RFR(M): 8139040: Fix initializations before ShouldNotReachHere() (original) (raw)

Lindenmaier, Goetz goetz.lindenmaier at sap.com
Thu Oct 22 06:31:58 UTC 2015

Hi Coleen,

that fix looks good, thanks. Sorry I can't avoid these. The warnings depend on the compiler optimizations, they probably differ slightly on the closed platform.

Best regards, Goetz.

-----Original Message----- From: hotspot-runtime-dev [mailto:hotspot-runtime-dev-_ _bounces at openjdk.java.net] On Behalf Of Coleen Phillimore Sent: Mittwoch, 21. Oktober 2015 21:26 To: hotspot-runtime-dev at openjdk.java.net Subject: Re: RFR(M): 8139040: Fix initializations before ShouldNotReachHere()

I had to add one more initialization for some internal platform for some reason I don't quite know. See threadService.hpp. open webrev at http://cr.openjdk.java.net/~coleenp/8139040.01/ Coleen On 10/20/15 3:24 AM, Lindenmaier, Goetz wrote: > Hi David, > > yes, I assume the compiler optimizes it. > If the compiler issues a warning, it has all information it needs to > optimize it. So I'm sure gcc does so. I don't know about the solaris > compiler, though. > > Other locations in that file also initialize at the declaration > (e.g., c1LIRAssemblerx86.cpp:3917). > I fixed these two anyways: > http://cr.openjdk.java.net/~goetz/webrevs/8139040-init/webrev.05/ > > Best regards, > Goetz. > > >> -----Original Message----- >> From: hotspot-runtime-dev [mailto:hotspot-runtime-dev-_ _>> bounces at openjdk.java.net] On Behalf Of David Holmes >> Sent: Dienstag, 20. Oktober 2015 02:46 >> To: hotspot-runtime-dev at openjdk.java.net >> Subject: Re: RFR(M): 8139040: Fix initializations before >> ShouldNotReachHere() >> >> Hi Goetz, >> >> src/cpu/x86/vm/c1LIRAssemblerx86.cpp >> >> I'd prefer to see the missing initialization added to the default of the >> switch: >> >> ! Assembler::Condition acond = Assembler::equal, ncond = >> Assembler::notEqual; >> switch (condition) { >> >> otherwise we're just wasting instructions (or assuming the compiler will >> optimize it away). >> >> Same with: >> >> 3189 Address::ScaleFactor scale = Address::noscale; >> >> More generally if something is complaining about an uninitialized >> variable on a ShouldNotReachHere/fatal path then put the initialization >> in that path, not the common code that will cause it to be unnecessarily >> executed. >> >> Cheers, >> David

More information about the hotspot-runtime-dev mailing list