microsoft / TileFusion
TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing. By providing a higher-level interface, algorithm developers can design hardware-aware algorithms without dealing with low-level hardware complexities.
☆67Updated this week
Alternatives and similar repositories for TileFusion:
Users that are interested in TileFusion are comparing it to the libraries listed below
- ☆24Updated 2 months ago
- GPTQ inference TVM kernel☆39Updated 10 months ago
- ☆19Updated 5 months ago
- Framework to reduce autotune overhead to zero for well known deployments.☆62Updated 2 weeks ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆105Updated 6 months ago
- ☆73Updated 4 months ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆59Updated last week
- High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.☆100Updated 8 months ago
- Implement Flash Attention using Cute.☆71Updated 2 months ago
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆176Updated last month
- Artifacts of EVT ASPLOS'24☆23Updated last year
- ☆42Updated last month
- extensible collectives library in triton☆83Updated 5 months ago
- ☆79Updated this week
- Several optimization methods of half-precision general matrix vector multiplication (HGEMV) using CUDA core.☆57Updated 6 months ago
- llama INT4 cuda inference with AWQ☆53Updated last month
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆89Updated 2 weeks ago
- ☆38Updated 9 months ago
- ☆87Updated 6 months ago
- An extention of TVMScript to write simple and high performance GPU kernels with tensorcore.☆51Updated 7 months ago
- ☆53Updated 2 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆103Updated this week
- FlexFlow Serve: Low-Latency, High-Performance LLM Serving☆25Updated this week
- play gemm with tvm☆89Updated last year
- PyTorch bindings for CUTLASS grouped GEMM.☆70Updated 4 months ago
- Tacker: Tensor-CUDA Core Kernel Fusion for Improving the GPU Utilization while Ensuring QoS☆19Updated last month
- ☆25Updated last week
- Benchmark tests supporting the TiledCUDA library.☆15Updated 3 months ago