560536 - sparc debugger builds running tests very slowly

Status: VERIFIED FIXED

(Tamarin Graveyard :: Virtual Machine, defect, P3)

Description

[Trevor Baker]

clean new bug for concern expressed in https://bugzilla.mozilla.org/show_bug.cgi?id=551173#c18

Found 2 issues in the testsuite:

debugger builds running the following testcases are EXTREMELY slow. I ended up
killing the run after 5 minutes
- as3/sampling/Callback.abc
- as3/sampling/FunctionEntry.abc

Callback does slowly spit out callback information, but after 5 minutes I
stopped it, should run much faster 
-bash-3.00$ $AVM as3/sampling/Callback.abc
Sampling api: callback
callback1 called at 360
callback1 called at 394
callback1 called at 394
callback1 called at 394
callback1 called at 394
callback1 called at 394
callback1 called at 394
callback1 called at 394

Comment 1

[Brent Baker]

The slow down could be a result of having to compile with -xO2 instead of -xO5 (https://bugzilla.mozilla.org/attachment.cgi?id=438652)

Comment 2

[Leon Sha]

Attachment: patch

It is not caused by the complier option. Solaris is using a different Performance Counter. That is not correct. Should use the same performance counter as linux.

Comment 3

[Edwin Smith]

Comment on attachment 446651 [details] [diff] [review]
patch

change looks okay to me as long as testing is successful.

Comment 4

[Edwin Smith]

Leon, any idea where the code to use gethrtime() crept in?  I'm wondering if it was there for an important reason that is no longer apparent since we factored VMPI_getPerformanceCounter() into the porting layer.

Comment 5

[Leon Sha]

At very beginning, there is no implementation on unix platform on for Performance Counter. Somehow empty function getPerformanceCounter() seems to work for linux but not for solaris. I add gethrtime() for solaris. The guy who implement this function for general unix later but didn't remove the solaris part. I remember I have removed it before. But seems I didn't do that or it is add back when porting to VMPI.
(In reply to comment #4)
> Leon, any idea where the code to use gethrtime() crept in?  I'm wondering if it
> was there for an important reason that is no longer apparent since we factored
> VMPI_getPerformanceCounter() into the porting layer.

Comment 6

[Leon Sha]

http://hg.mozilla.org/tamarin-redux/rev/eb054fb3ae2a

Comment 7

[Brent Baker]

Verified that the sampling testcases now run in a normal amount of time, seconds instead of minutes. Removed /skip/ from acceptance testconfig so they will again be executed during acceptance testing in the build system.

tr -> 4732:0effad8c47e9