llama.cpp
Add experimental ggml-hexagon backend for the Hexagon NPU
#16547
Merged

Add experimental ggml-hexagon backend for the Hexagon NPU #16547

max-krasnyansky merged 37 commits into ggml-org:master from CodeLinaro:hexagon
max-krasnyansky
github-actions github-actions added documentation
github-actions github-actions added ggml
max-krasnyansky max-krasnyansky requested a review from lhez lhez 64 days ago
jeffbolznv
max-krasnyansky
ggerganov
max-krasnyansky
ggerganov
max-krasnyansky max-krasnyansky force pushed 63 days ago
slaren
max-krasnyansky max-krasnyansky force pushed 62 days ago
max-krasnyansky
max-krasnyansky
github-actions github-actions added devops
max-krasnyansky max-krasnyansky force pushed 62 days ago
max-krasnyansky
CISC
max-krasnyansky max-krasnyansky force pushed 61 days ago
github-actions github-actions added script
max-krasnyansky
jeffbolznv
ggerganov
ggerganov commented on 2025-10-16
max-krasnyansky
max-krasnyansky max-krasnyansky marked this pull request as ready for review 59 days ago
max-krasnyansky max-krasnyansky requested a review from CISC CISC 59 days ago
max-krasnyansky max-krasnyansky requested a review from slaren slaren 59 days ago
max-krasnyansky
github-actions github-actions added python
max-krasnyansky max-krasnyansky force pushed 59 days ago
max-krasnyansky max-krasnyansky force pushed 57 days ago
max-krasnyansky
slaren
slaren commented on 2025-10-21
slaren
max-krasnyansky
max-krasnyansky
slaren
slaren commented on 2025-10-22
slaren
ggerganov
ggerganov commented on 2025-10-22
max-krasnyansky
max-krasnyansky model: add support for extra bufs for all devices
06253824
max-krasnyansky hexagon: add experimental ggml-hexagon backend for the Hexagon NPU
80dc8e80
max-krasnyansky hexagon: fix format checker errors
ec4436f7
max-krasnyansky hexagon: update readme and cmake presets
aa65f212
max-krasnyansky ci: add android-ndk-build jobs that build plain ARM64 and Snapdragon …
647fa3de
max-krasnyansky hexagon: add simple graph optimizer for stacking MUL_MAT ops with the…
da7caac4
max-krasnyansky hexagon: move ADB helper scripts into scripts/snapdragon/adb
bbbc8eae
max-krasnyansky hexagon: replace all f/printfs with GGML_LOG_...
cc7dbd4b
max-krasnyansky readme: add hexagon to the list supported backends
69a8047e
max-krasnyansky hexagon: stack malmuts with quantized inputs only
debdb3b4
max-krasnyansky hexagon: add TODO for fixing issues in hexagon_graph_optimize
3475e29b
max-krasnyansky hexagon: update to hex-sdk 6.4.0 and add scripts for running on QDC
1e750df0
max-krasnyansky scripts: fix lint errors
8e7d8b5a
max-krasnyansky scripts: update qdc pytest script to make linter happy
20aa6897
max-krasnyansky hexagon: add reduce sum in fp32
03e2b9c3
max-krasnyansky hexagon: reduce number of vector stores in matmul output
384164dc
max-krasnyansky hexagon: remove the need for vdelta in reduce-multiply-x8
a314eb69
max-krasnyansky hexagon: consistent use of reduce_sum_fp32 for row_sums
7f2d00bd
max-krasnyansky hexagon: some more matmul optimizations and comments
5de19f8b
max-krasnyansky hexagon: update cmake presets
cf0242e3
max-krasnyansky hexagon: add OPMASK support for run-bench.sh wrapper
250e3a66
max-krasnyansky hexagon: update to use GGML_BACKEND_API
08a97e63
max-krasnyansky hexagon: remove unused logic for setting tensor flags for the views
6d2d0bd2
max-krasnyansky hexagon: add asserts to set/get_tensor to make sure we handle complet…
18d7d204
max-krasnyansky hexagon: use cpy_tensor slow path for non-host buffers
26a90a0b
max-krasnyansky hexagon: error checks in the buffer allocator
a8e5ad82
max-krasnyansky cmake: move include(extProj) under ggml-hexagon
dc001b9f
max-krasnyansky hexagon: don't forget to delete the backend on free
c749b869
max-krasnyansky hexagon: set/get_tensor size assert apply only to quantized tensors
0c01229e
max-krasnyansky hexagon: reintroduce HEX_VERBOSE wrapper for GGML_LOG_DEBUG for now
62ef4eba
max-krasnyansky docs: typos in hexagon developer docs (libggm-...)
19041f7d
max-krasnyansky hexagon: overhaul error handling in the session/device allocation
3e4ff739
max-krasnyansky max-krasnyansky force pushed to 3e4ff739 54 days ago
max-krasnyansky
slaren
ggerganov
ggerganov commented on 2025-10-22
max-krasnyansky hexagon: update cmake presets to enable fp16 vectors
6acc2854
slaren
slaren commented on 2025-10-22
max-krasnyansky hexagon: remove unused time_usec function
dda466cf
slaren
slaren approved these changes on 2025-10-22
max-krasnyansky hexagon: don't forget to release buffer contexts
b0e5beb9
max-krasnyansky hexagon: fixed indents in hvx-utils (missed clang-format auto-format …
3049de50
max-krasnyansky hexagon: remove custom can_repeat function and use ggml_can_repeat
f7d74118
max-krasnyansky max-krasnyansky merged 63d2fc46 into master 54 days ago
DamonFool

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone