AriasCreations/LSL-PyOptimizer
3
0
Fork 0
mirror of https://github.com/Sei-Lisa/LSL-PyOptimizer synced 2025-07-01 23:58:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Improve the (in)accuracy of llSin, llCos, llTan

Browse source 
All three functions have in SL the Intel problem of inaccuracy with high values. Since we work in single precision, it's barely detectable usually, but for large input numbers the difference between the imprecise results of SL and the more accurate results of recent glibc become more obvious.

This change brings back the Intel inaccuracy, even across systems (different versions of Python/C might behave differently otherwise).

Reference:
https://randomascii.wordpress.com/2014/10/09/intel-underestimates-error-bounds-by-1-3-quintillion/
This commit is contained in:
Sei Lisa 2016-05-21 03:56:27 +02:00
parent e40c52075b
commit 9cf9478270
1 changed files with 9 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

12
lslopt/lslbasefuncs.py
View file

@ -780,6 +780,12 @@ def iskey(x):
def islist(x):
return type(x) == list
def reduce(t):
t = F32(t)
if not t.is_integer():
return t # Accurate-ish until big numbers come into play
return int(t * 18446744073709551616) % 115904311329233965478 / 18446744073709551616.
#
# LSL-compatible computation functions
#
@ -928,7 +934,7 @@ def llCeil(f):
def llCos(f):
assert isfloat(f)
f = ff(f)
f = reduce(ff(f))
if math.isinf(f):
return Indet
if -9223372036854775808.0 < f < 9223372036854775808.0:
@ -1556,7 +1562,7 @@ def llSHA1String(s):
def llSin(f):
assert isfloat(f)
f = ff(f)
f = reduce(ff(f))
if math.isinf(f):
return Indet
if -9223372036854775808.0 < f < 9223372036854775808.0:
@ -1600,7 +1606,7 @@ def llSubStringIndex(s, pattern):
def llTan(f):
assert isfloat(f)
f = ff(f)
f = reduce(ff(f))
if math.isinf(f):
return Indet
if -9223372036854775808.0 < f < 9223372036854775808.0: