Skip to content

[libc][math] Do not use float16 basic operations in hypotf16. #155177

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 2 commits into from
Aug 25, 2025
Merged

[libc][math] Do not use float16 basic operations in hypotf16. #155177

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
e718582
[libc][math] Do not use float16 basic operations in hypotf16 as compi…
lntue Aug 24, 2025
7d0656e
Omit hypot float16 check when targets do not have float16.
lntue Aug 24, 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
1 change: 1 addition & 0 deletions libc/src/__support/FPUtil/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -231,6 +231,7 @@ add_header_library(
Hypot.h
DEPENDS
.basic_operations
.cast
.fenv_impl
.fp_bits
.rounding_mode
Expand Down
15 changes: 13 additions & 2 deletions libc/src/__support/FPUtil/Hypot.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,7 @@
#include "BasicOperations.h"
#include "FEnvImpl.h"
#include "FPBits.h"
#include "cast.h"
#include "rounding_mode.h"
#include "src/__support/CPP/bit.h"
#include "src/__support/CPP/type_traits.h"
Expand Down Expand Up @@ -133,8 +134,18 @@ LIBC_INLINE T hypot(T x, T y) {
uint16_t a_exp = a_bits.get_biased_exponent();
uint16_t b_exp = b_bits.get_biased_exponent();

if ((a_exp - b_exp >= FPBits_t::FRACTION_LEN + 2) || (x == 0) || (y == 0))
return x_abs.get_val() + y_abs.get_val();
if ((a_exp - b_exp >= FPBits_t::FRACTION_LEN + 2) || (x == 0) || (y == 0)) {
#ifdef LIBC_TYPES_HAS_FLOAT16
if constexpr (cpp::is_same_v<T, float16>) {
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This fails on Windows, which doesn't have float16 type, so you may need to additionally gate it with float16 availability somehow.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed.

// Compiler runtime for basic operations of float16 might not be correctly
// rounded for all rounding modes.
float af = fputil::cast<float>(x_abs.get_val());
float bf = fputil::cast<float>(y_abs.get_val());
return fputil::cast<float16>(af + bf);
} else
#endif // LIBC_TYPES_HAS_FLOAT16
return x_abs.get_val() + y_abs.get_val();
}

uint64_t out_exp = a_exp;
StorageType a_mant = a_bits.get_mantissa();
Expand Down
13 changes: 6 additions & 7 deletions libc/src/math/generic/hypotf16.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -48,16 +48,15 @@ LLVM_LIBC_FUNCTION(float16, hypotf16, (float16 x, float16 y)) {
return a_bits.get_val();
}

// TODO: Investigate why replacing the return line below with:
// return x_bits.get_val() + y_bits.get_val();
// fails the hypotf16 smoke tests.
float af = fputil::cast<float>(a_bits.get_val());
float bf = fputil::cast<float>(b_bits.get_val());

// Compiler runtime basic operations for float16 might not be correctly
// rounded for all rounding modes.
if (LIBC_UNLIKELY(a_u - b_u >=
static_cast<uint16_t>((FPBits::FRACTION_LEN + 2)
<< FPBits::FRACTION_LEN)))
return a_bits.get_val() + b_bits.get_val();

float af = fputil::cast<float>(a_bits.get_val());
float bf = fputil::cast<float>(b_bits.get_val());
return fputil::cast<float16>(af + bf);

// These squares are exact.
float a_sq = af * af;
Expand Down
1 change: 1 addition & 0 deletions utils/bazel/llvm-project-overlay/libc/BUILD.bazel
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1230,6 +1230,7 @@ libc_support_library(
":__support_cpp_bit",
":__support_cpp_type_traits",
":__support_fputil_basic_operations",
":__support_fputil_cast",
":__support_fputil_fenv_impl",
":__support_fputil_fp_bits",
":__support_fputil_rounding_mode",
Expand Down
Loading