8g: fix conversion from float to uint64

The code for converting negative floats was
incorrectly loading an FP control word from
the stack without ever having stored it there.

Thanks to Lars Pensjö for reporting this bug.

R=golang-dev, rsc
CC=golang-dev
https://golang.org/cl/4515091

src/cmd/8g/gsubr.c

@@ -1429,11 +1429,11 @@ gmove(Node *f, Node *t)
 		split64(t, &tlo, &thi);
 		gins(AXORL, ncon(0x80000000), &thi);	// + 2^63
 		patch(p3, pc);
-		patch(p1, pc);
 		splitclean();
-
 		// restore rounding mode
 		gins(AFLDCW, &t1, N);
+
+		patch(p1, pc);
 		return;
 
 	/*

test/fixedbugs/bug341.go

+// $G $D/$F.go && $L $F.$A && ./$A.out || echo BUG: bug341
+
+// Copyright 2011 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.
+
+// Used to panic because 8g was generating incorrect
+// code for converting a negative float to a uint64.
+
+package main
+
+func main() {
+	var x float32 = -2.5
+
+	_ = uint64(x)
+	_ = float32(0)
+}
+/*
+panic: runtime error: floating point error
+
+[signal 0x8 code=0x6 addr=0x8048c64 pc=0x8048c64]
+*/