Module Hardcaml.Fifo
Synchronous FIFO.
type t
=
{
q : Signal.t;
full : Signal.t;
empty : Signal.t;
nearly_full : Signal.t;
nearly_empty : Signal.t;
used : Signal.t;
}
val sexp_of_t : t -> Ppx_sexp_conv_lib.Sexp.t
type create_fifo
= ?nearly_empty:Hardcaml__.Import.int -> ?nearly_full:Hardcaml__.Import.int -> ?overflow_check:Hardcaml__.Import.bool -> ?reset:Signal.t -> ?underflow_check:Hardcaml__.Import.bool -> ?ram_attributes:Rtl_attribute.t Hardcaml__.Import.list -> ?scope:Scope.t -> Hardcaml__.Import.unit -> capacity:Hardcaml__.Import.int -> clock:Signal.t -> clear:Signal.t -> wr:Signal.t -> d:Signal.t -> rd:Signal.t -> t
Base RTL FIFO
val create : ?showahead:Hardcaml__.Import.bool -> create_fifo
create ~clock ~clear ~wr ~d ~rd capacity
builds a FIFO withcapacity
elements which is written withd
whenwr
is high and read whenrd
is high.The default reset configuration is to use a synchronous
clr
signal. An asynchronousrst
may be optionally provided. One ofclr
orrst
must be non-empty.Optional overflow and underflow checking may be used. Data will not be written(/read) when the fifo is
full
(/empty
) regardles or thewr
/(rd
) signals.nearly_emtpy
andnearly_full
may be programmed to go high when the fifo is nearing an underflow or overflow state.The
showahead
mode changes the read behaviour of the FIFO. When showahead isfalse
read data is available 1 cycle afterrd
is high. With showaheadtrue
the data is available on the same cycle asrd
is high. To supportshowahead
behaviour the timing of thefull
/empty
flag also changes (although they still correctly indicate when it is safe to read or write to the FIFO).showahead
mode has some extra cost in terms of extra logic. The implementation ensures the output is registered and timing performance is good - nearly as fast as the underlying RAM allows..Note;
showahead
is sometimes referred to as "first word fall through".The
used
output indicates the number of elements currently in the FIFO.
Derived FIFO architectures.
val create_classic_with_extra_reg : create_fifo
Adds an extra output register to the non-showahead fifo. This delays the output, but ensures there is no logic data on the fifo output. Adds an extra cycle of latency (2 cycles from write to empty low).
val create_showahead_from_classic : create_fifo
Constructs a showahead fifo from a non-showahead fifo. Only modifies the control flags. Has 2 cycles of latency.
val create_showahead_with_extra_reg : create_fifo
Constructs a fifo similarly to
create_showahead_from_classic
and ensures the output data is registered. Has 3 cycles of latency, but is slightly faster thancreate ~showahead:true
- it seems to only be limited by the underlying RAM frequency.