ISPD24 Contest: GPU/ML-Enhanced Large Scale Global Routing

Contest Introduction

Global routing is a critical component of the VLSI design process, exerting a substantial influence on circuit timing, power consumption, and overall routability. The efficiency of global routing is of paramount importance, as a swift and scalable approach can guide optimizations in early design stages like floor-planning and placement.

Over the past decade, GPU accelerated computing platforms have been evolving into highly versatile and programmable systems capable of delivering immense parallel computing power. Recent studies have successfully leveraged GPUs to achieve over a 10x acceleration in global routing without compromising performance. Furthermore, machine learning (ML) techniques have been integrated into the global routing process, leading to enhanced routing solution quality.

The goal of this competition is to stimulate academic research aimed at developing a GPU/ML-enhanced global router tailored for industrial-level circuits. Notably, contemporary VLSI circuits often encompass tens of millions of cells, which is a significant departure from past global routing competitions that typically dealt with scenarios involving no more than 1 million cells. Due to the limitations of current routers, hierarchical or partitioning-based methods are commonly employed to manage large circuits, albeit at the risk of sacrificing a certain degree of optimality. It is of great importance to develop a scalable global router capable of handling circuits with tens of millions of cells, as it can greatly inform optimizations in the early design stages, such as floor-planning and placement. By fostering enthusiasm and innovation within the global routing research community, this competition aims to deliver substantial reductions in global routing runtime for these expansive industrial-grade circuits, harnessing the computational power of GPUs and the potential of machine learning techniques. Simultaneously, it strives to enhance the overall quality of routing results.

Input/Output Formats and Evaluation

To enable teams from diverse backgrounds to participate, we have extracted routing resource information and netlist data from LEF and DEF files and organized them in simplified formats. Consequently, participants can approach the contest as a mathematical optimization problem within the GCell grid graph. The desired outcome is global routing solutions described within the GCell grid graph. The evaluation process is centered on several key metrics, including total wirelength, via count and routing congestion of the global routing solution, as well as the execution runtime of the global router.

Please check Introduction of the contest for more details.

Submission Guidance

Teams are required to build a Docker image on top of the provided Dockerfile. Within the Docker environment, please create a directory named “router” under the “/workspace” folder and place the global router binary/scripts in this directory (/workspace/router). We expect that the global route can accept the following command line:

./route -cap $data.cap -net $data.net -output $data.output

During the evaluation process, the Docker images will be pulled and executed on a NVIDIA platform equipped with 4 NVIDIA A100 GPUs. Specifically, we will mount a “benchmarks” folder (containing the input files) to /workspace/benchmarks and an “evaluation” folder (containing the evaluation scripts) to /workspace/evaluation. The evaluation.sh script will be executed to run the submitted global router and evaluate the generated solutions.

Please kindly archive your Docker image to a tar file (refer to https://docs.docker.com/engine/reference/commandline/save/). Subsequently, upload the tar archive to Google Drive and share the link with us. Please grant access to the Docker image for anyone with the link. The global router binary/scripts are expected to be stored in /workspace/router within the Docker image. Therefore, there is no need to upload the global router binary/scripts separately. If you have any questions regarding this, please feel free to ask. Thank you!

Routing resource limit:

• RAM: 200 GB
• CPU Cores: 8 cores

• GPUs: 4 NVIDIA A100 GPUs

  design	runtime limit/s
  Ariane133_51	30
  Ariane133_68	30
  NVDLA	30
  BlackParrot	300
  MemPool-Tile	30
  MemPool-Group	1000
  MemPool-Cluster	10000
  TeraPool-Cluster	50000

Leaderboard

• Best scores on pubic benchmarks (alpha submission, updated on Jan 25, 2024)

design	runtime /s	WL cost	via cost	overflow cost	raw score	scaled score
Ariane133_51	12	9285586	2956924	209162 * 50	22700610	22599641.13
Ariane133_68	13	9435932	2883456	158823 * 50	20260538	20176881.88
BlackParrot	120	58702634	20208976	728858 * 50	115354510	115354510
MemPool-Tile	12	8484366	3501412	70477 * 50	15509628	15548566.75
MemPool-Group	362	271521737	94674144	1778913 * 50	455141531	448353488.1
MemPool-Cluster	14327	1103106843	285442676	5801104 * 50	1678604719	1752964443

• Best scores on pubic benchmarks (beta submission, updated on Feb 15, 2024)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane133_51	9335109	3060400	10369862	22765372	4	11	22100882
Ariane133_68	9443754	2981836	7825647	20251238	8	12	20014313
BlackParrot	58347098	19740536	35450029	113537664	27	70.5	110393434
Nvdla	21345766	4630872	23933347	49909986	4	20	47592238
MemPool-Tile	8407884	3425828	3569040	15402753	3	10	14867672
MemPool-Group	262817992	76146200	70345580	409309772	81	375	391210938
MemPool-Cluster	1094650057	268335040	300359611	1663344708	478	2048	1593513066
Tera-Cluster	12190406272	1542639724	6705414754	20438460750	4310	4584	20402113329

• Scores of #1 team (Main track) on hidden benchmarks (final submission, updated on March 19, 2024)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	11931752.82	2906688	7764435.24	22602876.06	6.238	9.535	22326149.32
BlackParrot_rank	55573070.65	19738252	37798297.79	113109620.4	16.551	34.3625	110725458.3
Nvdla_rank	21472300.92	4661336	17062342.24	43195979.16	7.504	10.432	42785371.97
MemPool-Tile_rank	7541038.845	3346788	2939315.209	13827142.05	5.093	5.3315	13808883.1
MemPool-Group_rank	247299729.1	73566828	62771095.09	383637652.2	45.455	144.5285	370833031.1
MemPool-Cluster_rank	1185865082	282548572	313778179.9	1782191834	217.099	765.475	1717391261
Tera-Cluster_rank	7963801630	1690296724	2867829285	12521927639	1940.183	7716.82	12023100694

• Scores of #2 team (Main track) on hidden benchmarks (final submission)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	12051686.81	2959392	8082422.507	23093501.32	2.234	9.535	22126528.5
BlackParrot_rank	55847570.15	19505444	35633766.41	110986780.6	9.573	34.3625	106894049.8
Nvdla_rank	21840287.42	4586500	17714764.37	44141551.79	3.295	10.432	42673699.17
MemPool-Tile_rank	7601380.712	3308184	3223704.168	14133268.88	2.085	5.3315	13750400.29
MemPool-Group_rank	250445949.5	74362960	70679949.95	395488859.5	27.603	144.5285	376596696.5
MemPool-Cluster_rank	1190412465	274145704	359968884.8	1824527054	124.417	765.475	1728879082
Tera-Cluster_rank	8039402094	1488566820	3148098101	12676067016	1352.127	7716.82	12039024729

• Scores of #3 team (Main track) on hidden benchmarks (final submission)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	12021241.28	2813744	7986303.865	22821289.14	1.443	9.535	21577911.53
BlackParrot_rank	55751772.56	19182208	36844605.46	111778586	8.689	34.3625	107344166.4
Nvdla_rank	21693203.41	4455428	17146460.17	43295091.59	4.102	10.432	42129054
MemPool-Tile_rank	7578562.281	3233024	3055538.132	13867124.41	1.057	5.3315	13219650.04
MemPool-Group_rank	249498393.4	70274012	68427932.19	388200337.6	22.408	144.5285	367320841
MemPool-Cluster_rank	1189656494	270441852	335426594.9	1795524941	275.123	765.475	1742511044
Tera-Cluster_rank	7970278976	1442567228	3184651822	12597498026	1862.479	7716.82	12080803508

• Scores of #1 team (Special Honor track) on hidden benchmarks (final submission)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	12051686.81	2959392	8082422.507	23093501.32	2.234	7.535	21068226.28
BlackParrot_rank	55847570.15	19505444	35633766.41	110986780.6	9.573	34.3625	100754953.7
Nvdla_rank	21840287.42	4586500	17714764.37	44141551.79	3.295	10.216	40538541.63
MemPool-Tile_rank	7601380.712	3308184	3223704.168	14133268.88	2.085	5.015	13238489.99
MemPool-Group_rank	250445949.5	74362960	70679949.95	395488859.5	27.603	121.879	353121075.5
MemPool-Cluster_rank	1190412465	274145704	359968884.8	1824527054	124.417	712.241	1594893733
Tera-Cluster_rank	8039402094	1488566820	3148098101	12676067016	1352.127	5331.15	11421633213

• Scores of #2 team (Special Honor track) on hidden benchmarks (final submission)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	12021241.28	2813744	7986303.865	22821289.14	1.443	7.535	20100379.62
BlackParrot_rank	55751772.56	19182208	36844605.46	111778586	8.689	34.3625	100692537.1
Nvdla_rank	21693203.41	4455428	17146460.17	43295091.59	4.102	10.216	40445341.48
MemPool-Tile_rank	7578562.281	3233024	3055538.132	13867124.41	1.057	5.015	12309656.14
MemPool-Group_rank	249498393.4	70274012	68427932.19	388200337.6	22.408	121.879	340774605.5
MemPool-Cluster_rank	1189656494	270441852	335426594.9	1795524941	275.123	712.241	1672326009
Tera-Cluster_rank	7970278976	1442567228	3184651822	12597498026	1862.479	5331.15	11641837581

• Scores of #3 team (Special Honor track) on hidden benchmarks (final submission)

design	WL cost	via cost	overflow cost	raw score	runtime /s	median runtime /s	scaled score
Ariane_rank	11932251.5	2908680	7769574.844	22610506.35	2.554	7.535	20845927.05
BlackParrot_rank	55573957.76	19763752	40253224.51	115590934.3	12.84	34.3625	107382881
Nvdla_rank	21469380.39	4696756	17389649.52	43555785.91	4.07	10.216	40664270.31
MemPool-Tile_rank	7541380.951	3347328	2937890.503	13826599.45	1.85	5.015	12831966.13
MemPool-Group_rank	247335596.2	73607632	62785396.86	383728625	40.457	121.879	353203237
MemPool-Cluster_rank	1185834292	283015216	315719710.8	1784569219	225.836	712.241	1636709628
Tera-Cluster_rank	7963550667	1695387868	2886683697	12545622231	2022.139	5331.15	11668329071

Anouncement

• We released the hidden benchmarks and contest results- March 26, 2024
• We released the slides for the contest results/summary - March 19, 2024
• We released the routing resource limit - Feb 2, 2024
• We released a testcase with around 50M cells and 60M nets! - Jan 29, 2024
• We evaluated the alpha submissioms on public benchmarks and created the leader board. - Jan 22, 2024
• We updated the submission guidance. -Jan 8, 2024
• We updated the benchmark input files. And a new design (“cluster”) with around 10 million cells is released. - Jan 4, 2024
• We postponed the alpha submission date to Jan 12, 2024. - Jan 2, 2024.
• We updated the global routing solution format and evaluation metrics. Please kinldy check the updated Introduction of the contest. - Jan 2, 2024.
• We released the updated version of evaluation script. - Jan 2, 2024.
• The evaluation platform is configured with CUDA version 11.7 and driver version 515.00. We have prepared a Dockerfile (https://github.com/liangrj2014/ISPD24_contest/blob/main/Dockerfile) that will install deep learning toolkits compatible with the CUDA version on our evaluation platform. Generally, participants are welcome to modify the Dockerfile as necessary, ensuring compatibility with our evaluation platform for utilizing the GPUs on the system. - Dec 02, 2023.
• We’ve identified some bugs in our evaluation script (thanks to the participants for bringing them to our attention!). An updated version will be released shortly, addressing these bugs and significantly improving runtime speed. Stay tuned for the latest updates! - Nov 30, 2023.
• We’ve released example global routing solutions on Oct 28, 2023.
• We’ve released the evaluation scripts on Oct 28, 2023.
• To simplify the entry process for the competition, we’ve extracted routing resource information and netlist data from LEF and DEF files and stored them in simplified formats. As a result, participants can tackle the contest challenge as a mathematic optimization problem within the GCell grid graph. - Oct 28, 2023.
• Registration opens on Sep 13, 2023!
• Released the first set of benchmarks on Sep 13, 2023
• Released the docker for environment setup on Sep 13, 2023

Registration

• Please fill in this online registration form

• Registration window: Sep 13, 2023 - Dec 1, 2023

• We confirm that we’ve received the registrations forms from the following teams. Please feel free to send us an email if we overlooked your registration forms or any related information.

  ID	Team Name	Affiliation
  1	Mamta	Unknown
  2	NianNianYouYu	Fudan University
  3	Binghamton	Binghamton University
  4	NTHU-TCLAB-GR	National Tsing Hua University
  5	Alpha-Team	AlphaCHIP LLC; National Research University of Electronic Technology
  6	HumbleRoute	University of South Florida; CMR Institute of Technology
  7	automl-for-global-routing	Texas A&M University
  8	Lira	Unknown
  9	SCAW	Sun Yat-sen University; Institute of Computing, Chinese Academy of Sciences
  10	TeamSD	Unknown
  11	NTUSTGR	National Taiwan University of Science and Technology
  12	SmartRoute	University of Science and Technology of China
  13	metaRoute	Fudan University
  14	IWantNvidia	Unknown
  15	GRFirExp	University of Science and Technology of China
  16	DIAG	University of Science and Technology of China
  17	Santhosh	Unknown
  18	VCISEDA	Beijing University of Posts and Telecommunications
  19	Puipui	Nara Institute of Science and Technology
  20	Team-XJWIN	Xi’an Jiaotong University
  21	Route 99	Tsinghua University; University of Science and Technology Beijing
  22	Team California	University of California , Santa Cruz
  23	AnxietyAttack	University of Peradeniya
  24	fzu _team	Unknown
  25	Jack’s team	Unknown
  26	iEDA-iRT	Institute of Computing Technology; Chinese Academy of Sciences; Shenzhen University; Peng Cheng Laboratory
  27	GGRouter	Pohang University of Science and Technology
  28	NTUEE+	National Taiwan University
  29	color	Chinese University of Hong Kong
  30	NNU_FZUTeam	Nanjing Normal University; Fuzhou University
  31	NYCU_GR	Unknown
  32	NTUEDA	National Taiwan University
  33	FZU_MathGR	Fuzhou University
  34	RL-Route	The Chinese University of Hong Kong
  35	HustForest	Huazhong University of Science and Technology
  36	ZJ	Unknown
  37	Marvin	Indian Institute of Technology Madras
  38	Tiger	Unknown
  39	NJUFZUglobalrouter	Nanjing Normal University
  40	GoToRoute	Fuzhou University
  41	NYCU-VDA-GR	National Yang Ming Chiao Tung University
  42	Kate	Scientific research institute for system analysis of Russian Academy of Sciences
  43	SEU-Router	Southeast University
  44	Easy_Router	Unknown
  45	etuoReL	Fudan University
  46	PY-Router	Unknown
  47	PampaRouting	Universidade Federal do Rio Grande do Sul
  48	Team Ondru	Indian Institute of Technology Madras
  49	XianBaoMing	National Tsing Hua University
  50	EA	University of Calgary
  51	CUEDA	The Chinese University of Hong Kong
  52	Team Hippo	Peking University

Important Dates

• Registration Open: Sep 13, 2023
• Registration Close: Dec 1, 2023
• Alpha Submission: Jan 5, 2024 Jan 12, 2024
• Beta Submission: Feb, 2, 2024
• Final Submission: Mar, 1, 2024
• Results Anouncement: March 15, 2024

Downloads

• Final contest result - March 19, 2024.
• Introduction of the contest Note that the introduction has been updated on Jan 29, 2024.
• First set of benchmarks with Nangate45 technology node Please note that all the essential input information for global routing is contained within the .cap files and .net files located in the “Simple_inputs” folder. We also release the LEF/DEF files of the circuits just for reference. Note that the benchmarks have been updated on Jan 29, 2024.
• Evaluation Scripts Note that the evaluation scripts have been updated on Jan 29, 2024.
• Example global routing solutions Note that the example solutions have been updated on Jan 03, 2024.
• Dockerfile for environment setup Note that the Dockerfile has been updated on Dec 02, 2023.

Q&A

• Please post your questions in GitHub Issues

Contact

• Email：ispd2024contest@gmail.com

Contest Prizes

• 1st place: $1000 + one NVIDIA GPU of similar value
• 2nd place: $500 + one NVIDIA GPU of similar value
• 3rd place: $250 + one NVIDIA GPU of similar value

Sponsor

• Sponsored by NVIDIA