-Xcomp test time regressions?

[David Holmes]

Hi Aleksey,

Seems like a question for the JIT folk on hotspot-compiler-dev.

Though ...

On 28/06/2021 9:19 pm, Aleksey Shipilev wrote:
> Hi,
> 
> I was doing the usual work, and then noticed -Xcomp tests got 
> considerably slower. I believe there are plenty of JBS bug reports 
> already with various test timeouts.
> 
> Running a sample test like:
>    $ CONF=linux-x86_64-server-fastdebug make run-test 
> TEST=gc/shenandoah/mxbeans/TestPauseNotifications.java 
> TEST_VM_OPTS="-Xcomp"
> 
> I initially thought it has something to do with JTReg upgrade, but it 
> does not look like it:
>    - 18, JTReg 6+1: 5m33s (!!!)
>    - 11u, JTReg 5.1-b01: 2m27s
>    - 11u, JTReg 6+1: 2m32s
> 
> Comparing the test logs between "slow" 18 and fast "11u", I can see that 
> javac tasks take a lot more (whole minutes more!) time in "slow" configs.

Did we use to handle the TEST_VM_OPTS differently, such that it didn't 
apply to the javac part?

David

> Before I go and bisect this more thoroughly, is this an already known 
> thing?
>