RFR: JDK-8157141 & JDK-8166454: Solaris getisax(2) and meminfo(2) cleanup

Alan Burlison Alan.Burlison at oracle.com
Thu Sep 29 17:42:21 UTC 2016

On 29/09/2016 00:15, Kim Barrett wrote:

> ------------------------------------------------------------------------------
> src/os_cpu/solaris_sparc/vm/vm_version_solaris_sparc.cpp
>  377   uint_t av1 = avs[AV_HW1_IDX];
>  378   if (av1 & AV_SPARC_MUL32)        features |= hardware_mul32_m;
> ...
> Good to use the symbolic index AV_HW1_IDX. Bad form to assume we have
> that many entries.  That should be
>   if (avn > AV_HW1_IDX) {
>     uint_t av1 = avs[AV_HW1_IDX];
>     if (av1 & AV_SPARC_MUL32)        features |= hardware_mul32_m;
>     ...
>   }

There is always at least one entry returned so the condition above will 
always be true - we will never reduce the number of HW capabilities bits 
so I believe the test is unnecessary. It's not there in the existing 
version of the code either. However if you want me to add it I'll do so.

>> A JPRT hotspot testset run is clean.
> Have you used logging to hand-verify the expected features are being set?

I've run it under the debugger and single stepped through the code on a 
T5 and T7. The features are set as expected.

Alan Burlison

More information about the hotspot-runtime-dev mailing list