From 628c88e335892d3472be946076d1507dc1f560b7 Mon Sep 17 00:00:00 2001 From: Bill Zorn Date: Tue, 24 Jul 2018 18:30:10 -0700 Subject: [PATCH] interface for posit8, quire8, posit16, and quire16 --- cposit.pxd | 136 +++++++++++++++++- posit.pyx | 405 ++++++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 536 insertions(+), 5 deletions(-) diff --git a/cposit.pxd b/cposit.pxd index b15cc10..913a96b 100644 --- a/cposit.pxd +++ b/cposit.pxd @@ -5,5 +5,139 @@ cdef extern from 'SoftPosit/source/include/softposit.h': ctypedef struct posit8_t: pass + ctypedef struct posit16_t: + pass + + ctypedef struct posit32_t: + pass + + ctypedef struct posit64_t: + pass + + ctypedef struct posit128_t: + pass + + ctypedef struct quire8_t: + pass + + ctypedef struct quire16_t: + pass + + ctypedef struct quire32_t: + pass + + # /*---------------------------------------------------------------------------- + # | Integer-to-posit conversion routines. + # *----------------------------------------------------------------------------*/ + + posit8_t ui32_to_p8( uint32_t ); + posit16_t ui32_to_p16( uint32_t ); + posit32_t ui32_to_p32( uint32_t ); + posit64_t ui32_to_p64( uint32_t ); + + posit8_t ui64_to_p8( uint64_t ); + posit16_t ui64_to_p16( uint64_t ); + posit32_t ui64_to_p32( uint64_t ); + posit64_t ui64_to_p64( uint64_t ); + + posit8_t i32_to_p8( int32_t ); + posit16_t i32_to_p16( int32_t ); + posit32_t i32_to_p32( int32_t ); + posit64_t i32_to_p64( int32_t ); + posit8_t i64_to_p8( int64_t ); - int_fast64_t p8_to_i64( posit8_t); + posit16_t i64_to_p16( int64_t ); + posit32_t i64_to_p32( int64_t ); + posit64_t i64_to_p64( int64_t ); + + # /*---------------------------------------------------------------------------- + # | 8-bit (quad-precision) posit operations. + # *----------------------------------------------------------------------------*/ + + bint isNaRP8UI( uint8_t ); + + uint_fast32_t p8_to_ui32( posit8_t ); + uint_fast64_t p8_to_ui64( posit8_t ); + int_fast32_t p8_to_i32( posit8_t ); + int_fast64_t p8_to_i64( posit8_t ); + + posit16_t p8_to_p16( posit8_t ); + #posit32_t p8_to_p32( posit8_t ); + #posit64_t p8_to_p64( posit8_t ); + + posit8_t p8_roundToInt( posit8_t ); + posit8_t p8_add( posit8_t, posit8_t ); + posit8_t p8_sub( posit8_t, posit8_t ); + posit8_t p8_mul( posit8_t, posit8_t ); + posit8_t p8_mulAdd( posit8_t, posit8_t, posit8_t ); + posit8_t p8_div( posit8_t, posit8_t ); + posit8_t p8_sqrt( posit8_t ); + bint p8_eq( posit8_t, posit8_t ); + bint p8_le( posit8_t, posit8_t ); + bint p8_lt( posit8_t, posit8_t ); + + # Quire 8 + quire8_t q8_fdp_add( quire8_t, posit8_t, posit8_t ); + quire8_t q8_fdp_sub( quire8_t, posit8_t, posit8_t ); + posit8_t q8_to_p8( quire8_t ); + + bint isNaRQ8( quire8_t ); + bint isQ8Zero( quire8_t ); + quire8_t q8_clr( quire8_t ); + quire8_t castQ8( uint32_t ); + posit8_t castP8( uint8_t ); + uint8_t castUI8( posit8_t ); + posit8_t negP8( posit8_t ); + + # Helper + double convertP8ToDouble( posit8_t ); + posit8_t convertDoubleToP8( double ); + + # /*---------------------------------------------------------------------------- + # | 16-bit (half-precision) posit operations. + # *----------------------------------------------------------------------------*/ + + bint isNaRP16UI( uint16_t ); + + uint_fast32_t p16_to_ui32( posit16_t ); + uint_fast64_t p16_to_ui64( posit16_t ); + int_fast32_t p16_to_i32( posit16_t); + int_fast64_t p16_to_i64( posit16_t ); + posit8_t p16_to_p8( posit16_t ); + #posit32_t p16_to_p32( posit16_t ); + #posit64_t p16_to_p64( posit16_t ); + + posit16_t p16_roundToInt( posit16_t); + posit16_t p16_add( posit16_t, posit16_t ); + posit16_t p16_sub( posit16_t, posit16_t ); + posit16_t p16_mul( posit16_t, posit16_t ); + posit16_t p16_mulAdd( posit16_t, posit16_t, posit16_t ); + posit16_t p16_div( posit16_t, posit16_t ); + posit16_t p16_sqrt( posit16_t ); + bint p16_eq( posit16_t, posit16_t ); + bint p16_le( posit16_t, posit16_t ); + bint p16_lt( posit16_t, posit16_t ); + + #Quire 16 + quire16_t q16_fdp_add( quire16_t, posit16_t, posit16_t ); + quire16_t q16_fdp_sub( quire16_t, posit16_t, posit16_t ); + #posit16_t convertQ16ToP16( quire16_t ); + posit16_t q16_to_p16( quire16_t ); + + bint isNaRQ16( quire16_t ); + bint isQ16Zero( quire16_t ); + #quire16_t q16_TwosComplement( quire16_t ); + + #void printBinary( uint64_t*, int ); + #void printHex( uint64_t ); + + quire16_t q16_clr( quire16_t ); + quire16_t castQ16( uint16_t, uint16_t ); + posit16_t castP16( uint16_t ); + uint16_t castUI16( posit16_t ); + posit16_t negP16( posit16_t ); + + # Helper + double convertP16ToDouble( posit16_t ); + #posit16_t convertFloatToP16( float ); + posit16_t convertDoubleToP16( double ); diff --git a/posit.pyx b/posit.pyx index 3bbeabf..ce3888c 100644 --- a/posit.pyx +++ b/posit.pyx @@ -1,11 +1,408 @@ cimport cposit cdef class Posit8: + + # the wrapped posit value cdef cposit.posit8_t _c_posit + # factory function constructors that bypass __init__ + + @staticmethod + cdef Posit8 from_c_posit(cposit.posit8_t f): + """Factory function to create a Posit8 object directly from + a C posit8_t. + """ + cdef Posit8 obj = Posit8.__new__(Posit8) + obj._c_posit = f + return obj + + @staticmethod + def from_bits(int value): + """Factory function to create a Posit8 object from a bit pattern + represented as an integer. + """ + cdef Posit8 obj = Posit8.__new__(Posit8) + obj._c_posit = cposit.castP8(value) + return obj + + @staticmethod + def from_double(float value): + """Factory function to create a Posit8 object from a double. + """ + cdef Posit8 obj = Posit8.__new__(Posit8) + obj._c_posit = cposit.convertDoubleToP8(value) + return obj + + # convenience interface for use inside Python + + def __init__(self, value): + if isinstance(value, int): + self._c_posit = cposit.castP8(value) + else: + f = float(value) + self._c_posit = cposit.convertDoubleToP8(f) + + def __str__(self): + return repr(cposit.convertP8ToDouble(self._c_posit)) + + def __repr__(self): + return 'Posit8(' + repr(cposit.convertP8ToDouble(self._c_posit)) + ')' + + cpdef get_bits(self): + cdef int u = cposit.castUI8(self._c_posit) + return u + bits = property(get_bits) + + # arithmetic + + cpdef round(self): + cdef cposit.posit8_t f = cposit.p8_roundToInt(self._c_posit) + return Posit8.from_c_posit(f) + + def __round__(self, Posit8 other): + return self.roundToInt(other) + + cpdef add(self, Posit8 other): + cdef cposit.posit8_t f = cposit.p8_add(self._c_posit, other._c_posit) + return Posit8.from_c_posit(f) + + def __add__(self, Posit8 other): + return self.add(other) + + cpdef sub(self, Posit8 other): + cdef cposit.posit8_t f = cposit.p8_sub(self._c_posit, other._c_posit) + return Posit8.from_c_posit(f) + + def __sub__(self, Posit8 other): + return self.sub(other) + + cpdef mul(self, Posit8 other): + cdef cposit.posit8_t f = cposit.p8_mul(self._c_posit, other._c_posit) + return Posit8.from_c_posit(f) + + def __mul__(self, Posit8 other): + return self.mul(other) + + cpdef fma(self, Posit8 a2, Posit8 a3): + cdef cposit.posit8_t f = cposit.p8_mulAdd(self._c_posit, a2._c_posit, a3._c_posit) + return Posit8.from_c_posit(f) + + cpdef div(self, Posit8 other): + cdef cposit.posit8_t f = cposit.p8_div(self._c_posit, other._c_posit) + return Posit8.from_c_posit(f) + + def __truediv__(self, Posit8 other): + return self.div(other) + + cpdef sqrt(self): + cdef cposit.posit8_t f = cposit.p8_sqrt(self._c_posit) + return Posit8.from_c_posit(f) + + # in-place arithmetic + + cpdef iround(self): + self._c_posit = cposit.p8_roundToInt(self._c_posit) + + cpdef iadd(self, Posit8 other): + self._c_posit = cposit.p8_add(self._c_posit, other._c_posit) + + def __iadd__(self, Posit8 other): + self.iadd(other) + return self + + cpdef isub(self, Posit8 other): + self._c_posit = cposit.p8_sub(self._c_posit, other._c_posit) + + def __isub__(self, Posit8 other): + self.isub(other) + return self + + cpdef imul(self, Posit8 other): + self._c_posit = cposit.p8_mul(self._c_posit, other._c_posit) + + def __imul__(self, Posit8 other): + self.imul(other) + return self + + cpdef ifma(self, Posit8 a2, Posit8 a3): + self._c_posit = cposit.p8_mulAdd(self._c_posit, a2._c_posit, a3._c_posit) + + cpdef idiv(self, Posit8 other): + self._c_posit = cposit.p8_div(self._c_posit, other._c_posit) + + def __itruediv__(self, Posit8 other): + self.idiv(other) + return self + + cpdef isqrt(self): + self._c_posit = cposit.p8_sqrt(self._c_posit) + + # comparison + + cpdef eq(self, Posit8 other): + cdef bint b = cposit.p8_eq(self._c_posit, other._c_posit) + return b + + cpdef le(self, Posit8 other): + cdef bint b = cposit.p8_le(self._c_posit, other._c_posit) + return b + + cpdef lt(self, Posit8 other): + cdef bint b = cposit.p8_lt(self._c_posit, other._c_posit) + return b + + def __lt__(self, Posit8 other): + return self.lt(other) + + def __le__(self, Posit8 other): + return self.le(other) + + def __eq__(self, Posit8 other): + return self.eq(other) + + def __ne__(self, Posit8 other): + return not self.eq(other) + + def __ge__(self, Posit8 other): + return other.le(self) + + def __gt__(self, Posit8 other): + return other.lt(self) + + # conversion to other posit types + + cpdef to_p16(self): + cdef cposit.posit16_t f = cposit.p8_to_p16(self._c_posit) + return Posit16.from_c_posit(f) + + +cdef class Quire8: + + # the wrapped quire value + cdef cposit.quire8_t _c_quire + + # limited interface for now + def __cinit__(self): - self._c_posit = cposit.i64_to_p8(1) + cposit.q8_clr(self._c_quire) + + def __str__(self): + return repr(cposit.convertP8ToDouble(cposit.q8_to_p8(self._c_quire))) + + def __repr__(self): + return 'Quire8(' + repr(cposit.convertP8ToDouble(cposit.q8_to_p8(self._c_quire))) + ')' + + cpdef fdp_add(self, Posit8 a2, Posit8 a3): + self._c_quire = cposit.q8_fdp_add(self._c_quire, a2._c_posit, a3._c_posit) + + cpdef fdp_sub(self, Posit8 a2, Posit8 a3): + self._c_quire = cposit.q8_fdp_sub(self._c_quire, a2._c_posit, a3._c_posit) + + cpdef get_p8(self): + cpdef cposit.posit8_t f = cposit.q8_to_p8(self._c_quire) + return Posit8.from_c_posit(f) + p8 = property(get_p8) + + +cdef class Posit16: + + # the wrapped posit value + cdef cposit.posit16_t _c_posit + + # factory function constructors that bypass __init__ + + @staticmethod + cdef Posit16 from_c_posit(cposit.posit16_t f): + """Factory function to create a Posit16 object directly from + a C posit16_t. + """ + cdef Posit16 obj = Posit16.__new__(Posit16) + obj._c_posit = f + return obj + + @staticmethod + def from_bits(int value): + """Factory function to create a Posit16 object from a bit pattern + represented as an integer. + """ + cdef Posit16 obj = Posit16.__new__(Posit16) + obj._c_posit = cposit.castP16(value) + return obj + + @staticmethod + def from_double(float value): + """Factory function to create a Posit16 object from a double. + """ + cdef Posit16 obj = Posit16.__new__(Posit16) + obj._c_posit = cposit.convertDoubleToP16(value) + return obj + + # convenience interface for use inside Python + + def __init__(self, value): + if isinstance(value, int): + self._c_posit = cposit.castP16(value) + else: + f = float(value) + self._c_posit = cposit.convertDoubleToP16(f) + + def __str__(self): + return repr(cposit.convertP16ToDouble(self._c_posit)) + + def __repr__(self): + return 'Posit16(' + repr(cposit.convertP16ToDouble(self._c_posit)) + ')' + + cpdef get_bits(self): + cdef int u = cposit.castUI16(self._c_posit) + return u + bits = property(get_bits) + + # arithmetic + + cpdef round(self): + cdef cposit.posit16_t f = cposit.p16_roundToInt(self._c_posit) + return Posit16.from_c_posit(f) + + def __round__(self, Posit16 other): + return self.roundToInt(other) + + cpdef add(self, Posit16 other): + cdef cposit.posit16_t f = cposit.p16_add(self._c_posit, other._c_posit) + return Posit16.from_c_posit(f) + + def __add__(self, Posit16 other): + return self.add(other) + + cpdef sub(self, Posit16 other): + cdef cposit.posit16_t f = cposit.p16_sub(self._c_posit, other._c_posit) + return Posit16.from_c_posit(f) + + def __sub__(self, Posit16 other): + return self.sub(other) + + cpdef mul(self, Posit16 other): + cdef cposit.posit16_t f = cposit.p16_mul(self._c_posit, other._c_posit) + return Posit16.from_c_posit(f) + + def __mul__(self, Posit16 other): + return self.mul(other) + + cpdef fma(self, Posit16 a2, Posit16 a3): + cdef cposit.posit16_t f = cposit.p16_mulAdd(self._c_posit, a2._c_posit, a3._c_posit) + return Posit16.from_c_posit(f) + + cpdef div(self, Posit16 other): + cdef cposit.posit16_t f = cposit.p16_div(self._c_posit, other._c_posit) + return Posit16.from_c_posit(f) + + def __truediv__(self, Posit16 other): + return self.div(other) + + cpdef sqrt(self): + cdef cposit.posit16_t f = cposit.p16_sqrt(self._c_posit) + return Posit16.from_c_posit(f) + + # in-place arithmetic + + cpdef iround(self): + self._c_posit = cposit.p16_roundToInt(self._c_posit) + + cpdef iadd(self, Posit16 other): + self._c_posit = cposit.p16_add(self._c_posit, other._c_posit) + + def __iadd__(self, Posit16 other): + self.iadd(other) + return self + + cpdef isub(self, Posit16 other): + self._c_posit = cposit.p16_sub(self._c_posit, other._c_posit) + + def __isub__(self, Posit16 other): + self.isub(other) + return self + + cpdef imul(self, Posit16 other): + self._c_posit = cposit.p16_mul(self._c_posit, other._c_posit) + + def __imul__(self, Posit16 other): + self.imul(other) + return self + + cpdef ifma(self, Posit16 a2, Posit16 a3): + self._c_posit = cposit.p16_mulAdd(self._c_posit, a2._c_posit, a3._c_posit) + + cpdef idiv(self, Posit16 other): + self._c_posit = cposit.p16_div(self._c_posit, other._c_posit) + + def __itruediv__(self, Posit16 other): + self.idiv(other) + return self + + cpdef isqrt(self): + self._c_posit = cposit.p16_sqrt(self._c_posit) + + # comparison + + cpdef eq(self, Posit16 other): + cdef bint b = cposit.p16_eq(self._c_posit, other._c_posit) + return b + + cpdef le(self, Posit16 other): + cdef bint b = cposit.p16_le(self._c_posit, other._c_posit) + return b + + cpdef lt(self, Posit16 other): + cdef bint b = cposit.p16_lt(self._c_posit, other._c_posit) + return b + + def __lt__(self, Posit16 other): + return self.lt(other) + + def __le__(self, Posit16 other): + return self.le(other) + + def __eq__(self, Posit16 other): + return self.eq(other) + + def __ne__(self, Posit16 other): + return not self.eq(other) + + def __ge__(self, Posit16 other): + return other.le(self) + + def __gt__(self, Posit16 other): + return other.lt(self) + + # conversion to other posit types + + cpdef to_p8(self): + cdef cposit.posit8_t f = cposit.p16_to_p8(self._c_posit) + return Posit8.from_c_posit(f) + + +cdef class Quire16: + + # the wrapped quire value + cdef cposit.quire16_t _c_quire + + # limited interface for now + + def __cinit__(self): + cposit.q16_clr(self._c_quire) + + def __str__(self): + return repr(cposit.convertP16ToDouble(cposit.q16_to_p16(self._c_quire))) + + def __repr__(self): + return 'Quire16(' + repr(cposit.convertP16ToDouble(cposit.q16_to_p16(self._c_quire))) + ')' + + cpdef fdp_add(self, Posit16 a2, Posit16 a3): + self._c_quire = cposit.q16_fdp_add(self._c_quire, a2._c_posit, a3._c_posit) + + cpdef fdp_sub(self, Posit16 a2, Posit16 a3): + self._c_quire = cposit.q16_fdp_sub(self._c_quire, a2._c_posit, a3._c_posit) - cpdef show(self): - print(cposit.p8_to_i64(self._c_posit)) - + cpdef get_p16(self): + cpdef cposit.posit16_t f = cposit.q16_to_p16(self._c_quire) + return Posit16.from_c_posit(f) + p16 = property(get_p16) -- 2.30.2