"src/emit.cpp" did not exist on "8fd372b0dba0c3a71815ad177e55dfde9eee36c3"
Commit a53d4d9e authored by Adam Osewski's avatar Adam Osewski
Browse files

Small refinements.

parent 98def248
...@@ -227,7 +227,6 @@ bool profile_ggemm_multid_splitk(int do_verification, ...@@ -227,7 +227,6 @@ bool profile_ggemm_multid_splitk(int do_verification,
auto argument_ptr = gemm_ptr->MakeArgumentPointer( auto argument_ptr = gemm_ptr->MakeArgumentPointer(
p_a, p_b, p_ds, p_c, gemm_descs, a_element_op, b_element_op, c_element_op); p_a, p_b, p_ds, p_c, gemm_descs, a_element_op, b_element_op, c_element_op);
auto invoker_ptr = gemm_ptr->MakeInvokerPointer(); auto invoker_ptr = gemm_ptr->MakeInvokerPointer();
DeviceMem gemm_arg_dev_mem(gptr->GetDeviceKernelArgSize(argument_ptr.get())); DeviceMem gemm_arg_dev_mem(gptr->GetDeviceKernelArgSize(argument_ptr.get()));
...@@ -263,7 +262,7 @@ bool profile_ggemm_multid_splitk(int do_verification, ...@@ -263,7 +262,7 @@ bool profile_ggemm_multid_splitk(int do_verification,
for(std::size_t i = 0; i < gemm_descs.size(); i++) for(std::size_t i = 0; i < gemm_descs.size(); i++)
c_device_buf[i]->SetZero(); c_device_buf[i]->SetZero();
invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, false}); invoker_ptr->Run(argument_ptr.get(), StreamConfig{nullptr, false, 1});
if(do_verification) if(do_verification)
{ {
...@@ -308,12 +307,10 @@ bool profile_ggemm_multid_splitk(int do_verification, ...@@ -308,12 +307,10 @@ bool profile_ggemm_multid_splitk(int do_verification,
<< std::endl; << std::endl;
pass = pass && instance_pass; pass = pass && instance_pass;
// std::cout << ">>>>>CPU verification end!" << std::endl;
} }
if(time_kernel) if(time_kernel)
{ {
// std::cout << ">>>>>GPU time profiling start!" << std::endl;
float avg_time = invoker_ptr->Run( float avg_time = invoker_ptr->Run(
argument_ptr.get(), argument_ptr.get(),
StreamConfig{nullptr, time_kernel, 0, warmup_iter, kernel_iter}); StreamConfig{nullptr, time_kernel, 0, warmup_iter, kernel_iter});
...@@ -342,7 +339,6 @@ bool profile_ggemm_multid_splitk(int do_verification, ...@@ -342,7 +339,6 @@ bool profile_ggemm_multid_splitk(int do_verification,
best_gb_per_sec = gb_per_sec; best_gb_per_sec = gb_per_sec;
best_kbatch = kbatch_curr; best_kbatch = kbatch_curr;
} }
// std::cout << ">>>>>GPU time profiling end!" << std::endl;
} }
} }
else else
......
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