summaryrefslogtreecommitdiff
path: root/README
diff options
context:
space:
mode:
authorSamuel Pitoiset <samuel.pitoiset@gmail.com>2018-04-06 10:23:56 +0200
committerSamuel Pitoiset <samuel.pitoiset@gmail.com>2018-04-06 10:23:56 +0200
commitacee7b888285c635f17b36cd13f8a6b6065b88e2 (patch)
tree88104e0ffc430b031acbe9fcf7b28ddf4b828488 /README
initial commit; import existing files
Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Diffstat (limited to 'README')
-rw-r--r--README44
1 files changed, 44 insertions, 0 deletions
diff --git a/README b/README
new file mode 100644
index 0000000..9335161
--- /dev/null
+++ b/README
@@ -0,0 +1,44 @@
+=== What ===
+
+vkpipeline-db is quite similar to shader-db, but for Vulkan. It allows to capture
+graphics and compute pipelines (including shaders and states) and to replay
+them in order to get shader stats.
+
+Currently vkpipeline-db can only report shader stats if VK_AMD_shader_info is
+exposed by the underlying Vulkan driver.
+
+=== Compiling ===
+
+$ mkdir build
+$ cd build
+$ cmake ..
+$ make
+
+=== Capturing shaders ===
+
+One component of vkpipeline-db is a Vulkan layer that allows to capture
+pipelines.
+
+# Set up the layer
+$ export VK_LAYER_PATH=$VK_LAYER_PATH:path_to_VkLayer_vkpipeline_db.json
+$ export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:path_to_libVkLayer_vkpipeline_db.so
+$ export VK_INSTANCE_LAYERS=VK_LAYER_vkpipeline_db
+
+# Create directory for pipeline files
+$ mkdir dirpath
+
+# Set up the pipeline capture directory.
+$ export VKPIPELINE_DB_CAPTURE_PATH=dirpath
+
+# Run your application
+$ yourapp
+
+All captured pipelines should be in 'dirpath'.
+
+=== Running shaders ===
+
+$ ./run dirpath
+
+=== Analysis ===
+
+$ ./radv-report.py old-run new-run