Kazeia-engine/dist/lib
Richard Loyer 0407ac8b71 dist: ROOT CAUSE crash dense-HTP = matmul HMX (fp16) ; fix = HVX pour le dense
Analyse complète (méthode SSM_CONV) du crash dense-HTP:
- bench HTP synthétique (pp512, +tg8) = OK ; cli/engine texte réel crashe 0x2e > ~14 tokens.
- OPFILTER bisect: SOFT_MAX/ROPE/RMS_NORM/GET_ROWS/CPY/CONT/ADD n'arrêtent PAS le crash ; MUL_MAT->CPU OUI.
- GGML_HEXAGON_USE_HMX=0 (matmul HVX) supprime le crash, =1 le reproduit. => le matmul HMX (fp16 tile)
  faute sur les activations réelles du dense (massive activations Qwen3 > plage fp16 ; bench synthétique
  ne déclenche pas l'overflow). Analogue du SSM_CONV pour la 3.5, mais ici c'est le matmul HMX coeur.

Câblage: lecture general.architecture dans le GGUF AVANT init backend -> hybride (qwen35/qwen3next):
HMX on + option C (préservé) ; dense: USE_HMX=0 + HTP contexte-unique (HVX, ~40 prefill =3x CPU, decode ~11,
stable). Repli CPU si pas de HTP. Validé test_jni_native: dense (HVX HTP) + 3.5 (option C) cohérents.
TODO backend: matmul HMX robuste aux activations hors plage fp16.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-27 12:42:15 +02:00
..
libc++_shared.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-base.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-cpu.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-hexagon.so dist: prefill HTP corrigé (SSM_CONV gate backend), JNI option C sans OPFILTER 2026-05-27 10:39:08 +02:00
libggml-htp-v68.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-htp-v69.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-htp-v73.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-htp-v75.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-htp-v79.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml-htp-v81.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libggml.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libkazeia_engine.so dist: ROOT CAUSE crash dense-HTP = matmul HMX (fp16) ; fix = HVX pour le dense 2026-05-27 12:42:15 +02:00
libllama-common.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00
libllama.so Engine sur fork qualcomm: decode dense 15.5>pte, 3.5 ~11, KVq8+t4+fa+lm_head-Q4; prefill 93/285; RAG cross 24-28k 2026-05-26 15:53:14 +02:00