New issue
Advanced search Search tips

Issue 621611 link

Starred by 1 user
Status: Fixed
Owner:
bmeu...@chromium.org
Closed: Jul 2016
Cc:
yangguo@chromium.org
rtoy@chromium.org
verwa...@chromium.org
Components:
EstimatedDays: ----
NextAction: ----
OS: All
Pri: 2
Type: Bug

Blocking:
issue v8:5086



Sign in to add a comment

v8 regression or user land bug around rounding numbers / exp form

Project Member Reported by pfeldman@chromium.org, Jun 20 2016 
This comes from PWA summit code labs...

https://tuinderlusten-jheronimusbosch.ntr.nl/en

throws in console upon zooming and panning, bisected to:

https://chromium.googlesource.com/chromium/src/+log/33cf8ebc9313936d8f9bb79748863821f0070824..88b4aebc7f3eaed5a961b82cb77a021c9927e1a7 and eventually to 

https://chromium.googlesource.com/v8/v8/+/d5f2ac5e3326f281cbe7ba57464a4b17a9938e37

Not sure if the user land bug or regression, but definitely an observable change in behavior.

Comment 1 by bmeu...@chromium.org, Jun 21 2016

Blocking: v8:5086
Cc: bmeu...@chromium.org yangguo@chromium.org
Components: -Blink>JavaScript Blink>JavaScript>Runtime
Labels: -Pri-3 -OS-Mac Hotlist-Fixit-PE2016 OS-All Pri-2
Owner: rtoy@chromium.org
Ok, so this particular issue is about Math.E not being equal to Math.exp(1).
I have a fix in flight that works around the problem. It seems like the fdlibm Math.exp gets rounding wrong in case of Math.exp(1).

rtoy@: Can you help us with the exp algorithm?

Comment 2 by rtoy@chromium.org, Jun 21 2016 

I think your fix for Math.E vs Math.exp(1) is about the best that can be done short of implementing correctly rounded special functions.

Comment 3 by bmeu...@chromium.org, Jul 1 2016

Cc: -bmeu...@chromium.org rtoy@chromium.org
Owner: bmeu...@chromium.org
Status: Fixed (was: Assigned)
►

Sign in to add a comment