ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 05:01:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add dynamic elementwise op (#1426)

Browse Source 
* Add dynamic elementwise op

Co-authored-by: ThruptiRajLakshmanaGowda <thruptiraj.lakshmanagowda@amd.com>

* CI issues fix

* Custom parameter value for dynamic functions - Comments addressed

---------

Co-authored-by: ThruptiRajLakshmanaGowda <thruptiraj.lakshmanagowda@amd.com>
Co-authored-by: ThruptiRajLakshmanaGowda <tlakshma@amd.com>
This commit is contained in:
Bartłomiej Kocot
2024-10-26 15:22:37 +02:00
committed by GitHub
parent 54f0e6f4bb
commit 31bf253aeb
32 changed files with 2194 additions and 170 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
include/ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_xdl_cshuffle.hpp
View File

@@ -85,9 +85,9 @@ __global__ void
BsPointer p_bs_grid,
DsPointer p_ds_grid,
EDataType* __restrict__ p_e_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CDEElementwiseOperation cde_element_op,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CDEElementwiseOperation cde_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_k0_m_k1,
const BGridDesc_BK0_N_BK1 b_grid_desc_k0_n_k1,
const DsGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
@@ -121,6 +121,19 @@ __global__ void
static_for<0, NumDTensor, 1>{}(
[&](auto i) { p_ds_grid_grp(i) = p_ds_grid[i] + ds_group_offset[i]; });
if constexpr(is_same_v<AElementwiseOperation, element_wise::DynamicUnaryOp>)
{
a_element_op.InitUnaryOpPtrOnDevice();
}
if constexpr(is_same_v<BElementwiseOperation, element_wise::DynamicUnaryOp>)
{
b_element_op.InitUnaryOpPtrOnDevice();
}
if constexpr(is_same_v<CDEElementwiseOperation, element_wise::DynamicUnaryOp>)
{
cde_element_op.InitUnaryOpPtrOnDevice();
}
if constexpr(isMultiA || isMultiB)
{
AsPointer p_as_grid_grp;