dmin@mylab-radxa:~/Llama3.2-1B-1024-qairt-v68$ fastrpc_test -a v68
Allocate 4000 bytes from ION heap
Creating sequence of numbers from 0 to 999
Compute sum on domain 3
Retry attempt unsuccessful. Timing out…
ERROR 0x80000600: Failed to compute sum on domain 3
ERROR 0x80000600: Failed to find max on domain 3
[FAIL] libcalculator.so (error code: 0x80000600)
ERROR 0x80000600: Unable to create FastRPC session on domain 3
hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector FARF_RUNTIME (id: 0): -2147482112
ERROR 0x80000600: Unable to create FastRPC session on domain 3
hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector MEM (id: 1): -2147482112
ERROR 0x80000600: Unable to create FastRPC session on domain 3
hap_example function FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
Error occurred with selector PERF (id: 2): -2147482112
[FAIL] libhap_example.so (error code: 0x80000600)
ERROR 0x80000600: Unable to create FastRPC session on domain 3
Exiting…
Test FAILED
Please look at the mini-dm logs or the adb logcat logs for DSP output
[FAIL] libmultithreading.so (error code: 0x80000600)
========================================
Test Summary:
Total tests run: 3
Passed: 0
Failed: 3
Skipped: 0
Hi @Mylab_Radxa your debugging is solid; I think the infrastructure is all up (remoteproc, fastrpc nodes, DMA, shell). IMHO the 0x80000600 at FASTRPC_IOCTL_INIT_CREATE isn’t a fastrpc / kernel / firmware-package regression, it’s more or less a QAIRT version mismatch with the DSP firmware.
You’re on QNN SDK 2.46.0, but the latest DSP firmware Radxa ships (CDSP.HT.2.5.c4-00004-KODIAK-1) only supports QNN interface 2.32.0 (QAIRT 2.42)in my experiences. libQnnHtpV68Skel.so from the 2.46 SDK calls DSP-side ABI the firmware doesn’t expose, so the DSP rejects the session at init-create, which surfaces as 0x80000600. FASTRPC_IOCTL_GET_DSP_INFO succeeding is the right call: the static shell loads fine.
Could you try pin QAIRT/QNN SDK to ≤ 2.42 and regenerate your Llama3.2-1B-1024-qairt-v68 context binary against that version? fastrpc_test -a v68 should then pass.