Skip to content

[libc][math] Fix signaling NaN handling for math functions. #133347

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

Merged
merged 9 commits into from
Apr 8, 2025
Merged

[libc][math] Fix signaling NaN handling for math functions. #133347

Changes from 1 commit
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
a3081fb
rebase
wldfngrs Mar 26, 2025
a8c6af9
fix issue #124812
wldfngrs Mar 27, 2025
831d103
clang-format
wldfngrs Mar 28, 2025
7c41335
:clang format again
wldfngrs Mar 28, 2025
d8c1756
fixed tests
wldfngrs Mar 28, 2025
c6dd13f
Merge branch 'main' into fix_sNaN
wldfngrs Mar 28, 2025
444c482
update powf check for signaling NaN
wldfngrs Mar 29, 2025
9d1518e
updated wrt comments
wldfngrs Apr 2, 2025
b7ed27d
updated atan2f.cpp header includes
wldfngrs Apr 5, 2025
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
20 changes: 13 additions & 7 deletions libc/src/math/generic/powf.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -657,18 +657,19 @@ LLVM_LIBC_FUNCTION(float, powf, (float x, float y)) {
uint32_t y_abs = ybits.abs().uintval();

///////// BEGIN - Check exceptional cases ////////////////////////////////////
// if x or y is signaling NaN
if (xbits.is_signaling_nan() || ybits.is_signaling_nan()) {
fputil::raise_except_if_required(FE_INVALID);
return FloatBits::quiet_nan().get_val();
}

// The single precision number that is closest to 1 is (1 - 2^-24), which has
// log2(1 - 2^-24) ~ -1.715...p-24.
// So if |y| > 151 * 2^24, and x is finite:
// |y * log2(x)| = 0 or > 151.
// Hence x^y will either overflow or underflow if x is not zero.
if (LIBC_UNLIKELY((y_abs & 0x0007'ffff) == 0) || (y_abs > 0x4f170000)) {
// y is signaling NaN
if (xbits.is_signaling_nan() || ybits.is_signaling_nan()) {
fputil::raise_except_if_required(FE_INVALID);
return FloatBits::quiet_nan().get_val();
}

// Exceptional exponents.
if (y == 0.0f)
return 1.0f;
Expand Down Expand Up @@ -741,8 +742,8 @@ LLVM_LIBC_FUNCTION(float, powf, (float x, float y)) {
}
}
if (y_abs > 0x4f17'0000) {
// y is NaN
if (y_abs > 0x7f80'0000) {
// y is NaN
if (x_u == 0x3f80'0000) { // x = 1.0f
// pow(1, NaN) = 1
return 1.0f;
Expand All @@ -764,6 +765,12 @@ LLVM_LIBC_FUNCTION(float, powf, (float x, float y)) {
// y is finite and non-zero.
if (LIBC_UNLIKELY(((x_u & 0x801f'ffffU) == 0) || x_u >= 0x7f80'0000U ||
x_u < 0x0080'0000U)) {
// if x is signaling NaN
if (xbits.is_signaling_nan()) {
fputil::raise_except_if_required(FE_INVALID);
return FloatBits::quiet_nan().get_val();
}

switch (x_u) {
case 0x3f80'0000: // x = 1.0f
return 1.0f;
Expand Down Expand Up @@ -804,7 +811,6 @@ LLVM_LIBC_FUNCTION(float, powf, (float x, float y)) {
}

if (x_abs > 0x7f80'0000) {
// x is NaN.
// pow (aNaN, 0) is already taken care above.
return x;
}
Expand Down
Loading