projects / platform / upstream / glibc.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 342a78c) | patch
tile: use ieee754/dbl-64 version of fma() and fmaf()
authorChris Metcalf <cmetcalf@tilera.com>
Thu, 17 May 2012 15:27:36 +0000 (11:27 -0400)
committerChris Metcalf <cmetcalf@tilera.com>
Thu, 17 May 2012 15:27:36 +0000 (11:27 -0400)
commitac9182e7532b7a903afa1b86545dd9f1cc308a74
treebadb3449727883358fff5edc1a8709328c43d40ftree | snapshot
parent342a78ccd537e6abd917eacac21ae2064ef39df9commit | diff
tile: use ieee754/dbl-64 version of fma() and fmaf()

It turns out that even if you stub out the rounding and exception
support and use the ieee754 version, it's still much better than
the generic version that just uses normal multiply and add.
The resulting functions have only 1 ULP of error according to the tests.
ChangeLog.tile diff | blob | history
sysdeps/tile/s_fma.c diff | blob | history
sysdeps/tile/s_fmaf.c diff | blob | history
Domain: System / Toolchain;