927389 - OdinMonkey: Assertion failure: max_exponent_ + canHaveFractionalPart_ >= mozilla::FloorLog2(mozilla::Abs(upper_)), at jit/RangeAnalysis.h

Status: RESOLVED FIXED

(Core :: JavaScript Engine: JIT, defect)

Description

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

Attachment: stack

(function() {
    "use asm"
    function f() {
        +~~1.1
    }
})()

asserts js debug shell on m-c changeset c14ca6b27b30 with --ion-gvn=off at Assertion failure: max_exponent_ + canHaveFractionalPart_ >= mozilla::FloorLog2(mozilla::Abs(upper_)), at jit/RangeAnalysis.h

My configure flags are:

CC="clang -Qunused-arguments" AR=ar CXX="clang++ -Qunused-arguments" sh ./configure --target=x86_64-apple-darwin12.5.0 --enable-optimize --enable-debug --enable-profiling --enable-gczeal --enable-debug-symbols --enable-methodjit --enable-type-inference --disable-tests --with-ccache

Comment 1

[Christian Holler (:decoder)]

JSBugMon: Cannot process bug: Unable to automatically reproduce, please track manually.

Comment 2

[Gary Kwong [:gkw] [:nth10sd] (NOT official MoCo now)]

autoBisect shows this is probably related to the following changeset:

The first bad revision is:
changeset:   http://hg.mozilla.org/mozilla-central/rev/42a20a0d4269
user:        Dan Gohman
date:        Tue Oct 15 20:49:44 2013 -0700
summary:     Bug 918607 - IonMonkey: Add a Range::setDouble which takes double arguments and use it to simplify and generalize several things.

Dan, is bug 918607 a likely regressor?

Comment 3

[Dan Gohman [:sunfish]]

Yes, it is.

The bug here is that Range::wrapAroundToInt32 is setting hasFractionalPart_ to false without checking whether this allows lower_/upper_ to be improved, and the new strict checking is catching the fact that an opportunity for improvement was missed.

Comment 4

[Dan Gohman [:sunfish]]

Attachment: range-refine-bounds-after-clearing-fractional.patch

This patch refactors out the code for refining lower_/upper_ using max_exponent_ from Range::intersect into a helper function so that it can be used by Range::wrapAroundToInt32 too.

Comment 5

[Nicolas B. Pierron [:nbp]]

Comment on attachment 819378 [details] [diff] [review]
range-refine-bounds-after-clearing-fractional.patch

Review of attachment 819378 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jit/RangeAnalysis.h
@@ +256,5 @@
>  
> +    // Compute the best lower and upper bounds that would be compatible with
> +    // max exponent value of e, if possible. If it is, set *impliedLower and
> +    // *impliedUpper and return true. Otherwise return false.
> +    static void refineInt32BoundsByExponent(uint16_t e, int32_t *l, int32_t *h) {

nit: Precise that this function is only useful when we get the exponent from a range which has a fractional part, as only in such case the exponent might be more precise than the int32 bound.

Comment 6

[Christian Holler (:decoder)]

No ASM.js required here, this reproduces for me as well:


(function () {
    var x = 0 || 1.1;
    x >> x;
})();

Comment 7

[Christian Holler (:decoder)]

Attachment: [crash-signature] Machine-readable crash signature

Comment 8

[Dan Gohman [:sunfish]]

(In reply to Nicolas B. Pierron [:nbp] from comment #5)
> Comment on attachment 819378 [details] [diff] [review]
> range-refine-bounds-after-clearing-fractional.patch
> 
> Review of attachment 819378 [details] [diff] [review]:
> -----------------------------------------------------------------
> 
> ::: js/src/jit/RangeAnalysis.h
> @@ +256,5 @@
> >  
> > +    // Compute the best lower and upper bounds that would be compatible with
> > +    // max exponent value of e, if possible. If it is, set *impliedLower and
> > +    // *impliedUpper and return true. Otherwise return false.
> > +    static void refineInt32BoundsByExponent(uint16_t e, int32_t *l, int32_t *h) {
> 
> nit: Precise that this function is only useful when we get the exponent from
> a range which has a fractional part, as only in such case the exponent might
> be more precise than the int32 bound.

Done.

https://hg.mozilla.org/integration/mozilla-inbound/rev/ce54c42790f6

Comment 9

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/ce54c42790f6