[Bug 218514] [LIBM] implementations of sinpi[fl], cospi[fl], and tanpi[fl]

From: <bugzilla-noreply_at_freebsd.org>
Date: Tue, 26 Oct 2021 19:27:47 +0000
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=218514

--- Comment #24 from Steve Kargl <kargl_at_FreeBSD.org> ---
For those curious, with older testing programs I'm see
Max ULP as shown in the tables below.

sinpif time: 0.0077 usec per call (average [0x1p-14,0.25] exhaustive)
cospif time: 0.0063 usec per call
tanpif time: 0.0078 usec per call

  Interval        sinpif     cospif     tanpif
[0x1p-14,0.25]   0.500074   0.500898   0.799714  (exhaustive)
[0.25,1]         0.500898   0.500893   0.799635  (exhaustive)
[100,104]        0.500893   0.500893   0.796762  (exhaustive)

sinpi time: 0.0159 usec per call (1M calls in [0x1p-29,0.25]) 
cospi time: 0.0145 usec per call
tanpi time: 0.0264 usec per call

  Interval       sinpi     cospi     tanpi
[0x1p-29,0.25]  0.735554  0.791773  0.863064  (total tested: 8388607)
[0.25,1]        0.767862  0.767862  0.868273  (total tested: 8388606)
[100,104]       0.769793  0.765948  0.855805  (total tested: 8388604)


ld80/sinpil time: 0.1046 usec per call (1M calls in [0x1p-34,0.25])
ld80/cospil time: 0.1014 usec per call
ld80/tanpil time: 0.1915 usec per call

  Interval       sinpil    cospil    tanpil
[0x1p-34,0.25]  0.710709  0.773680  0.870418 (total tested: 8388607)
[0.25,1]        0.769538  0.765667  0.863644 (total tested: 8388606)
[100,104]       0.765416  0.763284  0.861817 (total tested: 8388604)

-- 
You are receiving this mail because:
You are the assignee for the bug.
Received on Tue Oct 26 2021 - 19:27:47 UTC

Original text of this message