optimize floatscan: avoid excessive upscaling
authorRich Felker <dalias@aerifal.cx>
Wed, 11 Apr 2012 18:11:47 +0000 (14:11 -0400)
committerRich Felker <dalias@aerifal.cx>
Wed, 11 Apr 2012 18:11:47 +0000 (14:11 -0400)
commit1bdd5c8b9868ebc092074e078604acb80546e43f
treea9d2120c543db826cc9c30c902b62bd23267c10a
parent7ef1a9bba56aa756d8166c4c93cf4a178d6c0c0c
optimize floatscan: avoid excessive upscaling

upscaling by even one step too much creates 3-29 extra iterations for
the next loop. this is still suboptimal since it always goes by 2^29
rather than using a smaller upscale factor when nearing the target,
but performance on common, small-magnitude, few-digit values has
already more than doubled with this change.

more optimizations on the way...
src/internal/floatscan.c