8351515: C2 incorrectly removes double negation for double and float

src/hotspot/share/opto/subnode.cpp

@@ -53,11 +53,16 @@ Node* SubNode::Identity(PhaseGVN* phase) {
   assert(in(1) != this, "Must already have called Value");
   assert(in(2) != this, "Must already have called Value");
 
-  // Remove double negation
-  const Type *zero = add_id();
-  if( phase->type( in(1) )->higher_equal( zero ) &&
+  const Type* zero = add_id();
+
+  // Remove double negation if it is not a floating point number since negation
+  // is not the same as subtraction for floating point numbers
+  // (cf. JLS § 15.15.4). `0-(0-(-0.0))` must be equal to positive 0.0 according to
+  // JLS § 15.8.2, but would result in -0.0 if this folding would be applied.
+  if (phase->type(in(1))->higher_equal(zero) &&
       in(2)->Opcode() == Opcode() &&
-      phase->type( in(2)->in(1) )->higher_equal( zero ) ) {
+      phase->type(in(2)->in(1))->higher_equal(zero) &&
+      !phase->type(in(2)->in(2))->is_floatingpoint()) {
     return in(2)->in(2);
   }
 

test/hotspot/jtreg/compiler/floatingpoint/TestSubNodeFloatDoubleNegation.java

@@ -0,0 +1,80 @@
+/*
+ * Copyright (c) 2025, Oracle and/or its affiliates. All rights reserved.
+ * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License version 2 only, as
+ * published by the Free Software Foundation.
+ *
+ * This code is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * version 2 for more details (a copy is included in the LICENSE file that
+ * accompanied this code).
+ *
+ * You should have received a copy of the GNU General Public License version
+ * 2 along with this work; if not, write to the Free Software Foundation,
+ * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
+ * or visit www.oracle.com if you need additional information or have any
+ * questions.
+ */
+
+/**
+ * @test
+ * @bug 8351515
+ * @summary The compiler must not fold `0-(0-x)` for x: float | double
+ * @library /test/lib /
+ * @modules jdk.incubator.vector
+ * @run driver compiler.floatingpoint.TestSubNodeFloatDoubleNegation
+ */
+package compiler.floatingpoint;
+
+import compiler.lib.ir_framework.*;
+import jdk.incubator.vector.Float16;
+import jdk.test.lib.Asserts;
+
+public class TestSubNodeFloatDoubleNegation {
+
+    public static void main(String[] args) {
+        TestFramework.runWithFlags("--add-modules=jdk.incubator.vector", "-XX:CompileCommand=inline,jdk.incubator.vector.Float16::*");
+    }
+
+    @Run(test = { "testHalfFloat", "testFloat", "testDouble" })
+    public static void assertResults() {
+        short halfFloatRes = Float16.float16ToShortBits(testHalfFloat(Float16.valueOf(-0.0f)));
+        int floatRes = Float.floatToIntBits(testFloat(-0.0f));
+        long doubleRes = Double.doubleToLongBits(testDouble(-0.0));
+
+        Asserts.assertEQ((short) 0, halfFloatRes);
+        Asserts.assertEQ((int) 0, floatRes);
+        Asserts.assertEQ((long) 0, doubleRes);
+    }
+
+    @Test
+    // Match a generic SubNode, because scalar Float16 operations are often
+    // performed as float operations.
+    @IR(counts = { IRNode.SUB, "2" })
+    // Checks that the subtractions in 0 - (0 - hf) do not get eliminiated
+    public static Float16 testHalfFloat(Float16 hf) {
+        return Float16.subtract(
+                Float16.shortBitsToFloat16((short) 0),
+                Float16.subtract(Float16.shortBitsToFloat16((short) 0), hf));
+    }
+
+    @Test
+    @IR(counts = { IRNode.SUB_F, "2" })
+    // Checks that the subtractions in 0 - (0 - f) do not get eliminated
+    public static float testFloat(float f) {
+        return 0 - (0 - f);
+    }
+
+    @Test
+    @IR(counts = { IRNode.SUB_D, "2" })
+    // Checks that the subtractions in 0 - (0 - d) do not get eliminated
+    public static double testDouble(double d) {
+        return 0 - (0 - d);
+    }
+
+}

test/hotspot/jtreg/compiler/lib/ir_framework/IRNode.java

@@ -1833,7 +1833,7 @@ public class IRNode {
 
     public static final String SUB = PREFIX + "SUB" + POSTFIX;
     static {
-        beforeMatchingNameRegex(SUB, "Sub(I|L|F|D)");
+        beforeMatchingNameRegex(SUB, "Sub(I|L|F|D|HF)");
     }
 
     public static final String SUB_D = PREFIX + "SUB_D" + POSTFIX;