CUTLASS 2.3.0

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.

* Update splitk_gemm.cu

* Update gemm_bias_relu.cu

* Update mma_sm75.h

* Updated documentation of fused GEMM example and removed UNITY BUILD batch size. The default batch size when unity build is enabled tends to be favorable.

- Updated mma_sm80.h to avoid perf penalty due to reinterpret_cast<>.
- Enhancement to CUTLASS Utility Library's HostTensorPlanarComplex template to support copy-in and copy-out
- Added test_examples target to build and test all CUTLASS examples
- Minor edits to documentation to point to GTC 2020 webinar

Adds support for NVIDIA Ampere Architecture features. CUDA 11 Toolkit recommended.

#70 only updates the documentation. This commit reflects this bump in python version to the CMake configuration as well.

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

* Improved formatting, clarity, and content of several documents.

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.

* Fixed performance defect with indirect access to pointer array for Volta TensorCores TN arrangement.

* Updated patch version and changelog.

* Updated patch version and changelog.

* Added link to changelog in readme.

* Fixed markdown link

Make CUTLASS compileable with Clang.

Requires a recent clang build (r359248 or newer).

Enable compilation with clang with these options:
cmake -DCUDA_COMPILER=clang -DCMAKE_CXX_COMPILER=/path/to/clang++

CUTLASS 1.3.1 patch resolves failing text with NVRTC.

CUTLASS 1.3 Release
- Efficient GEMM kernel targeting Volta Tensor Cores via mma.sync instruction added in CUDA 10.1.

Removed patch version from README.

Resolved issue for incorrect SGEMM on Maxwell architecture.

CUTLASS 1.2

Fixed cutlass_utilities example. 

CUDA 9 lacks host-side conversions from float=>half. Instead, we must reinterpret_cast<> from cutlass::half_t => half.

Fixed typeo

Fixed typeo

Clarification to README

Updated CUTLASS.md

Cutlass 1.1