FrankChen021 commented on code in PR #19768: URL: https://github.com/apache/druid/pull/19768#discussion_r3871712989
########## processing/src/test/java/org/apache/druid/math/expr/vector/VectorTestAssertions.java: ########## @@ -0,0 +1,64 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, + * software distributed under the License is distributed on an + * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY + * KIND, either express or implied. See the License for the + * specific language governing permissions and limitations + * under the License. + */ + +package org.apache.druid.math.expr.vector; + +import org.apache.druid.java.util.common.StringUtils; +import org.junit.jupiter.api.Assertions; + +/** + * Shared assertions for tests that compare vectorized (SIMD or otherwise) expression outputs against a + * scalar reference and need to tolerate the small numeric differences the SIMD path can introduce for + * VO_MATHLIB-backed transcendentals — see {@code SimdVoMathlibParityTest} and the {@code useVectorMathApi} + * flag javadoc for context. + */ +public final class VectorTestAssertions +{ + private VectorTestAssertions() + { + // no instantiation + } + + /** + * Assert that two doubles agree to within {@code maxUlps} ulps of each other. NaN is treated as equal to + * NaN and bit-identical values pass trivially. On failure, the message identifies the expected/actual bits + * and the measured ulp delta. + */ + public static void assertDoublesEquivalent(String message, double expected, double actual, int maxUlps) + { + if (Double.isNaN(expected) && Double.isNaN(actual)) { + return; + } + if (Double.doubleToRawLongBits(expected) == Double.doubleToRawLongBits(actual)) { + return; + } + final double ulp = Math.ulp(expected); + final double diff = Math.abs(actual - expected); + if (Double.isNaN(diff) || diff > (double) maxUlps * ulp) { Review Comment: P2 Handle non-finite expected values before ULP comparison When expected is infinite and actual is finite or has the opposite sign, Math.ulp(expected) and the comparison threshold are infinite, so this assertion can incorrectly pass. Handle non-finite values with exact equality before calculating the ULP threshold so parity tests fail when an infinite expected result is not reproduced. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
