16:19 tomeu: finally! got ssdlite mobiledet working on the rk3588
16:20 tomeu: 32ms in a single 2 TOPS core
16:22 phh: cool
16:22 tomeu: the blob takes twice as much, don't know why...
16:23 phh: does that include exec() library loading, or is it like ioctl-to-ioctl?
16:23 tomeu: hmm, what do you mean by these?
16:24 tomeu: this is my Mesa driver, calling into their kernel driver
16:24 phh: I mean how do you measure the time taken by the blob?
16:25 tomeu: just measuring wall clock time around
16:25 tomeu: outputs = rknn.inference(inputs=[input], data_type="uint8")
16:26 phh: after previous runs in case they "compile" the model or something?
16:29 tomeu: ah, I would expect it to be ready to inference, but I haven't studied they API in depth (do they even have docs?)
16:30 phh: that's my expectation as well, but I think we've already established the blob isn't super reasonable...
16:31 tomeu: yeah...
18:56 f_: >finally! got ssdlite mobiledet working on the rk3588
18:56 f_: tomeu: \o/ !!
22:21 lvrp16_: wow...
22:22 lvrp16_: can't wait for the rest of tRK3588 to catch up