llama.cpp
CUDA/HIP: Share the same unified memory allocation logic.
#12934

Merged

CUDA/HIP: Share the same unified memory allocation logic. #12934

IMbackK merged 5 commits into ggml-org:master from hjc4869:hip-uma

CUDA/HIP: Share the same unified memory allocation logic.

493d9f79

github-actions added documentation

github-actions added Nvidia GPU

github-actions added ggml

hjc4869 changed the title ~~CUDA/HIP: Reduce the divergence of unified memory allocation between CUDA and HIP.~~ CUDA/HIP: Share the same unified memory allocation logic. 1 year ago

IMbackK commented on 2025-04-14

Remove comment

6fab0d4c

Add a fallback to hipMalloc() and print warning once when managed mem…

2428d672

Fix editorconfig check error

a3862d91

JohannesGaessler commented on 2025-04-14

Use snake case variable

e6956cae

JohannesGaessler approved these changes on 2025-04-14

IMbackK approved these changes on 2025-04-15

IMbackK merged 84778e97 into master 1 year ago

hjc4869 deleted the hip-uma branch 1 year ago

Reviewers

IMbackK

JohannesGaessler

Assignees

No one assigned

Labels

documentation Nvidia GPU ggml

Milestone

No milestone

llama.cpp CUDA/HIP: Share the same unified memory allocation logic. #12934 Merged

CUDA/HIP: Share the same unified memory allocation logic. #12934

llama.cpp
CUDA/HIP: Share the same unified memory allocation logic.
#12934

Merged