Skip to content

Update Windows ROCm build to 26.Q1#19810

Merged
CISC merged 1 commit into
ggml-org:masterfrom
superm1:update-windows-hip
Feb 25, 2026
Merged

Update Windows ROCm build to 26.Q1#19810
CISC merged 1 commit into
ggml-org:masterfrom
superm1:update-windows-hip

Conversation

@superm1

@superm1 superm1 commented Feb 22, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

This updates Windows to the latest stable release of ROCm just as Linux was just recently updated (#19433).

Here is a build in my fork demonstrating it passes.

CC @IMbackK , @slojosic-amd, @hammmmy

@superm1 superm1 requested a review from CISC as a code owner February 22, 2026 17:21
@github-actions github-actions Bot added the devops improvements to build systems and github actions label Feb 22, 2026
@superm1 superm1 force-pushed the update-windows-hip branch 2 times, most recently from 773bb87 to 4d50bae Compare February 23, 2026 16:12
slojosic-amd
slojosic-amd suggested changes Feb 23, 2026
View reviewed changes
Comment thread .github/workflows/release.yml
Comment thread .github/workflows/release.yml
Comment thread .github/workflows/release.yml
Comment thread .github/workflows/release.yml
@slojosic-amd

slojosic-amd commented Feb 24, 2026 •
edited
Loading

Copy link
Copy Markdown
Contributor

@superm1 Actually, --target ggml-hip and -DGGML_CPU=OFF should stay as they were before your latest changes, because true packaging comes at the end https://github.com/ggml-org/llama.cpp/blob/master/.github/workflows/release.yml#L867 during release job when all other important targets and missing files will be added to the new zip package. For example, llama-bin-win-hip-radeon-x64.zip is just intermediate artifact which is than used at the end to generate final llama-b8144-bin-win-hip-radeon-x64.zip package with all important .dll and .exe files.

slojosic-amd
slojosic-amd reviewed Feb 24, 2026
View reviewed changes
Comment thread .github/workflows/release.yml
slojosic-amd
slojosic-amd reviewed Feb 24, 2026
View reviewed changes
Comment thread .github/workflows/release.yml
slojosic-amd
slojosic-amd reviewed Feb 24, 2026
View reviewed changes
Comment thread .github/workflows/release.yml Outdated
@superm1
Update Windows ROCm build to 26.Q1
a092656 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
@superm1 superm1 requested a review from slojosic-amd February 24, 2026 15:00
@superm1

superm1 commented Feb 24, 2026

Copy link
Copy Markdown
Contributor Author

Thanks for digging into it @slojosic-amd . I've updated per your guidance. Here is the updated test run.

https://github.com/superm1/llama.cpp/actions/runs/22353759055/job/64687505235

slojosic-amd
slojosic-amd approved these changes Feb 24, 2026
View reviewed changes

@slojosic-amd slojosic-amd left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

IMbackK
IMbackK approved these changes Feb 25, 2026
View reviewed changes

@IMbackK IMbackK left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is fine, and i am not aware of anything blocking this, but i dont have a windows rocm setup to try the artifacts on.

@slojosic-amd

slojosic-amd commented Feb 25, 2026

Copy link
Copy Markdown
Contributor

@superm1 @IMbackK I tried https://github.com/superm1/llama.cpp/actions/runs/22314451817/artifacts/5620163961 on Windows on Strix Halo and llama.bench and llama-cli seems to work fine:

image image

CISC
CISC reviewed Feb 25, 2026
View reviewed changes
Comment thread .github/workflows/release.yml
include:
- name: "radeon"
gpu_targets: "gfx1151;gfx1200;gfx1201;gfx1100;gfx1101;gfx1102;gfx1030;gfx1031;gfx1032"
gpu_targets: "gfx1150;gfx1151;gfx1200;gfx1201;gfx1100;gfx1101;gfx1102;gfx1030;gfx1031;gfx1032"

@CISC CISC Feb 25, 2026

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why do we have different targets on Windows?

@IMbackK IMbackK Feb 25, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The Windows kernel driver dose not support the full range of gpus, gfx9 for instance was never supported at all on windows via the clr kernel interface windows hip uses

@CISC CISC merged commit 8fdf269 into ggml-org:master Feb 25, 2026
2 checks passed
@CISC

CISC commented Feb 25, 2026

Copy link
Copy Markdown
Member

@superm1 Can you also update the regular build and windows-setup-rocm action to resolve the installer cache conflict?

@slojosic-amd

slojosic-amd commented Feb 25, 2026

Copy link
Copy Markdown
Contributor

@superm1 Can you also update the regular build and windows-setup-rocm action to resolve the installer cache conflict?

@CISC I am preparing PR for this

@slojosic-amd slojosic-amd mentioned this pull request Feb 25, 2026
Merged
@slojosic-amd

slojosic-amd commented Feb 25, 2026

Copy link
Copy Markdown
Contributor

@superm1 Can you also update the regular build and windows-setup-rocm action to resolve the installer cache conflict?

@CISC I am preparing PR for this

@CISC FYI: #19891

bartowski1182 pushed a commit to bartowski1182/llama.cpp that referenced this pull request Mar 2, 2026
@superm1 @bartowski1182
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
9f8d3cc 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
ArberSephirotheca pushed a commit to ArberSephirotheca/llama.cpp that referenced this pull request Mar 3, 2026
@superm1 @ArberSephirotheca
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
786c3dc 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
aldehir pushed a commit to aldehir/llama.cpp that referenced this pull request Mar 6, 2026
@superm1 @aldehir
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
69dd0c6 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
Ethan-a2 pushed a commit to Ethan-a2/llama.cpp that referenced this pull request Mar 20, 2026
@superm1 @xiukeding
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
adc5043 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
@Jasdfgh Jasdfgh mentioned this pull request Mar 30, 2026
Open
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
a6dc826 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
ljubomirj pushed a commit to ljubomirj/llama.cpp that referenced this pull request May 6, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
3566376 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
e633f16 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
my-other-github-account pushed a commit to my-other-github-account/llama.cpp that referenced this pull request May 15, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
e5c4642 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
fewtarius pushed a commit to fewtarius/CachyLLama that referenced this pull request May 30, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
5c70003 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
fukuro-kun pushed a commit to fukuro-kun/fukuro-llama-cpp-turboquant that referenced this pull request Jul 5, 2026
@superm1
ci : update Windows ROCm build to 26.Q1 [no ci] (ggml-org#19810)
6d090cf 
* Update build command to build llama-* tools not just ggml-hip
* Update rocWMMA headers to 7.2
* Add GFX1150 target
* Correct library paths for AMD libraries in 26.Q1
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@CISC CISC CISC approved these changes
@IMbackK IMbackK IMbackK approved these changes
+1 more reviewer
@slojosic-amd slojosic-amd slojosic-amd approved these changes
Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

devops improvements to build systems and github actions

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@superm1 @slojosic-amd @CISC @IMbackK