Skip to content

FIX: Keep the intermediate values in float64 #414

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 2 commits into
base: master
Choose a base branch
from
Open

FIX: Keep the intermediate values in float64 #414

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5ffa5bf
FIX: Keep the intermediate values in float64
jkadbear Mar 19, 2022
3048df4
Merge branch 'master' into float64-intermediate
rdbisme Nov 11, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 9 additions & 3 deletions bottleneck/src/move_template.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -74,7 +74,9 @@ mover(char *name,
/* dtype = [['float64'], ['float32']] */
MOVE(move_sum, DTYPE0) {
Py_ssize_t count;
npy_DTYPE0 asum, ai, aold;
// Keep the intermediate values in float64 to reduce precision error.
// See issue pydata/bottleneck#164.
npy_float64 asum, ai, aold;
INIT(NPY_DTYPE0)
BN_BEGIN_ALLOW_THREADS
WHILE {
Expand Down Expand Up @@ -156,7 +158,9 @@ MOVE_MAIN(move_sum, 0)
/* dtype = [['float64'], ['float32']] */
MOVE(move_mean, DTYPE0) {
Py_ssize_t count;
npy_DTYPE0 asum, ai, aold, count_inv;
// Keep the intermediate values in float64 to reduce precision error.
// See issue pydata/bottleneck#164.
npy_float64 asum, ai, aold, count_inv;
INIT(NPY_DTYPE0)
BN_BEGIN_ALLOW_THREADS
WHILE {
Expand Down Expand Up @@ -244,7 +248,9 @@ MOVE_MAIN(move_mean, 0)
/* dtype = [['float64'], ['float32']] */
MOVE(NAME, DTYPE0) {
Py_ssize_t count;
npy_DTYPE0 delta, amean, assqdm, ai, aold, yi, count_inv, ddof_inv;
// Keep the intermediate values in float64 to reduce precision error.
// See issue pydata/bottleneck#164.
npy_float64 delta, amean, assqdm, ai, aold, yi, count_inv, ddof_inv;
INIT(NPY_DTYPE0)
BN_BEGIN_ALLOW_THREADS
WHILE {
Expand Down