* [WIP] GEMM StreamK w/ Fused Epilogue
* Adds Gemm Streamk with Fused Epilogue kernel level struct.
* Mostly based on Gemm with Fused Epilogue,
* Requires a new epilogue
* Work in progress
* [WIP] StreamK support for GemmUniversalWithBroadcast
* Just based off of how StreamK is allowed in GemmUniversal
* Untested and a work in progress
* Minor fixes
* [WIP] It compiles!
It is almost certainly incorrect, but we're past getting the templates
to match, so checkpointing.
* Correction to reference kernel
* Fix typo
* Added MSE measurement
* Switch back to reference kernel + host for loop
Still WIP. Now we're getting even a larger MSE, but it's both on
basic Split-K and Stream-K.
* Fix typos
* Fix broadcast vector + requested changes
* Comment typo
* Small int option and more
* Fix incorrect condition on source needed
* Requested changes
* I think I got it?
* Bias vector should be stride 0
* Two source added!
* Typos
* Merge examples
* Bring back vector row offset
Just to ensure consistency with universal gemm with fused epilogue
* Base arguments and params structs for StreamK
* StreamK epilogue with broadcast now inherits the original
* undo params_streamk_base.h
---------
Co-authored-by: Ali Hassani <ahassanijr@gmail.com>
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
* add two missing files
* fix bunch of bugs of gemm-reducek fusion and add a device interface
* small changes
Co-authored-by: Haicheng Wu <haichengw@nvidia.com>
* Fix the build of cutlass/gemm/device/gemm_array.h and add a demo for GemmArray
* Add a reference to GemmArray to the docs
Co-authored-by: Ivan Komarov <dfyz@yandex-team.ru>
CUTLASS 2.3 adds GEMMs targeting Sparse Tensor Cores on the NVIDIA Ampere Architecture, fast SGEMM, and small matrix classes, bug fixes, and performance enhancements.
CUTLASS 2.1 contributes:
- BLAS-style host-side API added to CUTLASS Library
- Planar Complex GEMM kernels targeting Volta and Turing Tensor Cores
- Minor enhancements and bug fixes
CUTLASS 2.0
Substantially refactored for
- Better performance, particularly for native Turing Tensor Cores
- Robust and durable templates spanning the design space
- Encapsulated functionality embodying modern C++11 programming techniques
- Optimized containers and data types for efficient, generic, portable device code
Updates to:
- Quick start guide
- Documentation
- Utilities
- CUTLASS Profiler
Native Turing Tensor Cores
- Efficient GEMM kernels targeting Turing Tensor Cores
- Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands
Coverage of existing CUTLASS functionality:
- GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs
- Volta Tensor Cores through native mma.sync and through WMMA API
- Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions
- Batched GEMM operations
- Complex-valued GEMMs
Note: this commit and all that follow require a host compiler supporting C++11 or greater.
