Commits

Commit history

1. Commits on Jan 7, 2026

   • flashattention e4m3 kernel & benchmark, potential implementation plans for vulkan, impl of device management & instance management for vulkan

     SuriyaaMM

     committed
     055b3ecView commit details

     Copy full SHA for 055b3ec

     Browse repository at this point
2. Commits on Dec 18, 2025

   • flash attention e5m2 kernel & benchmarks

     SuriyaaMM

     committed
     e9190e4View commit details

     Copy full SHA for e9190e4

     Browse repository at this point
3. Commits on Dec 14, 2025

   • e5m2 tiled attention & ncu benchmarks for gemv and attention kernels

     SuriyaaMM

     committed
     862f3b0View commit details

     Copy full SHA for 862f3b0

     Browse repository at this point

   • attention kernel for e5m2 & benchkmark

     SuriyaaMM

     committed
     32995f1View commit details

     Copy full SHA for 32995f1

     Browse repository at this point
4. Commits on Dec 12, 2025

   • implementation of FP8 E4M3 & E5M2 support for gemv and dot (partial), migrated to ml_dtypes for packing FP8 types

     SuriyaaMM

     committed
     0d7bdb7View commit details

     Copy full SHA for 0d7bdb7

     Browse repository at this point

   • cleared ambiguities from readme

     SuriyaaMM

     committed
     bd0a581View commit details

     Copy full SHA for bd0a581

     Browse repository at this point

   • partial implementation of gemv kernel for fp8 packing

     SuriyaaMM

     committed
     9bf3ef2View commit details

     Copy full SHA for 9bf3ef2

     Browse repository at this point
5. Commits on Dec 11, 2025

   • fp8 packing, unpacking & dot-product implementation

     SuriyaaMM

     committed
     4e9cf41View commit details

     Copy full SHA for 4e9cf41

     Browse repository at this point

   • small modif in readme

     SuriyaaMM

     committed
     79fbd80View commit details

     Copy full SHA for 79fbd80

     Browse repository at this point

   • gpu optimised impl of dot product for packed fp16s

     SuriyaaMM

     committed
     8b92ce4View commit details

     Copy full SHA for 8b92ce4

     Browse repository at this point

   • updated unpacking, packing & dotproduct support for FP32 -> FP16 packing

     SuriyaaMM

     committed
     0d9442aView commit details

     Copy full SHA for 0d9442a

     Browse repository at this point

   • repo init

     SuriyaaMM

     committed
     18b59f6View commit details

     Copy full SHA for 18b59f6

     Browse repository at this point

   • Initial commit

     SuriyaaMM

     authored
     bafa481View commit details

     Copy full SHA for bafa481

     Browse repository at this point