[jdk17u-dev] RFR: 8324123: aarch64: fix prfm literal encoding in assembler [v2]

[Liang Mao]

On Tue, 5 Mar 2024 08:33:34 GMT, Andrew Haley <aph at openjdk.org> wrote:

> > > > Could reviewers please help with the review? The risk is very low as the original wrong encoding is not used in code base.
> > > 
> > > 
> > > There is, therefore, no reason to backport this. It should not be approved.
> > 
> > 
> > Hi @theRealAph , I guess we might have enhancements in the future depending on this correct instruction encoding. 17u still accepts enhancements with substantial benefit.
> 
> ... to the users. I guess what you say is possible, and applying this patch now might avoid some future risk. Having said that, if we can speculate about any and every possible future we'd backport an awful lot of stuff, with the risk that implies.
> 
> As it stands, this backport is outside the current rules. The maintainers are allowed to make an exception for a special case, though.

Sure. We can still have a review before maintainer makes the decision. It's not just a normal pre-requisite enhancement which seems unnecessary as you described but an incorrect instruction encoding mistake which really makes compiler engineers nervous. The wrong encoding may not lead to crash but cannot perform the prefetch as expected and very difficult for engineers to find out the root cause.

-------------

PR Comment: https://git.openjdk.org/jdk17u-dev/pull/2256#issuecomment-1978233147