blob: bc83e04312d26f6f8aeb3e23f6624a9d5150123f (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
|
// This isn't actually testing much, because no warning is emitted in versions
// before the fix which comes with this because the method isn't even considered
// for inlining due to the bug.
class A {
private var debug = false
@inline private def ifelse[T](cond: => Boolean, ifPart: => T, elsePart: => T): T =
if (cond) ifPart else elsePart
final def bob1() = ifelse(debug, 1, 2)
final def bob2() = if (debug) 1 else 2
}
// Cool:
//
// % ls -1 /tmp/2901/
// A$$anonfun$bob1$1.class
// A$$anonfun$bob1$2.class
// A$$anonfun$bob1$3.class
// A.class
// % ls -1 /tmp/trunk
// A.class
//
// Observations:
//
// (1) The inlined version accesses the field: the explicit one calls the accessor.
// (2) The inlined version fails to eliminate boxing. With reference types it emits
// an unneeded checkcast.
// (3) The private var debug is mangled to A$$debug, but after inlining it is never accessed
// from outside of the class and doesn't need mangling.
// (4) We could forego emitting bytecode for ifelse entirely if it has been
// inlined at all sites.
//
// Generated bytecode for the above:
//
// public final int bob1();
// Code:
// Stack=1, Locals=1, Args_size=1
// 0: aload_0
// 1: getfield #11; //Field A$$debug:Z
// 4: ifeq 14
// 7: iconst_1
// 8: invokestatic #41; //Method scala/runtime/BoxesRunTime.boxToInteger:(I)Ljava/lang/Integer;
// 11: goto 18
// 14: iconst_2
// 15: invokestatic #41; //Method scala/runtime/BoxesRunTime.boxToInteger:(I)Ljava/lang/Integer;
// 18: invokestatic #45; //Method scala/runtime/BoxesRunTime.unboxToInt:(Ljava/lang/Object;)I
// 21: ireturn
//
// public final int bob2();
// Code:
// Stack=1, Locals=1, Args_size=1
// 0: aload_0
// 1: invokevirtual #48; //Method A$$debug:()Z
// 4: ifeq 11
// 7: iconst_1
// 8: goto 12
// 11: iconst_2
// 12: ireturn
|
