diff --git a/README.md b/README.md
index 1d26615e3d7bba13e6c25e1c6c2950db67ed7df5..ee69383a6e7147f704d01482dd0650e33b96b8c6 100644
--- a/README.md
+++ b/README.md
@@ -73,3 +73,4 @@ We are providing the following benchmarks with HPVM:
 
 We are also providing [unit tests](/hpvm/test/unitTests) and [regression tests](/hpvm/test/regressionTests).
 
+Benchmark descriptions and instructions on how to compile and run them are [here](/hpvm/test).
diff --git a/hpvm/test/README.md b/hpvm/test/README.md
index 59832c64452d1103d8ff0cd7148a9379611b810f..fbce8dce48510a872ccc5b09e4861c85df271eeb 100644
--- a/hpvm/test/README.md
+++ b/hpvm/test/README.md
@@ -4,28 +4,13 @@ The below benchmarks are provided with HPVM, along with a template Makefile for
 * CUDA_PATH: should point to your local CUDA installation.
 
 ## Parboil
-Several tests from the [parboil suite](http://impact.crhc.illinois.edu/parboil/parboil.aspx) have been ported to HPVM.
-To run one of these tests, navigate to its directory under `parboil/benchmarks/`.
-Tests may be built for the cpu or gpu with hpvm.
-```
-# sgemm example
-cd parboil/benchmarks/sgemm
-# HPVM cpu
-make TARGET=seq VERSION=hpvm
-make run TARGET=seq VERSION=hpvm
-# HPVM gpu
-make TARGET=gpu VERSION=hpvm
-make run TARGET=gpu VERSION=hpvm
-```
+Instructions to compile and run Parboil are provided in the following [README](/hpvm/test/parboil).
 
 ## Harvard Camera Pipeline (HPVM-CAVA)
-Instructions for HPVM-CAVA are provided in the following [README](/hpvm/test/hpvm-cava)
+Instructions to compile and run HPVM-CAVA are provided in the following [README](/hpvm/test/hpvm-cava).
 
-## Pipeline
-```
-make TARGET={seq, gpu}
- ./pipeline-{seq, gpu} datasets/formula1_scaled.mp4
-```
+## Edge Detection Pipeline
+Instructions to compile and run Pipeline are provided in the following [README](/hpvm/test/pipeline).
 
 ## Your own project
 See `template/` for an example Makefile and config.
diff --git a/hpvm/test/parboil/README.md b/hpvm/test/parboil/README.md
index 853b46ed515455fbcb206630a74d5490c79ffd88..f05088d6915e4ead7ec35830be39bd4e57656d77 100644
--- a/hpvm/test/parboil/README.md
+++ b/hpvm/test/parboil/README.md
@@ -1,3 +1,18 @@
+## Compiling and Running Parboil Benchmarks
+Several tests from the [parboil suite](http://impact.crhc.illinois.edu/parboil/parboil.aspx) have been ported to HPVM.
+To run one of these tests, navigate to its directory under `benchmarks/`.
+Tests may be built for the cpu or gpu with hpvm.
+```
+# sgemm example
+cd benchmarks/sgemm
+# HPVM cpu
+make TARGET=seq VERSION=hpvm
+make run TARGET=seq VERSION=hpvm
+# HPVM gpu
+make TARGET=gpu VERSION=hpvm
+make run TARGET=gpu VERSION=hpvm
+```
+
 # Current Benchmark Compatability
 
 | Benchmark | Version | Supported on CPU | Supported on GPU |
diff --git a/hpvm/test/pipeline/DESCRIPTION b/hpvm/test/pipeline/DESCRIPTION
deleted file mode 100644
index 3ecb1184ad24f09a5e23e52269c8959a9a99d87a..0000000000000000000000000000000000000000
--- a/hpvm/test/pipeline/DESCRIPTION
+++ /dev/null
@@ -1,10 +0,0 @@
-A register-tiled matrix-matrix multiplication, with default column-majored
-layout on matrix A and C, but B is transposed.
-
-src/cuda/
-        CUDA version of sgemm('N', 'T', ...)
-
-See also:
-
-Volkov, V., and Demmel, J. W. 2008. Benchmarking GPUs to tune dense linear      
-algebra, 2008 ACM/IEEE Conference on Supercomputing (SC08)
diff --git a/hpvm/test/pipeline/README.md b/hpvm/test/pipeline/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..4a224c6076fafa6b0dece05b921146f95f8355bb
--- /dev/null
+++ b/hpvm/test/pipeline/README.md
@@ -0,0 +1,25 @@
+# Edge Detection Pipeline
+
+HPVM benchmark performing edge detection on a stream of input images.
+
+## Dependencies
+
+Edge detection pipeline depends on `OpenCV==2.4`. To use OpenCV 3.4, go to line 13 in `src/main.cc`
+and replace 
+
+```
+#include "opencv2/ocl/ocl.hpp"
+```
+with
+```
+#include "opencv2/core/ocl.hpp"
+```
+
+Edge detection pipeline is not tested with other versions of OpenCV.
+
+## How to Build and Test
+
+```
+make TARGET={seq, gpu}
+./pipeline-{seq, gpu} datasets/formula1_scaled.mp4
+```