Skip to content

Correctly rounded floating point div_euclid. #134145

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

Closed
wants to merge 5 commits into from
Closed

Correctly rounded floating point div_euclid. #134145

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
3deff84
Correctly rounded floating point `div_euclid`.
tczajka Dec 10, 2024
4534005
Fix overflow bug in f16::div_ceil
tczajka Dec 11, 2024
cedc718
Software `div_euclid` refactoring.
tczajka Dec 11, 2024
009cf4b
div_euclid benchmarks
tczajka Dec 12, 2024
d0f8bae
div_euclid: add a fast path for overflow
tczajka Dec 12, 2024
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
Prev Previous commit
div_euclid: add a fast path for overflow
  • Loading branch information
@tczajka
tczajka committed Dec 12, 2024
commit d0f8bae821930eb0a4e34b0326f36bb3ff374ef5
12 changes: 10 additions & 2 deletions library/std/src/f128/soft.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -201,7 +201,7 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// `q` fits in `Bits` because `a.mantissa < 2 * b.mantissa`.
let q: Bits = (aa / bb).wrap_u128();
q as FP
} else {
} else if exp <= FP::MAX_EXP {
// exp >= Bits::BITS
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
Expand Down Expand Up @@ -256,6 +256,10 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q | 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}

Expand All @@ -282,7 +286,7 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// <= 2^Bits::BITS - 2^(Bits::BITS - SIGNIFICAND_BITS) + 1 <= Bits::MAX
let q = ((aa - U256::ONE) / bb).wrap_u128() + 1;
q as FP
} else {
} else if exp <= FP::MAX_EXP {
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
// e > 0
Expand Down Expand Up @@ -346,5 +350,9 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q + 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}
12 changes: 10 additions & 2 deletions library/std/src/f32/soft.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,7 +198,7 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// `q` fits in `Bits` because `a.mantissa < 2 * b.mantissa`.
let q = (aa / bb) as Bits;
q as FP
} else {
} else if exp <= FP::MAX_EXP {
// exp >= Bits::BITS
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
Expand Down Expand Up @@ -252,6 +252,10 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q | 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}

Expand All @@ -277,7 +281,7 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// <= 2^Bits::BITS - 2^(Bits::BITS - SIGNIFICAND_BITS) + 1 <= Bits::MAX
let q = ((aa - 1) / bb) as Bits + 1;
q as FP
} else {
} else if exp <= FP::MAX_EXP {
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
// e > 0
Expand Down Expand Up @@ -340,5 +344,9 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q + 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}
12 changes: 10 additions & 2 deletions library/std/src/f64/soft.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -198,7 +198,7 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// `q` fits in `Bits` because `a.mantissa < 2 * b.mantissa`.
let q = (aa / bb) as Bits;
q as FP
} else {
} else if exp <= FP::MAX_EXP {
// exp >= Bits::BITS
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
Expand Down Expand Up @@ -252,6 +252,10 @@ fn div_floor_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q | 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}

Expand All @@ -277,7 +281,7 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
// <= 2^Bits::BITS - 2^(Bits::BITS - SIGNIFICAND_BITS) + 1 <= Bits::MAX
let q = ((aa - 1) / bb) as Bits + 1;
q as FP
} else {
} else if exp <= FP::MAX_EXP {
let aa = DoubleBits::from(a.mantissa) << (Bits::BITS - 1);
let bb = DoubleBits::from(b.mantissa);
// e > 0
Expand Down Expand Up @@ -340,5 +344,9 @@ fn div_ceil_pos_finite(a: PositiveFinite, b: PositiveFinite) -> FP {
q + 1
};
q_adj as FP * pow2(e)
} else {
// a.mantissa / b.mantissa >= 1/2, hence
// a / b >= 1/2 * 2^(MAX_EXP+1) = 2^MAX_EXP
FP::INFINITY
}
}
Loading