IntenseWebs/openvino
3
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

[GPU] Bugfix for onednn post op optimization (#10416)

Browse Source 
When post-op has pattern like below, binary_mul was ignored previously.
1. binary_add
2. eltwise_linear
3. binary_mul
4. binary_add

It happens when prev_post_op_idx == 2, cur_post_op_idx == 4.
prev_post_op_idx was supposed to proceed to idx 3, but it did not.
This commit is contained in:
Mingyu Kim
2022-02-16 16:44:42 +09:00
committed by GitHub
parent fa4246d531
commit c0d54e48bb
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
src/plugins/intel_gpu/src/graph/program_node.cpp
View File

@@ -489,9 +489,9 @@ dnnl::post_ops program_node::try_optimize_post_ops(dnnl::post_ops& p_ops, const
auto prev_type = cur_post_ops[prev_post_op_idx].op_type;
// Ignore optimized operations for "previous" operation in our operation pair
while (type_is_any_optimized(prev_type) && cur_post_op_idx < post_ops_size - 1) {
while (type_is_any_optimized(prev_type) && prev_post_op_idx < post_ops_size - 1) {
prev_post_op_idx++;
if (prev_post_op_idx == cur_post_op_idx)
if (prev_post_op_idx == cur_post_op_idx && cur_post_op_idx < post_ops_size - 1)
cur_post_op_idx++;
prev_type = cur_post_ops[prev_post_op_idx].op_type;
cur_type = cur_post_ops[cur_post_op_idx].op_type;