[13]RFR: 8226627: assert(t->singleton()) failed: must be a constant

[Pengfei Li (Arm Technology China)]

Hi Jamsheed,

Thanks for finding this issue. Regarding on your question in https://bugs.openjdk.java.net/browse/JDK-8226627

> constant cases were disabled for this intrinsics. but still some cases it found later to be constants and tried to fold. 
> i tried implementing constant folding here [1] . but real fix should be to check why constant folding is disabled at first place. and do necessary actions. 
> [1] http://cr.openjdk.java.net/~jcm/8226627/webrev/ (some typos, updated inplace)

Constant folding for MinFD/MaxFD was left unimplemented just because we didn't know it's mandatory and it's a bit complex to handle +inf/-inf/NaN.

I'm not a reviewer but I could help test your code now.

--
Thanks,
Pengfei

> -----Original Message-----
> From: Jamsheed <jamsheed.c.m at oracle.com>
> Sent: Tuesday, July 2, 2019 14:24
> To: hotspot-compiler-dev at openjdk.java.net; Pengfei Li (Arm Technology
> China) <Pengfei.Li at arm.com>; Andrew Haley <aph at redhat.com>
> Subject: Re: [13]RFR: 8226627: assert(t->singleton()) failed: must be a
> constant
> 
> Hi Pengfei Li,  Andrew Haley,
> 
> could you please share your thoughts on this.
> 
> ( i had CC'ed  in initial mail, but somehow it doesn't work well)
> 
> Best regards,
> 
> Jamsheed
> 
> On 01/07/19 10:48 AM, Jamsheed wrote:
> > Hi,
> >
> > Request for review.
> >
> > with 8212043, new nodes were added for MaxFD,MinFD, from mails it
> > seems like constant folding part was not implemented and was partially
> > disabled here[1].
> >
> > implemented constant folding for MaxFD, MinFD
> >
> > webrev: http://cr.openjdk.java.net/~jcm/8226627/webrev.00/
> >
> > bug: https://bugs.openjdk.java.net/browse/JDK-8226627
> >
> > testing: mach1-5, mach1-5(-XX:CompileThresholdScaling=0.1) mach5
> > (hs-precheckin-comp).
> >
> > Best regards,
> >
> > Jamsheed
> >
> > [1]http://hg.openjdk.java.net/jdk/jdk/file/5ce081828739/src/hotspot/sh
> > are/opto/library_call.cpp#l6709
> >
> >