xref: /OK3568_Linux_fs/external/rknpu2/examples/rknn_matmul_api_demo/README.md (revision 4882a59341e53eb6f0b4789bf948001014eff981)
1<!--
2 * @Author: Chifred
3 * @Date: 2023-04-25 17:32:21
4 * @LastEditTime: 2023-04-25 17:47:01
5 * @Editors: Chifred
6 * @Description: TODO
7-->
8rknn_matmul_demo is a example which performs int8 matrix multiplication using the RKNPU matmul C API.
9
10Usage:
11
12./rknn_matmul_demo
13
14The following <TARGET_PLATFORM> represents RK356X or RK3588
15
16# Aarch64 Linux Demo
17## Build
18
19modify `GCC_COMPILER` on `build-linux_<TARGET_PLATFORM>.sh` for target platform,
20
21then execute
22
23```
24./build-linux_<TARGET_PLATFORM>.sh
25```
26
27## Install
28
29Copy install/rknn_matmul_demo_Linux to the devices under /userdata/.
30
31- If you use rockchip's evb board, you can use the following way:
32
33Connect device and push the program and rknn model to `/userdata`
34
35```
36adb push install/rknn_matmul_demo_Linux /userdata/
37```
38
39- If your board has sshd service, you can use scp or other methods to copy the program and rknn model to the board.
40
41## Run
42
43```
44adb shell
45cd /userdata/rknn_matmul_demo_Linux/
46```
47
48```
49export LD_LIBRARY_PATH=./lib
50./rknn_matmul_demo
51```
52
53# Android Demo
54## Build
55
56modify `ANDROID_NDK_PATH` on `build-android_<TARGET_PLATFORM>.sh` for target platform, then execute
57
58```
59./build-android_<TARGET_PLATFORM>.sh
60```
61
62## Install
63
64connect device and push build output into `/data`
65
66```
67adb push install/rknn_matmul_demo_Android /data/
68```
69
70## Run
71
72```
73adb shell
74cd /data/rknn_matmul_demo_Android/
75```
76
77```
78export LD_LIBRARY_PATH=./lib
79./rknn_matmul_demo
80```
81