8364407: [REDO] Consolidate Identity of self-inverse operations #26823

SirYwell · 2025-08-18T15:49:48Z

The previous approach was flawed for short and char as these are int-subtypes and truncate the result (see the backout issue https://bugs.openjdk.org/browse/JDK-8364409 for a reproducer).

This change now first ensures that the input type is small enough so no truncation gets lost when dropping the operations.

The previous implementation also used an InvolutionNode superclass with one Identity(...) implementation, but there were some reservations whether this is the right way to go. As we now have a ReverseBytesNode, there is also less benefit in having the supertype, as this covers 4 in 1 already.

I also added test cases on top of the original ones that ensure the nodes stay when we can't prove the input type is small enough.

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8364407: [REDO] Consolidate Identity of self-inverse operations (Enhancement - P4)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/26823/head:pull/26823
$ git checkout pull/26823

Update a local copy of the PR:
$ git checkout pull/26823
$ git pull https://git.openjdk.org/jdk.git pull/26823/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 26823

View PR using the GUI difftool:
$ git pr show -t 26823

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/26823.diff

Using Webrev

Link to Webrev Comment

SirYwell · 2025-08-18T15:50:37Z

test/hotspot/jtreg/compiler/c2/gvn/InvolutionIdentityTests.java

+    }
+
+    @Test
+    @IR(counts = {IRNode.REVERSE_BYTES_S, "2"})


I'm not sure if this is fine. The intrinsics might not apply to all platforms, in which case this would fail I think?

The only case I can immediately think of is riscv without -XX:+UseZbb. But you can easily disable the test for that platform.

bridgekeeper · 2025-08-18T15:50:45Z

👋 Welcome back hgreule! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2025-08-18T15:53:32Z

❗ This change is not yet ready to be integrated.
See the Progress checklist in the description for automated requirements.

openjdk · 2025-08-18T15:54:36Z

@SirYwell The following label will be automatically applied to this pull request:

hotspot-compiler

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

mlbridge · 2025-08-18T15:57:55Z

Webrevs

00: Full (a086a4de)

mhaessig

Thank you for noticing the bug and improving your original PR, @SirYwell! I like your new approach without a superclass and your extensive testing of all edge cases and random values. Nice work!

However, I do have some questions below.

mhaessig · 2025-08-18T16:30:40Z

test/hotspot/jtreg/compiler/c2/gvn/InvolutionIdentityTests.java

+ * @summary Test that Identity simplifications of Involution nodes are being performed as expected.
+ * @library /test/lib /


Suggested change

* @summary Test that Identity simplifications of Involution nodes are being performed as expected.

* @library /test/lib /

* @summary Test that Identity simplifications of Involution nodes are being performed as expected.

* @key randomness

* @library /test/lib /

Since you are using random inputs.

mhaessig · 2025-08-18T16:40:11Z

src/hotspot/share/opto/subnode.cpp

+    // Operations on sub-int types might not be "real" involutions for values outside their type range.
+    // Make sure not to drop potential truncations.
+    if (type == nullptr || involution->bottom_type()->is_int()->contains(type)) {
+      return involution->in(1)->in(1);
+    }


Instead of skipping the optimization, could you "clean" involution->in(1)->in(1) using mask_int_value()? That would follow the semantics of JVMS§6.5 ireturn.

I think this would need to happen in Ideal() then instead. Doing that would work too, but it further complicates things with little benefits imo.

mhaessig · 2025-08-18T16:50:12Z

test/hotspot/jtreg/compiler/c2/gvn/InvolutionIdentityTests.java

Why are you not testing involution on NegL/I nodes? Can this not be optimized?

NegL/I aren't used currently (see e.g. https://bugs.openjdk.org/browse/JDK-8262346).

jaskarth

This looks nice! I've just left some style comments.

jaskarth · 2025-08-18T16:57:58Z

src/hotspot/share/opto/subnode.cpp

+static Node* simplify_involution(PhaseGVN* phase, Node* involution) {
+  if (involution->in(1)->Opcode() == involution->Opcode()) {
+    Node* original = involution->in(1)->in(1);
+    const TypeInt *type = phase->type(original)->isa_int();


Suggested change

const TypeInt *type = phase->type(original)->isa_int();

const TypeInt* type = phase->type(original)->isa_int();

jaskarth · 2025-08-18T17:11:13Z

test/hotspot/jtreg/compiler/c2/gvn/InvolutionIdentityTests.java

+
+    }


There's an extra whitespace:

Suggested change

}

}

jaskarth · 2025-08-18T17:48:25Z

src/hotspot/share/opto/subnode.cpp

+  if (involution->in(1)->Opcode() == involution->Opcode()) {
+    Node* original = involution->in(1)->in(1);
+    const TypeInt *type = phase->type(original)->isa_int();
+    // Operations on sub-int types might not be "real" involutions for values outside their type range.


I think it would be helpful to state an example of the disallowed case in the comment, maybe something like:

Suggested change

// Operations on sub-int types might not be "real" involutions for values outside their type range.

// Operations on sub-int types might not be "real" involutions for values outside their type range, for example a ReverseBytesS node with an input larger than short.

SirYwell added 2 commits August 18, 2025 08:07

test

fbc4be7

redo with type check

a086a4d

SirYwell commented Aug 18, 2025

View reviewed changes

openjdk bot changed the title ~~8364407~~ 8364407: [REDO] Consolidate Identity of self-inverse operations Aug 18, 2025

openjdk bot added hotspot-compiler [email protected] rfr Pull request is ready for review labels Aug 18, 2025

mhaessig suggested changes Aug 18, 2025

View reviewed changes

jaskarth reviewed Aug 18, 2025

View reviewed changes

SirYwell marked this pull request as draft August 19, 2025 08:27

openjdk bot removed the rfr Pull request is ready for review label Aug 19, 2025

		* @summary Test that Identity simplifications of Involution nodes are being performed as expected.
		* @library /test/lib /

	const TypeInt *type = phase->type(original)->isa_int();
	const TypeInt* type = phase->type(original)->isa_int();

	// Operations on sub-int types might not be "real" involutions for values outside their type range.
	// Operations on sub-int types might not be "real" involutions for values outside their type range, for example a ReverseBytesS node with an input larger than short.

8364407: [REDO] Consolidate Identity of self-inverse operations #26823

Are you sure you want to change the base?

8364407: [REDO] Consolidate Identity of self-inverse operations #26823

Conversation

SirYwell commented Aug 18, 2025 • edited by openjdk bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Progress

Issue

Reviewing

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

bridgekeeper bot commented Aug 18, 2025

Uh oh!

openjdk bot commented Aug 18, 2025

Uh oh!

openjdk bot commented Aug 18, 2025

Uh oh!

mlbridge bot commented Aug 18, 2025

Webrevs

Uh oh!

mhaessig left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jaskarth left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

SirYwell commented Aug 18, 2025 •

edited by openjdk bot

Loading