3 Pull requests merged by 2 people

• chore(CI): Add issue templates.

  #124 merged May 17, 2025

• Refactor (kernels): Enhance JIT-compiled Flash Attention kernel implementations.

  #123 merged May 16, 2025

• feat(jit): Implement JIT Compilation for GEMM, Replacing Hardcoded Dispatch

  #121 merged May 12, 2025

1 Pull request opened by 1 person

• feat(bench): Verify pipelined GEMM performance.

  #122 opened May 15, 2025

2 Issues closed by 1 person

• Implement JIT Compiler for CUDA Kernels with Runtime NVCC Compilation

  #113 closed May 10, 2025

• Remove duplicated examples directory that overlaps with kernels

  #108 closed May 10, 2025

2 Issues opened by 1 person

• [Performance]: Testing the Performance of Pipelined GEMM

  #120 opened May 10, 2025

• [Enhancement]: JIT Compilation of Template Kernels Instead of Hardcoded Dispatch

  #119 opened May 10, 2025

4 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• [Enhancement]: Convert `swizzling` into a layout to decouple it from `SharedTile`.

  #52 commented on May 10, 2025 • 0 new comments

• [Enhancement]: Compile-time static calculation of `SwizzleOffset`.

  #88 commented on May 10, 2025 • 0 new comments

• [Feature]: Add CI for tests and examples

  #27 commented on May 10, 2025 • 0 new comments

• [Feature]: Add FP8 GEMM.

  #118 commented on May 11, 2025 • 0 new comments