Commit ae59a3b1 authored by Shucai Xiao's avatar Shucai Xiao
Browse files

remove unnecessary data copy related to rocblas api call

parent 9f06859b
...@@ -916,15 +916,15 @@ struct find_gemm_add ...@@ -916,15 +916,15 @@ struct find_gemm_add
auto inputs = gemm_ins->inputs(); auto inputs = gemm_ins->inputs();
inputs.pop_back(); inputs.pop_back();
auto copy_ins = c_ins; // auto copy_ins = c_ins;
// Insert copy // Insert copy
if(ins == p.end() or c_ins->outputs().size() > 1 or c_ins->inputs().empty()) // if(ins == p.end() or c_ins->outputs().size() > 1 or c_ins->inputs().empty())
{ // {
copy_ins = p.insert_instruction(ins, hip_copy{}, c_ins, ins->inputs().back()); // copy_ins = p.insert_instruction(ins, hip_copy{}, c_ins, ins->inputs().back());
} // }
inputs.push_back(copy_ins); inputs.push_back(c_ins);
inputs.push_back(copy_ins); inputs.push_back(gemm_ins->inputs().back());
gemm.beta = 1; gemm.beta = 1;
p.replace_instruction(ins, gemm, inputs); p.replace_instruction(ins, gemm, inputs);
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment