gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

removed shared conflicts for 8x32x4 and 32x8x4 configurations

Browse Source
This commit is contained in:
Philippe Tillet
2019-06-13 17:51:54 -07:00
parent 21a9b92c87
commit 36e3667a9a
3 changed files with 11 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
examples/python/tensorflow/dot.cpp
View File

@@ -119,7 +119,7 @@ class BlockSparseGemmOp : public OpKernel {
// just-in-time compile source-code
// jit.autotune("matmul", src, benchmark);
// jit.add_module("matmul", src, {4, 2, 8, 4, 2, 32, 1, 4, 1, 1, 8, 8, 8, 1});
jit.add_module("matmul", src, {16, 4, 128, 16, 4, 128, 2, 2, 2, 2, 8, 32, 8, 1});
jit.add_module("matmul", src, {16, 4, 128, 16, 4, 128, 1, 4, 2, 2, 8, 32, 8, 1});
// jit.add_module("matmul", src, {8, 8, 128, 16, 8, 128, 2, 2, 2, 2, 16, 32, 8, 1 });
triton::driver::kernel* kernel = jit.get_function("matmul");
triton::jit::launch_information info = jit.get_launch_info("matmul");