bpo-36044: Reduce number of unit tests run for PGO build (GH-14702)

Reduce the number of unit tests run for the PGO generation task.  This
speeds up the task by a factor of about 15x.  Running the full unit test
suite is slow.  This change may result in a slightly less optimized build
since not as many code branches will be executed.  If you are willing to
wait for the much slower build, the old behavior can be restored using
'./configure [..] PROFILE_TASK="-m test --pgo-extended"'.  We make no
guarantees as to which PGO task set produces a faster build.  Users who
care should run their own relevant benchmarks as results can depend on
the environment, workload, and compiler tool chain.
(cherry picked from commit 4e16a4a3112161a5c6981c0588142d4a4673a934)

Co-authored-by: Neil Schemenauer <nas-github@arctrix.com>

7 files changed, 96 insertions, 4 deletions

diff --git a/Lib/test/libregrtest/cmdline.py b/Lib/test/libregrtest/cmdline.py
index 9f1bf68..c8fedc7 100644
--- a/Lib/test/libregrtest/cmdline.py
+++ b/Lib/test/libregrtest/cmdline.py
@@ -264,7 +264,9 @@ def _create_parser():
                        help='only write the name of test cases that will be run'
                             ' , don\'t execute them')
     group.add_argument('-P', '--pgo', dest='pgo', action='store_true',
-                       help='enable Profile Guided Optimization training')
+                       help='enable Profile Guided Optimization (PGO) training')
+    group.add_argument('--pgo-extended', action='store_true',
+                       help='enable extended PGO training (slower training)')
     group.add_argument('--fail-env-changed', action='store_true',
                        help='if a test file alters the environment, mark '
                             'the test as failed')
@@ -344,6 +346,8 @@ def _parse_args(args, **kwargs):
         parser.error("-G/--failfast needs either -v or -W")
     if ns.pgo and (ns.verbose or ns.verbose2 or ns.verbose3):
         parser.error("--pgo/-v don't go together!")
+    if ns.pgo_extended:
+        ns.pgo = True  # pgo_extended implies pgo
 
     if ns.nowindows:
         print("Warning: the --nowindows (-n) option is deprecated. "
diff --git a/Lib/test/libregrtest/main.py b/Lib/test/libregrtest/main.py
index e227425..78b6790 100644
--- a/Lib/test/libregrtest/main.py
+++ b/Lib/test/libregrtest/main.py
@@ -17,6 +17,7 @@ from test.libregrtest.runtest import (
     INTERRUPTED, CHILD_ERROR, TEST_DID_NOT_RUN,
     PROGRESS_MIN_TIME, format_test_result, is_failed)
 from test.libregrtest.setup import setup_tests
+from test.libregrtest.pgo import setup_pgo_tests
 from test.libregrtest.utils import removepy, count, format_duration, printlist
 from test import support
 
@@ -214,6 +215,9 @@ class Regrtest:
 
         removepy(self.tests)
 
+        # add default PGO tests if no tests are specified
+        setup_pgo_tests(self.ns)
+
         stdtests = STDTESTS[:]
         nottests = NOTTESTS.copy()
         if self.ns.exclude:
diff --git a/Lib/test/libregrtest/pgo.py b/Lib/test/libregrtest/pgo.py
new file mode 100644
index 0000000..327f193
--- /dev/null
+++ b/Lib/test/libregrtest/pgo.py
@@ -0,0 +1,52 @@
+# Set of tests run by default if --pgo is specified.  The tests below were
+# chosen based on the following criteria: either they exercise a commonly used
+# C extension module or type, or they run some relatively typical Python code.
+# Long running tests should be avoided because the PGO instrumented executable
+# runs slowly.
+PGO_TESTS = [
+    'test_array',
+    'test_base64',
+    'test_binascii',
+    'test_binop',
+    'test_bisect',
+    'test_bytes',
+    'test_cmath',
+    'test_codecs',
+    'test_collections',
+    'test_complex',
+    'test_dataclasses',
+    'test_datetime',
+    'test_decimal',
+    'test_difflib',
+    'test_embed',
+    'test_float',
+    'test_fstring',
+    'test_functools',
+    'test_generators',
+    'test_hashlib',
+    'test_heapq',
+    'test_int',
+    'test_itertools',
+    'test_json',
+    'test_long',
+    'test_math',
+    'test_memoryview',
+    'test_operator',
+    'test_ordered_dict',
+    'test_pickle',
+    'test_pprint',
+    'test_re',
+    'test_set',
+    'test_statistics',
+    'test_struct',
+    'test_tabnanny',
+    'test_time',
+    'test_unicode',
+    'test_xml_etree',
+    'test_xml_etree_c',
+]
+
+def setup_pgo_tests(ns):
+    if not ns.args and not ns.pgo_extended:
+        # run default set of tests for PGO training
+        ns.args = PGO_TESTS[:]
diff --git a/Makefile.pre.in b/Makefile.pre.in
index 88abb56..6a9f4b5 100644
--- a/Makefile.pre.in
+++ b/Makefile.pre.in
@@ -255,9 +255,10 @@ TCLTK_INCLUDES=	@TCLTK_INCLUDES@
 TCLTK_LIBS=	@TCLTK_LIBS@
 
 # The task to run while instrumented when building the profile-opt target.
-# We exclude unittests with -x that take a rediculious amount of time to
-# run in the instrumented training build or do not provide much value.
-PROFILE_TASK=-m test.regrtest --pgo
+# To speed up profile generation, we don't run the full unit test suite
+# by default. The default is "-m test --pgo". To run more tests, use
+# PROFILE_TASK="-m test --pgo-extended"
+PROFILE_TASK=	@PROFILE_TASK@
 
 # report files for gcov / lcov coverage report
 COVERAGE_INFO=	$(abs_builddir)/coverage.info
diff --git a/Misc/NEWS.d/next/Build/2019-07-11-01-28-24.bpo-36044.gIgfiJ.rst b/Misc/NEWS.d/next/Build/2019-07-11-01-28-24.bpo-36044.gIgfiJ.rst
new file mode 100644
index 0000000..177c4cb
--- /dev/null
+++ b/Misc/NEWS.d/next/Build/2019-07-11-01-28-24.bpo-36044.gIgfiJ.rst
@@ -0,0 +1,9 @@
+Reduce the number of unit tests run for the PGO generation task.  This
+speeds up the task by a factor of about 15x.  Running the full unit test
+suite is slow.  This change may result in a slightly less optimized build
+since not as many code branches will be executed.  If you are willing to
+wait for the much slower build, the old behavior can be restored using
+'./configure [..] PROFILE_TASK="-m test --pgo-extended"'.  We make no
+guarantees as to which PGO task set produces a faster build.  Users who
+care should run their own relevant benchmarks as results can depend on
+the environment, workload, and compiler tool chain.
diff --git a/configure b/configure
index 6e7f277..cb5f130 100755
--- a/configure
+++ b/configure
@@ -686,6 +686,7 @@ target_vendor
 target_cpu
 target
 LLVM_AR
+PROFILE_TASK
 DEF_MAKE_RULE
 DEF_MAKE_ALL_RULE
 ABIFLAGS
@@ -856,6 +857,7 @@ LDFLAGS
 LIBS
 CPPFLAGS
 CPP
+PROFILE_TASK
 PKG_CONFIG
 PKG_CONFIG_PATH
 PKG_CONFIG_LIBDIR'
@@ -1559,6 +1561,8 @@ Some influential environment variables:
   CPPFLAGS    (Objective) C/C++ preprocessor flags, e.g. -I<include dir> if
               you have headers in a nonstandard directory <include dir>
   CPP         C preprocessor
+  PROFILE_TASK
+              Python args for PGO generation task
   PKG_CONFIG  path to pkg-config utility
   PKG_CONFIG_PATH
               directories to add to pkg-config's search path
@@ -6426,6 +6430,16 @@ else
   DEF_MAKE_RULE="all"
 fi
 
+
+{ $as_echo "$as_me:${as_lineno-$LINENO}: checking PROFILE_TASK" >&5
+$as_echo_n "checking PROFILE_TASK... " >&6; }
+if test -z "$PROFILE_TASK"
+then
+	PROFILE_TASK='-m test --pgo'
+fi
+{ $as_echo "$as_me:${as_lineno-$LINENO}: result: $PROFILE_TASK" >&5
+$as_echo "$PROFILE_TASK" >&6; }
+
 # Make llvm-relatec checks work on systems where llvm tools are not installed with their
 # normal names in the default $PATH (ie: Ubuntu).  They exist under the
 # non-suffixed name in their versioned llvm directory.
diff --git a/configure.ac b/configure.ac
index 324ce0b..b31ed24 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1293,6 +1293,14 @@ else
   DEF_MAKE_RULE="all"
 fi
 
+AC_ARG_VAR(PROFILE_TASK, Python args for PGO generation task)
+AC_MSG_CHECKING(PROFILE_TASK)
+if test -z "$PROFILE_TASK"
+then
+	PROFILE_TASK='-m test --pgo'
+fi
+AC_MSG_RESULT($PROFILE_TASK)
+
 # Make llvm-relatec checks work on systems where llvm tools are not installed with their
 # normal names in the default $PATH (ie: Ubuntu).  They exist under the
 # non-suffixed name in their versioned llvm directory.