Skip to content

Commit f139e39

Browse files
committed
[ET-VK] Minor fix to conv 2d op using wg_size from create_conv2d_global_wg_size to determine local wg size.
Pull Request resolved: #7450 This diff contains changes to the Convolution.cpp file in the Vulkan backend of Executorch. The changes involve updating the code to use the create_conv2d_global_wg_size function to determine the local workgroup size for the convolution operation. This is done to ensure that the correct workgroup size is used for the operation, which can improve performance. ghstack-source-id: 260166246 @exported-using-ghexport Differential Revision: [D67676422](https://our.internmc.facebook.com/intern/diff/D67676422/)
1 parent 7c31a5d commit f139e39

File tree

1 file changed

+4
-2
lines changed

1 file changed

+4
-2
lines changed

backends/vulkan/runtime/graph/ops/impl/Convolution.cpp

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -370,11 +370,13 @@ void add_conv2d_node(
370370
weight_data,
371371
clamp_out);
372372

373+
utils::uvec3 wg_size = create_conv2d_global_wg_size(graph, method, out);
374+
373375
graph.execute_nodes().emplace_back(new DispatchNode(
374376
graph,
375377
shader,
376-
create_conv2d_global_wg_size(graph, method, out),
377-
graph.create_local_wg_size(out),
378+
wg_size,
379+
graph.create_local_wg_size(wg_size),
378380
// Inputs and Outputs
379381
{{out, vkapi::MemoryAccessType::WRITE},
380382
{{in, arg_weight, arg_bias}, vkapi::MemoryAccessType::READ}},

0 commit comments

Comments
 (0)