Popular repositories Loading

1. StanleyCWestII.github.io StanleyCWestII.github.io Public

2. riscv-pipelined-core riscv-pipelined-core Public

   5-stage pipelined RV32I-subset core in SystemVerilog, running on a Nexys A7. 64-entry 2-bit branch predictor, 2-way set-associative D-cache, UART echo and VGA pattern generator over MMIO.

   SystemVerilog

3. cuda-tiled-sgemm cuda-tiled-sgemm Public

   Single-precision dense matrix multiplication written from scratch in CUDA, reaching 80% of cuBLAS on an RTX 4090 through 128x128 block tiles, 8x8 register tiles, and double-buffered shared memory.

   Cuda

4. flashattention-cuda flashattention-cuda Public

   FlashAttention forward pass written from scratch in CUDA: tiled keys and values, online softmax, and no materialized T x T attention matrix.

   Cuda