improve `cosc(::Float32)` and `cosc(::Float64)` accuracy #59087
nsajko
marked this pull request as ready for review 149 days ago
nsajko
force pushed
from
4f6ee4e7
to
b4b9f269
138 days ago
nsajko
force pushed
from
7d11d2c2
to
bfb32aac
137 days ago
nsajko
force pushed
from
a720e29b
to
1f08e7a6
134 days ago
nsajko
force pushed
from
5995cf90
to
8f5a476d
132 days ago
a bit of code deduplication
11134a8c
eliminate common subexpression
2e17039a
better accuracy for `cosc` for `Float32`, `Float64` around `x = 0`
0011a4f7
test accuracy
a3c4a735
add copyright notice
0ab1f6b0
ULPError simplification as per Oscar's suggestion
d991f1ef
ULPError common subexpression elimination
4459c67a
ULPError: handle `isinf(approximate)`
2592ee90
put comment in correct position
48b80928
style: get rid of the `let`s/nesting
7140f605
style: simplify `return`s
11c29f3d
fix `ULPError` edge case regarding zero/inf
fb2ac2f3
test `ULPError` edge cases
c65115a2
`ULPError`: simplify
33717f57
`ULPError`: delete dead code
f6770626
`ULPError`: eliminate variables only used once
c0340a84
`ULPError`: use `Float32` literals
85ace6ce
`ULPError`: guard the edge case handling behind a single `if`
0b05e22a
`ULPError`: use `Float64` instead of `BigFloat`
71126d58
`ULPError`: avoid unnecessary conversion to `BigFloat`
cc22fc97
delete debugging throw
ca57e60d
test: add missing edge case
ce8d25d6
test: `ULPError`: faitfhul approximation
8a60bf08
Revert "`ULPError`: avoid unnecessary conversion to `BigFloat`"
a95af773
handle precision better
92a93784
expand `ULPError` test suite
cee5a83a
`ulp_error_maximum`: use `Fix1`
e8319acb
use `ulp_error_maximum` instead of a loop
1ca04f36
`ULPError`: delete unnecessary method static parameters
4f86cf31
improve style
e9331157
get rid of the `struct`
57fea2b8
nsajko
force pushed
from
dc3d649f
to
e9331157
127 days ago
fix
b7adfd36
merge two methods into one
c1e72b0e
nsajko
deleted the reimplement_cosc_for_Float32_Float64 branch 113 days ago
Login to write a write a comment.
Login via GitHub