ac_types

[drom]

Add support for ac_types

ac_int<W,S>

• W - width of the number (non-negative interger number)
• S - signed number (bool flag)

Type	Numerical range	Quantum
{W: 1, S: false}	0..1	1
{W: 1, S: true}	-1..0	1
{W: 4, S: false}	0..15	1
{W: 4, S: true}	-8..7	1

ac_int<W, S> === ac_fixed<W, W, S>

ac_fixed<W, I, S, Q, O>

• W - width of the number (non-negative interger number)
• I - integer width (non-negative interger number)
• S - signed number (bool flag)
• Q - rounding (bool flag)
• O - overflow (bool flag)

Type	Numerical range	Quantum
{W: 4, I: 4, S: false}	0..15	1
{W: 4, I: 4, S: true}	-8..7	1
{W: 4, I: 6, S: false}	0..60	4
{W: 4, I: 6, S: true}	-32..28	4
{W: 4, I: 0, S: false}	0..15/16	1/16
{W: 4, I: 0, S: true}	-1/2..7/16	1/16
{W: 4, I: -1, S: false}	0..15/32	1/32
{W: 4, I: -1, S: true}	-0.25..7/32	1/32

ac_float<W, I, E, Q>

• W - width of mantissa (non-negative interger number)
• I - integer width (non-negative interger number)
• E - exponent (non-negative interger number)
• Q - rounding (bool flag)

Type	Numerical range	Quantum
{W: 54, I: 2, E: 11}	IEEE 754 double
{W: 25, I: 2, E: 8}	IEEE 754 float
{W: 12, I: 2, E: 5}	IEEE 754 half

❓ should width be W+E-1 ❓

ac_complex

C: true flag makes two part complex number.

Type	Numerical range	Quantum
{W: 8, I: 4, S: false, C: true}	0..15	1