RFR: JDK-8202384: Introduce altserver jvm variant with

Baesken, Matthias matthias.baesken at sap.com
Wed Jun 6 07:59:56 UTC 2018

Thanks for the clarification !

From: Erik Joelsson [mailto:erik.joelsson at oracle.com]
Sent: Dienstag, 5. Juni 2018 17:51
To: Baesken, Matthias <matthias.baesken at sap.com>; 'hotspot-dev at openjdk.java.net' <hotspot-dev at openjdk.java.net>; 'build-dev at openjdk.java.net' <build-dev at openjdk.java.net>
Cc: Zeller, Arno <arno.zeller at sap.com>
Subject: Re: RFR: JDK-8202384: Introduce altserver jvm variant with

Hello Matthias,

For GCC, you need 7.3.0 or later. For Microsoft you need VS2017 and I think some minimal update version (the option is called -Qspectre), we use 15.5.5.

I was not involved in the benchmarking so I don't know any details there, only the conclusion.


On 2018-06-05 01:30, Baesken, Matthias wrote:

Hi Erik , is there  some info available about  the performance impact when  disabling  disabling speculative execution  ?

And which compiler versions are needed for this ?

Best regards, Matthias

>We need to add compilation flags for disabling speculative execution to

>our native libraries and executables. In order to allow for users not

>affected by problems with speculative execution to run a JVM at full

>speed, we need to be able to ship two JVM libraries - one that is

>compiled with speculative execution enabled, and one that is compiled

>without. Note that this applies to the build time C++ flags, not the

>compiler in the JVM itself. Luckily adding these flags to the rest of

>the native libraries did not have a significant performance impact so

>there is no need for making it optional there.


>This patch defines flags for disabling speculative execution for GCC and

>Visual Studio and applies them to all binaries except libjvm when

>available in the compiler. It defines a new jvm feature

>no-speculative-cti, which is used to control whether to use the flags

>for libjvm. It also defines a new jvm variant "altserver" which is the

>same as server, but with this new feature added.


>For Oracle builds, we are changing the default for linux-x64 and

>windows-x64 to build both server and altserver, giving the choice to the

>user which JVM they want to use. If others would prefer this default, we

>could make it default in configure as well.


>The change in GensrcJFR.gmk fixes a newly introduced race that appears

>when building multiple jvm variants.


>Bug: https://bugs.openjdk.java.net/browse/JDK-8202384


>Webrev: http://cr.openjdk.java.net/~erikj/8202384/webrev.01<http://cr.openjdk.java.net/%7Eerikj/8202384/webrev.01>

More information about the hotspot-dev mailing list