Comment #22 on issue 3006 by [email protected]: Inaccurate sin/cos
values
http://code.google.com/p/v8/issues/detail?id=3006
@20: Does "at least as accurate and at least as fast as what we have now"
work for you? A small regression on the one metric in return for a huge
improvement on the other may be acceptable too, but I don't have a hard
rule for that -- it's best to avoid regressions entirely.
As for benchmarks, one that we unfortunately still have to look at is
Sunspider's "3d-morph", which computes 120 different sin values 120 times
each (but in a way that the compiler can't easily avoid the recomputations).
When Yang developed the current implementation, AFAIK he used a simple
handcrafted benchmark, something like "compute 1M sin values for random
inputs" (using a repeatable sequence of pseudo-random numbers, of course).
You can find our automated correctness/accuracy test at
https://code.google.com/p/v8/source/browse/branches/bleeding_edge/test/mjsunit/sin-cos.js .
--
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.
