[test-suite] r313711 - litsupport: Extend documentation

Tue Sep 19 20:00:05 PDT 2017

Author: matze
Date: Tue Sep 19 20:00:05 2017
New Revision: 313711

URL: http://llvm.org/viewvc/llvm-project?rev=313711&view=rev
Log:
litsupport: Extend documentation

- Write a paragraph about testing modules and how to enable/disable
  them.
- Write a paragraph on developing new testing modules.
- Improve API documentation for the testplan object and related
  functions.

Modified:
    test-suite/trunk/litsupport/README.md
    test-suite/trunk/litsupport/testplan.py

Modified: test-suite/trunk/litsupport/README.md
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/litsupport/README.md?rev=313711&r1=313710&r2=313711&view=diff
==============================================================================

--- test-suite/trunk/litsupport/README.md (original)
+++ test-suite/trunk/litsupport/README.md Tue Sep 19 20:00:05 2017
@@ -3,7 +3,7 @@ Introduction
 
 This is the benchmark runner for llvm test-suite. It is only available when the
 test-suite was built with cmake. It runs benchmarks, checks the results and
-collects various metrics such as runtime, compiletime or code size.
+collects various metrics such as runtime, compile time or code size.
 
 The runner is implemented as a custom test format for the llvm-lit tool.
 
@@ -48,3 +48,50 @@ install `lit` from the python package in
     # Run benchmarks with reduced lit output but save collected metrics to json
     # file. This format is used by LNT or viewable by test-suite/utils/compare.py.
     $ llvm-lit . -o result.json -s
+
+Testing Modules
+===============
+
+The benchmark runner behaviour is defined and enhanced by testing modules.
+Testing modules can modify the command lines to run the benchmark and register
+callbacks to collect metrics.
+
+The list of modules is defined in the `lit.site.cfg` file
+(`config.test_modules`) which in turn is generated by cmake. The module list is
+influenced by a number of cmake flags. For a complete list consult the cmake
+code; typical examples are:
+
+- `cmake -DTEST_SUITE_RUN_BENCHMARKS=Off` removes the `run` module, so no
+  benchmarks are actually run; this is useful as code size data, compiletime, or
+  compilation statistics can be collected anyway.
+- `cmake -DTEST_SUITE_RUN_UNDER=qemu` enable the `run_under` module and will
+  prefix all benchmark invocations with the specified command (here: `qemu`).
+- `cmake -DTEST_SUITE_REMOTE_HOST=xxx` enabled the `remote` module that uses
+  ssh to run benchmarks on a remote device (assuming shared file systems).
+- `cmake -DTEST_SUITE_PROFILE_GENERATE` compiles benchmark with
+  `-fprofile-instr-generate` and enables the `profilegen` module that runs
+  `llvm-profdata` after running the benchmarks.
+
+Developing New Modules
+----------------------
+
+Testing modules consist of a python module with a `mutatePlan(context, plan)`
+function. Modules can:
+
+- Modify the scripts used to execute the benchmark (`plan.preparescript`,
+	`plan.runscript`, `plan.verifyscript`). A script is a list of strings with
+	shell commands. Modifying the list is best done via the testplan.mutateScript
+	function which sets `context.tmpBase` to a unique for the command to be
+	modified. The `shellcommand` module helps analyzing and modifying posix shell
+	command lines.
+
+- Append a function to the `plan.metric_collectors` list.  The metric collector
+  functions are executed after the benchmark scripts ran.  Metrics must be
+  returned as a dictionary with int, float or string values (anything supported
+  by `lit.Test.toMetricValue()`).
+
+- The `context` object is passed to all testing modules and metric collectors
+  and may be used to communicate information between them. `context.executable`
+  contains the path to the benchmark executable; `context.config` contains the
+  lit configuration coming from a combination of `lit.site.cfg`, `lit.cfg` and
+  `lit.local.cfg` files for the benchmark.

Modified: test-suite/trunk/litsupport/testplan.py
URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/litsupport/testplan.py?rev=313711&r1=313710&r2=313711&view=diff
==============================================================================
--- test-suite/trunk/litsupport/testplan.py (original)
+++ test-suite/trunk/litsupport/testplan.py Tue Sep 19 20:00:05 2017
@@ -10,6 +10,12 @@ import subprocess
 
 
 class TestPlan(object):
+    """Describes how to execute a benchmark and how to collect metrics.
+    A script is a list of strings containing shell commands. The available
+    scripts are: preparescript, runscript, verifyscript, profilescript,
+    metricscripts and are executed in this order.
+    metric_collectors contains a list of functions executed after the scripts
+    finished."""
     def __init__(self):
         self.runscript = []
         self.verifyscript = []
@@ -20,6 +26,10 @@ class TestPlan(object):
 
 
 def mutateScript(context, script, mutator):
+    """Apply `mutator` function to every command in the `script` array of
+    strings. The mutator function is called with `context` and the string to
+    be mutated and must return the modified string. Sets `context.tmpBase`
+    to a path unique to every command."""
     previous_tmpbase = context.tmpBase
     i = 0
     mutated_script = []
@@ -36,6 +46,7 @@ def mutateScript(context, script, mutato
 
 
 def _executeScript(context, script, scriptBaseName, useExternalSh=True):
+    """Execute an array of strings with shellcommands (a script)."""
     if len(script) == 0:
         return "", "", 0, None
 
@@ -74,20 +85,8 @@ def _executeScript(context, script, scri
     return (out, err, exitCode, timeoutInfo)
 
 
-def check_output(commandline, *aargs, **dargs):
-    """Wrapper around subprocess.check_output that logs the command."""
-    logging.info(" ".join(commandline))
-    return subprocess.check_output(commandline, *aargs, **dargs)
-
-
-def check_call(commandline, *aargs, **dargs):
-    """Wrapper around subprocess.check_call that logs the command."""
-    logging.info(" ".join(commandline))
-    return subprocess.check_call(commandline, *aargs, **dargs)
-
-
 def _executePlan(context, plan):
-    """This is the main driver for executing a benchmark."""
+    """Executes a test plan (a TestPlan object)."""
     # Execute PREPARE: part of the test.
     _, _, exitCode, _ = _executeScript(context, plan.preparescript, "prepare")
     if exitCode != 0:
@@ -152,3 +151,15 @@ def executePlanTestResult(context, testp
     for key, value in context.result_metrics.items():
         result.addMetric(key, value)
     return result
+
+
+def check_output(commandline, *aargs, **dargs):
+    """Wrapper around subprocess.check_output that logs the command."""
+    logging.info(" ".join(commandline))
+    return subprocess.check_output(commandline, *aargs, **dargs)
+
+
+def check_call(commandline, *aargs, **dargs):
+    """Wrapper around subprocess.check_call that logs the command."""
+    logging.info(" ".join(commandline))
+    return subprocess.check_call(commandline, *aargs, **dargs)


