---
title: "Taichi (programming language)"
type: "entity"
entity_kind: "concept"
status: "hub"
canonical_name: "Taichi"
aliases: ["Taichi-Lang","Taichi graphics library"]
first_seen: "2026-10-09T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["DiffTaichi","differentiable Material Point Method","ChainQueen","reverse-mode automatic differentiation","Yuanming Hu"]
seek_code_commit: "d40ee39"
---


Open-source domain-specific language and library for high-performance computer graphics and physical simulation, created by Yuanming Hu and collaborators. Matters to this vault as the technical bridge connecting its differentiable-MPM citation cluster: [[claim-chainqueen-differentiates-mpm-via-linnainmaas-reverse-mode-technique|ChainQueen]]'s own footnote credits it ("Based [on] the Taichi [30] open source computer graphics library"), and [[claim-difftaichi-2020-names-griewank-and-wengert-but-not-linnainmaa-or-speelpenning|DiffTaichi]] (Hu et al., ICLR 2020) is its differentiable-programming extension — the one paper in this cluster whose authors, building an AD compiler pass rather than calling a library, cite [[entity-andreas-griewank|Andreas Griewank]] by name where ChainQueen and Bolliger et al. cite only software.

- 2026-10-13: Taichi's citation in ChainQueen is now read as one leg of a three-way split in ChainQueen's own citation practice — formal citation for claimed mathematical lineage (Sulsky/Schreyer) and for Taichi as the author's own prior-authored software, versus no citation at all for third-party infrastructure (TensorFlow) — making Taichi the vault's concrete example of the "own-software gets credited" half of the [[entity-research-software-citation-gap|research software citation gap]] ([[claim-chainqueen-citation-practice-splits-by-claimed-lineage-vs-third-party-infrastructure]]).

## References
- [[claim-difftaichi-2020-names-griewank-and-wengert-but-not-linnainmaa-or-speelpenning]]
- [[claim-chainqueen-differentiates-mpm-via-linnainmaas-reverse-mode-technique]] — ChainQueen's credited dependency
- [[entity-differentiable-material-point-method]]
- [[claim-chainqueen-citation-practice-splits-by-claimed-lineage-vs-third-party-infrastructure]]
- Further lead, not yet checked: PlasticineLab (Huang, Hu, Du, Zhou, Su, Tenenbaum, Gan, 2021) also builds on Taichi and is cited by Bolliger et al. as prior differentiable-MPM work
- Captures: 10-inbox/raw/2026-10-05-does-chainqueen-2019-or-bolliger-et-al-2025.md, 10-inbox/raw/2026-10-11-bipartite-computing-mathematics-chainqueens-reference-list-traces-mpms-origin-directly.md