From e282915a9c76964551d600bc01602dc53d7a4da1 Mon Sep 17 00:00:00 2001
From: Lianmin Zheng <mercy_zheng@sjtu.edu.cn>
Date: Thu, 16 Aug 2018 16:37:03 -0700
Subject: [PATCH] add -mattr=+neon for all arm cpu target (#1612)
---
apps/benchmark/README.md | 11 +++++++----
python/tvm/target.py | 16 ++++++++--------
tutorials/nnvm/deploy_model_on_rasp.py | 2 +-
3 files changed, 16 insertions(+), 13 deletions(-)
diff --git a/apps/benchmark/README.md b/apps/benchmark/README.md
index e83e47c46..ee22f90dc 100644
--- a/apps/benchmark/README.md
+++ b/apps/benchmark/README.md
@@ -63,8 +63,11 @@ python3 -m tvm.exec.rpc_tracker
python3 arm_cpu_imagenet_bench.py --device mate10pro --rpc-key mate10pro
```
- If your device has a same SoC of the above device, you can reuse these parameters
- (e.g. use `llvm -device=arm_cpu -mode=rk3399 -target=aarch64-linux-gnu` as target).
- Otherwise, you need to tune for your own device, please follow this
- [tutorial](https://docs.tvm.ai/tutorials/autotvm/tune_nnvm_arm.html).
+ If your device has a same or similar SoC of the above devices, you can reuse these parameters.
+ For example, if your SoC is similar to rasp3b, use
+ ```bash
+ python3 arm_cpu_imagenet_bench.py --device rasp3b --rpc-key your_custom_key
+ ```
+ For other devices, to get the best performance, it is recommended that you tune your network by yourself.
+ Please follow this [tutorial](https://docs.tvm.ai/tutorials/autotvm/tune_nnvm_arm.html).
diff --git a/python/tvm/target.py b/python/tvm/target.py
index fed20c391..e2d780f75 100644
--- a/python/tvm/target.py
+++ b/python/tvm/target.py
@@ -427,14 +427,14 @@ def arm_cpu(model='unknown', options=None):
from . import autotvm
trans_table = {
- "pixel2": ["-model=snapdragon835", "-target=arm64-linux-android"],
- "mate10": ["-model=kirin970", "-target=arm64-linux-android"],
- "mate10pro": ["-model=kirin970", "-target=arm64-linux-android"],
- "p20": ["-model=kirin970", "-target=arm64-linux-android"],
- "p20pro": ["-model=kirin970", "-target=arm64-linux-android"],
- "rasp3b": ["-model=bcm2837", "-target=armv7l-linux-gnueabihf"],
- "rk3399": ["-model=rk3399", "-target=aarch64-linux-gnu"],
- "pynq": ["-model=pynq", "-target=armv7a-linux-eabi"],
+ "pixel2": ["-model=snapdragon835", "-target=arm64-linux-android -mattr=+neon"],
+ "mate10": ["-model=kirin970", "-target=arm64-linux-android -mattr=+neon"],
+ "mate10pro": ["-model=kirin970", "-target=arm64-linux-android -mattr=+neon"],
+ "p20": ["-model=kirin970", "-target=arm64-linux-android -mattr=+neon"],
+ "p20pro": ["-model=kirin970", "-target=arm64-linux-android -mattr=+neon"],
+ "rasp3b": ["-model=bcm2837", "-target=armv7l-linux-gnueabihf -mattr=+neon"],
+ "rk3399": ["-model=rk3399", "-target=aarch64-linux-gnu -mattr=+neon"],
+ "pynq": ["-model=pynq", "-target=armv7a-linux-eabi -mattr=+neon"],
}
pre_defined_opt = trans_table.get(model, ["-model=%s" % model])
diff --git a/tutorials/nnvm/deploy_model_on_rasp.py b/tutorials/nnvm/deploy_model_on_rasp.py
index c11f202c1..40dbdaeb0 100644
--- a/tutorials/nnvm/deploy_model_on_rasp.py
+++ b/tutorials/nnvm/deploy_model_on_rasp.py
@@ -154,7 +154,7 @@ if local_demo:
else:
target = tvm.target.arm_cpu('rasp3b')
# The above line is a simple form of
- # target = tvm.target.create('llvm -devcie=arm_cpu -target=armv7l-linux-gnueabihf')
+ # target = tvm.target.create('llvm -devcie=arm_cpu -model=bcm2837 -target=armv7l-linux-gnueabihf -mattr=+neon')
with nnvm.compiler.build_config(opt_level=2, add_pass=['AlterOpLayout']):
graph, lib, params = nnvm.compiler.build(
--
GitLab