mirror of
https://github.com/golang/go
synced 2024-11-13 18:10:24 -07:00
c6118af558
That optimization is not valid if x == -0. The test is a bit tricky because 0 == -0. We distinguish 0 from -0 with 1/0 == inf, 1/-0 == -inf. This has been a bug since CL 24790 in Go 1.8. Probably doesn't warrant a backport. Fixes #27718 Note: the optimization x-0 -> x is actually valid. But it's probably best to take it out, so as to not confuse readers. Change-Id: I99f16a93b45f7406ec8053c2dc759a13eba035fa Reviewed-on: https://go-review.googlesource.com/135701 Reviewed-by: Cherry Zhang <cherryyz@google.com>
73 lines
1.1 KiB
Go
73 lines
1.1 KiB
Go
// run
|
|
|
|
// Copyright 2018 The Go Authors. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
// (-0)+0 should be 0, not -0.
|
|
|
|
package main
|
|
|
|
//go:noinline
|
|
func add64(x float64) float64 {
|
|
return x + 0
|
|
}
|
|
|
|
func testAdd64() {
|
|
var zero float64
|
|
inf := 1.0 / zero
|
|
negZero := -1 / inf
|
|
if 1/add64(negZero) != inf {
|
|
panic("negZero+0 != posZero (64 bit)")
|
|
}
|
|
}
|
|
|
|
//go:noinline
|
|
func sub64(x float64) float64 {
|
|
return x - 0
|
|
}
|
|
|
|
func testSub64() {
|
|
var zero float64
|
|
inf := 1.0 / zero
|
|
negZero := -1 / inf
|
|
if 1/sub64(negZero) != -inf {
|
|
panic("negZero-0 != negZero (64 bit)")
|
|
}
|
|
}
|
|
|
|
//go:noinline
|
|
func add32(x float32) float32 {
|
|
return x + 0
|
|
}
|
|
|
|
func testAdd32() {
|
|
var zero float32
|
|
inf := 1.0 / zero
|
|
negZero := -1 / inf
|
|
if 1/add32(negZero) != inf {
|
|
panic("negZero+0 != posZero (32 bit)")
|
|
}
|
|
}
|
|
|
|
//go:noinline
|
|
func sub32(x float32) float32 {
|
|
return x - 0
|
|
}
|
|
|
|
func testSub32() {
|
|
var zero float32
|
|
inf := 1.0 / zero
|
|
negZero := -1 / inf
|
|
if 1/sub32(negZero) != -inf {
|
|
panic("negZero-0 != negZero (32 bit)")
|
|
}
|
|
}
|
|
|
|
func main() {
|
|
testAdd64()
|
|
testSub64()
|
|
testAdd32()
|
|
testSub32()
|
|
}
|