Required:
Optional:
msvc
command prompt (e.g. VS2013 x64 Native Tools Command Prompt) and run cmake
command for configuration:
cd source-root
mkdir build
cd build
cmake -G “NMake Makefiles” -DCMAKE_BUILD_TYPE=Release -DARNOLD_ROOT=path-to-arnold-sdk -DCMAKE_INSTALL_PREFIX=install-dir ..cmake
command for configuration:
cd source-root
mkdir build
cd build
cmake -G “Unix Makefiles” -DCMAKE_BUILD_TYPE=Release -DARNOLD_ROOT=path-to-arnold-sdk -DCMAKE_INSTALL_PREFIX=install-dir ..python testsuite/runtest.py render –ai path-to-MtoA -l rlShaders/shaders –sn 1..10 This command would render test images of test case 1 to 10 and compare the differences against the reference image with idiff
tool from OpenImageIO.
Since I only have trial version of Arnold renderer, there are watermarks in all the reference images. Which means the images can’t be 100% identical due to the noise in wartermarks. Thus I use RMS error as a metric for comparison instead of the result directly returned from idiff
.
To view the reference and test images together, we could use display
sub-command:
python testsuite/runtest.py display –sn 1 3 5 This will popup an image viewer (iv
tool from OpenImageIO) loaded with reference/test image, we can use page-up/down to switch the image.