David Yue

This is my personal paper reading list. Most of it relates to Tensor optimization, GPU data layouts, and some are actually blog posts or articles. I started this list a little late, but I hope to put every paper I've ever read on here. If you are a student interested in the high performance computing area, consider joining our reading group

Another small portion of it are papers I read during my time at Huawei.

Papers to read

• Simple and Efficient Construction of Static Single Assignment Form
• Rematerialization
• Verifying Peephole Rewriting In SSA Compiler IRs
• The Deep Learning Compiler: A Comprehensive Survey
• Toward Accelerated Stencil Computation by Adapting Tensor Core Unit on GPU
• LoRAStencil: Low-Rank Adaptation of Stencil Computation on Tensor Cores
• Efficient and precise array access analysis
• PEAK: Generating High-Performance Schedules in MLIR
• Mixed-precision iterative refinement using tensor cores on GPUs to accelerate solution of linear systems

Papers I've read

• Iterated Register Coalescing
• ConvStencil: Transform Stencil Computation to Matrix Multiplication on Tensor Cores
• Accelerating Reduction and Scan Using Tensor Core Unit
• Triton: An Intermediate Language and Compiler for Tiled Neural Network Computations
• tcFFT: Accelerating Half-Precision FFT through Tensor Cores
• Linear Layouts: Robust Code Generation of Efficient Tensor Computation Using F2
• Graphene: An IR for Optimized Tensor Computations on GPUs
• First Class Verification Dialects for MLIR
• TensorIR: An Abstraction for Automatic Tensorized Program Optimization
• LORA: LOW-RANK ADAPTATION OF LARGE LANGUAGE MODELS
• WATERMARKING IMAGES IN SELF-SUPERVISED LATENT SPACES
• Efficient Memory Management for Large Language Model Serving with Paged Attention
• Video for teaching: classroom use, instructor self-production and teachers' preferences in presentation format
• How Video Production Affects Student Engagement: An Empirical Study of MOOC Videos
• Evaluating the Effectiveness of Flipped Classrooms for Teaching CS1
• Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computation
• Hexcute: A Tile-based Programming Language with Automatic Layout and Task-Mapping Synthesis
• Insum: Sparse GPU Kernels Simplified and Optimized with Indirect Einsums
• Resource Constrained Software Pipelining