### **Background:**
I've been following the auto-tuning procedure detailed in this documentation
link: https://tvm.apache.org/docs/how_to/tune_with_autotvm/tune_relay_arm.html
The LocalBuilder class used in that tutorial creates multiple BuildResult
objects from which the RPCRunner class ext
Hello,
I'm testing tvm on android phone via RPC. I'm wondering how to know which flags
I have to use, e.g.
I have phone with mt6762v platform, which has cortex A53 procesor and PowerVR
GPU. Android is aarch64, so I decided to you following flags: `llvm
-mtriple=aarch64-linux-gnu -device=arm_
Thanks @gilfr06 for reporting this. I don't have an answer why this happens at
the moment, but I have filed an issue for it and we will work on it.
https://github.com/apache/tvm/issues/10568
---
[Visit
Topic](https://discuss.tvm.apache.org/t/aot-ordering-of-functions-definition-is-random/1
Hi,
I noticed that, when generating AOT code, the order of the functions definition
is random.
So if I launch twice the same AOT generation, I get differences in produced C
files due to this random ordering that makes change tracking rather difficult.
Do you know why TVM AOT codegen behaves lik
Hello! I would like to ask a question about darknet. The version of TVM is
v0.9.dev0. When I run the following code, I get a "kernel died" error.
```python
from tvm.relay.testing.darknet import __darknetffi__
from cffi import FFI
lib_path = "/home/zhangmengbo/tvm/Lsw/tvmtest/deploy_detection/l
As far as I know this is still not directly possible. If your main goal is to
look at the generated tir have a look at my previous answer to:
https://discuss.tvm.apache.org/t/tvm-terms-relay-topi-tir-te/6474/6
If you really need to generate to implement somthing like a `relay.lower` you
could
