]> git.djapps.eu Git - pkg/ggml/sources/llama.cpp/commit
overview / pkg / ggml / sources / llama.cpp / commit
summary | shortlog | log | commit | commitdiff | tree
(parent: 14be5a3)
vulkan: Remove transfer_ctx, do everything in compute_ctx. (#18945)
authorJeff Bolz <redacted>
Wed, 21 Jan 2026 17:01:40 +0000 (11:01 -0600)
committerGitHub <redacted>
Wed, 21 Jan 2026 17:01:40 +0000 (18:01 +0100)
commitbd544c94a34366b009901b47d4850bf9a9bce4d1
tree6141fbcb2395520687784470bde0ca8ad351d4c0tree
parent14be5a39b1286fc514efd19119c8f1c2e545634fcommit | diff
vulkan: Remove transfer_ctx, do everything in compute_ctx. (#18945)

* vulkan: Remove transfer_ctx, do everything in compute_ctx.

We had a bug where a set_tensor_async (using transfer_ctx) didn't get
submitted before the graph_compute (using compute_ctx) that came after
it. To avoid this sort of issue, just do everything in compute_ctx.

Remove transfer_cmd_pool, which was already unused.

* fix crash with perf logger
ggml/src/ggml-vulkan/ggml-vulkan.cpp diff | blob | history
Packaging of ggml-org/llama.cpp