PR #6210 move BLAS to a separate backend

move BLAS to a separate backend #6210

slaren merged 15 commits into master from sl/blas-backend

mofosyne added Review Complexity : High

mofosyne added refactoring

github-actions added ggml

slaren force pushed from d7cc6bc0 to cde46d2e 1 year ago

github-actions added build

slaren force pushed from cde46d2e 1 year ago

slaren force pushed 1 year ago

slaren force pushed to ca912057 1 year ago

github-actions added Vulkan

github-actions added SYCL

github-actions added Kompute

move BLAS to a separate backend

7f58793c

rename GGML_USE_OPENBLAS to GGML_USE_BLAS

b88957e5

add support for out_prod

77f88e35

alloc : reuse same buffer when the same buffer type if used multiple …

845fa20f

slaren force pushed from ca912057 to 845fa20f 1 year ago

support multithreaded dequantization with std::async when openmp is n…

2bfdb7fe

fix apple build

0425305d

slaren marked this pull request as ready for review 1 year ago

ggml-ci

2dd049ed

reuse main thread

63e06d0c

Merge remote-tracking branch 'origin/master' into sl/blas-backend

fa2ec607

set number of threads automatically for openblas and blis

a8a1bf79

sched : print assignments when GGML_SCHED_DEBUG env variable is set

ecb75b5f

fixes

e0665981

github-actions added examples

ggerganov requested a review from

ggerganov 1 year ago

ggerganov commented on 2024-06-12

Apply suggestions from code review

a54b7912

ggerganov approved these changes on 2024-06-12

fix metal being used in layers not offloaded

ae9cd856

sched : allow ops with weights on an incompatible buffer type

211fb045

slaren merged f578b86b into master 1 year ago

slaren deleted the sl/blas-backend branch 1 year ago

Reviewers

ggerganov

Assignees

No one assigned

Labels

build refactoring Vulkan examples Review Complexity : High ggml SYCL Kompute

Milestone

No milestone

llama.cpp move BLAS to a separate backend #6210 Merged

move BLAS to a separate backend #6210

llama.cpp
move BLAS to a separate backend
#6210

Merged