Skip to content

Commit

Permalink
Further streamlining of overflow checks
Browse files Browse the repository at this point in the history
  • Loading branch information
@ckormanyos
ckormanyos committed Dec 30, 2024
1 parent 7faf35e commit 47cbf53
Showing 1 changed file with 34 additions and 42 deletions.
76 changes: 34 additions & 42 deletions include/boost/multiprecision/cpp_double_fp.hpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -606,9 +606,7 @@ class cpp_double_fp_backend

if (cpp_df_qf_detail::ccmath::isinf(C))
{
// In this case, data.first is evidently so large that multipllication
// with the split has overflowed to infinity. Take a last-chance try
// for multiplication by scaling down (and then back up) with the split.
// Handle overflow by scaling down (and then back up) with the split.

C = data.first - cpp_df_qf_detail::ccmath::ldexp(data.first, -cpp_df_qf_detail::split_maker<float_type>::n_shl);

Expand All @@ -623,6 +621,8 @@ class cpp_double_fp_backend

if (cpp_df_qf_detail::ccmath::isinf(C))
{
// Handle overflow.

const bool b_neg { (isneg() != v.isneg()) };

*this = cpp_double_fp_backend::my_value_inf();
Expand All @@ -641,9 +641,7 @@ class cpp_double_fp_backend

if (cpp_df_qf_detail::ccmath::isinf(c))
{
// In this case, v.data.first is evidently so large that multipllication
// with the split has overflowed to infinity. Take a last-chance try
// for multiplication by scaling down (and then back up) with the split.
// Handle overflow by scaling down (and then back up) with the split.

c = v.data.first - cpp_df_qf_detail::ccmath::ldexp(v.data.first, -cpp_df_qf_detail::split_maker<float_type>::n_shl);

Expand Down Expand Up @@ -756,48 +754,42 @@ class cpp_double_fp_backend

float_type c { cpp_df_qf_detail::split_maker<float_type>::value * C };

const float_type hc { c - float_type { c - C } };

float_type u { cpp_df_qf_detail::split_maker<float_type>::value * v.data.first };

const float_type hv { u - float_type { u - v.data.first } };
float_type hc;

if (cpp_df_qf_detail::ccmath::isinf(u) || cpp_df_qf_detail::ccmath::isinf(c))
if (cpp_df_qf_detail::ccmath::isinf(c))
{
// Evidently we have some very large operands. Take a last-chance try
// for finite division. Use the ratio of square roots and subsequently
// square the ratio, handling the sign of the result separately.

// TBD: Is there a more sensible catch (and handling) for this case?

const bool u_neg { isneg() };
const bool v_neg { v.isneg() };
const bool b_neg { u_neg != v_neg };

cpp_double_fp_backend uu { *this };
cpp_double_fp_backend vv { v };
// Handle overflow by scaling down (and then back up) with the split.

cpp_double_fp_backend sqrt_u { };
cpp_double_fp_backend sqrt_v { };

if(u_neg) { uu.negate(); }
if(v_neg) { vv.negate(); }

eval_sqrt(sqrt_u, uu);
eval_sqrt(sqrt_v, vv);

cpp_double_fp_backend sqrt_ratio { sqrt_u / sqrt_v };
hc =
cpp_df_qf_detail::ccmath::ldexp
(
C - float_type { C - cpp_df_qf_detail::ccmath::ldexp(C, -cpp_df_qf_detail::split_maker<float_type>::n_shl) },
cpp_df_qf_detail::split_maker<float_type>::n_shl
);
}
else
{
hc = c - float_type { c - C };
}

*this = sqrt_ratio;
float_type u { cpp_df_qf_detail::split_maker<float_type>::value * v.data.first };

eval_multiply(*this, sqrt_ratio);
float_type hv;

if (b_neg)
{
negate();
}
if (cpp_df_qf_detail::ccmath::isinf(u))
{
// Handle overflow by scaling down (and then back up) with the split.

return *this;
hv =
cpp_df_qf_detail::ccmath::ldexp
(
v.data.first - float_type { v.data.first - cpp_df_qf_detail::ccmath::ldexp(v.data.first, -cpp_df_qf_detail::split_maker<float_type>::n_shl) },
cpp_df_qf_detail::split_maker<float_type>::n_shl
);
}
else
{
hv = u - float_type { u - v.data.first };
}

{
Expand All @@ -818,7 +810,7 @@ class cpp_double_fp_backend
u = cpp_df_qf_detail::ccmath::unsafe::fma(tc, tv, u);
}

c = float_type { c - u } + data.second;
c = float_type { c - u } + data.second;

c = (c - float_type { C * v.data.second }) / v.data.first;

Expand Down

0 comments on commit 47cbf53

Please sign in to comment.