[Webkit-unassigned] [Bug 182788] Test fails with fused multiply-add instructions in parseInt

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=182788

--- Comment #9 from Filip Pizlo <fpizlo at apple.com> ---
(In reply to Dominik Inführ from comment #0)
> Created attachment 333786 [details]
> Patch
> 
> This patch allows to pass multiple allowed values to verify. On ARMv7 this
> test fails because GCC compiles parseInt with fused multiply-add
> instructions. The same thing happens on x64 when compiling with e.g.
> -march=skylake. Using fused multiply-add seems to be allowed by the standard
> for this radix: https://tc39.github.io/ecma262/#sec-parseint-string-radix.
> Another option would be to compile JSC with "-ffp-contract=off" to disable
> the use of fused multiply-add. I am not sure what's your preferred way of
> dealing with this test failure, see this patch as an initial proposal.

JSC is supposed to be a portable VM.  I don't think we ever want to allow something as fundamental as parseInt to behave differently depending on whether multiply and add can be fused.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20180307/bf62002d/attachment.html>