pkg std = const flt64bits : (flt : flt64 -> uint64) const flt32bits : (flt : flt32 -> uint32) const flt64inf : (-> flt64) const flt64nan : (-> flt64) const flt32inf : (-> flt32) const flt32nan : (-> flt32) generic isnan : (f : @a -> bool) ::floating @a const flt64frombits : (bits : uint64 -> flt64) const flt32frombits : (bits : uint32 -> flt32) const flt64explode : (flt : flt64 -> (bool, int64, uint64)) const flt32explode : (flt : flt32 -> (bool, int32, uint32)) const flt64assem : (sign : bool, exp : int64, mant : uint64 -> flt64) const flt32assem : (sign : bool, exp : int32, mant : uint32 -> flt32);; const flt64bits = {flt; -> (&flt : uint64#)#}const flt32bits = {flt; -> (&flt : uint32#)#}const flt64frombits = {bits; -> (&bits : flt64#)#}const flt32frombits = {bits; -> (&bits : flt32#)#} const Dblbias = 1023const Fltbias = 127 const flt64explode = {flt var bits, isneg, mant, uexp, exp bits = flt64bits(flt) isneg = (bits >> 63) != 0 /* msb is sign bit */ uexp = (bits >> 52) & 0x7ff /* exp is in bits [52..63] */ mant = bits & ((1ul << 52) - 1) /* mant is in bits [..51] */ /* add back the implicit bit if this is not a denormal */ if uexp != 0 mant |= 1ul << 52 ;; /* adjust for exponent bias */ exp = (uexp : int64) - Dblbias -> (isneg, exp, mant)} const flt32explode = {flt var bits, isneg, mant, uexp, exp bits = flt32bits(flt) isneg = (bits >> 31) != 0 /* msb is sign bit */ uexp = (bits >> 23) & 0xff /* exp is in bits [23..30] */ mant = bits & ((1 << 23) - 1) /* mant is in bits [0..22] */ /* add back the implicit bit if this is not a denormal */ if uexp != 0 mant |= 1 << 23 ;; /* adjust for exponent bias */ exp = (uexp : int32) - Fltbias -> (isneg, exp, mant)} const flt64assem = {sign, exp, mant var s, m, e if exp <= -Dblbias && (mant & (1ul << 52) != 0) var roundup = false var shift : uint64 = ((1 - Dblbias - exp) : uint64) var firstcut = mant & (1 << shift) var restcut = mant & ((1 << shift) - 1) var lastkept = mant & (1 << (shift + 1)) roundup = firstcut != 0 && (lastkept != 0 || restcut != 0) mant >>= shift exp = -Dblbias if roundup mant++ if (mant & (1ul << 52) != 0) exp++ ;; ;; ;; s = (sign : uint64) e = (exp + Dblbias : uint64) & 0x7ff m = (mant : uint64) & ((1ul<<52) - 1) -> std.flt64frombits((s << 63) | (e << 52) | m)} const flt32assem = {sign, exp, mant var s, m, e if exp <= -Fltbias && (mant & (1 << 23) != 0) var roundup = false var shift : uint32 = ((1 - Fltbias - exp) : uint32) var firstcut = mant & (1 << shift) var restcut = mant & ((1 << shift) - 1) var lastkept = mant & (1 << (shift + 1)) roundup = firstcut != 0 && (lastkept != 0 || restcut != 0) mant >>= shift exp = -Fltbias if roundup mant++ if (mant & (1 << 23) != 0) exp++ ;; ;; ;; s = (sign : uint32) e = (exp + Fltbias : uint32) & 0xff m = (mant : uint32) & ((1<<23) - 1) -> std.flt32frombits(s << 31 | e << 23 | m) } generic isnan = {f var b b = flt64bits((f : flt64)) -> (b >> 52) & 0x7fful == 0x7fful && \ b & ~(0xffful << 52) != 0} const flt64inf = { -> std.flt64frombits(0x7ff0000000000000ul)} const flt64nan = { -> std.flt64frombits(0x7ff8000000000000ul)} const flt32inf = { -> std.flt32frombits(0x7f800000)} const flt32nan = { -> std.flt32frombits(0x7fc00000)}
pkg std =
const flt64bits : (flt : flt64 -> uint64)
const flt32bits : (flt : flt32 -> uint32)
const flt64inf : (-> flt64)
const flt64nan : (-> flt64)
const flt32inf : (-> flt32)
const flt32nan : (-> flt32)
generic isnan : (f : @a -> bool) ::floating @a
const flt64frombits : (bits : uint64 -> flt64)
const flt32frombits : (bits : uint32 -> flt32)
const flt64explode : (flt : flt64 -> (bool, int64, uint64))
const flt32explode : (flt : flt32 -> (bool, int32, uint32))
const flt64assem : (sign : bool, exp : int64, mant : uint64 -> flt64)
const flt32assem : (sign : bool, exp : int32, mant : uint32 -> flt32)
;;
const flt64bits = {flt; -> (&flt : uint64#)#}
const flt32bits = {flt; -> (&flt : uint32#)#}
const flt64frombits = {bits; -> (&bits : flt64#)#}
const flt32frombits = {bits; -> (&bits : flt32#)#}
const Dblbias = 1023
const Fltbias = 127
const flt64explode = {flt
var bits, isneg, mant, uexp, exp
bits = flt64bits(flt)
isneg = (bits >> 63) != 0 /* msb is sign bit */
uexp = (bits >> 52) & 0x7ff /* exp is in bits [52..63] */
mant = bits & ((1ul << 52) - 1) /* mant is in bits [..51] */
/* add back the implicit bit if this is not a denormal */
if uexp != 0
mant |= 1ul << 52
;;
/* adjust for exponent bias */
exp = (uexp : int64) - Dblbias
-> (isneg, exp, mant)
}
const flt32explode = {flt
var bits, isneg, mant, uexp, exp
bits = flt32bits(flt)
isneg = (bits >> 31) != 0 /* msb is sign bit */
uexp = (bits >> 23) & 0xff /* exp is in bits [23..30] */
mant = bits & ((1 << 23) - 1) /* mant is in bits [0..22] */
/* add back the implicit bit if this is not a denormal */
if uexp != 0
mant |= 1 << 23
;;
/* adjust for exponent bias */
exp = (uexp : int32) - Fltbias
-> (isneg, exp, mant)
}
const flt64assem = {sign, exp, mant
var s, m, e
if exp <= -Dblbias && (mant & (1ul << 52) != 0)
var roundup = false
var shift : uint64 = ((1 - Dblbias - exp) : uint64)
var firstcut = mant & (1 << shift)
var restcut = mant & ((1 << shift) - 1)
var lastkept = mant & (1 << (shift + 1))
roundup = firstcut != 0 && (lastkept != 0 || restcut != 0)
mant >>= shift
exp = -Dblbias
if roundup
mant++
if (mant & (1ul << 52) != 0)
exp++
;;
;;
;;
s = (sign : uint64)
e = (exp + Dblbias : uint64) & 0x7ff
m = (mant : uint64) & ((1ul<<52) - 1)
-> std.flt64frombits((s << 63) | (e << 52) | m)
}
const flt32assem = {sign, exp, mant
var s, m, e
if exp <= -Fltbias && (mant & (1 << 23) != 0)
var roundup = false
var shift : uint32 = ((1 - Fltbias - exp) : uint32)
var firstcut = mant & (1 << shift)
var restcut = mant & ((1 << shift) - 1)
var lastkept = mant & (1 << (shift + 1))
roundup = firstcut != 0 && (lastkept != 0 || restcut != 0)
mant >>= shift
exp = -Fltbias
if roundup
mant++
if (mant & (1 << 23) != 0)
exp++
;;
;;
;;
s = (sign : uint32)
e = (exp + Fltbias : uint32) & 0xff
m = (mant : uint32) & ((1<<23) - 1)
-> std.flt32frombits(s << 31 | e << 23 | m)
}
generic isnan = {f
var b
b = flt64bits((f : flt64))
-> (b >> 52) & 0x7fful == 0x7fful && \
b & ~(0xffful << 52) != 0
}
const flt64inf = {
-> std.flt64frombits(0x7ff0000000000000ul)
}
const flt64nan = {
-> std.flt64frombits(0x7ff8000000000000ul)
}
const flt32inf = {
-> std.flt32frombits(0x7f800000)
}
const flt32nan = {
-> std.flt32frombits(0x7fc00000)
}