RFR(M): 8136525 Generate interpreter entries only once and avoid unnecessary jump to jump

[Coleen Phillimore]

I see the fixes in the new webrev.  It looks good.

Coleen

On 9/18/15 4:56 AM, Doerr, Martin wrote:
> Hi Colleen,
>
> thanks for testing this.
> The problem is that the CRC32C method is not native while the CRC32 methods are native.
> I just fixed it in this new webrev:
> http://cr.openjdk.java.net/~mdoerr/8136525_interp_entries/webrev.04/
>
> Please note that this was also wrong in "8134553: CRC32C implementations for Nehalem x86/amd64	& Westmere+ x86/amd64".
> template_interpreter_x86_32/64 used generate_native_entry for CRC32C which is wrong.
> However, my new code replaces the incorrect entries.
>
> Best regards,
>     Martin
>
>
> -----Original Message-----
> From: Coleen Phillimore [mailto:coleen.phillimore at oracle.com]
> Sent: Freitag, 18. September 2015 04:42
> To: Doerr, Martin; Andrew Haley; hotspot-runtime-dev at openjdk.java.net
> Subject: Re: RFR(M): 8136525 Generate interpreter entries only once and avoid unnecessary jump to jump
>
> When I ran this through JPRT as a test, I get the assertion in
>
> #  Internal Error
> (/opt/jprt/T/P1/002711.cphillim/s/hotspot/src/cpu/sparc/vm/macroAssembler_sparc.cpp:1599),
> pid=15540, tid=40
> #  assert(false) failed: DEBUG MESSAGE: tried to execute non-native
> method as native
> #
>
> on solaris/sparc fastdebug with test
> hotspot/test/compiler/intrinsics/crc32c/TestCRC32C.java
>
> thanks,
> Coleen
>
> On 9/17/15 9:18 AM, Doerr, Martin wrote:
>> Hi,
>>
>> unfortunately, there was a conflict with 8134553: CRC32C implementations for x86/x64 targets
>> I just rebased and updated the webrev:
>> http://cr.openjdk.java.net/~mdoerr/8136525_interp_entries/webrev.02/
>>
>> Please use this one.
>>
>> We tested the change in our nightly runs on x86, sparc and ppc64. I also tested the new webrev on x86 with some CRC32 code.
>>
>> Thanks,
>>     Martin
>>
>>
>> -----Original Message-----
>> From: hotspot-runtime-dev [mailto:hotspot-runtime-dev-bounces at openjdk.java.net] On Behalf Of Andrew Haley
>> Sent: Donnerstag, 17. September 2015 13:23
>> To: hotspot-runtime-dev at openjdk.java.net
>> Subject: Re: RFR(M): 8136525 Generate interpreter entries only once and avoid unnecessary jump to jump
>>
>> On 09/16/2015 07:38 PM, Coleen Phillimore wrote:
>>> This looks like a nice cleanup and I can sponsor it.
>>>
>>> Can someone test this on aarch64?
>> I'm getting a lot of crashes.  Part of the problem is that it does not
>> apply to current hs-comp so I had to hack the patch and I might have
>> messed it up.  I don't want to do any more until we get a patch
>> against current hs-comp.
>>
>> Andrew.
>>
>>