[Piglit] [PATCH 2/2] all.py: filter directories traversed to find shader tests

Dylan Baker dylan at pnwbakers.com
Fri Nov 10 00:29:12 UTC 2017


Quoting Brian Paul (2017-11-09 15:16:33)
> The script searches all files under tests/ and generated_tests/ for
> files ending in ".shader_test".  For each match, a ShaderTest() object
> is created and added to the test list.
> 
> For GL extensions or versions not supported by the driver, this is
> wasted effort.
> 
> This patch looks for directories under spec/ and tries to determine if
> the extension/feature is actually supported by the current driver.  If
> not, it's skipped.
> 
> This somewhat reduces Piglit start up time, but substantially more
> time is spent in the ShaderTest() constructor which actually opens
> and parses each file to determine its GL/ES dependencies.  I'll try
> to address that in the future.
> 
> v2:
> - replace os.path.walk() with my own walk_dir_tree() which avoids
>   decending into subdirs when the parent directory/feature is not supported.
> - Use env var to enable/disable shader directory filtering
> - Also, fix naming conventions and minor formatting issues.
> ---
>  tests/all.py | 73 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 72 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/all.py b/tests/all.py
> index ae4a995..93f66cf 100644
> --- a/tests/all.py
> +++ b/tests/all.py
> @@ -8,12 +8,15 @@ import collections
>  import itertools
>  import os
>  import platform
> +import re
>  
>  import six
>  from six.moves import range
>  
>  from framework import grouptools
> +from framework.test import opengl
>  from framework import options
> +from framework import wflinfo
>  from framework.profile import TestProfile
>  from framework.driver_classifier import DriverClassifier
>  from framework.test import (PiglitGLTest, GleanTest, PiglitBaseTest,
> @@ -210,9 +213,77 @@ profile = TestProfile()  # pylint: disable=invalid-name
>  
>  shader_tests = collections.defaultdict(list)
>  
> +wfl_info = wflinfo.WflInfo()
> +
> +
> +def gl_extension_supported(ext_name):
> +    """Is the named OpenGL extension supported?"""
> +    return ext_name in wfl_info.gl_extensions

There should be one extra newline here.

> +
> +def is_feature_directory_supported(dir_name):
> +    """Determine if dir_name specifies an OpenGL feature (extension or GL
> +    version) which is supported by the host.  If we return False, it means
> +    the extension/version is definitely not supported.  If we return True,
> +    it means the extension/version is possibly suppported.  We're a little
> +    fuzzy because we don't yet parse all the directory name possibilities
> +    (like ES tests).
> +    """
> +    if dir_name[:4] in {"amd_", "arb_", "ati_", "ext_", "khr_", "oes_"}:
> +        # The directory is a GL extension name, but of the format "arb_foo_bar"
> +        # instead of "GL_ARB_foo_bar".  We convert the former into the later
> +        # and check if the extension is supported.
> +        ext_name = "GL_" + dir_name[0:4].upper() + dir_name[4:]
> +        return gl_extension_supported(ext_name)
> +    elif re.match("gl-\d\.\d+", dir_name):

You're using `re` here because you're not checking gl[sl]-es, right? Does it
make more sense to handle ES before desktop GL so you don't need to use re?

> +        # The directory is a GL version
> +        version = dir_name[3:]
> +        return float(version) <= float(wfl_info.gl_version)
> +    elif re.match("glsl-\d\.\d+", dir_name):
> +        # The directory is a GLSL version
> +        version = dir_name[5:]
> +        return float(version) <= float(wfl_info.glsl_version)
> +    else:
> +        # The directory is something else.  Don't skip it.
> +        return True
> +
> +
> +def walk_filter_dir_tree(root):
> +    """Recursively walk the directory tree rooted at 'root'.
> +    If we find a directory path of the form ".../spec/foo/" we'll check if
> +    'foo' is a supported extension/feature/version.  If not, we do not
> +    traverse foo/.  Otherwise, we add continue traversing.
> +    The return value is a list of (dirpath, filename) tuples.
> +    """
> +    curdir = os.path.split(root)[1]
> +    files = []
> +    retval = []
> +
> +    for entry in os.listdir(root):
> +        full_path = os.path.join(root, entry)
> +        if os.path.isdir(full_path):
> +            # Check if we're in a "spec/" direcotry
> +            if curdir == "spec" and not is_feature_directory_supported(entry):
> +                # The directory's tests aren't supported by the driver.
> +                print("Skipping spec/{}".format(entry))
> +            else:
> +                # recursively walk the subdirectory
> +                retval += walk_filter_dir_tree(full_path)
> +        elif os.path.isfile(full_path):
> +            # Add the file to the files list
> +            files += [entry]
> +
> +    retval += [(root, None, files)]

It wont affect anything in our use, but to address potential future problems,
None should be [] here.

> +
> +    return retval

Just a comment, os.walk is a generator, it yields values instead of returning
a sequence. The way you're using them they're interchangeable, but they are not
universally interchangeable. It also means that this will much more memory
intensive than os.walk.

> +
> +
>  # Find and add all shader tests.
>  for basedir in [TESTS_DIR, GENERATED_TESTS_DIR]:
> -    for dirpath, _, filenames in os.walk(basedir):
> +    if os.environ.get("PIGLIT_FILTER_DIRECTORIES"):
> +        files = walk_filter_dir_tree(basedir)
> +    else:
> +        files = os.walk(basedir)
> +    for dirpath, _, filenames in files:
>          groupname = grouptools.from_path(os.path.relpath(dirpath, basedir))
>          for filename in filenames:
>              testname, ext = os.path.splitext(filename)
> -- 
> 1.9.1
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: signature
URL: <https://lists.freedesktop.org/archives/piglit/attachments/20171109/91856045/attachment.sig>


More information about the Piglit mailing list