infernap12/jdk
1
0
Fork 0
mirror of https://github.com/openjdk/jdk.git synced 2026-01-28 03:58:21 +00:00
Code Issues Packages Projects Releases Wiki Activity

8374540: Add comment describing implementation choices of Math.fma

Browse Source 
Reviewed-by: rgiulietti
This commit is contained in:
Joe Darcy 2026-01-08 18:51:25 +00:00
parent 9fd86e3749
commit 8212993ac3
1 changed files with 17 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
src/java.base/share/classes/java/lang/Math.java
View File

@ -1,5 +1,5 @@
/*
* Copyright (c) 1994, 2025, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 1994, 2026, 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
@ -2378,6 +2378,20 @@ public final class Math {
*/
@IntrinsicCandidate
public static double fma(double a, double b, double c) {
// Implementation note: this method is intentionally coded in
// a straightforward manner relying on BigDecimal for the
// heavy-lifting of the numerical computation. It would be
// possible for the computation to be done solely using binary
// floating-point and integer operations, at the cost of more
// complicated logic. Since most processors have hardware
// support for fma and this method is an intrinsic candidate,
// the software implementation below would only be used on
// processors without native fma support (and also possibly on
// processors with native fma support while running in the
// interpreter). Therefore, the direct performance of the code
// is less of a concern than the code's simplicity,
// maintainability, and testability.
/*
* Infinity and NaN arithmetic is not quite the same with two
* roundings as opposed to just one so the simple expression
@ -2492,6 +2506,8 @@ public final class Math {
*/
@IntrinsicCandidate
public static float fma(float a, float b, float c) {
// See implementation note in fma(double, double, double).
if (Float.isFinite(a) && Float.isFinite(b) && Float.isFinite(c)) {
if (a == 0.0 || b == 0.0) {
return a * b + c; // Handled signed zero cases