upstream-archive/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggml-org/llama.cpp.git synced 2026-10-02 10:57:33 -05:00
Code Issues Packages Projects Releases Wiki Activity

add TODO for apple RDMA

Browse Source
This commit is contained in:
Aman Gupta
2026-09-26 18:53:11 +08:00
parent 4ee77c31b2
commit f8bc970dfe
1 changed files with 1 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/ggml-rpc/transport-apple.cpp
+1
View File
@@ -26,6 +26,7 @@
// so every SEND posts a whole 128KiB stride over the wire, even when partially filled.
// (In testing 128KiB was the best performing among 32, 64, 128, 256)
//TODO: add mechanism similar to https://github.com/ggml-org/llama.cpp/pull/29440 to prevent idle CPU from spinning
static constexpr uint32_t RDMA_SEG_MAGIC = 0x52534547u; // "RSEG"
static constexpr int RDMA_NBUF = 16; // ring depth (frames per direction)
static constexpr size_t RDMA_FRAME = 4096; // Thunderbolt frame (fixed on Apple)