ci : enable Vulkan workflow on Mac by ggerganov · Pull Request #16194 · ggml-org/llama.cpp

ggerganov · 2025-09-23T07:32:42Z

No description provided.

CISC

Looks good, but no idea what's going on here:
https://github.com/ggml-org/llama.cpp/actions/runs/17938949707/job/51010682770?pr=16194#step:3:4196

ggerganov · 2025-09-23T08:50:28Z

Looks good, but no idea what's going on here: https://github.com/ggml-org/llama.cpp/actions/runs/17938949707/job/51010682770?pr=16194#step:3:4196

I'm not sure either.

@0cc4m Let me know if you are interested in adding MoltenVK workflow to the CI. Atm, it seems to require some fixes in the IM2COL and CONV kernels to be able to run correctly.

Feel free to push into this branch if you find a fix.

ericcurtin · 2025-09-23T08:59:11Z

Do people have hope KosmicKrisp will be even better than MoltenVK?

Tagging @kpouget here, I think he built this on macOS before.

0cc4m · 2025-09-23T09:10:03Z

@ggerganov we reported that problem to MoltenVK and it was fixed in 1.4. Are you sure the MVK version is up to date?

kpouget · 2025-09-23T09:11:38Z

@0cc4m Let me know if you are interested in adding MoltenVK workflow to the CI. Atm, it seems to require some fixes in the IM2COL and CONV kernels to be able to run correctly.

I have a nightly performance testing running everyday, and it builds correctly the Vulkan version

31/36 Test #31: test-backend-ops ..................***Exception: SegFault 33.16 sec

hum but I see now that it's the test-backend-ops that failed, this one isn't enabled in my nightly. I'll give it a shot to see if I have the same failure

ggerganov · 2025-09-23T09:46:03Z

@0cc4m The runner is using the v1.4.321.0 SDK from this site: https://vulkan.lunarg.com/sdk/home. I think this is the latest version.

ggml@ggml-100-mac-m4 ~ % vulkaninfo --summary 
==========
VULKANINFO
==========

Vulkan Instance Version: 1.4.321


Instance Extensions: count = 17
-------------------------------
VK_EXT_debug_report                    : extension revision 10
VK_EXT_debug_utils                     : extension revision 2
VK_EXT_headless_surface                : extension revision 1
VK_EXT_layer_settings                  : extension revision 2
VK_EXT_metal_surface                   : extension revision 1
VK_EXT_surface_maintenance1            : extension revision 1
VK_EXT_swapchain_colorspace            : extension revision 5
VK_KHR_device_group_creation           : extension revision 1
VK_KHR_external_fence_capabilities     : extension revision 1
VK_KHR_external_memory_capabilities    : extension revision 1
VK_KHR_external_semaphore_capabilities : extension revision 1
VK_KHR_get_physical_device_properties2 : extension revision 2
VK_KHR_get_surface_capabilities2       : extension revision 1
VK_KHR_portability_enumeration         : extension revision 1
VK_KHR_surface                         : extension revision 25
VK_LUNARG_direct_driver_loading        : extension revision 1
VK_MVK_macos_surface                   : extension revision 3

Instance Layers: count = 7
--------------------------
VK_LAYER_KHRONOS_profiles         Khronos Profiles layer                     1.4.321  version 1
VK_LAYER_KHRONOS_shader_object    Khronos Shader object layer                1.4.321  version 1
VK_LAYER_KHRONOS_synchronization2 Khronos Synchronization2 layer             1.4.321  version 1
VK_LAYER_KHRONOS_validation       Khronos Validation Layer                   1.4.321  version 1
VK_LAYER_LUNARG_api_dump          LunarG API dump layer                      1.4.321  version 2
VK_LAYER_LUNARG_gfxreconstruct    GFXReconstruct Capture Layer Version 1.0.5 1.4.321  version 4194309
VK_LAYER_LUNARG_screenshot        LunarG image capture layer                 1.4.321  version 1

Devices:
========
GPU0:
	apiVersion         = 1.3.313
	driverVersion      = 0.2.2108
	vendorID           = 0x106b
	deviceID           = 0xf010209
	deviceType         = PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU
	deviceName         = Apple M4
	driverID           = DRIVER_ID_MOLTENVK
	driverName         = MoltenVK
	driverInfo         = 1.3.0
	conformanceVersion = 1.3.8.0
	deviceUUID         = 0000106b-0f01-0209-0000-000000000000
	driverUUID         = 4d564b00-0000-283c-0f01-020900000000

0cc4m · 2025-09-23T10:13:52Z

I'm using a homebrew version and it's just fine. Seems to be 2 minor versions further, though:

Devices:
========
GPU0:
        apiVersion         = 1.4.323
        driverVersion      = 0.2.2208
        vendorID           = 0x106b
        deviceID           = 0xf060209
        deviceType         = PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU
        deviceName         = Apple M4 Max
        driverID           = DRIVER_ID_MOLTENVK
        driverName         = MoltenVK
        driverInfo         = 1.4.0
        conformanceVersion = 1.4.2.0
        deviceUUID         = 0000106b-0f06-0209-0000-000000000000
        driverUUID         = 4d564b00-0000-28a0-0f06-020900000000

With that, it passes test-backend-ops fully.

0cc4m · 2025-09-23T10:15:13Z

Actually no, your version is far behind:

driverName         = MoltenVK
driverInfo         = 1.3.0

That is the problem.

ggerganov · 2025-09-23T10:44:33Z

Thanks, that fixed it.

@danbev

* origin/master: (39 commits) ci : disable AMD workflows + update NVIDIA workflows (ggml-org#16200) ci : enable Vulkan workflow on Mac (ggml-org#16194) ggml-cpu: Respect cpumask settings (ggml-org#16164) ggml : fix uninitialized is_on_grid in quantize_row_iq3_xxs_impl (ggml-org#15928) zdnn: refactor codebase + add docs (ggml-org#16178) codeowners : add @danbev to model-conversion example [no ci] (ggml-org#16190) devops: add s390x containers (ggml-org#15915) ggml-cpu : fix typo in gemm comments [no ci] (ggml-org#16189) feat: Add conversion support in GraniteHybrid for non-hybrid (all attn) (ggml-org#16177) clang-tidy : disable warning about performance enum size (ggml-org#16127) ggml : implement set_rows with i32 index (ggml-org#16159) codeowners : update + cleanup (ggml-org#16174) common : enable `--offline` mode without curl support (ggml-org#16137) webui : fix handling incomplete chunks (ggml-org#16107) embedding : fix typos in README (ggml-org#16171) common : remove unused local variables (ggml-org#16140) ggml : extend ggml_can_fuse to work with non-sequential nodes (ggml-org#16123) ggml : add ggml_op_is_empty (ggml-org#16122) codeowners : update ownership for @ngxson and @allozuar (ggml-org#16128) Vulkan: add conv_transpose_2d operation (ggml-org#16022) ...

ci : enable Vulkan workflow on Mac

f126b6f

ggerganov requested a review from CISC as a code owner September 23, 2025 07:32

github-actions bot added the devops improvements to build systems and github actions label Sep 23, 2025

CISC approved these changes Sep 23, 2025

View reviewed changes

ggerganov merged commit 0889589 into master Sep 23, 2025
103 of 113 checks passed

ggerganov deleted the gg/ci-mac-vulkan branch September 23, 2025 10:44

struct pushed a commit to struct/llama.cpp that referenced this pull request Sep 26, 2025

ci : enable Vulkan workflow on Mac (ggml-org#16194)

de11e0c

pwilkin pushed a commit to pwilkin/llama.cpp that referenced this pull request Oct 23, 2025

ci : enable Vulkan workflow on Mac (ggml-org#16194)

0917083

Anico2 added a commit to Anico2/llama.cpp that referenced this pull request Jan 15, 2026

ci : enable Vulkan workflow on Mac (ggml-org#16194)

10b639d

blime4 referenced this pull request in blime4/llama.cpp Feb 5, 2026

ci : enable Vulkan workflow on Mac (#16194)

83b548f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ci : enable Vulkan workflow on Mac#16194

ci : enable Vulkan workflow on Mac#16194
ggerganov merged 1 commit intomasterfrom
gg/ci-mac-vulkan

ggerganov commented Sep 23, 2025

Uh oh!

CISC left a comment

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

ericcurtin commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

kpouget commented Sep 23, 2025

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

ggerganov commented Sep 23, 2025

Uh oh!

CISC left a comment

Choose a reason for hiding this comment

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

ericcurtin commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

kpouget commented Sep 23, 2025

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

0cc4m commented Sep 23, 2025

Uh oh!

Uh oh!

ggerganov commented Sep 23, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants