Commit Graph
2 Commits
Author SHA1 Message Date
Rail Chabdarov e1cdce46c5 hip: Avoid compiler bug in RDNA code generation during debug builds on Windows (llama/20655) 2026-03-29 15:04:36 +03:00
Rail Chabdarov 46aad766f5 Fix data race in CUDA's "cpy" kernel (influences GGML's DUP, CONT operations). (llama/20507)
* Fix datarace in CUDA's "cpy" kernel.

* Remove extra barrier by using more of shared memory.
2026-03-16 13:10:15 +02:00