mattesgroeger / mgbenchmark Goto Github PK

View Code? Open in Web Editor NEW

295.0 295.0 17.0 478 KB

Library to measure code execution times easily

License: MIT License

Ruby 8.96% Objective-C 91.04%

mgbenchmark's People

Contributors

Stargazers

Watchers

Forkers

tuantmdev rajdn workwithnano wpsteak lukelan sisufo vinod1988 stephanpartzsch mzaks hzwzw ioslabs hailongiphone rolandkakonyi meghs91 dodubaby muhitsarwar

mgbenchmark's Issues

Should not use NSAssert when stopping never started session

Bump version number to 0.4.0

When you get a chance, do you mind bumping the version number?

Add short notation for benchmarking

If you just have one session to measure at a time this macros would be even simpler to use:

benchmark_start("session name")
benchmark_step("step name")
benchmark_total()
benchmark_finish()

<Enhancement> Designate target when starting a new session.

[MGBenchmark start:@"sessionName"] uses the defaultTarget for all sessions.

+ (MGBenchmarkSession *)start:(NSString *)sessionName
{
    __block MGBenchmarkSession *session = [[MGBenchmarkSession alloc] initWithName:sessionName andTarget:defaultTarget];

    dispatch_sync(benchmarkQueue, ^
    {
        sessions[sessionName] = session;
    });

    return session;
}

Consider adding the following, to allow specifying the target when starting a new session:

MGBenchmark.h

/**
 * Starts a new session and assigns the target. An instance
 * of the session is returned.
 */
+ (MGBenchmarkSession *)start:(NSString *)sessionName target:(id<MGBenchmarkTarget>)target;

MGBenchmark.m

+ (MGBenchmarkSession *)start:(NSString *)sessionName
{
    return [self start:sessionName target:defaultTarget];
}

+ (MGBenchmarkSession *)start:(NSString *)sessionName target:(id<MGBenchmarkTarget>)target
{
    __block MGBenchmarkSession *session = [[MGBenchmarkSession alloc] initWithName:sessionName andTarget:target];

    dispatch_sync(benchmarkQueue, ^
    {
        sessions[sessionName] = session;
    });

    return session;
}

Use alternative enable/disable Macro

Currently, the enabling/disabling of benchmarking is done via the presence of the DEBUG macro. In my usage of this tool, I've wanted to build for Release with benchmarking enabled.

Would you be open to using a different flag?

#ifdef MGBENCHMARK
...

SIGSEGV

Occasionally (0.03%) benchmarking a step in our game causes this crash:

Exception Type:  SIGSEGV
Exception Codes: SEGV_ACCERR at 0x220826ed
Crashed Thread:  0

Thread 0 Crashed:
0   libobjc.A.dylib                     0x392525b0 _objc_msgSend + 16
1   PocketUniverse                      0x0020cae5 -[MGBenchmarkSession timePassedSince:] (MGBenchmarkSession.m:82)
2   PocketUniverse                      0x0020c9b9 -[MGBenchmarkSession step:] (MGBenchmarkSession.m:60)
[...]

It's called from various different places. No repeating pattern to see.

Allow to disable benchmarking

You might not want to have the benchmark running in the live environment. The lib should provide an easy way to disable the execution.

How to properly initialize the dispatch_queue

DISPATCH_QUEUE_SERIAL or DISPATCH_QUEUE_CONCURRENT

Add custom target for listing steps by execution time

If you measure different steps for a session, you might be interested in what actually takes the longest. Therefore a target that just collects data for each step and then prints out a sorted list of steps on demand would be helpful.

Starting nested sessions corrupts output

I didn't test this myself yet but I got this report from someone else. I need to check this.