Name and Version
version: 8532 (0a524f2)
built with Clang 17.0.2 for Android aarch64
Operating systems
Other? (Please let us know in description)
GGML backends
Hexagon
Hardware
sa8295p
Models
MAI-UI-2B.Q8_0.gguf
Problem description & steps to reproduce
I am attempting to build and run llama.cpp with the Hexagon backend on a Qualcomm SA8295P platform (Snapdragon Digital Chassis / Cockpit) running Android.
I followed the official build instructions provided in docs/backend/snapdragon.md. While the build process completed successfully, the application crashes during runtime initialization when calling dspqueue_create.
First Bad Commit
No response
Relevant log output
atlas:/data/local/tmp/llamacpp # LD_LIBRARY_PATH=$PWD/lib ADSP_LIBRARY_PATH=$PWD/lib GGML_HEXAGON_ARCH=v68 ./llama-cli -m MAI-UI-2B.Q8_0.gguf -p "hello, who are you" -c 4096 --device HTP0
ggml_opencl: selected platform: 'QUALCOMM Snapdragon(TM)'
ggml_opencl: device: 'QUALCOMM Adreno(TM) (OpenCL 2.0 Adreno(TM) 690)'
ggml_opencl: OpenCL driver: OpenCL 3.0 QUALCOMM build: commit unknown Compiler E031.40.08.02
ggml_opencl: vector subgroup broadcast support: false
ggml_opencl: device FP16 support: true
ggml_opencl: mem base addr align: 128
ggml_opencl: max mem alloc size: 1024 MB
ggml_opencl: device max image buffer size (pixels): 134217728
ggml_opencl: device max workgroup size: 1024
ggml_opencl: SVM coarse grain buffer support: true
ggml_opencl: SVM fine grain buffer support: false
ggml_opencl: SVM fine grain system support: false
ggml_opencl: SVM atomics support: false
ggml_opencl: flattening quantized weights representation as struct of arrays (GGML_OPENCL_SOA_Q)
ggml_opencl: using kernels optimized for Adreno (GGML_OPENCL_USE_ADRENO_KERNELS)
ggml_opencl: loading OpenCL kernels......................................................................................................
ggml_opencl: default device: 'QUALCOMM Adreno(TM) (OpenCL 2.0 Adreno(TM) 690)'
ggml-hex: Loading driver libcdsprpc.so
ggml-hex: Hexagon backend (experimental) : allocating new registry : ndev 1
ggml-hex: forcing ndev to 1 for SoCs archs lower than v75.
ggml-hex: Hexagon Arch version v68
ggml-hex: allocating new session: HTP0
ggml-hex: new session: HTP0 : session-id 0 domain-id 3 uri file:///libggml-htp-v68.so?htp_iface_skel_handle_invoke&_modver=1.0&_dom=cdsp&_session=0 handle 0xb40000777dbee190
ggml-hex: HTP0 dspqueue_create failed: 0x00000014
ggml-hex: releasing session: HTP0
ggml-hex: failed to create device/session 0
/workspace/code/llama.cpp/ggml/src/ggml-backend.cpp:483: GGML_ASSERT(device) failed
0: 0x7964d6b9d0
1: 0x7964d6b8dc ggml_print_backtrace
2: 0x7964d6bbe0 ggml_abort
3: 0x7964d871e4 ggml_backend_dev_description
4: 0x79619a50d0 ggml_backend_dev_by_name
5: 0x57dfc50cd8
6: 0x57dfc49e4c
7: 0x57dfc68664
8: 0x57dfc5f808
9: 0x57dfba49a4
10: 0x796231387c __libc_init
Aborted
Name and Version
version: 8532 (0a524f2)
built with Clang 17.0.2 for Android aarch64
Operating systems
Other? (Please let us know in description)
GGML backends
Hexagon
Hardware
sa8295p
Models
MAI-UI-2B.Q8_0.gguf
Problem description & steps to reproduce
I am attempting to build and run llama.cpp with the Hexagon backend on a Qualcomm SA8295P platform (Snapdragon Digital Chassis / Cockpit) running Android.
I followed the official build instructions provided in docs/backend/snapdragon.md. While the build process completed successfully, the application crashes during runtime initialization when calling dspqueue_create.
First Bad Commit
No response
Relevant log output
atlas:/data/local/tmp/llamacpp # LD_LIBRARY_PATH=$PWD/lib ADSP_LIBRARY_PATH=$PWD/lib GGML_HEXAGON_ARCH=v68 ./llama-cli -m MAI-UI-2B.Q8_0.gguf -p "hello, who are you" -c 4096 --device HTP0
ggml_opencl: selected platform: 'QUALCOMM Snapdragon(TM)'
ggml_opencl: device: 'QUALCOMM Adreno(TM) (OpenCL 2.0 Adreno(TM) 690)'
ggml_opencl: OpenCL driver: OpenCL 3.0 QUALCOMM build: commit unknown Compiler E031.40.08.02
ggml_opencl: vector subgroup broadcast support: false
ggml_opencl: device FP16 support: true
ggml_opencl: mem base addr align: 128
ggml_opencl: max mem alloc size: 1024 MB
ggml_opencl: device max image buffer size (pixels): 134217728
ggml_opencl: device max workgroup size: 1024
ggml_opencl: SVM coarse grain buffer support: true
ggml_opencl: SVM fine grain buffer support: false
ggml_opencl: SVM fine grain system support: false
ggml_opencl: SVM atomics support: false
ggml_opencl: flattening quantized weights representation as struct of arrays (GGML_OPENCL_SOA_Q)
ggml_opencl: using kernels optimized for Adreno (GGML_OPENCL_USE_ADRENO_KERNELS)
ggml_opencl: loading OpenCL kernels......................................................................................................
ggml_opencl: default device: 'QUALCOMM Adreno(TM) (OpenCL 2.0 Adreno(TM) 690)'
ggml-hex: Loading driver libcdsprpc.so
ggml-hex: Hexagon backend (experimental) : allocating new registry : ndev 1
ggml-hex: forcing ndev to 1 for SoCs archs lower than v75.
ggml-hex: Hexagon Arch version v68
ggml-hex: allocating new session: HTP0
ggml-hex: new session: HTP0 : session-id 0 domain-id 3 uri file:///libggml-htp-v68.so?htp_iface_skel_handle_invoke&_modver=1.0&_dom=cdsp&_session=0 handle 0xb40000777dbee190
ggml-hex: HTP0 dspqueue_create failed: 0x00000014
ggml-hex: releasing session: HTP0
ggml-hex: failed to create device/session 0
/workspace/code/llama.cpp/ggml/src/ggml-backend.cpp:483: GGML_ASSERT(device) failed
0: 0x7964d6b9d0
1: 0x7964d6b8dc ggml_print_backtrace
2: 0x7964d6bbe0 ggml_abort
3: 0x7964d871e4 ggml_backend_dev_description
4: 0x79619a50d0 ggml_backend_dev_by_name
5: 0x57dfc50cd8
6: 0x57dfc49e4c
7: 0x57dfc68664
8: 0x57dfc5f808
9: 0x57dfba49a4
10: 0x796231387c __libc_init
Aborted