On Fri, 12 Dec 2014, Joe Landman wrote: > On 12/12/2014 11:20 AM, Sage Weil wrote: > > > We can't use floating point. The code needs to run in the kernel. We also > > need the rseults to be perfectly deterministic and consistent across all > > architectures; I'm not sure if all floating point implementations (and log > > implementations) will do that? > > Have a look at this (using the code I pointed to before): > https://gist.github.com/joelandman/ec6f3abef9bc5f1c7b0e > > Running on my desktop box, library double function (with casts) is about 3x > slower than the local log_2 version.
It looks like the log_2 is only giving 5 or 6 bits of precision (counting 0's), but we want a fixed-point output with more like 16 bits. The link Milosz sent looks like the most promising candidate I've seen so far... sage -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
