]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 56666fa)
ggml-webgpu: Update register tiling matmul to use f32 accumulation (#21644)
authorReese Levine <redacted>
Tue, 14 Apr 2026 10:46:41 +0000 (03:46 -0700)
committerGitHub <redacted>
Tue, 14 Apr 2026 10:46:41 +0000 (13:46 +0300)
commit5a23695d5a78d2000a674fe54e5c4ecdff0a7089
tree238b7d16e67507580f48a847a689c617ea9ffcb1tree
parent56666fa6072f150fcc48f138433f82ad3fb76d79commit | diff
ggml-webgpu: Update register tiling matmul to use f32 accumulation (#21644)

* Update register tiling matmul to use f32 accumulation

* fix profiling code

* Fix register tiling matmul for chrome, i'm blaming dawn

* Update batch tuning value for iOS

* compile fix

* Fix use of new load function
ggml/src/ggml-webgpu/ggml-webgpu.cpp diff | blob | history
ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_decls.tmpl diff | blob | history
ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_reg_tile.wgsl diff | blob | history
ggml/src/ggml-webgpu/wgsl-shaders/mul_mat_subgroup_matrix.wgsl diff | blob | history
Packaging of ggml-org/llama.cpp