Skip to content

[ET-VK] Performance improvement to few indexing functions. #9138

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
Mar 11, 2025
Merged

[ET-VK] Performance improvement to few indexing functions. #9138

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
036f707
[ET-VK] Performance improvement to few indexing functions.
trivedivivek Mar 11, 2025
7bb3118
Update on "[ET-VK] Performance improvement to few indexing functions."
trivedivivek Mar 11, 2025
c92e417
Update on "[ET-VK] Performance improvement to few indexing functions."
trivedivivek Mar 11, 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
17 changes: 10 additions & 7 deletions backends/vulkan/runtime/graph/ops/glsl/indexing_utils.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,16 +104,19 @@ ivec4 tidx_to_4bufi(
}

ivec4 nchwi_to_tidx(const int nchwi, const ivec4 sizes) {
const int nchwi_div_x = nchwi / sizes.x;
const int nchwi_div_y = nchwi_div_x / sizes.y;
return ivec4(
nchwi % sizes.x,
(nchwi / (sizes.x)) % sizes.y,
(nchwi / (sizes.x * sizes.y)) % sizes.z,
(nchwi / (sizes.x * sizes.y * sizes.z)));
nchwi_div_x % sizes.y,
nchwi_div_y % sizes.z,
nchwi_div_y / sizes.z);
}

int tidx_to_nchwi(const ivec4 tidx, const ivec4 sizes) {
return tidx.w * sizes.x * sizes.y * sizes.z + tidx.z * sizes.x * sizes.y +
tidx.y * sizes.x + tidx.x;
const int sizes_xy = sizes.x * sizes.y;
return tidx.w * sizes_xy * sizes.z + tidx.z * sizes_xy + tidx.y * sizes.x +
tidx.x;
}

// TODO(ssjia): make this function use dim order so that it can work with any
Expand Down Expand Up @@ -360,8 +363,8 @@ ivec4 to_texture_elem_pos(ivec4 idx, ivec4 sizes, int packed_dim) {
// pos[4] is set to a placeholder value
ivec4 pos = idx.xyzx;
pos[BATCH_AXIS] += idx.w * sizes[BATCH_AXIS];
pos[packed_dim] /= 4;
pos.w = idx[packed_dim] % 4;
pos[packed_dim] >>= 2;
pos.w = idx[packed_dim] & 0x3;
return pos;
}

Expand Down
Loading