ci: build and run the Vulkan backend, under Lavapipe

[JustVugg]

grep -rn "VK=1\|glslc\|vulkan" .github/workflows/
(nothing)

The Vulkan backend has never been compiled by any job in this repository. Not built, not linked, not run. backend_vulkan.c and the four GLSL shaders could be broken on dev right now and every check would stay green.

Three Vulkan PRs are open as this lands — #891, #729, #789 — all written by people without the hardware and reviewed by reading the diff. #892 and #887 are Vulkan bug reports from users whose cards nobody here has.

What Lavapipe buys

Mesa's software Vulkan runs on any x86 runner. Locally:

[VK] VRAM pressure-proofing: memory_priority on, memory_budget on
[VK] ready: llvmpipe (LLVM 20.1.2, 256 bits), compute qfam 0, memtype 0,
     fused gate+up, absorb attention
[VK] expert tier active: routed quantized experts on the GPU (budget 320)

So it proves the .comp shaders compile to SPIR-V, the loader finds a device, queue-family and memory-type selection work, extension negotiation works, and the expert tier initialises.

What it does not

Nothing about performance. It is a CPU rasteriser and will be slower than the ordinary CPU path; any tok/s measured under it is meaningless.

No driver-specific behaviour. The VK_EXT_memory_budget under-reporting on RADV RX 6000 that @MasterCATZ documented in #891 will not appear here. Real cards still need @Limalski, @BranBushes, @MasterCATZ and @krusherpt.

This is a compile-and-execute gate, not a benchmark, and the job name says so.

The job

Two steps that fail independently:

1. make colibri VK=1, plus assertions that all four .spv exist and the binary links libvulkan. glslc can reject a .comp while backend_vulkan.c compiles perfectly, so shader compilation is its own failure mode and gets its own check.
2. Run against Lavapipe with a fabricated config.json — enough to reach coli_vk_init(), which runs before any weight is read. The run then fails on the missing model, which is expected and ignored; the [VK] banner is what is asserted.

Verified in both directions

Passing, with the exact commands the job runs:

27532 shaders/attention_absorb.spv     linked against libvulkan  OK
19120 shaders/qmatmul.spv              PASS: [VK] ready
16924 shaders/qmatmul_gate_up.spv      PASS: expert tier active
 4180 shaders/rmsnorm.spv

Negative control, one shader removed:

[VK] cannot open shaders/qmatmul.spv
[VK] Vulkan backend unavailable (tried shaders/qmatmul.spv; ...)
PASS: the gate FAILS, as it must

A job that cannot fail is worse than no job, and this repo shipped one of those a week ago: #868's release check was named "coli would not resolve these next to itself" and asserted file existence, so it passed while the launcher structurally could not select the engine — which is #879.

🤖 Generated with Claude Code