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

bug fixes

Browse Source
This commit is contained in:
Manikandan Ananth
2021-06-02 10:08:25 -07:00
parent ad9486684f
commit 47ebfccbec
2 changed files with 17 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
include/cutlass/epilogue/thread/linear_combination_clamp.h
View File

@ -223,11 +223,14 @@ public:
intermediate = mul_accumulator(alpha_, converted_accumulator); // D = alpha * Accum
/// Clamping constant value
ElementCompute const kClamp =
ElementCompute((1U << (sizeof_bits<ElementOutput>::value - 1)) - 1);
ElementCompute const kClampMax =
ElementCompute(platform::numeric_limits<ElementOutput>::max());
intermediate = max_accumulator(intermediate, -kClamp - ElementCompute(1));
intermediate = min_accumulator(intermediate, kClamp);
ElementCompute const kClampMin =
ElementCompute(platform::numeric_limits<ElementOutput>::lowest());
intermediate = max_accumulator(intermediate, kClampMin);
intermediate = min_accumulator(intermediate, kClampMax);
// Convert to destination numeric type
NumericArrayConverter<ElementOutput, ElementCompute, kCount, Round> destination_converter;