ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-19 04:19:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

Padded Generic Kernel Instance (#730)

Browse Source 
* Add NumReduceDim template parameter to DeviceSoftmax and Softmax client API to simplify instances collecting

* Move the generic kernel instance to be the first of the instance list for elementwise op of normalization

* Add GetGenericInstance() interface for DeviceOperationInstanceFactory class of DeviceSoftmax

* Add testing of GetGenericInstance() in client_example of Softmax

* Revert "Add testing of GetGenericInstance() in client_example of Softmax"

This reverts commit f629cd9a93.

* Revert "Add GetGenericInstance() interface for DeviceOperationInstanceFactory class of DeviceSoftmax"

This reverts commit a9f0d000eb.

* Support generic kernel instance to be the first instance returned by GetInstances() for GroupNorm

* Move generic kernel instance to separate tuple for elementwise op of normalization

* Remove un-used files for softmax instance

* Store generic kernel instance to separate tuple for softmax

* Add IsSupported checking for generic instance to client example of softmax

* Replace the get_device_normalize_from_mean_meansquare_instances() by the DeviceOperationInstanceFactory class for elementwise-normalization

* clang-format fix

* Remove int8 from softmax instances

---------

Co-authored-by: zjing14 <zhangjing14@gmail.com>

[ROCm/composable_kernel commit: 0d9118226b]
This commit is contained in:
Qianfeng
2023-06-17 12:43:11 +08:00
committed by GitHub
parent 033da551a7
commit eebebd33c6
76 changed files with 552 additions and 790 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
client_example/18_groupnorm/groupnorm_swish.cpp
View File

@@ -72,6 +72,30 @@ int main(int argc, char* argv[])
std::cout << "found " << op_ptrs.size() << " instances" << std::endl;
const auto& generic_op_ptr = op_ptrs[0];
auto generic_argument_ptr =
generic_op_ptr->MakeArgumentPointer({N, H, W, G, C}, // lengths
xy_strides, // xStrides
gamma_beta_strides, // gammaStrides
gamma_beta_strides, // betaStrides
xy_strides, // yStrides
{1, 2, 4}, // reduceDims
1e-6,
x_device_buf.GetDeviceBuffer(),
gamma_device_buf.GetDeviceBuffer(),
beta_device_buf.GetDeviceBuffer(),
y_device_buf.GetDeviceBuffer(),
nullptr,
nullptr,
Swish{});
if(!generic_op_ptr->IsSupportedArgument(generic_argument_ptr.get()))
{
throw std::runtime_error(
"The generic kernel instance should be able to support any input shapes");
};
std::string best_op_name;
bool found = false;
int best_op_id = -1;