Module Hardcaml_kernel.SignalSource

logic 1

logic 0

create signal from a constant

convert binary string to constant

Convert integer to constant. Negative values are sign extended up to width. Input values which are larger or smaller than those representable in width bits are truncated.

Same as of_int_trunc

• deprecated [since 2024-11] use [of_int_trunc]

• deprecated [since 2024-11] use [of_int32_trunc]

• deprecated [since 2024-11] use [of_int64_trunc]

Convert non-negative values to constant. Input values which are not representable in width bits will raise.

Convert signed values to constant. Input values which are not representable in width bits will raise.

convert hex string to a constant. If the target width is greater than the hex length and signedness is Signed then the result is sign extended. Otherwise the result is zero padded.

convert octal string to a constant. If the target width is greater than the octal length and signedness is Signed then the result is sign extended. Otherwise the result is zero padded.

Convert an arbitrarily wide integer value to a constant.

convert verilog style or binary string to constant

convert IntbitsList to constant

convert a char to an 8 bit constant

convert a bool to vdd or gnd

convert a string to a series of 8-bit characters and concatenate them into a signal, with the first character of the string being in the least-significant bits

convert a string to a series of 8-bit characters and concatenate them into a signal, with the first character of the string being in the most-significant bits

create random constant vector of given width

zero w makes a the zero valued constant of width w

ones w makes a constant of all ones of width w

one w makes a one valued constant of width w

the empty signal

returns true if the signal is empty

names a signal

let a = a -- "a" in ...

signals may have multiple names.

returns the width (number of bits) of a signal.

let w = width s in ...

addess_bits_for num_elements returns the address width required to index num_elements.

It is the same as Int.ceil_log2, except it wll return a minimum value of 1 (since you cannot have 0 width vectors). Raises if num_elements is < 0.

num_bits_to_represent x returns the number of bits required to represent the number x, which should be >= 0.

convert the signal to a constant

concat_msb ts concatenates a list of signals - the msb of the head of the list will become the msb of the result.

let c = concat_msb [ a; b; c ] in ...

concat_msb raises if ts is empty or if any t in ts is empty.

Similar to concat_msb except the lsb of the head of the list will become the lsb of the result.

Split signal into a list of signals with width equal to part_width. The least significant bits are at the head of the returned list. If exact is true the input signal width must be exactly divisable by part_width. When exact is false and the input signal width is not exactly divisible by part_width, the last element will contains residual bits.

eg:

  split_lsb ~part_width:4 16b0001_0010_0011_0100 =
    [ 4b0100; 4b0011; 4b0010; 4b0001 ]

  split_lsb ~exact:false ~part_width:4 17b11_0001_0010_0011_0100 =
    [ 4b0100; 4b0011; 4b0010; 4b0001; 2b11 ]

Like split_lsb except the most significant bits are at the head of the returned list. Residual bits when exact is false goes to the last element of the list, so in the general case split_lsb is not necessarily equivalent to split_msb |> List.rev.

multiplexer.

let m = mux sel inputs in ...

Given l = List.length inputs and w = width sel the following conditions must hold:

l <= Int.pow 2 w, l >= 1

If l < Int.pow 2 w, the last input is repeated.

All inputs provided must have the same width, which will in turn be equal to the width of m.

Same as mux except requires:

l = Int.pow 2 w

concatenate two signals.

let c = a @: b in ...

equivalent to concat [ a; b ]

• deprecated [since 2025-03] use [to_bool]

• deprecated [since 2025-03] use [not to_bool]

select t ~high ~low selects from t bits in the range high...low, inclusive. select raises unless high and low fall within 0 .. width t - 1 and high >= lo.

select a single bit

get most significant bit

get least significant bits

get least significant bit

get most significant bits

drop_bottom s ~width drop bottom width bits of s

drop_top s ~width drop top width bits of s

sel_bottom s ~width select bottom width bits of s

sel_top s ~width select top width bits of s

x.:[hi, lo] == select x hi lo

x.:+[lo, width] == select x (lo + width - 1) lo. If width is None it selects all remaining msbs of the vector ie x.:+[lo,None] == drop_bottom x lo

x.:-[hi, width] == select x hi (hi - width + 1). If hi is None it defaults to the msb of the vector ie x.:-[None, width] == sel_top x width

x.(i) == bit x i

insert ~into:t x ~at_offset insert x into t at given offet

mux2 c t f 2 input multiplexer. Selects t if c is high otherwise f.

t and f must have same width and c must be 1 bit.

Equivalent to mux c [f; t]

cases ~default select [(match0, value0); (match0, value0)...] compares select with matchN and outputs valueN. If nothing matches default is output. This construct maps to a case statement.

The matchN values must be constants.

logical and

a <>:. 0 &: b <>:. 0

logical or

 a <>:. 0 |: b <>:. 0

logic xor

logical not

addition

subtraction

negation

unsigned multiplication

signed multiplication

equality

inequality

less than

greater than

less than or equal to

greater than or equal to

signed less than

signed greater than

signed less than or equal to

signed greated than or equal to

Propositional logic implication operator

create string from signal

to_int t treats t as unsigned and resizes it to fit exactly within an OCaml Int.t.

• If width t > Int.num_bits then the upper bits are truncated.
• If width t >= Int.num_bits and bit t (Int.num_bits-1) = vdd (i.e. the msb of the resulting Int.t is set), then the result is negative.
• If t is Signal.t and not a constant value, an exception is raised.

Same as to_int_trunc.

• deprecated [since 2024-11] use [to_int_trunc]

• deprecated [since 2024-11] use [to_int32_trunc]

• deprecated [since 2024-11] use [to_int64_trunc]

to_signed_int t treats t as signed and resizes it to fit exactly within an OCaml Int.t.

• An exception is raised if the value cannot fit in the resulting int.
• If t is Signal.t and not a constant value, an exception is raised.

to_unsigned_int t treats t as unsigned and resizes it to fit exactly within an OCaml Int.t.

• An exception is raised if the value cannot fit in the resulting int.
• The value int value must be unsigned (so the max width is Int.num_bits - 1).
• If t is Signal.t and not a constant value, an exception is raised.

Convert signal to a bool. The signal must be 1 bit wide.

Convert signal to a char. The signal must be 8 bits wide.

Convert into a string, with the first character being from the least-significant bits of the signal. The signal width must be a multiple of 8 bits.

Convert into a string, with the first character being from the most-significant bits of the signal. The signal width must be a multiple of 8 bits.

create binary string from signal

Convert bits to a Zarith.t

convert signal to a list of bits with msb at head of list

convert signal to a list of bits with lsb at head of list

repeat signal count times

Split signal in half. The most significant bits will be in the left half of the returned tuple.

• If msbs is not provided, the signal will be split in half with the MSB part possibly containing one more bit.
• If msbs is provided, msbs most significant bits will be split off.

Same as split_in_half_msb, but

• If lsbs is not provided, the LSB part might have one more bit.
• If lsbs is provided, lsbs least significant bits will be split off.

The most significant bits will still be in the left half of the tuple.

shift left logical

shift right logical

shift right arithmetic

rotate left

rotate right

shift by variable amount

uresize t ~width:w returns the unsigned resize of t to width w. If w = width t, this is a no-op. If w < width t, this selects the w low bits of t. If w > width t, this extends t with zero (w - width t).

sresize t ~width:w returns the signed resize of t to width w. If w = width t, this is a no-op. If w < width t, this selects the w low bits of t. If w > width t, this extends t with w - width t copies of msb t.

uextend t ~width:w behaves similarly to uresize t ~width:w but asserts that w >= width to prevent inadvertently chopping the signal.

sextend t ~width:w behaves similarly to sresize t ~width:w but asserts that w >= width to prevent inadvertently chopping the signal.

unsigned resize by +1 bit

signed resize by +1 bit

resize_list ~resize l finds the maximum width in l and applies resize el max to each element.

resize_op2 ~resize f a b applies resize x w to a and b where w is the maximum of their widths. It then returns f a b

fold 'op' though list

reverse bits

mod_counter max t is if t = max then 0 else (t + 1), and can be used to count from 0 to max then from zero again. If max == (1<<n - 1), then a comparator is not generated and overflow arithmetic is used instead.

compute_arity ~steps num_values computes the tree arity required to reduce num_values in steps. steps<=0 raises.

compute_tree_branches ~steps num_values returns a list of length steps of branching factors required to reduce num_values. This tends to produce a slightly more balanced sequence than just applying compute_arity at every step.

tree ~arity ~f input creates a tree of operations. The arity of the operator is configurable. tree raises if input = [].

priority_select cases returns the value associated with the first case whose valid signal is high. valid will be set low in the returned with_valid if no case is selected.

Same as priority_select except returns default if no case matches.

Select a case where one and only one valid signal is enabled. If more than one case is valid then the return value is undefined. If no cases are valid, 0 is returned by the current implementation, though this should not be relied upon.

popcount t returns the number of bits set in t.

binary_to_byte_qualifier ~num_bytes t returns a byte qualifier encoding of a binary size for use when encoding contiguous byte enables or byte keep signals.

is_pow2 t returns a bit to indicate if t is a power of 2.

leading_ones t returns the number of consecutive 1s from the most significant bit of t down.

trailing_ones t returns the number of consecutive 1s from the least significant bit of t up.

leading_zeros t returns the number of consecutive 0s from the most significant bit of t down.

trailing_zeros t returns the number of consecutive 0s from the least significant bit of t up.

floor_log2 x returns the floor of log-base-2 of x. x is treated as unsigned and an error is indicated by valid = gnd in the return value if x = 0.

ceil_log2 x returns the ceiling of log-base-2 of x. x is treated as unsigned and an error is indicated by valid = gnd in the return value if x = 0.

convert binary to onehot

convert onehot to binary

convert binary to gray code

convert gray code to binary

Increment a gray code value. The implementation converts to binary, increments the binary value, then converts back to gray code.

any_bit_set t returns vdd if at least one bit if set in t and gnd otherwise.

all_bits_set t returns vdd if all bits in t are set and gnd otherwise.

no_bits_set t returns vdd if no bits in t are set and gnd otherwise.

Increment (defaults to 1). Wrap on overflow.

Decrement (defaults to 1). Wrap on overflow.

Concatention, selection and resizing functions for signals encoded as an option where None means zero width.

Unsigned vector operations (ie may operate on Bits.t or Signal.t directly).

Signed vector operations (ie may operate on Bits.t or Signal.t directly).