From 71e1011e64f65b230ecf9ecc74c73b514044eb84 Mon Sep 17 00:00:00 2001
From: zhanghb97 <hongbin2019@iscas.ac.cn>
Date: Wed, 10 Nov 2021 13:18:42 +0800
Subject: [PATCH] Update README.
---
README.md | 24 +++++++++++++++++++++++-
1 file changed, 23 insertions(+), 1 deletion(-)
diff --git a/README.md b/README.md
index 3ee98c2..9d0720f 100644
--- a/README.md
+++ b/README.md
@@ -27,7 +27,7 @@ $ mkdir build && cd build
$ cmake -G Ninja .. \
-DIMAGE_PROCESSING_BENCHMARKS=ON \
-DOpenCV_DIR=/path/to/opencv/build/ \
- -DBUDDY_CONV_OPT_BUILD_DIR=/path/to/buddy-mlir/build \
+ -DBUDDY_CONV_OPT_BUILD_DIR=/path/to/buddy-mlir/build/ \
-DBUDDY_CONV_OPT_STRIP_MINING=<strip mining size, default: 256> \
-DBUDDY_CONV_OPT_ATTR=<ISA vector extension, default: avx512f>
$ ninja image-processing-benchmark
@@ -35,3 +35,25 @@ $ cd bin && ./image-processing-benchmark
```
Note : The convolution implementation in buddy mlir is not feature complete at the moment and it may produce output which differs to some extent from the frameworks used in comparison.
+
+## Deep Learning Benchmark
+
+The deep learning benchmark includes the following e2e models:
+
+- MobileNet
+
+NOTE: We generated the model code with IREE and made appropriate modifications, and then compiled it with the MLIR tool chain.
+
+Run the image processing benchmark:
+
+```
+$ cd buddy-benchmark
+$ mkdir build && cd build
+$ cmake -G Ninja .. \
+ -DDEEP_LEARNING_BENCHMARKS=ON \
+ -DOpenCV_DIR=/path/to/opencv/build/ \
+ -DBUDDY_CONV_OPT_BUILD_DIR=/path/to/buddy-mlir/build/ \
+ -DBUDDY_CONV_OPT_ATTR=<ISA vector extension, default: avx512f>
+$ ninja deep-learning-benchmark
+$ cd bin && ./deep-learning-benchmark
+```
--
GitLab