Bug 134873 - Allow for Int52Rep to see things other than Int32, and make this testable
Summary: Allow for Int52Rep to see things other than Int32, and make this testable
Status: RESOLVED FIXED
Alias: None
Product: WebKit
Classification: Unclassified
Component: JavaScriptCore (show other bugs)
Version: 528+ (Nightly build)
Hardware: All All
: P2 Normal
Assignee: Filip Pizlo
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2014-07-13 14:47 PDT by Filip Pizlo
Modified: 2014-07-14 17:42 PDT (History)
7 users (show)

See Also:


Attachments
work in progress (16.65 KB, patch)
2014-07-13 14:48 PDT, Filip Pizlo
no flags Details | Formatted Diff | Diff
more (33.34 KB, patch)
2014-07-13 16:48 PDT, Filip Pizlo
no flags Details | Formatted Diff | Diff
the patch (55.80 KB, patch)
2014-07-14 13:25 PDT, Filip Pizlo
ggaren: review+
Details | Formatted Diff | Diff

Note You need to log in before you can comment on or make changes to this bug.
Description Filip Pizlo 2014-07-13 14:47:17 PDT
This would only occur if prediction propagation messes up.  But, currently, prediction propagation *is* allowed to mess up - in all cases except for Int52Rep.  We should just make it so that the cases that lead to Int52Rep can also involve messed up prediction propagation.

<rdar://problem/17641915>
Comment 1 Filip Pizlo 2014-07-13 14:48:39 PDT
Created attachment 234832 [details]
work in progress

Still have to write the FTL lowering for MachineIntUse and DoubleRepMachineIntUse, and then I have to write tests that use FiatInt52.
Comment 2 Filip Pizlo 2014-07-13 16:48:37 PDT
Created attachment 234834 [details]
more

Still have a bunch of tests to write.
Comment 3 Filip Pizlo 2014-07-14 13:25:35 PDT
Created attachment 234875 [details]
the patch
Comment 4 Geoffrey Garen 2014-07-14 13:50:11 PDT
Comment on attachment 234875 [details]
the patch

View in context: https://bugs.webkit.org/attachment.cgi?id=234875&action=review

r=me

> Source/JavaScriptCore/ChangeLog:9
> +        wants and we'll still valid valid IR after Fixup. This previously didn't work with Int52s.

valid valid => produce valid
Comment 5 Mark Hahnenberg 2014-07-14 14:03:09 PDT
Comment on attachment 234875 [details]
the patch

View in context: https://bugs.webkit.org/attachment.cgi?id=234875&action=review

r=me too

>> Source/JavaScriptCore/ChangeLog:9
>> +        wants and we'll still valid valid IR after Fixup. This previously didn't work with Int52s.
> 
> valid valid => produce valid

have valid IR?
Comment 6 Filip Pizlo 2014-07-14 17:42:26 PDT
Landed in http://trac.webkit.org/changeset/171096