| // NB: transitionary, de-mode-ing. |
| #[forbid(deprecated_mode)]; |
| #[forbid(deprecated_pattern)]; |
| |
| //! Operations and constants for `f32` |
| |
| pub use cmath::c_float::*; |
| pub use cmath::c_float_targ_consts::*; |
| |
| // These are not defined inside consts:: for consistency with |
| // the integer types |
| |
| pub const NaN: f32 = 0.0_f32/0.0_f32; |
| |
| pub const infinity: f32 = 1.0_f32/0.0_f32; |
| |
| pub const neg_infinity: f32 = -1.0_f32/0.0_f32; |
| |
| pub pure fn is_NaN(f: f32) -> bool { f != f } |
| |
| pub pure fn add(x: f32, y: f32) -> f32 { return x + y; } |
| |
| pub pure fn sub(x: f32, y: f32) -> f32 { return x - y; } |
| |
| pub pure fn mul(x: f32, y: f32) -> f32 { return x * y; } |
| |
| pub pure fn div(x: f32, y: f32) -> f32 { return x / y; } |
| |
| pub pure fn rem(x: f32, y: f32) -> f32 { return x % y; } |
| |
| pub pure fn lt(x: f32, y: f32) -> bool { return x < y; } |
| |
| pub pure fn le(x: f32, y: f32) -> bool { return x <= y; } |
| |
| pub pure fn eq(x: f32, y: f32) -> bool { return x == y; } |
| |
| pub pure fn ne(x: f32, y: f32) -> bool { return x != y; } |
| |
| pub pure fn ge(x: f32, y: f32) -> bool { return x >= y; } |
| |
| pub pure fn gt(x: f32, y: f32) -> bool { return x > y; } |
| |
| // FIXME (#1999): replace the predicates below with llvm intrinsics or |
| // calls to the libmath macros in the rust runtime for performance. |
| |
| /// Returns true if `x` is a positive number, including +0.0f320 and +Infinity |
| pub pure fn is_positive(x: f32) -> bool |
| { return x > 0.0f32 || (1.0f32/x) == infinity; } |
| |
| /// Returns true if `x` is a negative number, including -0.0f320 and -Infinity |
| pub pure fn is_negative(x: f32) -> bool |
| { return x < 0.0f32 || (1.0f32/x) == neg_infinity; } |
| |
| /** |
| * Returns true if `x` is a negative number, including -0.0f320 and -Infinity |
| * |
| * This is the same as `f32::is_negative`. |
| */ |
| pub pure fn is_nonpositive(x: f32) -> bool { |
| return x < 0.0f32 || (1.0f32/x) == neg_infinity; |
| } |
| |
| /** |
| * Returns true if `x` is a positive number, including +0.0f320 and +Infinity |
| * |
| * This is the same as `f32::is_positive`.) |
| */ |
| pub pure fn is_nonnegative(x: f32) -> bool { |
| return x > 0.0f32 || (1.0f32/x) == infinity; |
| } |
| |
| /// Returns true if `x` is a zero number (positive or negative zero) |
| pub pure fn is_zero(x: f32) -> bool { |
| return x == 0.0f32 || x == -0.0f32; |
| } |
| |
| /// Returns true if `x`is an infinite number |
| pub pure fn is_infinite(x: f32) -> bool { |
| return x == infinity || x == neg_infinity; |
| } |
| |
| /// Returns true if `x`is a finite number |
| pub pure fn is_finite(x: f32) -> bool { |
| return !(is_NaN(x) || is_infinite(x)); |
| } |
| |
| // FIXME (#1999): add is_normal, is_subnormal, and fpclassify. |
| |
| /* Module: consts */ |
| pub mod consts { |
| // FIXME (requires Issue #1433 to fix): replace with mathematical |
| // constants from cmath. |
| /// Archimedes' constant |
| pub const pi: f32 = 3.14159265358979323846264338327950288_f32; |
| |
| /// pi/2.0 |
| pub const frac_pi_2: f32 = 1.57079632679489661923132169163975144_f32; |
| |
| /// pi/4.0 |
| pub const frac_pi_4: f32 = 0.785398163397448309615660845819875721_f32; |
| |
| /// 1.0/pi |
| pub const frac_1_pi: f32 = 0.318309886183790671537767526745028724_f32; |
| |
| /// 2.0/pi |
| pub const frac_2_pi: f32 = 0.636619772367581343075535053490057448_f32; |
| |
| /// 2.0/sqrt(pi) |
| pub const frac_2_sqrtpi: f32 = 1.12837916709551257389615890312154517_f32; |
| |
| /// sqrt(2.0) |
| pub const sqrt2: f32 = 1.41421356237309504880168872420969808_f32; |
| |
| /// 1.0/sqrt(2.0) |
| pub const frac_1_sqrt2: f32 = 0.707106781186547524400844362104849039_f32; |
| |
| /// Euler's number |
| pub const e: f32 = 2.71828182845904523536028747135266250_f32; |
| |
| /// log2(e) |
| pub const log2_e: f32 = 1.44269504088896340735992468100189214_f32; |
| |
| /// log10(e) |
| pub const log10_e: f32 = 0.434294481903251827651128918916605082_f32; |
| |
| /// ln(2.0) |
| pub const ln_2: f32 = 0.693147180559945309417232121458176568_f32; |
| |
| /// ln(10.0) |
| pub const ln_10: f32 = 2.30258509299404568401799145468436421_f32; |
| } |
| |
| pub pure fn signbit(x: f32) -> int { |
| if is_negative(x) { return 1; } else { return 0; } |
| } |
| |
| pub pure fn logarithm(n: f32, b: f32) -> f32 { |
| return log2(n) / log2(b); |
| } |
| |
| impl f32: num::Num { |
| pure fn add(other: &f32) -> f32 { return self + *other; } |
| pure fn sub(other: &f32) -> f32 { return self - *other; } |
| pure fn mul(other: &f32) -> f32 { return self * *other; } |
| pure fn div(other: &f32) -> f32 { return self / *other; } |
| pure fn modulo(other: &f32) -> f32 { return self % *other; } |
| pure fn neg() -> f32 { return -self; } |
| |
| pure fn to_int() -> int { return self as int; } |
| static pure fn from_int(n: int) -> f32 { return n as f32; } |
| } |
| |
| // |
| // Local Variables: |
| // mode: rust |
| // fill-column: 78; |
| // indent-tabs-mode: nil |
| // c-basic-offset: 4 |
| // buffer-file-coding-system: utf-8-unix |
| // End: |
| // |