llama.cpp releases · May 22, 2026 · 2 min read

b9279

Mirrored from llama.cpp releases for archival readability. Support the source by reading on the original site.

vulkan: fuse snake activation (mul, sin, sqr, mul, add) (#22855)

vulkan: fuse snake activation (mul, sin, sqr, mul, add)

Add snake.comp shader with F32 / F16 / BF16 pipelines and
ggml_vk_snake_dispatch_fused. The matcher recognizes the naive 5 op
decomposition emitted by audio decoders (BigVGAN, Vocos) for snake
activation y = x + sin(a*x)^2 * inv_b and rewrites it to a single
elementwise kernel.

test_snake_fuse from the CUDA PR now also compares CPU naive vs
Vulkan fused across F32 / F16 / BF16.

vulkan: address jeffbolznv review for fused snake activation

Rename T / C to ne0 / ne1 in the shader and push constants to match
the standard naming convention used across the Vulkan backend.

Tighten ggml_vk_can_fuse_snake: require x and dst to be contiguous
(the shader uses idx = i0 + i1 * ne0) and require a / inv_b to be
tightly packed on the broadcast dim (the shader reads data_a[i1]).

vulkan: tighten snake fusion type checks for all operands (address jeffbolznv review)
vulkan: reject snake fusion when ne[2] or ne[3] > 1 (address jeffbolznv review)
vulkan: address 0cc4m review for fused snake activation

snake.comp is renamed to follow the ggml DATA_A_* / A_TYPE convention.
A_TYPE now applies to the activation tensor data_a instead of the
broadcast multiplier, and the bindings become data_a (A_TYPE), data_b
(float), data_c (float) and data_d (D_TYPE). A header at the top of
the shader maps each buffer to its role in y = x + sin(b * x)^2 * c.

On the C++ side, ggml_vk_can_fuse_snake reuses the existing snake_pattern
constant instead of duplicating the op list, sin_node is extracted as a
named local alongside the other chain nodes, and the broadcast operands
a and inv_b are now required to be GGML_TYPE_F32 to match the hardcoded
float bindings on data_b and data_c (the previous a->type == x->type
would silently reject any future BF16 or F16 chain once the supports_op
gate for SIN / SQR is lifted). ggml_vk_snake_dispatch_fused gets an
explicit GGML_TYPE_F32 case and GGML_ABORT on default in place of the
silent f32 fallback, and a stale comment about data_a[i1] / data_inv_b[i1]
is refreshed to match the new binding names.

macOS/iOS:

Linux:

Android:

Android arm64 (CPU)

Windows:

openEuler:

Discussion (0)

No comments yet. Sign in and be the first to say something.

Discussion (0)

More from llama.cpp releases