Skip to content

ggml-webgpu: support for SSM_SCAN and disable set_rows error checking#22327

Merged
ggerganov merged 6 commits intoggml-org:masterfrom
reeselevine:ssm_scan
Apr 25, 2026
Merged

ggml-webgpu: support for SSM_SCAN and disable set_rows error checking#22327
ggerganov merged 6 commits intoggml-org:masterfrom
reeselevine:ssm_scan

Conversation

@reeselevine
Copy link
Copy Markdown
Contributor

@reeselevine reeselevine commented Apr 24, 2026

Overview

  • Adds support for SSM_SCAN, so that models like Granite can run fully-offloaded to GPU in WebGPU. Uses a subgroup reduction if subgroups supported, otherwise falls back to shared memory reduction.
  • Disables the error checking for i64 set rows indexes, because practically the WebGPU backend isn't being run on models big enough for it to matter and I think they would break anyways due to WebGPU's lack of 64-bit types. This also allows graph_compute to be fully non-blocking, CPU/GPU synchronization now only needs to happen in get_tensor.
  • Also updated ops support for WebGPU since it hadn't been updated in a bit.

Requirements

  • I have read and agree with the contributing guidelines
  • AI usage disclosure: yes, to investigate ssm_scan kernels and prototype the WebGPU version.

@reeselevine reeselevine requested a review from a team as a code owner April 24, 2026 18:01
@github-actions github-actions Bot added ggml changes relating to the ggml tensor library for machine learning WebGPU labels Apr 24, 2026
@github-actions github-actions Bot added the documentation Improvements or additions to documentation label Apr 24, 2026
@reeselevine reeselevine added the merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. label Apr 24, 2026
@ggerganov ggerganov merged commit dd2914d into ggml-org:master Apr 25, 2026
43 of 47 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

documentation Improvements or additions to documentation ggml changes relating to the ggml tensor library for machine learning merge ready A maintainer can use this label to indicate that they consider the changes final and ready to merge. WebGPU

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants