digitalmars.D.bugs - [Issue 9581] New: Regression (2.053): Exception takes too much time to show up
- d-bugmail puremagic.com (48/50) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (6/6) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (12/12) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (24/28) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (19/19) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (59/70) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (7/7) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (12/14) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (15/15) Feb 23 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (7/8) Feb 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (11/11) Feb 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (8/10) Feb 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (13/14) Feb 24 2013 Of course, though this involves a lot more than simply throwing and catc...
- d-bugmail puremagic.com (11/14) Feb 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (10/10) Feb 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (9/13) Feb 26 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (7/8) Feb 26 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (8/8) Mar 31 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (11/14) Mar 31 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (13/19) Apr 16 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (11/11) Apr 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (9/11) Apr 24 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (7/13) May 19 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (8/9) May 19 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (9/15) May 19 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (9/12) May 19 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
- d-bugmail puremagic.com (11/11) May 19 2013 http://d.puremagic.com/issues/show_bug.cgi?id=9581
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Summary: Regression (2.053): Exception takes too much time to show up Product: D Version: D2 Platform: All OS/Version: Windows Status: NEW Keywords: EH, performance Severity: regression Priority: P2 Component: DMD AssignedTo: nobody puremagic.com ReportedBy: andrej.mitrovich gmail.com 18:35:50 PST --- test.d: ------ module test; void main() { assert(0); } ------ timer.d (or use your favorite timer app): ------ module timer; import std.stdio; import std.process; import std.datetime; void main() { auto sw1 = StopWatch(AutoStart.yes); system("dmd test.d"); system("test"); sw1.stop(); writefln("%s msecs.", sw1.peek.msecs); } ------ DMD 2.052: $ dmd timer.d $ timer81 msecs.DMD 2.053: $ dmd timer.d $ timer4127 msecs.This is using XP SP3. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 18:36:24 PST --- And this can be recreated all the way up to 2.062 and 2.063 git-head. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Steven Schveighoffer <schveiguy yahoo.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |schveiguy yahoo.com 19:24:45 PST --- Doesn't this just test the compile time? Not that 4 second compile time is really acceptable, but I don't know if this test does what you think it does. What happens when you remove the compilation line? -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 19:27:56 PST ---Doesn't this just test the compile time? Not that 4 second compile time is really acceptable, but I don't know if this test does what you think it does. What happens when you remove the compilation line?Sorry about that, it was a misplaced system call. Without timing the compilation its faster but still 4050 msecs. Here's the newer script: timer.d: --------- module timer; import std.stdio; import std.process; import std.datetime; void main() { system("dmd test.d"); auto sw1 = StopWatch(AutoStart.yes); system("test"); sw1.stop(); writefln("%s msecs.", sw1.peek.msecs); } --------- -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Jonathan M Davis <jmdavisProg gmx.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |jmdavisProg gmx.com PST --- If what you're testing is the amount of time that it takes an exception to be thrown and caught, then you shouldn't be doing anything like call system, since then that's testing what system's up to. And even worse, you're starting a different program, meaning that you're testing the amount of time it takes to run the program (including the startup time for druntime). Please test only one thing at a time. If system's too slow, then create a test just for that and a bug report for it. If exceptions are too slow, then cerate a test just for that and bug report for it. If druntime takes too long to startup, then create a test just for that and create a bug report for it. You're testing too many things at once here. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 19:47:32 PST ---If what you're testing is the amount of time that it takes an exception to be thrown and caught, then you shouldn't be doing anything like call system, since then that's testing what system's up to. And even worse, you're starting a different program, meaning that you're testing the amount of time it takes to run the program (including the startup time for druntime). Please test only one thing at a time. If system's too slow, then create a test just for that and a bug report for it. If exceptions are too slow, then cerate a test just for that and bug report for it. If druntime takes too long to startup, then create a test just for that and create a bug report for it. You're testing too many things at once here.This is the test-case: -------- module test; void main() { assert(0); } --------- You can use `timeit` or some other timing app you may have installed on the system, the 'timer' was just there in case someone doesn't have an alternative to test it with. Anyway you can get 'timeit' from http://www.microsoft.com/en-us/download/details.aspx?id=17657, then the test runs are: 2.052: $ D:\foobar>dmd test.d $ D:\foobar>timeit test core.exception.AssertError test(5): Assertion failure Version Number: Windows NT 5.1 (Build 2600) Exit Time: 4:45 am, Sunday, February 24 2013 Elapsed Time: 0:00:00.078 Process Time: 0:00:00.031 System Calls: 2121 Context Switches: 456 Page Faults: 678 Bytes Read: 142408 Bytes Written: 0 Bytes Other: 124382 2.053: D:\foobar>dmd test.d D:\foobar>timeit test core.exception.AssertError test(5): Assertion failure ---------------- 40CE70 40CCE7 40201A 4025C7 4021C3 4120C9 7C817067 ---------------- Version Number: Windows NT 5.1 (Build 2600) Exit Time: 4:46 am, Sunday, February 24 2013 Elapsed Time: 0:00:04.093 Process Time: 0:00:01.265 System Calls: 72733 Context Switches: 38975 Page Faults: 18821 Bytes Read: 17758445 Bytes Written: 8283 Bytes Other: 55317 The system calls should give a hint here, it's an enormous increase from 2121 calls in 2.052 to 72733 calls in 2.053. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 20:09:28 PST --- Also note, the new version prints a stack trace, while the previous doesn't. Interesting. I wonder if that is what is taking so long. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 20:14:06 PST ---Also note, the new version prints a stack trace, while the previous doesn't. Interesting. I wonder if that is what is taking so long.Yeah. Considering how long it takes I really wonder if we can improve the speed in some way. E.g. we could provide a switch to disable the stack-trace. Or maybe the real issue is spending time finding dbghelp.dll. If that's the case, we could try to search for a specific environment variable that the user would set so the DLL is quickly found. Anyway that's only guessing that's the actual issue. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 PST --- Exceptions in general are overly slow. As I understand it, one of the major things that could be done would be to make it so that it doesn't actually determine the strings for the stacktrace when it's constructed but rather when toString is called on it. We currently incur that extra regardless of whether the exception's toString method is ever used. Now, I don't think that it would help any in this case, because the toString method _is_ being used, but it is something that should be done to improve the performance of exceptions. And I'm sure that there are plenty of other things that could be done as well, but I don't know the inner workings of druntime well enough to know what they might be. Java beats the pants of D in exception speed though. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 23 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 06:52:00 PST ---Exceptions in general are overly slow.Yeah, but not *4 seconds* slow, that is unacceptable. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Walter Bright <bugzilla digitalmars.com> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |bugzilla digitalmars.com 14:34:52 PST --- This may have something to do with an earlier complaint that dbghelp.dll was searching one's entire hard disk looking for a source file. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 14:42:31 PST ---This may have something to do with an earlier complaint that dbghelp.dll was searching one's entire hard disk looking for a source file.That specific bug was fixed though (and I can confirm it was fixed). But there's still this delay of several seconds before the message shows up. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 PST ---Yeah, but not *4 seconds* slow, that is unacceptable.Of course, though this involves a lot more than simply throwing and catching an exception, which changes the equation a bit (though 4 seconds is still too slow). By the way, I just tried this on Linux for comparison, and it took about 8ms on my machine. So, this is clearly a Windows-only issue (though I just used the time command to time it rather than writing a program; Does Windows not have anything similar? Probably not, given how limited its command line capabilities are). -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 14:55:05 PST ---By the way, I just tried this on Linux for comparison, and it took about 8ms on my machine.I can confirm that, it only takes 3ms in Linux Mint in a VBox. It's a Windows issue.Does Windows not have anything similar?It has `timeit` available from http://www.microsoft.com/en-us/download/details.aspx?id=17657 -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 PST --- I couldn't find a bug report dealing with the general bad performance of So, that's related to this, but something here is making it far worse in this particular situation. In general, they're taking milliseconds when they should be taking microseconds, whereas here they're taking seconds. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 13:24:26 PST ---DMD 2.053: $ dmd timer.d $ timerYou know what? I think my hard drive is dying. It seems just running RDMD itself can sometimes take several seconds. This could be an invalid report. I'll try to boot up into another system on another hard-drive to test this. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------4127 msecs.
Feb 26 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 13:42:41 PST ---I'll try to boot up into another system on another hard-drive to test this.Nope, still an issue even on another hard drive. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Feb 26 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 03:56:24 PDT --- Well, I'm on Win7 now and things are pretty blazing fast now. I don't experience the speed issue anymore. Maybe we could close this down unless someone else has the same issue. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Mar 31 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 bearophile_hugs eml.cc changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |bearophile_hugs eml.ccWell, I'm on Win7 now and things are pretty blazing fast now. I don't experience the speed issue anymore. Maybe we could close this down unless someone else has the same issue.On Windows Vista 32bit the D exceptions are still very slow. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Mar 31 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Damian <damianday hotmail.co.uk> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |damianday hotmail.co.uk ---Just food for thought but, since this is a Windows issue, could it be anti-virus or malware scanner or some such application causing slowdown on access to dbghelp.dll? -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------Well, I'm on Win7 now and things are pretty blazing fast now. I don't experience the speed issue anymore. Maybe we could close this down unless someone else has the same issue.On Windows Vista 32bit the D exceptions are still very slow.
Apr 16 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Martin Nowak <code dawg.eu> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |code dawg.eu Does disabling Stacktrace resolve the problem? https://github.com/D-Programming-Language/druntime/blob/4c8fd9dd7b91e735eece5ffc746a0667ccc13dd6/src/core/runtime.d#L585 -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Apr 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581Does disabling Stacktrace resolve the problem? https://github.com/D-Programming-Language/druntime/blob/4c8fd9dd7b91e735eece5ffc746a0667ccc13dd6/src/core/runtime.d#L585And what about completely disabling the dbghelp initalization? I think commenting out the import would suffice. https://github.com/D-Programming-Language/druntime/blob/4c8fd9dd7b91e735eece5ffc746a0667ccc13dd6/src/core/runtime.d#L50 -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
Apr 24 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 04:52:35 PDT ---Are they over 4 seconds like they were for me before I switched? -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------Well, I'm on Win7 now and things are pretty blazing fast now. I don't experience the speed issue anymore. Maybe we could close this down unless someone else has the same issue.On Windows Vista 32bit the D exceptions are still very slow.
May 19 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581Are they over 4 seconds like they were for me before I switched?I have a small benchmark for exceptions, and they are much much slower than equivalent exceptions in Java. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
May 19 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 10:53:41 PDT ---Ok, but is it a regression from an earlier release? We should mark this as a normal bug if not to avoid blocking the 2.063 release (I personally can't recreate the huge slowness anymore). -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------Are they over 4 seconds like they were for me before I switched?I have a small benchmark for exceptions, and they are much much slower than equivalent exceptions in Java.
May 19 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581Ok, but is it a regression from an earlier release? We should mark this as a normal bug if not to avoid blocking the 2.063 release (I personally can't recreate the huge slowness anymore).I think this bug should not block 2.063, because probably will require a fair amount of work to be fixed, and the situation is probably not a regression any more. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
May 19 2013
http://d.puremagic.com/issues/show_bug.cgi?id=9581 Andrej Mitrovic <andrej.mitrovich gmail.com> changed: What |Removed |Added ---------------------------------------------------------------------------- Severity|regression |normal 17:17:33 PDT --- If someone has had their exceptions slow down significantly since 2.053, re-set it as a regression. But I'm lowering it to normal for now. -- Configure issuemail: http://d.puremagic.com/issues/userprefs.cgi?tab=email ------- You are receiving this mail because: -------
May 19 2013