Top all-time · #8802
Colfax Research
Company blog · research.colfax-intl.comprogramminghardware
Colfax Research is the company blog at research.colfax-intl.com, writing about programming and hardware. Since 2024, 9 different Hacker News readers have linked it 22 times, most often “FlashAttention-4: Algorithm and Kernel Pipelining”. It is still writing: the last post was on 2026-09-07, and you can follow it by feed.
Too few sampled comments (2) to say how often HN recommends it.?
Why HN links it
HN readers link Colfax Research's GPU tutorials, such as writing a fast matrix multiplication from scratch with tensor cores and a matrix transpose in Mojo, as technical references on kernel optimization.
Summarised by Claude from a sample of 2 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Most-linked posts?
- FlashAttention-4: Algorithm and Kernel Pipelining2 people · 7 mentions · first linked 2026-03-05
- Categorical Foundations for Cute Layouts2 people · 2 mentions · first linked 2025-09-22
- Cutlass Tutorial: Writing GEMM Kernels Using Tensor Memory for Blackwell GPUs2 people · 2 mentions · first linked 2025-04-19
What Hacker News says?
thanks so much for reading!! >I think around that microarchitecture you want at least a two-deep pipeline where global memory loads to shared memory two iterations head I agree, if I were to spend more time on this, I would add another layer of pipelining like you say, and also tweak it so that each threadblock is persistent on a particular SM, and processes multiple tiles. This allows you to hide the write latency…
Comment linking /cutlass-tutorial-writing-gemm-kernels-using-tensor-memory-for-nvidia-blackwell-gpus/
on “How to Write a Fast Matrix Multiplication from Scratch with Tensor Cores (2024)”Submitted Categorical Foundations for Cute Layouts as a story · 39 points
HN discussionMatrix transpose is a canonical example of a memory bound operation and often used to showcase optimization in a particular programming language or library. See for example the cutlass matrix transpose tutorial from Jay Shah of flash attention 3 paper: https://research.colfax-intl.com/tutorial-matrix-transpose-i...
Comment linking /tutorial-matrix-transpose-in-cutlass/
on “Highly efficient matrix transpose in Mojo”Submitted FlashAttention-4: Algorithm and Kernel Pipelining as a story · 2 points
HN discussionSubmitted FlashAttention-4: Algorithm and Kernel Pipelining Co-Design as a story · 2 points
HN discussionSubmitted Cutlass Tutorial: Efficient GEMM Kernel Designs with Pipelining as a story · 2 points
HN discussionSubmitted Cutlass Tutorial: Sub-Byte GEMM on Nvidia Blackwell GPUs as a story · 2 points
HN discussionSubmitted GEMM with Thread Block Clusters on Nvidia Blackwell GPUs as a story · 2 points
HN discussionSubmitted Cutlass Tutorial: Writing GEMM Kernels Using Tensor Memory for Blackwell GPUs as a story · 2 points
HN discussionSubmitted DeepSeek-R1 and FP8 Mixed-Precision Training as a story · 2 points
HN discussionSubmitted DeepSeek-R1 and FP8 Mixed-Precision Training as a story · 2 points
HN discussionSubmitted Optimizing an NVFP4 Blockscaled GEMM on RTX Pro 6000 Blackwell GPU (SM120) as a story · 1 points
HN discussionSubmitted FlashAttention-4: Algorithm and Kernel Pipelining Co-Design as a story · 1 points
HN discussionSubmitted FlashAttention-4: Algorithm and Kernel Pipelining Co-Design as a story · 1 points
HN discussionSubmitted FlashAttention-4: Algorithm and Kernel Pipelining as a story · 1 points
HN discussionSubmitted FlashAttention-4: Algorithm and Kernel Pipelining Co-Design as a story · 1 points
HN discussionSubmitted NVFP4 Blockscaled GEMM on NVIDIA RTX Pro Blackwell GPUs (SM12x) as a story · 1 points
HN discussionSubmitted Dynamic Persistent Tile Scheduling w/ Cluster Launch Control (CLC) on Blackwell as a story · 1 points
HN discussionSubmitted FlashAttention-4 as a story · 1 points
HN discussionSubmitted Categorical Foundations for CuTe Layouts as a story · 1 points
HN discussionSubmitted DeepSeek-R1 and FP8 Mixed-Precision Training as a story · 1 points
HN discussionSubmitted Cutlass Tutorial: Fast Matrix-Multiplication with Wgmma on Nvidia Hopper GPUs as a story · 1 points
HN discussion
