cutlass 1.3.1 (#46)
CUTLASS 1.3.1 patch resolves failing text with NVRTC.
This commit is contained in:
@@ -2,7 +2,7 @@
|
||||
|
||||
# CUTLASS 1.3
|
||||
|
||||
_CUTLASS 1.3.0 - March 2019_
|
||||
_CUTLASS 1.3.1 - April 2019_
|
||||
|
||||
CUTLASS is a collection of CUDA C++ template abstractions for implementing
|
||||
high-performance matrix-multiplication (GEMM) at all levels and scales within CUDA.
|
||||
@@ -28,6 +28,10 @@ CUTLASS 1.3 is described in the [CUTLASS Documentation](CUTLASS.md) and the acco
|
||||
We describe the structure of an efficient GEMM in our talk at the
|
||||
[GPU Technology Conference 2018](http://on-demand.gputechconf.com/gtc/2018/presentation/s8854-cutlass-software-primitives-for-dense-linear-algebra-at-all-levels-and-scales-within-cuda.pdf).
|
||||
|
||||
# What's New in CUTLASS 1.3.1
|
||||
_April 2019_
|
||||
* CUTLASS 1.3.1 corrected NVRTC unit tests..
|
||||
|
||||
# What's New in CUTLASS 1.3
|
||||
_March 2019_
|
||||
* CUTLASS 1.3 includes an efficient GEMM implementation with the `mma.sync` instruction added in CUDA 10.1.
|
||||
|
||||
Reference in New Issue
Block a user