an issue after: 8224234: compiler/codegen/TestCharVect2.java fails in test_mulc

[Doerr, Martin]

Hi everybody,

I think the appropriate effect is "TEMP_DEF dst" when you modify the dst register before you read one of the input registers.
"DEF dst" or "USE src" don't need to get specified explicitly if these registers occur in the match rule as destination or source.

Best regards,
Martin


> -----Original Message-----
> From: hotspot-compiler-dev <hotspot-compiler-dev-
> bounces at openjdk.java.net> On Behalf Of Vladimir Kozlov
> Sent: Dienstag, 11. Juni 2019 18:16
> To: Viswanathan, Sandhya <sandhya.viswanathan at intel.com>; Deshpande,
> Vivek R <vivek.r.deshpande at intel.com>
> Cc: hotspot-compiler-dev at openjdk.java.net
> Subject: Re: an issue after: 8224234: compiler/codegen/TestCharVect2.java
> fails in test_mulc
> 
> Thanks you, Sandhya
> 
> I had run the test on the same machine for 100 iterations but without
> reproducing the failure. It
> seems very rare case.
> 
> Vladimir
> 
> On 6/11/19 9:11 AM, Viswanathan, Sandhya wrote:
> > Hi Vladimir,
> >
> > We will take a look.
> >
> > Thanks for reporting the issue.
> >
> > Best Regards,
> > Sandhya
> >
> >
> > -----Original Message-----
> > From: Vladimir Kozlov [mailto:vladimir.kozlov at oracle.com]
> > Sent: Monday, June 10, 2019 2:44 PM
> > To: Deshpande, Vivek R <vivek.r.deshpande at intel.com>; Viswanathan,
> Sandhya <sandhya.viswanathan at intel.com>
> > Cc: hotspot-compiler-dev at openjdk.java.net
> > Subject: an issue after: 8224234: compiler/codegen/TestCharVect2.java
> fails in test_mulc
> >
> > Hi Sandhya and Vivek,
> >
> > I got new failure after pulling these changes. The failure is intermittent, I
> saw it once in my testing.
> >
> > The failed test compiler/codegen/TestShortFloatVect.java ran without any
> other flags with only ones specified in test's
> > @run.  I got:
> >
> > ----------System.out:(3/50)----------
> > Testing Short + Float vectors
> > Warmup
> > Verification
> > ----------System.err:(17/546)----------
> > test_ci: b1[968] = 0.0 != -103.0
> > test_ci: b1[969] = 0.0 != -103.0
> > test_ci: b1[970] = 0.0 != -103.0
> > test_ci: b1[971] = 0.0 != -103.0
> > test_ci: b1[976] = 0.0 != -103.0
> > test_ci: b1[977] = 0.0 != -103.0
> > test_ci: b1[978] = 0.0 != -103.0
> > test_ci: b1[979] = 0.0 != -103.0
> > test_ci: b1[984] = 0.0 != -103.0
> > test_ci: b1[985] = 0.0 != -103.0
> > test_ci: b1[986] = 0.0 != -103.0
> > test_ci: b1[987] = 0.0 != -103.0
> > test_ci: b1[992] = 0.0 != -103.0
> > test_ci: b1[993] = 0.0 != -103.0
> > test_ci: b1[994] = 0.0 != -103.0
> > test_ci: b1[995] = 0.0 != -103.0
> > FAILED: 16 errors
> >
> > test_ci() method simple assign constant value to short[] and float[] arrays
> in the same loop:
> >
> >     static void test_ci(short[] a, float[] b) {
> >       for (int i = 0; i < a.length; i+=1) {
> >         a[i] = -123;
> >         b[i] = -103.f;
> >       }
> >     }
> >
> > The HW is MacPro with Xeon E5-1620_v2 3.70GHz
> >
> > FPU VME DE PSE TSC MSR PAE MCE CX8 APIC SEP MTRR PGE MCA CMOV
> PAT PSE36 CLFSH DS ACPI MMX FXSR SSE SSE2 SS HTT TM PBE
> > SSE3 PCLMULQDQ DTES64 MON DSCPL VMX SMX EST TM2 SSSE3 CX16 TPR
> PDCM SSE4.1 SSE4.2 x2APIC POPCNT AES PCID XSAVE OSXSAVE
> > TSCTMR AVX1.0 RDRAND F16C SMEP ERMS RDWRFSGS SYSCALL XD
> 1GBPAGE EM64T LAHF RDTSCP TSCI
> >
> > Please, look.
> >
> > Thanks,
> > Vladimir
> >
> > On 6/7/19 3:48 PM, Deshpande, Vivek R wrote:
> >> Thank you All.
> >> Pushed the changes.
> >>
> >> Regards,
> >> Vivek
> >>
> >> -----Original Message-----
> >> From: Viswanathan, Sandhya
> >> Sent: Friday, June 7, 2019 11:45 AM
> >> To: Vladimir Ivanov <vladimir.x.ivanov at oracle.com>; Tobias Hartmann
> <tobias.hartmann at oracle.com>
> >> Cc: hotspot-compiler-dev at openjdk.java.net; Deshpande, Vivek R
> <vivek.r.deshpande at intel.com>
> >> Subject: RE: RFR (XS) 8224234: compiler/codegen/TestCharVect2.java fails
> in test_mulc
> >>
> >> Hi Vladimir/Tobias,
> >>
> >> Thanks a lot for the review.
> >>
> >> Best Regards,
> >> Sandhya
> >>
> >> -----Original Message-----
> >> From: Vladimir Ivanov [mailto:vladimir.x.ivanov at oracle.com]
> >> Sent: Friday, June 07, 2019 3:41 AM
> >> To: Viswanathan, Sandhya <sandhya.viswanathan at intel.com>; hotspot-
> compiler-dev at openjdk.java.net
> >> Subject: Re: RFR (XS) 8224234: compiler/codegen/TestCharVect2.java fails
> in test_mulc
> >>
> >> Looks good.
> >>
> >> Best regards,
> >> Vladimir Ivanov
> >>
> >> On 07/06/2019 03:41, Viswanathan, Sandhya wrote:
> >>> Please find below a small patch which fixes the issue:
> >>>
> >>> JBS: https://bugs.openjdk.java.net/browse/JDK-8224234
> >>>
> >>> Patch: http://cr.openjdk.java.net/~sviswanathan/8224234/
> >>>
> >>> Multiplication is implemented as shift, add sequence in certain cases.
> >>>
> >>> The problem was occurring because the "shift by" register was being
> >>> overwritten due to missing effect statement in the shift rules.
> >>>
> >>> The problem was introduced as part of
> >>> https://bugs.openjdk.java.net/browse/JDK-8222074.
> >>>
> >>> The patch adds the missing effect statements.
> >>>
> >>> Best Regards,
> >>>
> >>> Sandhya
> >>>