llama.cpp
ggml-cpu: Add IME2 Instruction Support for the SpacemiT Backend
#22863
Merged
Merged
ggml-cpu: Add IME2 Instruction Support for the SpacemiT Backend #22863
taronaeo
merged
16 commits into
ggml-org:master
from
spacemit-com:add-spacemit-backend-ime2
feat(spacemit): add spine memory pool and barrier implementations
5a99c96e
fix(spacemit): remove unused dst_scale variable and optimize gemm_ker…
98dce8f0
feature:add i8i4_hp_m4_v2
41884214
feature:add i8i4_hp_m4_v3
5217b98d
fix(spacemit): simplify TCM initialization logic by removing ime2 check
d467a575
fix(spacemit): enhance logging for allocation failures and errors
76fb6de0
fix(spacemit): restore rvv_kernels.cpp inclusion and update CMake con…
2d80b97c
fix(spacemit): update quantization support table and performance metrics
08a9f623
fix(spacemit): correct duplicate entry in quantization support table
f1b794a0
fix(spacemit): update CMake configuration to disable GGML_OPENMP and …
5ea8132b
fix(spacemit): improve code formatting and consistency in header files
5b3753a1
fix(spacemit): rename thread_id parameter to thread_n in NUMA thread …
5c920fdf
ggerganov
commented
on 2026-05-09
- Simplified the handling of tensor types in compute_forward for GGML…
61cfd798
- Changed function signatures in repack.cpp and repack.h to use ggml_…
247244ee
refactor(spacemit): simplify memory calculation logic in ggml_backend…
75d5f8d6
refactor(spacemit): simplify error handling in tensor operations
58745bdb
taronaeo
approved these changes
on 2026-05-13
xctan
approved these changes
on 2026-05-14
Reviewers
Assignees
No one assigned
Labels
documentation
build
devops
ggml
merge ready
Milestone
No milestone
Login to write a write a comment.
Login via GitHub