Next

This fixes the 32-bit divide bugs

This version includes support for P32. Exhaustive tests are currently still in progress and is expected to require weeks to months to complete.

Support C++ - with wrapper

Support posit8_t and posit16_t

Release Version 0.2

Full posit16_t (16-bit posit with 1 exponent bit) support.

Added four functions:

1. convert unsigned int to posit16_t
2. convert unsigned long int to posit16_t
3. convert int to posit16_t
4. convert long int to posit16_t

Bug fix:

1. Fixed p16_mulAdd bug (#1)

This version supports a 16-bit with one exponent bit (posit16_t). This code is tested on

• GNU gcc (SUSE Linux) 4.8.5
• Apple LLVM version 9.1.0 (clang-902.0.39.2)

An example on how to use the code to add:

union ui16_p16 uA, uB, uZ;

uA.ui = 0x42;

uB.ui = 0x23;

uZ.p = p16_add(uA.p, uB.p);

//To check answer

double dZ = convertP16ToDec_f(uZ.p);

//To print result in binary

printBinary(&uZ.ui, 16);

Build - softposit.a

cd SoftPosit/build/Linux-x86_64-GCC

make -j6 all

Please link to the archive to use.

Features

Main Posit Functionalities:

1. Add posit16_t p16_add(posit16_t, posit_16)

2. Subtract posit16_t p16_sub(posit16_t, posit_16)

3. Divide posit16_t p16_div(posit16_t, posit_16)

4. Multiply posit16_t p16_mul(posit16_t, posit_16)

5. Fused Multiply Add posit16_t p16_mulAdd(posit16_t, posit_16, posit_16)

   p16_mulAdd(a, b, c) <=> a*b + c

Main Quire Functionalities

1. Fused dot product-add quire16_t q16_fdp_add(quire16_t, posit_16, posit_16)

2. Fused dot product-subtract quire16_t q16_fdp_sub(quire16_t, posit_16, posit_16)

3. Set quire variable to zero quire16_t q16_clr(quire16_t)

4. Convert quire to posit posit_16 convertQ16ToP16(quire16_t)

Additional Posit Functionalites

1. Square root posit16_t p16_sqrt(posit16_t)

2. Round to nearest integer posit16_t p16_roundToInt(posit16_t)

3. Check equal bool p16_eq( posit16_t, posit16_t )

4. Check less than equal bool p16_le( posit16_t, posit16_t )

5. Check less than bool p16_lt( posit16_t, posit16_t )

6. Convert p16 to integer (32 bits) int_fast32_t p16_to_i32( posit16_t )

7. Convert p16 to long integer (64 bits) int_fast64_t p16_to_i64( posit16_t)

8. Convert p16 to unsigned integer (32 bits) uint_fast32_t p16_to_ui32( posit16_t )

9. Convert p16 to unsigned long integer (64 bits) uint_fast64_t p16_to_ui64( posit16_t)

10. Convert p16 to p32 (posit<32, 2>) posit32_t p16_to_p32( posit16_t )

11. Convert p16 to double (64 bits) double convertP16ToDec_f(posit16_t)

12. Convert double (64 bits) to posit posit16_t convertdoubleToP16(double)