Skip to content

[SYCL] Fix corner case in imf function 'fdiv_rn' #20427

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
Oct 23, 2025
Merged

[SYCL] Fix corner case in imf function 'fdiv_rn' #20427

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
69aa484
[SYCL] Fix corner case in imf function 'fdiv_rn'
jinge90 Oct 22, 2025
bfc7930
Merge remote-tracking branch 'upstream/sycl' into imf_fdiv_div_by_zer…
jinge90 Oct 23, 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
16 changes: 14 additions & 2 deletions libdevice/imf_rounding_op.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -670,12 +670,24 @@ template <typename Ty> Ty __fp_div(Ty x, Ty y, int rd) {
const UTy sig_off_mask = (one_bits << (sizeof(UTy) * 8 - 1)) - 1;

if (((x_exp == __iml_fp_config<Ty>::exp_mask) && (x_fra != 0x0)) ||
((y_exp == __iml_fp_config<Ty>::exp_mask) && (y_fra != 0x0)) ||
((y_bit & sig_off_mask) == 0x0)) {
((y_exp == __iml_fp_config<Ty>::exp_mask) && (y_fra != 0x0))) {
UTy tmp = __iml_fp_config<Ty>::nan_bits;
return __builtin_bit_cast(Ty, tmp);
}

// 0.f / 0.f ----> NAN
if ((y_bit & sig_off_mask) == 0x0) {
if ((x_bit & sig_off_mask) == 0x0) {
UTy tmp = __iml_fp_config<Ty>::nan_bits;
return __builtin_bit_cast(Ty, tmp);
} else {
// return +inf if x_sig and y_sig are same otherwise return -inf
UTy tmp = (z_sig == 0) ? __iml_fp_config<Ty>::pos_inf_bits
: __iml_fp_config<Ty>::neg_inf_bits;
return __builtin_bit_cast(Ty, tmp);
}
}

if ((x_exp == __iml_fp_config<Ty>::exp_mask) && (x_fra == 0x0)) {
if ((y_exp == __iml_fp_config<Ty>::exp_mask) && (y_fra == 0x0)) {
UTy tmp = __iml_fp_config<Ty>::nan_bits;
Expand Down
Binary file added sycl/test-e2e/DeviceLib/imf/a.out
View file
Open in desktop
Binary file not shown.
30 changes: 18 additions & 12 deletions sycl/test-e2e/DeviceLib/imf/fp32_rounding_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,18 +92,24 @@ int main(int, char **) {
}

{
std::initializer_list<float> input_vals1 = {0x1p-1, 0x1.8bd054p+6,
0x1.fcd686p+0, -0x1.7f9abp+3};
std::initializer_list<float> input_vals2 = {-0x1.a8p+2, -0x1.674a3cp+5,
0x1.f3d6aep+10, 0x1.d6bf48p+10};
std::initializer_list<uint32_t> ref_vals_rd = {0xbd9a90e8, 0xc00d030d,
0x3a824df9, 0xbbd09c3a};
std::initializer_list<uint32_t> ref_vals_rn = {0xbd9a90e8, 0xc00d030c,
0x3a824df9, 0xbbd09c39};
std::initializer_list<uint32_t> ref_vals_ru = {0xbd9a90e7, 0xc00d030c,
0x3a824dfa, 0xbbd09c39};
std::initializer_list<uint32_t> ref_vals_rz = {0xbd9a90e7, 0xc00d030c,
0x3a824df9, 0xbbd09c39};
std::initializer_list<float> input_vals1 = {
0x1p-1, 0x1.8bd054p+6, 0x1.fcd686p+0, -0x1.7f9abp+3,
0x1p+0, -0x1p+0, 0x0p+0};
std::initializer_list<float> input_vals2 = {
-0x1.a8p+2, -0x1.674a3cp+5, 0x1.f3d6aep+10, 0x1.d6bf48p+10,
0x0p+0, 0x0p+0, 0x0p+0};
std::initializer_list<uint32_t> ref_vals_rd = {
0xbd9a90e8, 0xc00d030d, 0x3a824df9, 0xbbd09c3a,
0x7F800000, 0xFF800000, 0x7FC00000};
std::initializer_list<uint32_t> ref_vals_rn = {
0xbd9a90e8, 0xc00d030c, 0x3a824df9, 0xbbd09c39,
0x7F800000, 0xFF800000, 0x7FC00000};
std::initializer_list<uint32_t> ref_vals_ru = {
0xbd9a90e7, 0xc00d030c, 0x3a824dfa, 0xbbd09c39,
0x7F800000, 0xFF800000, 0x7FC00000};
std::initializer_list<uint32_t> ref_vals_rz = {
0xbd9a90e7, 0xc00d030c, 0x3a824df9, 0xbbd09c39,
0x7F800000, 0xFF800000, 0x7FC00000};
test2(device_queue, input_vals1, input_vals2, ref_vals_rd,
F2T(uint32_t, sycl::ext::intel::math::fdiv_rd));
std::cout << "sycl::ext::intel::math::fdiv_rd passes." << std::endl;
Expand Down
32 changes: 22 additions & 10 deletions sycl/test-e2e/DeviceLib/imf/fp64_rounding_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,24 +115,36 @@ int main(int, char **) {
}

{
std::initializer_list<double> input_vals1 = {
0x1.5ef3da7bf609ap+4, 0x1.fbd37afb0f8edp-1, 0x1.9238e38e38e35p+6,
0x1.7p+3};
std::initializer_list<double> input_vals2 = {
-0x1.bc7db6de6d33fp+9, 0x1.2f638fa4e71a6p+10, 0x1.08e38e38e38e3p+4,
-0x1.94p+3};
std::initializer_list<double> input_vals1 = {0x1.5ef3da7bf609ap+4,
0x1.fbd37afb0f8edp-1,
0x1.9238e38e38e35p+6,
0x1.7p+3,
0x1p+0,
-0x1p+0,
0x0p+0};
std::initializer_list<double> input_vals2 = {-0x1.bc7db6de6d33fp+9,
0x1.2f638fa4e71a6p+10,
0x1.08e38e38e38e3p+4,
-0x1.94p+3,
0x0p+0,
0x0p+0,
0x0p+0};
std::initializer_list<uint64_t> ref_vals_rd = {
0xbf994414312c26ab, 0x3f4ac811fc63acd9, 0x40184b98e9aa180a,
0xbfed260511be1959};
0xbfed260511be1959, 0x7FF0000000000000, 0xFFF0000000000000,
0x7FF8000000000000};
std::initializer_list<uint64_t> ref_vals_rn = {
0xbf994414312c26ab, 0x3f4ac811fc63acd9, 0x40184b98e9aa180b,
0xbfed260511be1959};
0xbfed260511be1959, 0x7FF0000000000000, 0xFFF0000000000000,
0x7FF8000000000000};
std::initializer_list<uint64_t> ref_vals_ru = {
0xbf994414312c26aa, 0x3f4ac811fc63acda, 0x40184b98e9aa180b,
0xbfed260511be1958};
0xbfed260511be1958, 0x7FF0000000000000, 0xFFF0000000000000,
0x7FF8000000000000};
std::initializer_list<uint64_t> ref_vals_rz = {
0xbf994414312c26aa, 0x3f4ac811fc63acd9, 0x40184b98e9aa180a,
0xbfed260511be1958};
0xbfed260511be1958, 0x7FF0000000000000, 0xFFF0000000000000,
0x7FF8000000000000};
test2(device_queue, input_vals1, input_vals2, ref_vals_rd,
F2T(uint64_t, sycl::ext::intel::math::ddiv_rd));
std::cout << "sycl::ext::intel::math::ddiv_rd passes." << std::endl;
Expand Down
27 changes: 27 additions & 0 deletions sycl/test-e2e/DeviceLib/imf/imf_utils.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,6 +14,11 @@ typedef _Float16 _iml_half_internal;
typedef uint16_t _iml_half_internal;
#endif

#define EXP_MASK32 0xFF
#define EXP_MASK64 0x7FF
#define FRA_MASK32 0x7FFFFF
#define FRA_MASK64 0xFFFFFFFFFFFFF

template <class Ty> class imf_utils_default_equ {
public:
bool operator()(Ty x, Ty y) {
Expand All @@ -24,6 +29,28 @@ template <class Ty> class imf_utils_default_equ {
};
};

template <> class imf_utils_default_equ<uint32_t> {
public:
bool operator()(uint32_t x, uint32_t y) {
bool x_is_nan =
(((x >> 23) & EXP_MASK32) == EXP_MASK32) && ((x & FRA_MASK32) != 0);
bool y_is_nan =
(((y >> 23) & EXP_MASK32) == EXP_MASK32) && ((y & FRA_MASK32) != 0);
return (x_is_nan && y_is_nan) || (x == y);
}
};

template <> class imf_utils_default_equ<uint64_t> {
public:
bool operator()(uint64_t x, uint64_t y) {
bool x_is_nan =
(((x >> 52) & EXP_MASK64) == EXP_MASK64) && ((x & FRA_MASK64) != 0);
bool y_is_nan =
(((y >> 52) & EXP_MASK64) == EXP_MASK64) && ((y & FRA_MASK64) != 0);
return (x_is_nan && y_is_nan) || (x == y);
}
};

// Used to test half precision utils
template <class InputTy, class OutputTy, class FuncTy,
class EquTy = imf_utils_default_equ<OutputTy>>
Expand Down
Loading