Comment #26 on issue 3006 by [email protected]: Inaccurate sin/cos values
http://code.google.com/p/v8/issues/detail?id=3006
I did some testing on reduction (in C). For x < 1.5, V8 and fdlibm are
equivalent. As x gets larger fdblim performs worse. At x = 8, fdlibm takes
1.5 times longer with accuracy slightly better than V8. For x = 1e36,
fdlibm takes 9 times longer, but V8 is hopelessly inaccurate now.
If we use a different implementation (Apache's CodyWaite routine), at x =
8, it's 1.4 times slower. At x = 3e5, it's the performance. Note that
Apache's CodyWaite returns different (but correct) values from the results
of fdlibm.
I find this is a reasonable tradeoff: If you use huge args, you have to
pay the price. If you're not willing to, the developer can easily do
simple pi reduction himself.
And can't V8 just call fdlibm's C implementation of reduction? It would be
quite a bit of work and testing to translate that to javascript.
--
You received this message because this project is configured to send all
issue notifications to this address.
You may adjust your notification preferences at:
https://code.google.com/hosting/settings
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.
