From 83f7531364cb2e489f27b44bbd9436e56f6a02d3 Mon Sep 17 00:00:00 2001 From: rocking Date: Wed, 18 May 2022 08:38:30 +0800 Subject: [PATCH] Let it be the multiple number of CU --- .../tensor_operation/gpu/device/device_binary_elementwise.hpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/include/ck/tensor_operation/gpu/device/device_binary_elementwise.hpp b/include/ck/tensor_operation/gpu/device/device_binary_elementwise.hpp index a818009f9a..198bf42ce7 100644 --- a/include/ck/tensor_operation/gpu/device/device_binary_elementwise.hpp +++ b/include/ck/tensor_operation/gpu/device/device_binary_elementwise.hpp @@ -86,7 +86,7 @@ struct DeviceBinaryElementwise : public BaseOperator p_c_(p_c), functor_(functor), threadPerBlock_(threadPerBlock), - gridSize_(128) // FIXME - Calculate the grid size by number of CU in the future + gridSize_(120) // FIXME - Calculate the grid size by number of CU in the future { a_grid_desc_m0_ = MakeDescriptor_M0(shape, stride_a, gridSize_, threadPerBlock_); b_grid_desc_m0_ = MakeDescriptor_M0(shape, stride_b, gridSize_, threadPerBlock_);