yosys/docs/source/cell/word_arith.rst

94 lines
3.0 KiB
ReStructuredText

Coarse arithmetics
------------------
Multiply-accumulate
~~~~~~~~~~~~~~~~~~~
The `$macc` cell type represents a generalized multiply and accumulate
operation. The cell is purely combinational. It outputs the result of summing up
a sequence of products and other injected summands.
.. code-block::
Y = 0 +- a0factor1 * a0factor2 +- a1factor1 * a1factor2 +- ...
+ B[0] + B[1] + ...
The A port consists of concatenated pairs of multiplier inputs ("factors"). A
zero length factor2 acts as a constant 1, turning factor1 into a simple summand.
In this pseudocode, ``u(foo)`` means an unsigned int that's foo bits long.
.. code-block::
struct A {
u(CONFIG.mul_info[0].factor1_len) a0factor1;
u(CONFIG.mul_info[0].factor2_len) a0factor2;
u(CONFIG.mul_info[1].factor1_len) a1factor1;
u(CONFIG.mul_info[1].factor2_len) a1factor2;
...
};
The cell's ``CONFIG`` parameter determines the layout of cell port ``A``. The
CONFIG parameter carries the following information:
.. code-block::
struct CONFIG {
u4 num_bits;
struct mul_info {
bool is_signed;
bool is_subtract;
u(num_bits) factor1_len;
u(num_bits) factor2_len;
}[num_ports];
};
B is an array of concatenated 1-bit-wide unsigned integers to also be summed up.
Arbitrary logic functions
~~~~~~~~~~~~~~~~~~~~~~~~~
The `$lut` cell type implements a single-output LUT (lookup table). It
implements an arbitrary logic function with its ``\LUT`` parameter to map input
port ``\A`` to values of ``\Y`` output port values. In psuedocode: ``Y =
\LUT[A]``. ``\A`` has width set by parameter ``\WIDTH`` and ``\Y`` has a width
of 1. Every logic function with a single bit output has a unique `$lut`
representation.
The `$sop` cell type implements a sum-of-products expression, also known as
disjunctive normal form (DNF). It implements an arbitrary logic function. Its
structure mimics a programmable logic array (PLA). Output port ``\Y`` is the sum
of products of the bits of the input port ``\A`` as defined by parameter
``\TABLE``. ``\A`` is ``\WIDTH`` bits wide. The number of products in the sum is
set by parameter ``\DEPTH``, and each product has two bits for each input bit -
for the presence of the unnegated and negated version of said input bit in the
product. Therefore the ``\TABLE`` parameter holds ``2 * \WIDTH * \DEPTH`` bits.
For example:
Let ``\WIDTH`` be 3. We would like to represent ``\Y =~\A[0] + \A[1]~\A[2]``.
There are 2 products to be summed, so ``\DEPTH`` shall be 2.
.. code-block::
~A[2]-----+
A[2]----+|
~A[1]---+||
A[1]--+|||
~A[0]-+||||
A[0]+|||||
|||||| product formula
010000 ~\A[0]
001001 \A[1]~\A[2]
So the value of ``\TABLE`` will become ``010000001001``.
Any logic function with a single bit output can be represented with ``$sop`` but
may have variously minimized or ordered summands represented in the ``\TABLE``
values.
.. autocellgroup:: arith
:members:
:source:
:linenos: