llvm-project/lldb/tools/lldb-perf/darwin/sketch/sketch.cpp

381 lines
11 KiB
C++
Raw Normal View History

//===-- sketch.cpp ----------------------------------------------*- C++ -*-===//
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
//
// The LLVM Compiler Infrastructure
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
//
//===----------------------------------------------------------------------===//
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
#include <CoreFoundation/CoreFoundation.h>
#include "lldb-perf/lib/Timer.h"
#include "lldb-perf/lib/Metric.h"
#include "lldb-perf/lib/Measurement.h"
#include "lldb-perf/lib/TestCase.h"
#include "lldb-perf/lib/Xcode.h"
#include <iostream>
#include <unistd.h>
#include <fstream>
#include <getopt.h>
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
using namespace lldb_perf;
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
static struct option g_long_options[] = {
{ "verbose", no_argument, NULL, 'v' },
{ "sketch", required_argument, NULL, 'c' },
{ "foobar", required_argument, NULL, 'f' },
{ "out-file", required_argument, NULL, 'o' },
{ NULL, 0, NULL, 0 }
};
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
class SketchTest : public TestCase
{
public:
SketchTest () :
m_fetch_frames_measurement ([this] () -> void
{
Xcode::FetchFrames (GetProcess(),false,false);
}, "fetch-frames", "time to dump backtrace for every frame in every thread"),
m_file_line_bp_measurement([this] (const char* file, uint32_t line) -> void
{
Xcode::CreateFileLineBreakpoint(GetTarget(), file, line);
}, "file-line-bkpt", "time to set a breakpoint given a file and line"),
m_fetch_modules_measurement ([this] () -> void
{
Xcode::FetchModules(GetTarget());
}, "fetch-modules", "time to get info for all modules in the process"),
m_fetch_vars_measurement([this] (int depth) -> void
{
SBProcess process (GetProcess());
auto threads_count = process.GetNumThreads();
for (size_t thread_num = 0; thread_num < threads_count; thread_num++)
{
SBThread thread(process.GetThreadAtIndex(thread_num));
SBFrame frame(thread.GetFrameAtIndex(0));
Xcode::FetchVariables(frame,depth,GetVerbose());
}
}, "fetch-vars", "time to dump variables for the topmost frame in every thread"),
m_run_expr_measurement([this] (SBFrame frame, const char* expr) -> void
{
SBValue value(frame.EvaluateExpression(expr, lldb::eDynamicCanRunTarget));
Xcode::FetchVariable (value, 0, GetVerbose());
}, "run-expr", "time to evaluate an expression and display the result")
{
m_app_path.clear();
m_out_path.clear();
m_doc_path.clear();
m_print_help = false;
}
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
virtual
~SketchTest ()
{
}
virtual bool
ParseOption (int short_option, const char* optarg)
{
switch (short_option)
{
case 0:
return false;
case -1:
return false;
case '?':
case 'h':
m_print_help = true;
break;
case 'v':
SetVerbose(true);
break;
case 'c':
{
SBFileSpec file(optarg);
if (file.Exists())
SetExecutablePath(optarg);
else
fprintf(stderr, "error: file specified in --sketch (-c) option doesn't exist: '%s'\n", optarg);
}
break;
case 'f':
{
SBFileSpec file(optarg);
if (file.Exists())
SetDocumentPath(optarg);
else
fprintf(stderr, "error: file specified in --foobar (-f) option doesn't exist: '%s'\n", optarg);
}
break;
case 'o':
SetResultFilePath(optarg);
break;
default:
m_print_help = true;
fprintf (stderr, "error: unrecognized option %c\n", short_option);
break;
}
return true;
}
virtual struct option*
GetLongOptions ()
{
return g_long_options;
}
virtual bool
Setup (int& argc, const char**& argv)
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
{
TestCase::Setup(argc,argv);
bool error = false;
if (GetExecutablePath() == NULL)
{
// --sketch is mandatory
error = true;
fprintf (stderr, "error: the '--sketch=PATH' option is mandatory\n");
}
if (GetDocumentPath() == NULL)
{
// --foobar is mandatory
error = true;
fprintf (stderr, "error: the '--foobar=PATH' option is mandatory\n");
}
if (error || GetPrintHelp())
{
puts(R"(
NAME
lldb_perf_sketch -- a tool that measures LLDB peformance while debugging sketch.
SYNOPSIS
lldb_perf_sketch --sketch=PATH --foobar=PATH [--out-file=PATH --verbose]
DESCRIPTION
Runs a set of static timing and memory tasks against sketch and outputs results
to a plist file.
)");
}
if (error)
{
exit(1);
}
lldb::SBLaunchInfo launch_info = GetLaunchInfo();
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
m_target = m_debugger.CreateTarget(m_app_path.c_str());
m_file_line_bp_measurement("SKTDocument.m",245);
m_file_line_bp_measurement("SKTDocument.m",283);
m_file_line_bp_measurement("SKTText.m",326);
return Launch (launch_info);
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
}
lldb::SBLaunchInfo
GetLaunchInfo ()
{
const char* file_arg = m_doc_path.c_str();
const char* persist_arg = "-ApplePersistenceIgnoreState";
const char* persist_skip = "YES";
const char* empty = nullptr;
const char* args[] = {file_arg,persist_arg,persist_skip,empty};
return SBLaunchInfo(args);
}
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
void
DoTest ()
{
m_fetch_frames_measurement();
m_fetch_modules_measurement();
m_fetch_vars_measurement(1);
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
}
virtual void
TestStep (int counter, ActionWanted &next_action)
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
{
static int launch = 1;
switch (counter % 10)
{
case 0:
{
DoTest ();
if (counter == 0)
m_file_line_bp_measurement("SKTDocument.m",254);
next_action.Continue();
}
break;
case 1:
{
DoTest ();
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"properties");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"[properties description]");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"typeName");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"data");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"[data description]");
next_action.Continue();
}
break;
case 2:
{
DoTest ();
next_action.Continue();
}
break;
case 3:
{
DoTest ();
next_action.StepOver(m_thread);
}
break;
case 4:
{
DoTest ();
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"layoutManager");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"contents");
next_action.StepOver(m_thread);
}
break;
case 5:
{
DoTest ();
next_action.StepOver(m_thread);
}
break;
case 6:
{
DoTest ();
next_action.StepOver(m_thread);
}
break;
case 7:
{
DoTest ();
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"@\"an NSString\"");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"[(id)@\"an NSString\" description]");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"@[@1,@2,@3]");
next_action.StepOut(m_thread);
}
break;
case 8:
{
DoTest ();
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"[graphics description]");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"[selectionIndexes description]");
m_run_expr_measurement(m_thread.GetFrameAtIndex(0),"(BOOL)NSIntersectsRect(rect, graphicDrawingBounds)");
}
next_action.CallNext();
break;
case 9:
if (++launch < 10)
next_action.Relaunch(GetLaunchInfo());
else
next_action.Kill();
break;
default:
{
next_action.Kill();
}
break;
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
}
}
virtual void
WriteResults (Results &results)
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
{
m_fetch_frames_measurement.WriteAverageAndStandardDeviation(results);
m_file_line_bp_measurement.WriteAverageAndStandardDeviation(results);
m_fetch_modules_measurement.WriteAverageAndStandardDeviation(results);
m_fetch_vars_measurement.WriteAverageAndStandardDeviation(results);
m_run_expr_measurement.WriteAverageAndStandardDeviation(results);
results.Write(GetResultFilePath());
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
}
void
SetExecutablePath (const char* str)
{
if (str)
m_app_path.assign(str);
}
const char*
GetExecutablePath ()
{
if (m_app_path.empty())
return NULL;
return m_app_path.c_str();
}
void
SetDocumentPath (const char* str)
{
if (str)
m_doc_path.assign(str);
}
const char*
GetDocumentPath ()
{
if (m_doc_path.empty())
return NULL;
return m_doc_path.c_str();
}
void
SetResultFilePath (const char* str)
{
if (str)
m_out_path.assign(str);
}
const char*
GetResultFilePath ()
{
if (m_out_path.empty())
return "/dev/stdout";
return m_out_path.c_str();
}
bool
GetPrintHelp ()
{
return m_print_help;
}
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
private:
Measurement<lldb_perf::TimeGauge, std::function<void()>> m_fetch_frames_measurement;
Measurement<lldb_perf::TimeGauge, std::function<void(const char*, uint32_t)>> m_file_line_bp_measurement;
Measurement<lldb_perf::TimeGauge, std::function<void()>> m_fetch_modules_measurement;
Measurement<lldb_perf::TimeGauge, std::function<void(int)>> m_fetch_vars_measurement;
Measurement<lldb_perf::TimeGauge, std::function<void(SBFrame, const char*)>> m_run_expr_measurement;
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
std::string m_app_path;
std::string m_doc_path;
std::string m_out_path;
bool m_print_help;
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
};
int main(int argc, const char * argv[])
{
SketchTest test;
return TestCase::Run(test, argc, argv);
Initial checkin of a new project: LLDB Performance Testing Infrastructure This is a very basic implementation of a library that easily allows to drive LLDB.framework to write test cases for performance This is separate from the LLDB testsuite in test/ in that: a) this uses C++ instead of Python to avoid measures being affected by SWIG b) this is in very early development and needs lots of tweaking before it can be considered functionally complete c) this is not meant to test correctness but to help catch performance regressions There is a sample application built against the library (in darwin/sketch) that uses the famous sample app Sketch as an inferior to measure certain basic parameters of LLDB's behavior. The resulting output is a PLIST much like the following: <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <array> <dict> <key>fetch-frames</key> <real>0.13161715522222225</real> </dict> <dict> <key>file-line-bkpt</key> <real>0.029111678750000002</real> </dict> <dict> <key>fetch-modules</key> <real>0.00026376766666666668</real> </dict> <dict> <key>fetch-vars</key> <real>0.17820429311111111</real> </dict> <dict> <key>run-expr</key> <real>0.029676525769230768</real> </dict> </array> </plist> Areas for improvement: - code cleanups (I will be out of the office for a couple days this coming week, but please keep ideas coming!) - more metrics and test cases - better error checking This toolkit also comprises a simple event-loop-driven controller for LLDB, similar yet much simpler to what the Driver does to implement the lldb command-line tool. llvm-svn: 176715
2013-03-09 04:29:13 +08:00
}