From: Keith Randall <khr@golang.org>
Date: Mon, 19 Oct 2015 23:13:58 +0000 (-0700)
Subject: [dev.ssa] cmd/compile: Don't rematerialize getg
X-Git-Tag: go1.7beta1~1623^2^2~137
X-Git-Url: http://www.git.cypherpunks.su/?a=commitdiff_plain;h=2dc88eead8d28c6d84f60e30746d90a76f920be6;p=gostls13.git

[dev.ssa] cmd/compile: Don't rematerialize getg

It isn't safe in functions that also call setg.

Change-Id: I76a7bf0401b4b6c8a129c245b15a2d6f06080e94
Reviewed-on: https://go-review.googlesource.com/16095
Reviewed-by: Todd Neal <todd@tneal.org>
---

diff --git a/src/cmd/compile/internal/ssa/regalloc.go b/src/cmd/compile/internal/ssa/regalloc.go
index 6418bb375d..72b056cd8d 100644
--- a/src/cmd/compile/internal/ssa/regalloc.go
+++ b/src/cmd/compile/internal/ssa/regalloc.go
@@ -982,7 +982,11 @@ func (v *Value) rematerializeable() bool {
 		// which can't be moved.
 		return false
 	}
-	// TODO: maybe not OpAMD64LoweredGetG?
+	if v.Op == OpAMD64LoweredGetG {
+		// It would almost always be ok to rematerialize this op.
+		// The annoying exception is functions that call runtime.setg.
+		return false
+	}
 	if len(v.Args) == 0 {
 		return true
 	}