Skip to content

[ET-VK] Changing texture access pattern for conv2d pw op to improve performance. #7476

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jan 3, 2025
Merged

[ET-VK] Changing texture access pattern for conv2d pw op to improve performance. #7476

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
4d9679f
[ET-VK] Changing texture access pattern for conv2d pw op to improve p…
trivedivivek Jan 2, 2025
cf18bd4
Update on "[ET-VK] Changing texture access pattern for conv2d pw op t…
trivedivivek Jan 3, 2025
69e93a6
Update on "[ET-VK] Changing texture access pattern for conv2d pw op t…
trivedivivek Jan 3, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 4 additions & 4 deletions backends/vulkan/runtime/graph/ops/glsl/conv2d_pw.glsl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -43,13 +43,13 @@ shared u16vec2 pos_shared[gl_WorkGroupSize.x * gl_WorkGroupSize.y * gl_WorkGroup
* size is only 1x1, making it easier to re-use loaded texels from t_kernel.
*/
void main() {
const uint16_t out_limits_y_scaled = uint16_t((out_limits.y + TILE_SIZE - 1) / TILE_SIZE);
const uvec2 out_limits_scaled = (out_limits.xy + TILE_SIZE - 1) / TILE_SIZE;
const uint shared_mem_stride = gl_WorkGroupSize.x * gl_WorkGroupSize.y * gl_WorkGroupSize.z;

const u16vec3 gpos = u16vec3(
gl_GlobalInvocationID.x / (out_limits_y_scaled * out_limits.z),
(gl_GlobalInvocationID.x / out_limits.z) % out_limits_y_scaled,
gl_GlobalInvocationID.x % out_limits.z);
gl_GlobalInvocationID.x % out_limits_scaled.x,
(gl_GlobalInvocationID.x / out_limits_scaled.x) % out_limits_scaled.y,
gl_GlobalInvocationID.x / (out_limits_scaled.x * out_limits_scaled.y));

// Output position for TILE_SIZE = 2
// +--------+--------+
Expand Down