diff options
| author | Wilco Dijkstra <wdijkstr@arm.com> | 2018-02-07 12:24:43 +0000 |
|---|---|---|
| committer | Wilco Dijkstra <wdijkstr@arm.com> | 2018-02-07 12:24:43 +0000 |
| commit | b7c83ca30ef8e85b6642151d95600a36535f8d97 (patch) | |
| tree | 96bc33835429ccfe40cd3c257f183b57926ec40c /manual/probes.texi | |
| parent | 388ff7bd0d57d7061fdd39a2f26f65687e8058da (diff) | |
| download | glibc-b7c83ca30ef8e85b6642151d95600a36535f8d97.tar glibc-b7c83ca30ef8e85b6642151d95600a36535f8d97.tar.gz glibc-b7c83ca30ef8e85b6642151d95600a36535f8d97.tar.bz2 glibc-b7c83ca30ef8e85b6642151d95600a36535f8d97.zip | |
Remove slow paths from log
Remove the slow paths from log. Like several other double precision math
functions, log is exactly rounded. This is not required from math functions
and causes major overheads as it requires multiple fallbacks using higher
precision arithmetic if a result is close to 0.5ULP. Ridiculous slowdowns
of up to 100000x have been reported when the highest precision path triggers.
Interestingly removing the slow paths makes hardly any difference in practice:
the worst case error is still ~0.502ULP, and exp(log(x)) shows identical results
before/after on many millions of random cases. All GLIBC math tests pass on
AArch64 and x64 with no change in ULP error. A simple test over a few hundred
million values shows log is now 18% faster on average.
* manual/probes.texi (slowlog): Delete documentation of removed probe.
(slowlog_inexact): Likewise
* sysdeps/ieee754/dbl-64/e_log.c (__ieee754_log): Remove slow paths.
* sysdeps/ieee754/dbl-64/ulog.h: Remove unused declarations.
Diffstat (limited to 'manual/probes.texi')
| -rw-r--r-- | manual/probes.texi | 17 |
1 files changed, 0 insertions, 17 deletions
diff --git a/manual/probes.texi b/manual/probes.texi index 8ab67562d7..e99b7f3cb4 100644 --- a/manual/probes.texi +++ b/manual/probes.texi @@ -288,23 +288,6 @@ input that results in multiple precision computation with precision and @code{$arg4} is the final accurate value. @end deftp -@deftp Probe slowlog (int @var{$arg1}, double @var{$arg2}, double @var{$arg3}) -This probe is triggered when the @code{log} function is called with an -input that results in multiple precision computation. Argument -@var{$arg1} is the precision with which the computation succeeded. -Argument @var{$arg2} is the input and @var{$arg3} is the computed -output. -@end deftp - -@deftp Probe slowlog_inexact (int @var{$arg1}, double @var{$arg2}, double @var{$arg3}) -This probe is triggered when the @code{log} function is called with an -input that results in multiple precision computation and none of the -multiple precision computations result in an accurate result. -Argument @var{$arg1} is the maximum precision with which computations -were performed. Argument @var{$arg2} is the input and @var{$arg3} is -the computed output. -@end deftp - @deftp Probe slowatan2 (int @var{$arg1}, double @var{$arg2}, double @var{$arg3}, double @var{$arg4}) This probe is triggered when the @code{atan2} function is called with an input that results in multiple precision computation. Argument |