squall/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

Merge pull request #9 from NVIDIA/cutlass_v1.0_rel

Browse Source 
Updated URL to Doxygen and modified usage statement
This commit is contained in:
Andrew Kerr 2018-05-17 11:12:37 -07:00 committed by GitHub
commit 68aaee8773
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
README.md
View File

@ -26,7 +26,7 @@ post. We have decomposed the structure of the GEMM computation into deeper, stru
primitives for loading data, computing predicate masks, streaming data at each level of primitives for loading data, computing predicate masks, streaming data at each level of
the GEMM hierarchy, and updating the output matrix. the GEMM hierarchy, and updating the output matrix.
CUTLASS 1.0 is described in the [Doxygen documentation](https://github.com/NVIDIA/cutlass/docs) CUTLASS 1.0 is described in the [Doxygen documentation](https://nvidia.github.io/cutlass)
and 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). and 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).
# Performance # Performance
@ -169,7 +169,7 @@ Program usage:
--m=<height>[:max height[:step]] Height of GEMM problem (number of rows of C). May specify a range with optional step size. --m=<height>[:max height[:step]] Height of GEMM problem (number of rows of C). May specify a range with optional step size.
--n=<width>[:max width[:step]] Width of GEMM problem (number of columns of C). May specify a range with optional step size. --n=<width>[:max width[:step]] Width of GEMM problem (number of columns of C). May specify a range with optional step size.
--k=<depth>[:max depth[:step]] Size of inner dimension of A and B. May specify a range with optional step size. --k=<depth>[:max depth[:step]] Size of inner dimension of A and B. May specify a range with optional step size.
--kernels=<{s|d|h|i|wmma}gemm_{nn,nt,tn,tt}> Select GEMM datatype and layout to use for tests --kernels=<{s|d|h|i|wmma}_gemm_{nn,nt,tn,tt}> Select GEMM datatype and layout to use for tests
--peak=<bool> If true, only reports peak performance per kernel after profiling specified problem space. --peak=<bool> If true, only reports peak performance per kernel after profiling specified problem space.
--save_workspace={*never,incorrect,always} Specifies when to save the GEMM inputs and results to the filesystem. --save_workspace={*never,incorrect,always} Specifies when to save the GEMM inputs and results to the filesystem.
--seed=<seed> Random seed used by the random number generator in initializing input matrices. --seed=<seed> Random seed used by the random number generator in initializing input matrices.

2
tools/test/perf/testbench_options.h
View File

@ -546,7 +546,7 @@ struct TestbenchOptions {
<< " --k=<depth>[:max depth[:step]] " << " --k=<depth>[:max depth[:step]] "
<< " Size of inner dimension of A and B. May specify a range with optional step size.\n" << " Size of inner dimension of A and B. May specify a range with optional step size.\n"
<< " --kernels=<{s|d|h|i|wmma}gemm_{nn,nt,tn,tt}> " << " --kernels=<{s|d|h|i|wmma}_gemm_{nn,nt,tn,tt}> "
<< " Select GEMM datatype and layout to use for tests\n" << " Select GEMM datatype and layout to use for tests\n"
<< " --peak=<bool> " << " --peak=<bool> "