TileLang
A hardware-neutral programming language originating from Peking University, designed to describe AI training operators. It supports compiling unified computation logic to different hardware instruction sets, with the goal of reducing dependence on the CUDA ecosystem.
Core Facts
TileLang has been validated and proven mature on the NVIDIA platform, and carries the implementation of most operators in the training of the DeepSeek V4 series models.
DeepSeek open-sourced its trained model weights, training data recipe, and technical report.
DeepSeek has open-sourced infrastructure components for the Ascend platform, covering the TileLang compiler, compute library, and communication library.
Timeline (last 90 days)
The author predicts that in the coming years, AI low-level languages such as TileLang, Triton, Mojo, AscendC, and CUDA will enter a phase of intense competition.
TileLang has made dedicated optimizations for the operator requirements of the large-model era, such as Flash Attention variants and MoE routing.
Compared to Triton, TileLang operates at a more explicit level of abstraction, emphasizing exposing blocking strategies explicitly to developers in order to preserve fine-grained control space beyond automatic optimization.
TileLang's most direct application scenarios are concentrated on core deep learning operators such as FlashAttention, GEMM (General Matrix Multiplication), and quantization and dequantization.
TileLang targets GPUs, CPUs, and various accelerators simultaneously, supporting operator logic reuse across hardware platforms.
TileLang's GitHub repository address is tile-ai/tilelang.
TileLang is a domain-specific language (DSL) that aims to simplify high-performance kernel development targeting GPUs, CPUs, and various accelerators, using Python as the development language and tile-based blocking as its core abstraction.
TileLang gained approximately 800 new stars in a single week.
TileLang uses Python-style syntax to express computation logic and data movement, which is then mapped to Ascend hardware through a compilation toolchain.
The TileLang development workflow for Ascend is roughly divided into three steps: expressing computation logic and data movement in TileLang, compiling via the toolchain to generate code targeting the Ascend 950, and running and benchmarking on the hardware.
23 more timeline events
All Facts (20)
TileLang has been validated and proven mature on the NVIDIA platform, and carries the implementation of most operators in the training of the DeepSeek V4 series models.
65%VerifiedDeepSeek open-sourced its trained model weights, training data recipe, and technical report.
65%VerifiedDeepSeek has open-sourced infrastructure components for the Ascend platform, covering the TileLang compiler, compute library, and communication library.
70%VerifiedOn September 30, 2025, DeepSeek open-sourced a batch of underlying infrastructure components, including operators, a programming language, compute libraries, and communication libraries
65%UnverifiedCUDA Tile并非要取代现有的CUDA编程方式，而是在其基础上提供更高层次的选择
90%UnverifiedTileLang's GitHub repository address is tile-ai/tilelang.
50%UnverifiedTileLang gained approximately 800 new stars in a single week.
50%UnverifiedTileLang is a domain-specific language (DSL) that aims to simplify high-performance kernel development targeting GPUs, CPUs, and various accelerators, using Python as the development language and tile-based blocking as its core abstraction.
50%UnverifiedTileLang has made dedicated optimizations for the operator requirements of the large-model era, such as Flash Attention variants and MoE routing.
50%UnverifiedCompared to Triton, TileLang operates at a more explicit level of abstraction, emphasizing exposing blocking strategies explicitly to developers in order to preserve fine-grained control space beyond automatic optimization.
50%UnverifiedTileLang's most direct application scenarios are concentrated on core deep learning operators such as FlashAttention, GEMM (General Matrix Multiplication), and quantization and dequantization.
50%UnverifiedTileLang targets GPUs, CPUs, and various accelerators simultaneously, supporting operator logic reuse across hardware platforms.
50%UnverifiedTileLang is a compiled DSL whose code goes through stages including frontend parsing, intermediate representation optimization, and backend code generation, ultimately producing low-level code that can run on specific hardware.
50%UnverifiedThe TileLang development workflow for Ascend is roughly divided into three steps: expressing computation logic and data movement in TileLang, compiling via the toolchain to generate code targeting the Ascend 950, and running and benchmarking on the hardware.
50%UnverifiedTileLang's support for the Ascend 950 makes it possible to rapidly validate new operator ideas and test performance on Ascend hardware.
50%UnverifiedThe supporting resources developed around TileLang for Ascend include the TileLang main repository, the TileKernels operator library, an Ascend-oriented operator library, LearningHub development courses, and AI-assisted development tools.
50%UnverifiedTileLang is positioned similarly to Triton in the CUDA ecosystem or TVM/TIR in AI compilers, seeking to strike a balance between development efficiency and generating high-performance code.
50%UnverifiedTileLang's compilation toolchain automatically translates high-level computation descriptions into corresponding TBE primitives or AscendC code, and handles the timing and synchronization of data movement.
50%UnverifiedTileLang uses Python-style syntax to express computation logic and data movement, which is then mapped to Ascend hardware through a compilation toolchain.
50%UnverifiedTileLang belongs to the Tile-level programming abstraction, allowing developers to describe computation logic in units of data blocks (Tiles), with the compiler automatically handling memory transfer, pipeline scheduling, and instruction mapping.
50%