• YAP+IO_Assign is a Python-based yield modeling and simulation tool for advanced packaging that supports yield analysis of arbitrary I/O pad layouts. Currently, the model is specifically designed for wafer-to-wafer (W2W) and die-to-wafer (D2W) hybrid bonding.
• A GUI of YAP and the user guide video is available.

Current active D2W design families in this repository are:

• design_1_p5
• design_2_p10
• HBM_A
• HBM_B

Legacy design_1 / design_2 assets are still present under D2W/configs/old_configs/ or older handoff folders, but the examples below are written for the current active flow.

.
├── D2W/      # Code for D2W hybrid bonding
│   ├── configs/    # Golden and per-design configuration files
│   │   ├── GOLDEN.yaml
│   │   ├── design_1_p5/
│   │   ├── design_1_p10_proj/
│   │   ├── design_1_p20_proj/
│   │   ├── design_2_p10/
│   │   ├── HBM_A/
│   │   ├── HBM_B/
│   │   └── old_configs/
│   ├── input/      # Per-design 3dblox inputs, bump maps, and criticality files
│   │   ├── design_1_p5/
│   │   │   └── c10_r0_pg60_dm30/
│   │   │       ├── Center_IO/
│   │   │       ├── Edge_IO/
│   │   │       ├── Random_IO/
│   │   │       └── <chiplet_A>_to_<chiplet_B>_shared_nets.txt
│   │   ├── design_1_p10_proj/
│   │   │   └── c30_r0_pg50_dm20/
│   │   │       └── Edge_IO/
│   │   ├── design_1_p20_proj/
│   │   ├── design_2_p10/
│   │   ├── HBM_A/
│   │   │   ├── Original/
│   │   │   ├── Center_IO/
│   │   │   ├── Edge_IO/
│   │   │   └── Random_IO/
│   │   ├── HBM_B/
│   │   └── old_bmap/
│   └── utils/      # Helper scripts for bump map / criticality processing
├── W2W/      # Code for W2W hybrid bonding
├── LICENSE
├── README.md
└── requirements.txt    # Requirements of Python packages

1. Clone the repository

git clone -b yap+IO_assign https://github.com/Chen-Zhichao/YAP.git
cd ./YAP

2. (Optional) Create and activate a virtual environment:

conda create -n yap_env python=3.12
conda activate yap_env

3. Install dependencies:

pip install -r requirements.txt

• Generate criticality file from bump map

  python D2W/utils/generate_criticality.py --force
  

  Generate both criticality profiles for explicit bump maps:

  python D2W/utils/generate_criticality.py --file D2W/input/design_1_p5/c10_r0_pg60_dm30/Center_IO/Compute_Small_From_Substrate_Silicon.bmap --profiles both --force
  python D2W/utils/generate_criticality.py --file D2W/input/design_2_p10/c10_r0_pg60_dm30/Center_IO/Compute_Large_0_From_Substrate_Organic.bmap --profiles both --force
  

  The supported profiles are:

  • default: replicated redundant nets tolerate R-1 ESD failures and R-1 mechanical failures
  • esd_strict: replicated redundant nets tolerate 0 ESD failures and R-1 mechanical failures

• Run the yield simulator and model for D2W hybrid bonding.

  cd D2W
  

  Common variants are:

  • Original
  • Center_IO
  • Edge_IO
  • Random_IO

  Example command to run the pad risk map calculator for D2W hybrid bonding for a single design

  python pad_risk_map_calculator.py --config configs/design_1_p10_proj/design_1_p10_proj.yaml --mode d2w_modeling --ds_name design_1_p10_proj/c30_r0_pg50_dm20/Edge_IO --ds_dir input/design_1_p10_proj/c30_r0_pg50_dm20/Edge_IO --verbose
  
  python pad_risk_map_calculator.py --config configs/design_1_p20_proj/design_1_p20_proj.yaml --mode d2w_modeling --ds_name design_1_p20_proj/c30_r0_pg50_dm20/Edge_IO --ds_dir input/design_1_p20_proj/c30_r0_pg50_dm20/Edge_IO --verbose
  
  python pad_risk_map_calculator.py --config configs/design_1_p5/design_1_p5.yaml --mode d2w_modeling --ds_name design_1_p5/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_1_p5/c10_r0_pg60_dm30/Center_IO --verbose
  
  python pad_risk_map_calculator.py --config configs/design_2_p10/design_2_p10.yaml --mode d2w_modeling --ds_name design_2_p10/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_2_p10/c10_r0_pg60_dm30/Center_IO --verbose
  

  Notes for large-pad-count analytical ESD maps:

  • ESD pad risk map generation now applies two automatic accelerations by default.
  • A coarse-grid ESD evaluation is computed and then interpolated back to the full pad map.
  • Candidate-pad pruning evaluates only pads near the deterministic first-touch edge when that pruning is safe.
  • Risk-map saving is also optimized: text .risk.map output is written with a vectorized writer, and per-mechanism risk-map PNGs are now saved by default.
  • --plot is only needed if you also want the extra interactive mechanism plots shown during modeling.
  • You can override the coarse-grid factor with ESD_PAD_MAP_SUB_FACTOR in the YAML.
  • If ESD_PAD_MAP_SUB_FACTOR is unset or 0, a factor is chosen automatically from the active pad count.
  • Optional candidate-pruning knobs:
  • ESD_ANALYTICAL_CANDIDATE_SIGMA_WINDOW
  • ESD_ANALYTICAL_CANDIDATE_MIN_PADS
  • ESD_ANALYTICAL_CANDIDATE_DISABLE_FRACTION

  Example command to run the pad risk map calculator with the strict-ESD criticality profile

  python pad_risk_map_calculator.py --config configs/design_1_p5/design_1_p5.yaml --mode d2w_modeling --ds_name design_1_p5/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_1_p5/c10_r0_pg60_dm30/Center_IO --criticality-profile esd_strict --verbose
  
  python pad_risk_map_calculator.py --config configs/design_2_p10/design_2_p10.yaml --mode d2w_modeling --ds_name design_2_p10/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_2_p10/c10_r0_pg60_dm30/Center_IO --criticality-profile default --verbose
  
  python pad_risk_map_calculator.py --config configs/HBM_A/HBM_A.yaml --mode d2w_modeling --ds_name HBM_A/Center_IO --ds_dir input/HBM_A/Center_IO --verbose
  
  python pad_risk_map_calculator.py --config configs/HBM_A/HBM_A_overlay_pessimistic.yaml --mode d2w_modeling --ds_name HBM_A/Center_IO --ds_dir input/HBM_A/Center_IO --verbose
  
  

  Example command to run the pad risk map calculator for all variants of one or more designs

  ./run_design_pad_risk_maps.sh --ratio c10_r0_pg60_dm30 design_1_p5
  ./run_design_pad_risk_maps.sh --ratio c10_r0_pg60_dm30 design_1_p5 design_2_p10 HBM_A HBM_B
  ./run_design_pad_risk_maps.sh HBM_A HBM_B
  

  Example command to run pad risk map generation in parallel

  ./run_all_pad_risk_maps_parallel.sh --jobs 16 design_1_p5 design_2_p10 HBM_A HBM_B
  ./run_all_pad_risk_maps_parallel.sh --jobs 16 --skip-existing design_1_p5 design_2_p10 HBM_A HBM_B
  

  Notes:

  • Ratio-based designs such as design_1_p5 and design_2_p10 should be run with --ratio.
  • Current active ratios are c0_r20_pg60_dm20, c5_r10_pg60_dm25, and c10_r0_pg60_dm30.
  • HBM_A and HBM_B use direct variant folders and do not require a ratio.
  • The legacy wrapper run_design_1_2_hbm_pad_risk_maps_parallel.sh still exists in the repo but points to older design names and is not the recommended entry point for the current active datasets.

  Example command to run the simulator main for D2W hybrid bonding for a single design

  python simulator_main.py --config configs/design_1_p10_proj/design_1_p10_proj.yaml --mode d2w_simulation --ds_name design_1_p10_proj/c30_r0_pg50_dm20/Edge_IO --ds_dir input/design_1_p10_proj/c30_r0_pg50_dm20/Edge_IO --criticality-profile default --verbose
  
  python simulator_main.py --config configs/design_1_p20_proj/design_1_p20_proj.yaml --mode d2w_simulation --ds_name design_1_p20_proj/c30_r0_pg50_dm20/Edge_IO --ds_dir input/design_1_p20_proj/c30_r0_pg50_dm20/Edge_IO --criticality-profile default --verbose
  
  python simulator_main.py --config configs/design_1_p5/design_1_p5.yaml --mode d2w_simulation --ds_name design_1_p5/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_1_p5/c10_r0_pg60_dm30/Center_IO --criticality-profile default --verbose
  
  python simulator_main.py --config configs/design_2_p10/design_2_p10.yaml --mode d2w_simulation --ds_name design_2_p10/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_2_p10/c10_r0_pg60_dm30/Center_IO --criticality-profile default --verbose
  

  Example command to run the simulator with the strict-ESD criticality profile

  python simulator_main.py --config configs/design_1_p5/design_1_p5.yaml --mode d2w_simulation --ds_name design_1_p5/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_1_p5/c10_r0_pg60_dm30/Center_IO --criticality-profile esd_strict --verbose
  
  python simulator_main.py --config configs/design_2_p10/design_2_p10.yaml --mode d2w_simulation --ds_name design_2_p10/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_2_p10/c10_r0_pg60_dm30/Center_IO --criticality-profile default --verbose
  
  python simulator_main.py --config configs/HBM_A/HBM_A.yaml --mode d2w_simulation --ds_name HBM_A/Original --ds_dir input/HBM_A/Original --criticality-profile default --verbose
  
  python simulator_main.py --config configs/HBM_A/HBM_A_particle_pessimistic.yaml --mode d2w_simulation --ds_name HBM_A/Center_IO --ds_dir input/HBM_A/Center_IO --criticality-profile default --verbose
  
  python simulator_main.py --config configs/design_1_p5/design_1_p5.yaml --mode d2w_simulation --ds_name design_1_p5/c10_r0_pg60_dm30/Center_IO --ds_dir input/design_1_p5/c10_r0_pg60_dm30/Center_IO --criticality-profile default --save-failure-maps --verbose
  

  Example command to run D2W simulation for all variants of one or more designs

  ./run_design_simulations.sh --ratio c10_r0_pg60_dm30 design_1_p5
  ./run_design_simulations.sh --ratio c10_r0_pg60_dm30 --verbose design_1_p5 design_2_p10 HBM_A HBM_B
  ./run_design_simulations.sh HBM_A HBM_B
  

  Example command to run simulation experiments in parallel

  ./run_design_1_p5_design_2_p10_hbm_parallel.sh --jobs 16
  ./run_design_1_p5_design_2_p10_hbm_parallel.sh --jobs 16 --skip-existing
  ./run_design_1_p5_design_2_p10_hbm_parallel.sh --dry-run --jobs 4
  

  Example command to run the full design_1_p5 / design_2_p10 / HBM_A / HBM_B sweep used for result tables

  ./run_design_1_p5_design_2_p10_hbm_parallel.sh --jobs 16
  ./run_design_1_p5_design_2_p10_hbm_parallel.sh --dry-run --jobs 4
  

  Notes for the parallel simulation launcher:

  • Each experiment is forced to single-threaded execution.
  • Jobs sharing the same ds_dir are serialized; only different ds_dir values run in parallel.
  • .bmap inputs are no longer sorted in place during runtime. A sorted copy is created under output/<ds_name>/temp/.
  • Mechanical dishing caches are isolated by ds_name + config + criticality_profile, so different configs can run on the same ds_dir without sharing temp files.
  • Per-run generated interface YAML files are written under the design config folder (for example D2W/configs/design_1_p5/) with the __<config_stem>__<criticality_profile> suffix in the filename to avoid cross-run overwrites.
  • Runtime temp files are cleaned automatically after each experiment finishes.
  • Per-experiment logs are written to output/<ds_name>/parallel_simulation__<config_stem>__<criticality_profile>.log.
  • run_design_1_p5_design_2_p10_hbm_parallel.sh defaults to design_1_p5 and design_2_p10 across all active ratios with Center_IO / Edge_IO / Random_IO, plus HBM_A and HBM_B across Original / Center_IO / Edge_IO / Random_IO.
  • The legacy wrapper run_design_1_2_hbm_simulations_parallel.sh still exists in the repo but points to older design names and is not the recommended entry point for the current active datasets.

  Example commands for the current sensitivity-study helpers

  ./run_design1_p5_esd_sensitivity.sh
  ./run_design1_p5_esd_sensitivity_parallel.sh --jobs 8
  ./run_design2_p10_particle_sensitivity.sh
  ./run_design2_p10_particle_sensitivity_parallel.sh --jobs 16
  
  python D2W/utils/paper/plot_esd_sensitivity.py --input D2W/output/esd_sensitivity_check/esd_sensitivity_design_1_p5.tsv --output D2W/output/esd_sensitivity_check/esd_sensitivity_design_1_p5.png --ylim 85,100
  python D2W/utils/paper/plot_particle_sensitivity.py --input D2W/output/particle_sensitivity_design_2_p10.tsv --output D2W/output/particle_sensitivity_design_2_p10_bar.png
  

  Example command to package the current design_1_p5 / design_2_p10 / HBM_A / HBM_B configs and inputs for handoff

  tar -czf configs_input_design_1_p5_design_2_p10_HBM_A_HBM_B_20260414.tar.gz \
    D2W/configs/design_1_p5 D2W/configs/design_2_p10 D2W/configs/HBM_A D2W/configs/HBM_B \
    D2W/input/design_1_p5 D2W/input/design_2_p10 D2W/input/HBM_A D2W/input/HBM_B
  

1. Bump Map (.bmap):

Format: <instance> <bump_type> <x> <y> <port> <net>

Example: Bump_0 uBUMP 115 1610 txdatasb txdatasb

2. Risk Map (.map):

Format: <x> <y> <esd_failure_probability> <overlay_failure_probability> <particle_failure_probability> <mechanical_failure_probability>

Example: 115 1610 0.15 0.05 0.03 0.20

Note: Probabilities are float values between 0 and 1

NOTE: ESD criticality is multiplied by esd_failure_probability. Mechanical criticality is multiplied by overlay_failure_probability, particle_failure_probability, and mechanical_failure_probability. All four failure modes are considered in the optimization objective.

3. Criticality (.txt):

Current Format: <net1> [net2] [net3] ... <group_size> <tolerated_esd_failures> <tolerated_mechanical_failures>

Where:

• group_size: Total number of pads/bumps in the redundancy group
• tolerated_esd_failures: Number of ESD failures the group can tolerate before failing
• tolerated_mechanical_failures: Number of mechanical failures the group can tolerate before failing

Two filename variants are supported:

• *_criticality.txt
  • Default profile
  • Replicated redundant signal nets tolerate R-1 ESD failures and R-1 mechanical failures
• *_criticality_esd_strict.txt
  • Strict ESD profile
  • Replicated redundant signal nets tolerate 0 ESD failures and R-1 mechanical failures
  • PG and dummy nets are unchanged relative to the default file

Criticality values are calculated when reading the file:

• esd_criticality = (group_size - tolerated_esd_failures) / group_size
• mechanical_criticality = (group_size - tolerated_mechanical_failures) / group_size

Examples:

Single net with 5 pads, tolerates 4 ESD failures and 4 mechanical failures: vccfwdio 5 4 4 (Results in esd_criticality = 0.2, mechanical_criticality = 0.2)

Redundancy group with 4 pads, tolerates 1 ESD failure and 1 mechanical failure: rxckRD rxckn rxckp rxtrk 4 1 1 (Results in esd_criticality = 0.75, mechanical_criticality = 0.75)

Redundancy group with 34 pads, tolerates 2 ESD failures and 2 mechanical failures: rxdata0 rxdata1 rxdata2 ... rxdata31 34 2 2 (Results in esd_criticality = 0.941, mechanical_criticality = 0.941)

Legacy format (deprecated but still supported): <net> <esd_criticality> <mechanical_criticality> Example: txdatasb 0.8 0.7

Note:

• Criticality values range from 0 (non-critical) to 1 (critical)
• Values between 0 and 1 indicate redundancy where multiple failures can be tolerated
• Multiple nets listed on the same line form a redundancy group sharing the same failure tolerance
• Each net name should appear only once in the entire file
• See UCIe_advanced_criticality.txt for a complete example of the current format

4. 3dbv File (.3dbv):

Input file in 3dblox format. This file contains info including the die size and the path of 3dbf file.

5. 3dbf File (.3dbf):

Input file in 3dblox format. This file contains info including bump pitch and bump size.

1.risk_<config_stem>__<criticality_profile>.map

The risk map of the interface in a text format. Each line corresponds to a pad and contains the x and y coordinates of the pad, followed by the failure probabilities of different failure mechanisms. The __<config_stem>__<criticality_profile> suffix distinguishes baseline and each pessimistic case.

1a._risk_map_<config_stem>__<criticality_profile>.png

Per-mechanism pad risk maps are written by default for pad_risk_map_calculator.py. --plot only controls the extra interactive plots shown during modeling.

2.assembly_yield_summary__<config_stem>__<criticality_profile>.txt

The simulation summary in a text format. It includes:

• simulation settings
• runtime information
• overall assembly yield
• per-interface yield

3.assembly_yield_per_interface__<config_stem>__<criticality_profile>.txt

The simulated assembly yield of each interface in a text format. Each line corresponds to an interface and contains the interface name and the simulated yield.

4.assembly_fail_map_per_interface_dict__<config_stem>__<criticality_profile>.npz

The average failure count (across all simulation samples) of each pad in a pad map format for all failure mechanisms. This file is only written when both --verbose and --save-failure-maps are enabled.

5.assembly_fail_vec_per_interface_dict__<config_stem>__<criticality_profile>.npz

The failure vector of the survival scenario of each die samples for all failure mechanisms. This file is written in verbose simulation mode even when failure-map PNG/NPZ saving is disabled.

Example: die A, B, C, D, and E are simulated. A, B and D pass, and C and E fail. The failure vector of this failure mechanism is : 0, 0, 1, 0, 1.

6.simulation_failure_map_<config_stem><criticality_profile>.png

Per-interface simulation failure heatmaps for:

• overlay
• particle
• mechanical
• ESD
• overall

These PNGs are only written when --save-failure-maps is enabled for simulator_main.py.

Four helper scripts are provided to quickly generate starter files for testing:

• assign_bump_names.py: Assign net names and port names to bump maps to raw bump maps.
• generate_criticality.py: Generate criticality files from bump maps.

To be continued...