from nmigen import Signal, Cat, Const, Mux, Module, Value
from nmigen.cli import verilog, rtlil
-from nmigen.lib.fifo import SyncFIFO
+from nmigen.lib.fifo import SyncFIFO, SyncFIFOBuffered
from nmigen.hdl.ast import ArrayProxy
from nmigen.hdl.rec import Record, Layout
from abc import ABCMeta, abstractmethod
from collections.abc import Sequence
+from queue import Queue
class RecordObject(Record):
return eqs
+ def _postprocess(self, i): # XXX DISABLED
+ return i # RETURNS INPUT
+ if hasattr(self.stage, "postprocess"):
+ return self.stage.postprocess(i)
+ return i
+
def set_input(self, i):
""" helper function to set the input data
"""
# data pass-through conditions
with self.m.If(npnn):
+ o_data = self._postprocess(result)
self.m.d.sync += [self.n.o_valid.eq(p_i_valid), # valid if p_valid
- eq(self.n.o_data, result), # update output
+ eq(self.n.o_data, o_data), # update output
]
# buffer flush conditions (NOTE: can override data passthru conditions)
with self.m.If(nir_por_n): # not stalled
# Flush the [already processed] buffer to the output port.
+ o_data = self._postprocess(r_data)
self.m.d.sync += [self.n.o_valid.eq(1), # reg empty
- eq(self.n.o_data, r_data), # flush buffer
+ eq(self.n.o_data, o_data), # flush buffer
]
# output ready conditions
self.m.d.sync += self.p._o_ready.eq(nir_novn | por_pivn)
+--process->--^
Truth Table
- Inputs Temporary Output
- ------- ---------- -----
- P P N N PiV& ~NiV& N P
+ Inputs Temporary Output Data
+ ------- ---------- ----- ----
+ P P N N PiV& ~NiR& N P
i o i o PoR NoV o o
V R R V V R
------- - - - -
- 0 0 0 0 0 0 >0 0
- 0 0 0 1 0 1 >1 0
- 0 0 1 0 0 0 0 1
- 0 0 1 1 0 0 0 1
+ 0 0 0 0 0 0 >0 0 reg
+ 0 0 0 1 0 1 >1 0 reg
+ 0 0 1 0 0 0 0 1 process(i_data)
+ 0 0 1 1 0 0 0 1 process(i_data)
------- - - - -
- 0 1 0 0 0 0 >0 0
- 0 1 0 1 0 1 >1 0
- 0 1 1 0 0 0 0 1
- 0 1 1 1 0 0 0 1
+ 0 1 0 0 0 0 >0 0 reg
+ 0 1 0 1 0 1 >1 0 reg
+ 0 1 1 0 0 0 0 1 process(i_data)
+ 0 1 1 1 0 0 0 1 process(i_data)
------- - - - -
- 1 0 0 0 0 0 >0 0
- 1 0 0 1 0 1 >1 0
- 1 0 1 0 0 0 0 1
- 1 0 1 1 0 0 0 1
+ 1 0 0 0 0 0 >0 0 reg
+ 1 0 0 1 0 1 >1 0 reg
+ 1 0 1 0 0 0 0 1 process(i_data)
+ 1 0 1 1 0 0 0 1 process(i_data)
------- - - - -
- 1 1 0 0 1 0 1 0
- 1 1 0 1 1 1 1 0
- 1 1 1 0 1 0 1 1
- 1 1 1 1 1 0 1 1
+ 1 1 0 0 1 0 1 0 process(i_data)
+ 1 1 0 1 1 1 1 0 process(i_data)
+ 1 1 1 0 1 0 1 1 process(i_data)
+ 1 1 1 1 1 0 1 1 process(i_data)
------- - - - -
"""
# previous valid and ready
with m.If(p_i_valid_p_o_ready):
+ o_data = self._postprocess(result)
m.d.sync += [r_busy.eq(1), # output valid
- eq(self.n.o_data, result), # update output
+ eq(self.n.o_data, o_data), # update output
]
# previous invalid or not ready, however next is accepting
with m.Elif(n_i_ready):
- m.d.sync += [eq(self.n.o_data, result)]
+ o_data = self._postprocess(result)
+ m.d.sync += [eq(self.n.o_data, o_data)]
# TODO: could still send data here (if there was any)
#m.d.sync += self.n.o_valid.eq(0) # ...so set output invalid
m.d.sync += r_busy.eq(0) # ...so set output invalid
Truth Table
- Inputs Temp Output
- ------- - -----
+ Inputs Temp Output Data
+ ------- - ----- ----
P P N N ~NiR& N P
i o i o NoV o o
V R R V V R
------- - - -
- 0 0 0 0 0 0 1
- 0 0 0 1 1 1 0
- 0 0 1 0 0 0 1
- 0 0 1 1 0 0 1
+ 0 0 0 0 0 0 1 reg
+ 0 0 0 1 1 1 0 reg
+ 0 0 1 0 0 0 1 reg
+ 0 0 1 1 0 0 1 reg
------- - - -
- 0 1 0 0 0 0 1
- 0 1 0 1 1 1 0
- 0 1 1 0 0 0 1
- 0 1 1 1 0 0 1
+ 0 1 0 0 0 0 1 reg
+ 0 1 0 1 1 1 0 reg
+ 0 1 1 0 0 0 1 reg
+ 0 1 1 1 0 0 1 reg
------- - - -
- 1 0 0 0 0 1 1
- 1 0 0 1 1 1 0
- 1 0 1 0 0 1 1
- 1 0 1 1 0 1 1
+ 1 0 0 0 0 1 1 reg
+ 1 0 0 1 1 1 0 reg
+ 1 0 1 0 0 1 1 reg
+ 1 0 1 1 0 1 1 reg
------- - - -
- 1 1 0 0 0 1 1
- 1 1 0 1 1 1 0
- 1 1 1 0 0 1 1
- 1 1 1 1 0 1 1
+ 1 1 0 0 0 1 1 process(i_data)
+ 1 1 0 1 1 1 0 process(i_data)
+ 1 1 1 0 0 1 1 process(i_data)
+ 1 1 1 1 0 1 1 process(i_data)
------- - - -
Note: PoR is *NOT* involved in the above decision-making.
# some temporaries
p_i_valid = Signal(reset_less=True)
pv = Signal(reset_less=True)
+ buf_full = Signal(reset_less=True)
m.d.comb += p_i_valid.eq(self.p.i_valid_test)
m.d.comb += pv.eq(self.p.i_valid & self.p.o_ready)
+ m.d.comb += buf_full.eq(~self.n.i_ready_test & data_valid)
m.d.comb += self.n.o_valid.eq(data_valid)
m.d.comb += self.p._o_ready.eq(~data_valid | self.n.i_ready_test)
- m.d.sync += data_valid.eq(p_i_valid | \
- (~self.n.i_ready_test & data_valid))
+ m.d.sync += data_valid.eq(p_i_valid | buf_full)
+
with m.If(pv):
m.d.sync += eq(r_data, self.stage.process(self.p.i_data))
- m.d.comb += eq(self.n.o_data, r_data)
+ o_data = self._postprocess(r_data)
+ m.d.comb += eq(self.n.o_data, o_data)
return self.m
-
class UnbufferedPipeline2(ControlBase):
""" A simple pipeline stage with single-clock synchronisation
and two-way valid/ready synchronised signalling.
A temporary (buffered) copy of a valid output
This is HELD if the output is not ready. It is updated
SYNCHRONOUSLY.
+
+ Inputs Temp Output Data
+ ------- - -----
+ P P N N ~NiR& N P (buf_full)
+ i o i o NoV o o
+ V R R V V R
+
+ ------- - - -
+ 0 0 0 0 0 0 1 process(i_data)
+ 0 0 0 1 1 1 0 reg (odata, unchanged)
+ 0 0 1 0 0 0 1 process(i_data)
+ 0 0 1 1 0 0 1 process(i_data)
+ ------- - - -
+ 0 1 0 0 0 0 1 process(i_data)
+ 0 1 0 1 1 1 0 reg (odata, unchanged)
+ 0 1 1 0 0 0 1 process(i_data)
+ 0 1 1 1 0 0 1 process(i_data)
+ ------- - - -
+ 1 0 0 0 0 1 1 process(i_data)
+ 1 0 0 1 1 1 0 reg (odata, unchanged)
+ 1 0 1 0 0 1 1 process(i_data)
+ 1 0 1 1 0 1 1 process(i_data)
+ ------- - - -
+ 1 1 0 0 0 1 1 process(i_data)
+ 1 1 0 1 1 1 0 reg (odata, unchanged)
+ 1 1 1 0 0 1 1 process(i_data)
+ 1 1 1 1 0 1 1 process(i_data)
+ ------- - - -
+
+ Note: PoR is *NOT* involved in the above decision-making.
"""
def elaborate(self, platform):
m.d.comb += self.p._o_ready.eq(~buf_full)
m.d.sync += buf_full.eq(~self.n.i_ready_test & self.n.o_valid)
- odata = Mux(buf_full, buf, self.stage.process(self.p.i_data))
- m.d.comb += eq(self.n.o_data, odata)
+ o_data = Mux(buf_full, buf, self.stage.process(self.p.i_data))
+ o_data = self._postprocess(o_data)
+ m.d.comb += eq(self.n.o_data, o_data)
m.d.sync += eq(buf, self.n.o_data)
return self.m
class PassThroughHandshake(ControlBase):
""" A control block that delays by one clock cycle.
+
+ Inputs Temporary Output Data
+ ------- ------------------ ----- ----
+ P P N N PiV& PiV| NiR| pvr N P (pvr)
+ i o i o PoR ~PoR ~NoV o o
+ V R R V V R
+
+ ------- - - - - - -
+ 0 0 0 0 0 1 1 0 1 1 odata (unchanged)
+ 0 0 0 1 0 1 0 0 1 0 odata (unchanged)
+ 0 0 1 0 0 1 1 0 1 1 odata (unchanged)
+ 0 0 1 1 0 1 1 0 1 1 odata (unchanged)
+ ------- - - - - - -
+ 0 1 0 0 0 0 1 0 0 1 odata (unchanged)
+ 0 1 0 1 0 0 0 0 0 0 odata (unchanged)
+ 0 1 1 0 0 0 1 0 0 1 odata (unchanged)
+ 0 1 1 1 0 0 1 0 0 1 odata (unchanged)
+ ------- - - - - - -
+ 1 0 0 0 0 1 1 1 1 1 process(in)
+ 1 0 0 1 0 1 0 0 1 0 odata (unchanged)
+ 1 0 1 0 0 1 1 1 1 1 process(in)
+ 1 0 1 1 0 1 1 1 1 1 process(in)
+ ------- - - - - - -
+ 1 1 0 0 1 1 1 1 1 1 process(in)
+ 1 1 0 1 1 1 0 0 1 0 odata (unchanged)
+ 1 1 1 0 1 1 1 1 1 1 process(in)
+ 1 1 1 1 1 1 1 1 1 1 process(in)
+ ------- - - - - - -
+
"""
def elaborate(self, platform):
self.m = m = ControlBase._elaborate(self, platform)
+ r_data = self.stage.ospec() # output type
+
# temporaries
p_i_valid = Signal(reset_less=True)
pvr = Signal(reset_less=True)
m.d.comb += self.p.o_ready.eq(~self.n.o_valid | self.n.i_ready_test)
m.d.sync += self.n.o_valid.eq(p_i_valid | ~self.p.o_ready)
- odata = Mux(pvr, self.stage.process(self.p.i_data), self.n.o_data)
- m.d.sync += eq(self.n.o_data, odata)
+ odata = Mux(pvr, self.stage.process(self.p.i_data), r_data)
+ m.d.sync += eq(r_data, odata)
+ r_data = self._postprocess(r_data)
+ m.d.comb += eq(self.n.o_data, r_data)
return m
i_data -> fifo.din -> FIFO -> fifo.dout -> o_data
"""
- def __init__(self, depth, iospecfn):
- """ * iospecfn: specification for incoming and outgoing data
- * depth : number of entries in the FIFO
+ def __init__(self, depth, stage, in_multi=None, stage_ctl=False,
+ fwft=True, buffered=False, pipe=False):
+ """ FIFO Control
+
+ * depth: number of entries in the FIFO
+ * stage: data processing block
+ * fwft : first word fall-thru mode (non-fwft introduces delay)
+ * buffered: use buffered FIFO (introduces extra cycle delay)
+
+ NOTE 1: FPGAs may have trouble with the defaults for SyncFIFO
+ (fwft=True, buffered=False)
+
+ NOTE 2: i_data *must* have a shape function. it can therefore
+ be a Signal, or a Record, or a RecordObject.
- NOTE: FPGAs may have trouble with the defaults for SyncFIFO
+ data is processed (and located) as follows:
+
+ self.p self.stage temp fn temp fn temp fp self.n
+ i_data->process()->result->flatten->din.FIFO.dout->flatten(o_data)
+
+ yes, really: flatten produces a Cat() which can be assigned to.
+ this is how the FIFO gets de-flattened without needing a de-flatten
+ function
"""
+ assert not (fwft and buffered), "buffered cannot do fwft"
+ if buffered:
+ depth += 1
+ self.fwft = fwft
+ self.buffered = buffered
+ self.pipe = pipe
self.fdepth = depth
- stage = PassThroughStage(iospecfn)
- ControlBase.__init__(self, stage=stage)
+ ControlBase.__init__(self, stage, in_multi, stage_ctl)
def elaborate(self, platform):
self.m = m = ControlBase._elaborate(self, platform)
- (fwidth, _) = self.p.i_data.shape()
- fifo = SyncFIFO(fwidth, self.fdepth)
+ # make a FIFO with a signal of equal width to the o_data.
+ (fwidth, _) = self.n.o_data.shape()
+ if self.buffered:
+ fifo = SyncFIFOBuffered(fwidth, self.fdepth)
+ else:
+ fifo = Queue(fwidth, self.fdepth, fwft=self.fwft, pipe=self.pipe)
m.submodules.fifo = fifo
- # prev: make the FIFO "look" like a PrevControl...
- fp = PrevControl()
- fp.i_valid = fifo.we
- fp._o_ready = fifo.writable
- fp.i_data = fifo.din
- m.d.comb += fp._connect_in(self.p, True, fn=flatten)
-
- # next: make the FIFO "look" like a NextControl...
- fn = NextControl()
- fn.o_valid = fifo.readable
- fn.i_ready = fifo.re
- fn.o_data = fifo.dout
- # ... so we can do this!
- m.d.comb += fn._connect_out(self.n, fn=flatten)
-
- # err... that should be all!
+ # store result of processing in combinatorial temporary
+ result = self.stage.ospec()
+ m.d.comb += eq(result, self.stage.process(self.p.i_data))
+
+ # connect previous rdy/valid/data - do flatten on i_data
+ # NOTE: cannot do the PrevControl-looking trick because
+ # of need to process the data. shaaaame....
+ m.d.comb += [fifo.we.eq(self.p.i_valid_test),
+ self.p.o_ready.eq(fifo.writable),
+ eq(fifo.din, flatten(result)),
+ ]
+
+ # connect next rdy/valid/data - do flatten on o_data
+ connections = [self.n.o_valid.eq(fifo.readable),
+ fifo.re.eq(self.n.i_ready_test),
+ ]
+ if self.fwft or self.buffered:
+ m.d.comb += connections
+ else:
+ m.d.sync += connections # unbuffered fwft mode needs sync
+ o_data = flatten(self.n.o_data).eq(fifo.dout)
+ o_data = self._postprocess(o_data)
+ m.d.comb += o_data
+
return m
+
+# aka "RegStage".
+class UnbufferedPipeline(FIFOControl):
+ def __init__(self, stage, in_multi=None, stage_ctl=False):
+ FIFOControl.__init__(self, 1, stage, in_multi, stage_ctl,
+ fwft=True, pipe=False)
+
+# aka "BreakReadyStage" XXX had to set fwft=True to get it to work
+class PassThroughHandshake(FIFOControl):
+ def __init__(self, stage, in_multi=None, stage_ctl=False):
+ FIFOControl.__init__(self, 1, stage, in_multi, stage_ctl,
+ fwft=True, pipe=True)
+
+# this is *probably* BufferedHandshake, although test #997 now succeeds.
+class BufferedHandshake(FIFOControl):
+ def __init__(self, stage, in_multi=None, stage_ctl=False):
+ FIFOControl.__init__(self, 2, stage, in_multi, stage_ctl,
+ fwft=True, pipe=False)
+
+
+# this is *probably* SimpleHandshake (note: memory cell size=0)
+class SimpleHandshake(FIFOControl):
+ def __init__(self, stage, in_multi=None, stage_ctl=False):
+ FIFOControl.__init__(self, 0, stage, in_multi, stage_ctl,
+ fwft=True, pipe=False)
projects / ieee754fpu.git / blobdiff