Comment #22 on issue 3903 by [email protected]: Math.log10 is unexpectedly
slow
https://code.google.com/p/v8/issues/detail?id=3903
#21: Some points there:
1. In #3 it's stated the Chrome 41 implementation of Math.log10 is slower
than the approximation polyfill in Chrome 40, which explains why our
results differ.
2. In that test, I am obtaining log10 via reciprocal estimate, as a mirror
of the first test with the native Math.log; I did not reimplement Chrome
41's new implementation of Math.log10 that wraps Math.log. Hence, the
earlier comparison to Math.log only.
3. I was not able to account for native V8 inlined intrinsics or native
code context switching costs in my tests, so the actual integrated
performance of your log port as a component of Math.log10 could not be
directly measured by my test.
4. Thus, some of the performance difference you see is that a reciprocal
estimate is faster than Chrome 41's Math.log10 code *before* invoking
Math.log. But most of the cost, as post #3 alludes to, is likely the
overhead from the context switches / lack of inlining / etc.
tl;dr: It's probably faster than Chrome 41 but that would need to be
verified with integration testing.
--
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/d/optout.
