[aarch64-port-dev ] [11u] RFR (S) 8212043: AArch64: Add floating-point Math.min/max intrinsics

[Dmitry Chuyko]

Hi Paul,

Thank you. Here are three pieces for aarch64 part (diffs do not depend 
on each other):

Intrinsics: http://cr.openjdk.java.net/~dchuyko/8212043/webrev.11u.01/

Graal test: http://cr.openjdk.java.net/~dchuyko/8215687/webrev.11u.00/

jtreg test: http://cr.openjdk.java.net/~dchuyko/8218550/webrev.11u.00/

Testing: aarch64 build, tier1, tier2, Graal HotspotTest (with Graal 
enabled) - fails because of ECB, TestFpMinMaxIntrinsics.

I'd be grateful if you take x86 part.

-Dmitry

On 1/22/21 10:34 PM, Hohensee, Paul wrote:
> This looks fine to me, but please split the patch into two separate ones for 8212043 and 8215687, and push them together.
>
> Looking at the 8212043 links, we should also backport this test issue.
>
> https://bugs.openjdk.java.net/browse/JDK-8218550: Add test omitted from JDK-8212043
>
> And, if we're going to do this for aarch64, I'd want to see the Intel version backported too, see
>
> https://bugs.openjdk.java.net/browse/JDK-8217561: X86: Add floating-point Math.min/max intrinsics
>
> and the issues linked from it.
>
> https://bugs.openjdk.java.net/browse/JDK-8220407: compiler/intrinsics/math/TestFpMinMaxIntrinsics.java timedout
> https://bugs.openjdk.java.net/browse/JDK-8227222: vmTestbase/jit/FloatingPoint/gen_math/Loops04/Loops04.java hits assert(((!attributes->uses_vl()) || (attributes->get_vector_len() == AVX_512bit) || (!_legacy_mode_vl) || (attributes->is_legacy_mode()))) failed: XMM register should be 0-15
>
> I'd be willing to take these on if the this aarch64 backport is approved.
>
> Thanks,
> Paul
>
> -----Original Message-----
> From: jdk-updates-dev <jdk-updates-dev-retn at openjdk.java.net> on behalf of Dmitry Chuyko <dmitry.chuyko at bell-sw.com>
> Date: Tuesday, January 19, 2021 at 7:10 AM
> To: "jdk-updates-dev at openjdk.java.net" <jdk-updates-dev at openjdk.java.net>
> Cc: "aarch64-port-dev at openjdk.java.net" <aarch64-port-dev at openjdk.java.net>
> Subject: [11u] RFR (S) 8212043: AArch64: Add floating-point Math.min/max intrinsics
>
> Hello,
>
> Original RFE: https://bugs.openjdk.java.net/browse/JDK-8212043
> Original post-fix: https://bugs.openjdk.java.net/browse/JDK-8215687
>
> Original patch mostly applies cleanly except macro() additions in
> classes.hpp that were missing surroundings - they are redone manually.
> Initial post-fix for CheckGraalIntrinsics (HotspotTest) Graal test was
> modified to mark new intrinsics as toBeInvestigated in JDK 11+ and this
> change is included in the changeset.
>
> 11u webrev: http://cr.openjdk.java.net/~dchuyko/8212043/webrev.11u.00/
>
> Testing: tier1, tier2; compiler/graalunit with Graal on -
> CheckGraalIntrinsics fails because of ECB intrinsics (JDK-8229848,
> jdk11u-jvmci-defer) but not not because of min/max.
>
> --
> Thanks,
> -Dmitry
>
>