![3D pareto from](screenshots/optuna_dse_results.html)
 
 ## Quick Start
 
 ### Main Tool: Parameter Optimization (Recommended)
 
 ```bash
-# Single-objective: Minimize area
-bazelisk run //optuna:run -- --n-trials 20
-
-# Multi-objective: Minimize area and power (Pareto frontier)
-bazelisk run //optuna:run -- --multi-objective --n-trials 30
+# Minimize all objectives (Pareto frontier)
+bazelisk run //optuna:run -- --n-trials 30
 
 # Custom parameter ranges
 bazelisk run //optuna:run -- \
@@ -56,7 +53,6 @@ bazelisk run //optuna:run -- --help
 - `--max-density FLOAT`: Maximum PLACE_DENSITY (default: 0.70)
 - `--n-trials INT`: Number of optimization trials (default: 20)
 - `--seed INT`: Random seed for reproducibility (default: 42)
-- `--multi-objective`: Enable multi-objective optimization (area + power)
 - `--output-dir PATH`: Output directory for results (default: optuna/results)
 
 ## Experiment Flow

optuna/extract_ppa.tcl

@@ -1,5 +1,8 @@
 source $::env(SCRIPTS_DIR)/load.tcl
-load_design 4_cts.odb 4_cts.sdc
+# from optuna/results/asap7/mock_cpu/base/3_place.odb, we want 3_place
+# stracted from $::env(ODB_FILE)
+set file_stem [file rootname [file tail $::env(ODB_FILE)]]
+load_design ${file_stem}.odb ${file_stem}.sdc
 
 # ============================================================
 # Performance Metrics