onnxruntime
[Native WebGPU] Support shared memory version of ReduceOps
#24399

Merged

[Native WebGPU] Support shared memory version of ReduceOps #24399

satyajandhyala merged 10 commits into main from sajandhy/webgpu-ep-shader-reduction-op

Added shared memory version of ReduceOps

076cc6cb

Use the output from the context instead of a local tensor

cdf35e4b

github-actions commented on 2025-04-11

Use naive version conditionally.

6e313e26

satyajandhyala marked this pull request as ready for review 256 days ago

satyajandhyala added ep:WebGPU

Current shared implementation does not support ArgMin/ArgMax

8728cfa8

Added workgroup_size to hint.

61407903

guschmue dismissed these changes on 2025-04-14

Added SetWorkgroupSize call.

2351268d

satyajandhyala dismissed their stale review via 2351268d 253 days ago

fs-eire dismissed these changes on 2025-04-14

Fix coner case when the input is empty.

5442d44b

satyajandhyala dismissed their stale review via 5442d44b 253 days ago

Simplified ReduceMean

0ed04a84

Removed GetOpSpecificCode function. Using table lookup instead.

6c7bf340

ArgMax/ArgMin constructor does not know the attributes before the bas…

76a9fe16

satyajandhyala requested a review from

fs-eire 253 days ago

satyajandhyala requested a review from

guschmue 253 days ago

guschmue approved these changes on 2025-04-15

satyajandhyala merged ff607b48 into main 252 days ago

satyajandhyala deleted the sajandhy/webgpu-ep-shader-reduction-op branch 252 days ago

Reviewers

guschmue

fs-eire

github-actions

Assignees

No one assigned

Labels

ep:WebGPU

Milestone

No milestone

onnxruntime [Native WebGPU] Support shared memory version of ReduceOps #24399 Merged

[Native WebGPU] Support shared memory version of ReduceOps #24399

onnxruntime
[Native WebGPU] Support shared memory version of ReduceOps
#24399

Merged