github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 08:02:14 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[PR #13872] [MERGED] llama: fix fattn-tile shared memory overflow on sm_50/52 #14429

New Issue
Closed
opened 2026-04-13 00:53:48 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-13 00:53:48 -05:00
Owner
Copy Link

📋 Pull Request Information

Original PR: https://github.com/ollama/ollama/pull/13872
Author: @jmorganca
Created: 1/24/2026
Status: ✅ Merged
Merged: 1/24/2026
Merged by: @jmorganca

Base: main ← Head: fix-cuda12-fattn-shmem

📝 Commits (1)

  • e3fdc39 ggml-cuda: fix fattn-tile shared memory overflow on sm_50/52

📊 Changes

2 files changed (+18 additions, -19 deletions)

View changed files

📝 llama/patches/0032-ggml-enable-MLA-flash-attention-for-GLM-4.7-flash.patch (+11 -12)
📝 ml/backend/ggml/ggml/src/ggml-cuda/fattn-tile.cuh (+7 -7)

📄 Description

Use nthreads=128 for ncols=4 configurations in flash attention tile kernel to reduce shared memory usage below 48KB limit on Maxwell architectures (sm_50/52).

With nthreads=256 and ncols=4, np=2 which caused shared memory to exceed 48KB. With nthreads=128 and ncols=4, np=1 keeps shared memory under the limit.

This should fix the release CI.

🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.

## 📋 Pull Request Information **Original PR:** https://github.com/ollama/ollama/pull/13872 **Author:** [@jmorganca](https://github.com/jmorganca) **Created:** 1/24/2026 **Status:** ✅ Merged **Merged:** 1/24/2026 **Merged by:** [@jmorganca](https://github.com/jmorganca) **Base:** `main` ← **Head:** `fix-cuda12-fattn-shmem` --- ### 📝 Commits (1) - [`e3fdc39`](https://github.com/ollama/ollama/commit/e3fdc3928dbabf9cd03e2634479006a4342ea9e7) ggml-cuda: fix fattn-tile shared memory overflow on sm_50/52 ### 📊 Changes **2 files changed** (+18 additions, -19 deletions) <details> <summary>View changed files</summary> 📝 `llama/patches/0032-ggml-enable-MLA-flash-attention-for-GLM-4.7-flash.patch` (+11 -12) 📝 `ml/backend/ggml/ggml/src/ggml-cuda/fattn-tile.cuh` (+7 -7) </details> ### 📄 Description Use nthreads=128 for ncols=4 configurations in flash attention tile kernel to reduce shared memory usage below 48KB limit on Maxwell architectures (sm_50/52). With nthreads=256 and ncols=4, np=2 which caused shared memory to exceed 48KB. With nthreads=128 and ncols=4, np=1 keeps shared memory under the limit. This should fix the release CI. --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
GiteaMirror added the pull-request label 2026-04-13 00:53:48 -05:00
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label pull-request
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#14429
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?