September 2023 - leyden-dev

git: openjdk/leyden: premain: Fixed CDS jtreg tests to use new command line options like -XX:+StoreCachedCode
by duke 06 Sep '23

06 Sep '23

Changeset: 265c226c Author: iklam <ilam3388(a)gmail.com> Date: 2023-09-05 20:46:38 +0000 URL: https://git.openjdk.org/leyden/commit/265c226c9a190b42bf8d6aaf0ff870dca6275… Fixed CDS jtreg tests to use new command line options like -XX:+StoreCachedCode ! test/hotspot/jtreg/runtime/cds/appcds/indy/IndyTestBase.java

1 0

git: openjdk/leyden: premain: Work around for JDK-8315719: [premain] -XX:+ArchiveInvokeDynamic is incompatible with old classes
by duke 05 Sep '23

05 Sep '23

Changeset: d960fb15 Author: iklam <ilam3388(a)gmail.com> Date: 2023-09-05 11:08:52 +0000 URL: https://git.openjdk.org/leyden/commit/d960fb15258cc99a1bf7f0b1e94bd8be06605… Work around for JDK-8315719: [premain] -XX:+ArchiveInvokeDynamic is incompatible with old classes ! src/hotspot/share/oops/instanceKlass.cpp ! src/hotspot/share/runtime/arguments.cpp + test/hotspot/jtreg/runtime/cds/appcds/indy/IndyMiscTests.java

1 0

git: openjdk/leyden: computed-constants: Remove @ValueBased from ODCCL
by duke 05 Sep '23

05 Sep '23

Changeset: 97db0130 Author: Per Minborg <pminborg(a)gmail.com> Date: 2023-09-05 09:19:06 +0000 URL: https://git.openjdk.org/leyden/commit/97db01308fac4ee3fe81c2d58013fe43f5049… Remove @ValueBased from ODCCL ! src/java.base/share/classes/jdk/internal/constant/OnDemandComputedConstantList.java ! src/java.base/share/classes/jdk/internal/constant/State.java

1 0

git: openjdk/leyden: computed-constants: Refactor to use @Stable state field
by duke 05 Sep '23

05 Sep '23

Changeset: ccf9155b Author: Per Minborg <pminborg(a)gmail.com> Date: 2023-09-05 09:05:05 +0000 URL: https://git.openjdk.org/leyden/commit/ccf9155b7de572ac003e10c302a71425796c3… Refactor to use @Stable state field ! src/java.base/share/classes/jdk/internal/constant/AbstractComputedConstant.java - src/java.base/share/classes/jdk/internal/constant/ConstantUtil.java ! src/java.base/share/classes/jdk/internal/constant/ListElementComputedConstant.java ! src/java.base/share/classes/jdk/internal/constant/MethodHandleComputedConstant.java ! src/java.base/share/classes/jdk/internal/constant/StandardComputedConstant.java + src/java.base/share/classes/jdk/internal/constant/State.java

1 0

Computed Constants API question
by - 02 Sep '23

02 Sep '23

Hello Per and Leyden subscribers, First, I am glad that we are finally adding an API that exposes one of core libraries' favorite feature, `@Stable`, to common users! For the API design, however, I have a request: Can we have a ComputedConstant factory that creates a List<V> in addition to one that creates a List<ComputedConstant<V>>? I think using the List<ComputedConstant<V>> is confusing. The example usages in the JEP [1] and in the API specification [2] are already wrong: we need an extra ConputedConstant.get() call to unwrap the ComputedConstant after List.get(index) call, which currently returns a computed constant than the actual value. The current List<ComputedConstant<V>> is to be kept in case users want fine-grained control over each constant's resolution failure, etc. and covers the new factory's functionality. But I believe the new factory will see wider usage: 1. None of the 2 old patterns in the "Motivation" section uses any of these exception handling or initialization state detection. 2. Returning a List<V> allows users to conveniently pass the list in usages instead of using streams or writing custom wrappers. A follow up to a previous request [3], I believe having a map (of type Map<K, V> instead of Map<K, ComputedConstant<V>>) would be feasible too. Finally, a side comment about the current OnDemandComputedConstantList: it computes ComputedConstant wrappers in addition to the actual constants on demand, which... seems a bit overkill, when ComputedConstant itself is already a lightweight wrapper of a heavy computation? Best, Chen Liang [1]: https://openjdk.org/jeps/8312611 "var kbd = lbl.labels.get(3);" [2]: https://cr.openjdk.org/~pminborg/computed-constant/api/java.base/java/lang/… "return PO2_CACHE.get(n);" [3]: https://mail.openjdk.org/pipermail/leyden-dev/2023-August/000277.html

6 15

git: openjdk/leyden: premain: 2 new changesets
by duke 01 Sep '23

01 Sep '23

Changeset: fa813084 Author: Vladimir Kozlov <vladimir.kozlov(a)oracle.com> Date: 2023-09-01 13:59:09 +0000 URL: https://git.openjdk.org/leyden/commit/fa8130845b7e26a4c11a95f35a2871d0a976a… Renames in cached code. ! src/hotspot/cpu/aarch64/c1_LIRAssembler_aarch64.cpp ! src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp ! src/hotspot/cpu/x86/c1_LIRAssembler_x86.cpp ! src/hotspot/cpu/x86/gc/g1/g1BarrierSetAssembler_x86.cpp ! src/hotspot/cpu/x86/gc/shared/cardTableBarrierSetAssembler_x86.cpp ! src/hotspot/cpu/x86/macroAssembler_x86.cpp ! src/hotspot/cpu/x86/sharedRuntime_x86_64.cpp ! src/hotspot/cpu/x86/stubGenerator_x86_64.cpp ! src/hotspot/cpu/x86/x86_64.ad ! src/hotspot/share/adlc/main.cpp ! src/hotspot/share/asm/codeBuffer.hpp ! src/hotspot/share/c1/c1_Compiler.cpp ! src/hotspot/share/c1/c1_LIRGenerator.cpp ! src/hotspot/share/cds/metaspaceShared.cpp ! src/hotspot/share/ci/ciEnv.cpp ! src/hotspot/share/ci/ciEnv.hpp ! src/hotspot/share/ci/ciField.cpp ! src/hotspot/share/ci/ciMethod.cpp ! src/hotspot/share/ci/ciObject.cpp ! src/hotspot/share/classfile/vmClasses.cpp ! src/hotspot/share/code/SCArchive.cpp ! src/hotspot/share/code/SCArchive.hpp ! src/hotspot/share/code/codeCache.cpp ! src/hotspot/share/code/compiledMethod.hpp ! src/hotspot/share/code/exceptionHandlerTable.hpp ! src/hotspot/share/code/nmethod.cpp ! src/hotspot/share/code/nmethod.hpp ! src/hotspot/share/code/relocInfo.cpp ! src/hotspot/share/compiler/compilationPolicy.cpp ! src/hotspot/share/compiler/compileBroker.cpp ! src/hotspot/share/compiler/compileBroker.hpp ! src/hotspot/share/compiler/compileTask.cpp ! src/hotspot/share/compiler/compileTask.hpp ! src/hotspot/share/compiler/compiler_globals.hpp ! src/hotspot/share/compiler/oopMap.hpp ! src/hotspot/share/jvmci/jvmciRuntime.cpp ! src/hotspot/share/logging/logTag.hpp ! src/hotspot/share/oops/method.cpp ! src/hotspot/share/oops/method.hpp ! src/hotspot/share/opto/c2compiler.cpp ! src/hotspot/share/opto/rootnode.cpp ! src/hotspot/share/runtime/globals.hpp ! src/hotspot/share/runtime/init.cpp ! src/hotspot/share/runtime/java.cpp ! src/hotspot/share/runtime/threads.cpp ! test/hotspot/jtreg/premain/javac/javac-test.sh ! test/hotspot/jtreg/premain/javac_helloworld/run.sh ! test/hotspot/jtreg/premain/jmh/run.sh ! test/hotspot/jtreg/premain/lib/premain-run.sh Changeset: fd82682f Author: Vladimir Kozlov <vladimir.kozlov(a)oracle.com> Date: 2023-09-01 14:30:23 +0000 URL: https://git.openjdk.org/leyden/commit/fd82682ff5f555407fbb9769cf4796af5e218… Rename class SCArchive to SCCache ! src/hotspot/cpu/aarch64/c1_LIRAssembler_aarch64.cpp ! src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp ! src/hotspot/cpu/x86/c1_LIRAssembler_x86.cpp ! src/hotspot/cpu/x86/gc/g1/g1BarrierSetAssembler_x86.cpp ! src/hotspot/cpu/x86/gc/shared/cardTableBarrierSetAssembler_x86.cpp ! src/hotspot/cpu/x86/macroAssembler_x86.cpp ! src/hotspot/cpu/x86/sharedRuntime_x86_64.cpp ! src/hotspot/cpu/x86/stubGenerator_x86_64.cpp ! src/hotspot/share/adlc/main.cpp ! src/hotspot/share/c1/c1_Compiler.cpp ! src/hotspot/share/c1/c1_LIRGenerator.cpp ! src/hotspot/share/cds/metaspaceShared.cpp ! src/hotspot/share/ci/ciEnv.cpp ! src/hotspot/share/ci/ciField.cpp ! src/hotspot/share/ci/ciObject.cpp = src/hotspot/share/code/SCCache.cpp = src/hotspot/share/code/SCCache.hpp ! src/hotspot/share/code/codeCache.cpp ! src/hotspot/share/code/nmethod.cpp ! src/hotspot/share/code/relocInfo.cpp ! src/hotspot/share/compiler/compilationPolicy.cpp ! src/hotspot/share/compiler/compileBroker.cpp ! src/hotspot/share/compiler/compileTask.cpp ! src/hotspot/share/opto/c2compiler.cpp ! src/hotspot/share/opto/rootnode.cpp ! src/hotspot/share/runtime/init.cpp ! src/hotspot/share/runtime/java.cpp ! src/hotspot/share/runtime/threads.cpp

1 0

ComputedConstant condenser
by Remi Forax 01 Sep '23

01 Sep '23

Hi Per, hi all, When coming back from the JVM Languages Summit, I've written a bytecode rewriter [1] (using ASM) that removes the private static final ConputedConstant inside the static block (<clinit>) and uses invokedynamic + constant dynamic when the methods ComputedConstant.get()/orElse()/orElseThrow() are called to initialize the value of the computed constant (once). If the static final ComputedConstant is needed for other methods or thing like an acmp, a shim is produced on the spot. I believe it's a valid semantics given that a ComputedConstant is a value based class. I've used the idea of John to group all the initializations inside a static method ($staticInit$) that takes the name of the constant and call the corresponging desugared lambda body (bypassing the lambda creation). The metafactory for the invokedynamic and constant dynamic is here [2]. For example for the code, public class Main { private static final ComputedConstant<String> TEXT = ComputedConstant.of(() -> "Hello"); public static String message() { return TEXT.get(); } static class Nested { public static String message2() { return TEXT.get(); } } public static void main(String[] args) { System.out.println(message()); System.out.println(TEXT.isBound()); // this one need a shim ComputedConstant System.out.println(Nested.message2()); } } The bytecode rewriter produces the following bytecode (you can notice that the <clinit> is empty !) Compiled from "Main.java" public class com.github.forax.concurrent.constant.Main { public com.github.forax.concurrent.constant.Main(); Code: 0: aload_0 1: invokespecial #1 // Method java/lang/Object."<init>":()V 4: return public static java.lang.String message(); Code: 0: invokedynamic #114, 0 // InvokeDynamic #2:get:()Ljava/lang/Object; 5: checkcast #19 // class java/lang/String 8: areturn public static void main(java.lang.String[]); Code: 0: getstatic #21 // Field java/lang/System.out:Ljava/io/PrintStream; 3: invokestatic #27 // Method message:()Ljava/lang/String; 6: invokevirtual #31 // Method java/io/PrintStream.println:(Ljava/lang/String;)V 9: getstatic #21 // Field java/lang/System.out:Ljava/io/PrintStream; 12: invokestatic #118 // Method java/lang/invoke/MethodHandles.lookup:()Ljava/lang/invoke/MethodHandles$Lookup; 15: ldc #8 // class com/github/forax/concurrent/constant/Main 17: ldc #119 // String TEXT 19: invokestatic #123 // Method com/github/forax/concurrent/constant/ComputedConstantMetafactory.ofShim:(Ljava/lang/invoke/MethodHandles$Lookup;Ljava/lang/Class;Ljava/lang/String;)Lcom/github/forax/concurrent/constant/ComputedConstant; 22: invokeinterface #37, 1 // InterfaceMethod com/github/forax/concurrent/constant/ComputedConstant.isBound:()Z 27: invokevirtual #41 // Method java/io/PrintStream.println:(Z)V 30: getstatic #21 // Field java/lang/System.out:Ljava/io/PrintStream; 33: invokestatic #44 // Method com/github/forax/concurrent/constant/Main$Nested.message2:()Ljava/lang/String; 36: invokevirtual #31 // Method java/io/PrintStream.println:(Ljava/lang/String;)V 39: return private static java.lang.String lambda$static$0(); Code: 0: ldc #49 // String Hello 2: areturn static {}; Code: 0: return public static java.lang.Object $staticInit$(java.lang.String); Code: 0: aload_0 1: invokevirtual #127 // Method java/lang/String.hashCode:()I 4: lookupswitch { // 1 2571565: 24 default: 40 } 24: aload_0 25: ldc #119 // String TEXT 27: invokevirtual #131 // Method java/lang/String.equals:(Ljava/lang/Object;)Z 30: ifeq 37 33: invokestatic #77 // Method lambda$static$0:()Ljava/lang/String; 36: areturn 37: goto 40 40: new #133 // class java/lang/AssertionError 43: dup 44: aload_0 45: invokespecial #136 // Method java/lang/AssertionError."<init>":(Ljava/lang/Object;)V 48: athrow } I still think that using a keyword is better than using an API. It's too easy to forget "final" or "private" when declaring a static computed constant or to think that a computed constant is a real object always available at runtime. The API of ComputedConstant also shows too much. All the methods is* should not be part of the public API so a condenser has more freedom in term of implementation and the user has less chance to see that a computed constant is perhaps not equals to itself (at least until the first part of Valhalla lands). The rewriter is pretty basic, it's one pass on the bytecode (no computed constant propagation) so for example storing a computed constant in a local variable inside the <clinit> is enough to avoid the computed consatnt to be optimized. I'm also pretty sure that this condenser may not work if a private static final computed constant is access by a nestmate class, but I have not taken a look yet. regards, Rémi [1] https://github.com/forax/computed-constant/blob/master/constant/src/main/ja… [2] https://github.com/forax/computed-constant/blob/master/constant/src/main/ja…

2 1

Re: [External] : ComputedConstant: rename to Lazy?
by Per-Ake Minborg 01 Sep '23

01 Sep '23

Hi Chen, In the beginning, we used Lazy but it became apparent that CCs are more than that. In fact, computation might occur even BEFORE the JVM starts and then the name Lazy becomes confusing. It is all but lazy in this case. So, we arrived at the more general name ComputedConstant. It does not carry any indication as to when the constant is computed. Best, Per ________________________________ From: liangchenblue(a)gmail.com <liangchenblue(a)gmail.com> Sent: Friday, September 1, 2023 6:59 AM To: leyden-dev(a)openjdk.org <leyden-dev(a)openjdk.org>; Per-Ake Minborg <per-ake.minborg(a)oracle.com> Subject: [External] : ComputedConstant: rename to Lazy? Hello, I noticed that ComputedConstant, in addition to ensuring constant-folding of lazy values, is a general-purpose lazy wrapper, and can be used even outside of constant-folding scenarios to offer thread safety and error handling in lazy computations. Thus, I recommend renaming it into Lazy, or some other name more simple and concise than ComputedConstant, to the ease of users. A similar scenario exists in JDK already: List.of() factory wraps an array with stable annotation, thereby offering constant-folded arrays/lists that's faster to iterate (random access via List.get() in an index loop, not using iterators) by index compared to a plain array when both are stored in static final fields. However, we are not calling such a factory list a "Constant List," but simply an immutable list. This proposal only changes the name of CC; it will not affect the specification (thread safety, error handling, constant-foldability) in any way, and it does not propose to change the method names. Best, Chen Liang

2 2