Revision control

Copy as Markdown

Other Tools

#![cfg(feature = "compact")]
mod bellerophon;
use bellerophon::{bellerophon_test, compute_float32, compute_float64};
use minimal_lexical::num::Float;
#[test]
fn halfway_round_down_test() {
// Halfway, round-down tests
bellerophon_test::<f64>(9007199254740992, 0, false, 0, 1076);
bellerophon_test::<f64>(
9007199254740993,
0,
false,
9223372036854776832,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9007199254740994, 0, false, 1, 1076);
bellerophon_test::<f64>(18014398509481984, 0, false, 0, 1077);
bellerophon_test::<f64>(
18014398509481986,
0,
false,
9223372036854776832,
1066 + f64::INVALID_FP,
);
bellerophon_test::<f64>(18014398509481988, 0, false, 1, 1077);
bellerophon_test::<f64>(9223372036854775808, 0, false, 0, 1086);
bellerophon_test::<f64>(
9223372036854776832,
0,
false,
9223372036854776832,
1075 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9223372036854777856, 0, false, 1, 1086);
// Add a 0 but say we're truncated.
bellerophon_test::<f64>(9007199254740992000, -3, true, 0, 1076);
bellerophon_test::<f64>(
9007199254740993000,
-3,
true,
9223372036854776832,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9007199254740994000, -3, true, 1, 1076);
}
#[test]
fn halfway_round_up_test() {
// Halfway, round-up tests
bellerophon_test::<f64>(9007199254740994, 0, false, 1, 1076);
bellerophon_test::<f64>(
9007199254740995,
0,
false,
9223372036854778880,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9007199254740996, 0, false, 2, 1076);
bellerophon_test::<f64>(18014398509481988, 0, false, 1, 1077);
bellerophon_test::<f64>(
18014398509481990,
0,
false,
9223372036854778880,
1066 + f64::INVALID_FP,
);
bellerophon_test::<f64>(18014398509481992, 0, false, 2, 1077);
bellerophon_test::<f64>(9223372036854777856, 0, false, 1, 1086);
bellerophon_test::<f64>(
9223372036854778880,
0,
false,
9223372036854778880,
1075 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9223372036854779904, 0, false, 2, 1086);
// Add a 0 but say we're truncated.
bellerophon_test::<f64>(9007199254740994000, -3, true, 1, 1076);
bellerophon_test::<f64>(
9007199254740994990,
-3,
true,
9223372036854778869,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(
9007199254740995000,
-3,
true,
9223372036854778879,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(
9007199254740995010,
-3,
true,
9223372036854778890,
1065 + f64::INVALID_FP,
);
bellerophon_test::<f64>(9007199254740995050, -3, true, 2, 1076);
bellerophon_test::<f64>(9007199254740996000, -3, true, 2, 1076);
}
#[test]
fn extremes_test() {
// Need to check we get proper results with rounding for near-infinity
// and near-zero and/or denormal floats.
bellerophon_test::<f64>(5, -324, false, 1, 0);
bellerophon_test::<f64>(10, -324, false, 2, 0);
// This is very close to 2.4703282292062327206e-342.
bellerophon_test::<f64>(
2470328229206232720,
-342,
false,
18446744073709551608,
-64 + f64::INVALID_FP,
);
bellerophon_test::<f64>(2470328229206232721, -342, false, 9223372036854775808, -32831);
bellerophon_test::<f64>(2470328229206232725, -342, false, 9223372036854775824, -32831);
bellerophon_test::<f64>(2470328229206232726, -342, false, 1, 0);
bellerophon_test::<f64>(2470328229206232730, -342, false, 1, 0);
// Check very close to literal infinity.
// 17.976931348623155
// 1.797693134862315508561243283845062402343434371574593359244049e+308
// 1.797693134862315708145274237317043567980705675258449965989175e+308
bellerophon_test::<f64>(17976931348623155, 292, false, 4503599627370494, 2046);
bellerophon_test::<f64>(17976931348623156, 292, false, 4503599627370494, 2046);
bellerophon_test::<f64>(1797693134862315605, 290, false, 4503599627370494, 2046);
bellerophon_test::<f64>(1797693134862315607, 290, false, 4503599627370494, 2046);
bellerophon_test::<f64>(1797693134862315608, 290, false, 18446744073709548540, -30733);
bellerophon_test::<f64>(1797693134862315609, 290, false, 18446744073709548550, -30733);
bellerophon_test::<f64>(179769313486231561, 291, false, 4503599627370495, 2046);
bellerophon_test::<f64>(17976931348623157, 292, false, 4503599627370495, 2046);
// Check existing issues and underflow.
bellerophon_test::<f64>(2470328229206232726, -343, false, 0, 0);
bellerophon_test::<f64>(2470328229206232726, -342, false, 1, 0);
bellerophon_test::<f64>(1, -250, false, 1945308223406668, 192);
bellerophon_test::<f64>(1, -150, false, 2867420733609077, 524);
bellerophon_test::<f64>(1, -45, false, 1924152549665465, 873);
bellerophon_test::<f64>(1, -40, false, 400386103400348, 890);
bellerophon_test::<f64>(1, -20, false, 2142540351554083, 956);
bellerophon_test::<f64>(1, 0, false, 0, 1023);
bellerophon_test::<f64>(1, 20, false, 1599915997629504, 1089);
bellerophon_test::<f64>(1, 40, false, 3768206498159781, 1155);
bellerophon_test::<f64>(1, 150, false, 999684479948463, 1521);
bellerophon_test::<f64>(1, 250, false, 1786584717939204, 1853);
// Minimum positive normal float.
bellerophon_test::<f64>(22250738585072014, -324, false, 0, 1);
// Maximum positive subnormal float.
bellerophon_test::<f64>(2225073858507201, -323, false, 4503599627370495, 0);
// Next highest subnormal float.
bellerophon_test::<f64>(22250738585072004, -324, false, 4503599627370494, 0);
bellerophon_test::<f64>(22250738585072006, -324, false, 4503599627370494, 0);
bellerophon_test::<f64>(22250738585072007, -324, false, 4503599627370495, 0);
bellerophon_test::<f64>(222507385850720062, -325, false, 4503599627370494, 0);
bellerophon_test::<f64>(222507385850720063, -325, false, 4503599627370494, 0);
bellerophon_test::<f64>(222507385850720064, -325, false, 4503599627370494, 0);
bellerophon_test::<f64>(2225073858507200641, -326, false, 18446744073709545462, -32779);
bellerophon_test::<f64>(2225073858507200642, -326, false, 18446744073709545472, -32779);
bellerophon_test::<f64>(222507385850720065, -325, false, 4503599627370495, 0);
}
#[test]
fn compute_float_f32_test() {
// These test near-halfway cases for single-precision floats.
assert_eq!(compute_float32(0, 16777216), (151, 0));
assert_eq!(compute_float32(0, 16777217), (111 + f32::INVALID_FP, 9223372586610589696));
assert_eq!(compute_float32(0, 16777218), (151, 1));
assert_eq!(compute_float32(0, 16777219), (111 + f32::INVALID_FP, 9223373686122217472));
assert_eq!(compute_float32(0, 16777220), (151, 2));
// These are examples of the above tests, with
// digits from the exponent shifted to the mantissa.
assert_eq!(compute_float32(-10, 167772160000000000), (151, 0));
assert_eq!(
compute_float32(-10, 167772170000000000),
(111 + f32::INVALID_FP, 9223372586610589696)
);
assert_eq!(compute_float32(-10, 167772180000000000), (151, 1));
// Let's check the lines to see if anything is different in table...
assert_eq!(
compute_float32(-10, 167772190000000000),
(111 + f32::INVALID_FP, 9223373686122217472)
);
assert_eq!(compute_float32(-10, 167772200000000000), (151, 2));
}
#[test]
fn compute_float_f64_test() {
// These test near-halfway cases for double-precision floats.
assert_eq!(compute_float64(0, 9007199254740992), (1076, 0));
assert_eq!(compute_float64(0, 9007199254740993), (1065 + f64::INVALID_FP, 9223372036854776832));
assert_eq!(compute_float64(0, 9007199254740994), (1076, 1));
assert_eq!(compute_float64(0, 9007199254740995), (1065 + f64::INVALID_FP, 9223372036854778880));
assert_eq!(compute_float64(0, 9007199254740996), (1076, 2));
assert_eq!(compute_float64(0, 18014398509481984), (1077, 0));
assert_eq!(
compute_float64(0, 18014398509481986),
(1066 + f64::INVALID_FP, 9223372036854776832)
);
assert_eq!(compute_float64(0, 18014398509481988), (1077, 1));
assert_eq!(
compute_float64(0, 18014398509481990),
(1066 + f64::INVALID_FP, 9223372036854778880)
);
assert_eq!(compute_float64(0, 18014398509481992), (1077, 2));
// These are examples of the above tests, with
// digits from the exponent shifted to the mantissa.
assert_eq!(compute_float64(-3, 9007199254740992000), (1076, 0));
assert_eq!(
compute_float64(-3, 9007199254740993000),
(1065 + f64::INVALID_FP, 9223372036854776832)
);
assert_eq!(compute_float64(-3, 9007199254740994000), (1076, 1));
assert_eq!(
compute_float64(-3, 9007199254740995000),
(1065 + f64::INVALID_FP, 9223372036854778879)
);
assert_eq!(compute_float64(-3, 9007199254740996000), (1076, 2));
}