Skip to content

GitLab

"vscode:/vscode.git/clone" did not exist on "ef6d5a73a24d0a0522f7b5b07284f62b17cb8662"
Commit 5a7d99ce authored by iq136boy's avatar iq136boy
Browse files

add compute type check for fp16 in forward convolution instances

parent f46a6ffa
Show whitespace changes
Inline Side-by-side
Showing with 5 additions and 5 deletions
library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward.hpp
View file @ 5a7d99ce
......@@ -935,7 +935,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
#endif
#ifdef CK_ENABLE_FP16
if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
is_same_v<OutDataType, half_t>)
is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
{
add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances(op_ptrs);
}
......@@ -977,7 +977,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
#ifdef CK_ENABLE_FP16
if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
is_same_v<OutDataType, half_t>)
is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
{
add_device_grouped_conv2d_fwd_xdl_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
......@@ -989,7 +989,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
is_same_v<OutDataType, half_t>)
is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
{
add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(op_ptrs);
}
......@@ -1045,7 +1045,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
#if(defined(CK_ENABLE_FP16) && defined(DL_KERNELS))
if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
is_same_v<OutDataType, half_t>)
is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
{
add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(op_ptrs);
}
......@@ -1082,7 +1082,7 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
#endif
#ifdef CK_ENABLE_FP16
if constexpr(is_same_v<InDataType, half_t> && is_same_v<WeiDataType, half_t> &&
is_same_v<OutDataType, half_t>)
is_same_v<OutDataType, half_t> && is_same_v<ComputeType, half_t>)
{
add_device_grouped_conv3d_fwd_xdl_gndhwc_gkzyxc_gndhwk_f16_instances(op_ptrs);
add_device_grouped_conv3d_fwd_wmma_gndhwc_gkzyxc_gndhwk_f16_instances(op_ptrs);
......
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