On Sun, 1 Oct 2006, Linus Torvalds wrote:
>
> See? Even from a purely optimization angle, we're actually better off just
> noticing that the whole comparison can be done in the narrower type.
So here's a test-case:
typedef unsigned short le16;
int cmp(le16 a, le16 b)
{
return a == (a & ~b);
}
and doing a "./test-linearize" on this gets us:
cmp:
.L0xf7fa500c:
<entry-point>
dead %arg1
scast.32 %r2 <- (16) %arg1
dead %arg2
scast.32 %r6 <- (16) %arg2
dead %r6
not.32 %r7 <- %r6
dead %r7
and.32 %r8 <- %r2, %r7
dead %r8
dead %r2
seteq.32 %r9 <- %r2, %r8
dead %r9
scast.32 %r10 <- (1) %r9
dead %r10
ret.32 %r10
which is just horrible code anyway.
I think it _should_ be
cmp:
.L0xf7fa500c:
<entry-point>
dead %arg2
not %r4 <- %arg2
dead %r4
and %r5 <- %arg1, %r4
dead %r5
dead %arg1
seteq.16 %r6 <- %arg1, %r5
dead %r6
scast.32 %r7 <- (1) %r6
dead %r7
ret.32 %r7
instead, which is obviously smaller and avoids the two unnecessary casts.
(Now, it's entirely possible that we warn so early that we can't
reasonably do this optimization until after we've already warned, but I've
been able to remove these kinds of warnings before, so I think it should
be possible to do it here too).
I much prefer (if possible) the "make sparse so much smarter that it sees
that it's ok" approach over "let's add a magic special case". For example,
a lot of the work I did to make the lock acquire/release logic useful was
very much about teaching sparse to simplify code-flow rather than anything
else.
But hey, maybe this case is too hard or nasty..
Linus
-
To unsubscribe from this list: send the line "unsubscribe linux-sparse" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
