* i_data : an input - added by the user of this class
"""
- def __init__(self, i_width=1):
+ def __init__(self, i_width=1, stage_ctl=False):
+ self.stage_ctl = stage_ctl
self.i_valid = Signal(i_width, name="p_i_valid") # prev >>in self
- self.o_ready = Signal(name="p_o_ready") # prev <<out self
+ self._o_ready = Signal(name="p_o_ready") # prev <<out self
self.i_data = None # XXX MUST BE ADDED BY USER
+ if stage_ctl:
+ self.s_o_ready = Signal(name="p_s_o_rdy") # prev <<out self
+
+ @property
+ def o_ready(self):
+ """ public-facing API: indicates (externally) that stage is ready
+ """
+ if self.stage_ctl:
+ return self.s_o_ready # set dynamically by stage
+ return self._o_ready # return this when not under dynamic control
def _connect_in(self, prev):
""" internal helper function to connect stage to an input source.
do not use to connect stage-to-stage!
"""
- return [self.i_valid.eq(prev.i_valid),
+ return [self.i_valid.eq(prev.i_valid_test),
prev.o_ready.eq(self.o_ready),
eq(self.i_data, prev.i_data),
]
- def i_valid_logic(self):
+ @property
+ def i_valid_test(self):
vlen = len(self.i_valid)
- if vlen > 1: # multi-bit case: valid only when i_valid is all 1s
+ if vlen > 1:
+ # multi-bit case: valid only when i_valid is all 1s
all1s = Const(-1, (len(self.i_valid), False))
- return self.i_valid == all1s
- # single-bit i_valid case
- return self.i_valid
+ i_valid = (self.i_valid == all1s)
+ else:
+ # single-bit i_valid case
+ i_valid = self.i_valid
+
+ # when stage indicates not ready, incoming data
+ # must "appear" to be not ready too
+ if self.stage_ctl:
+ i_valid = i_valid & self.s_o_ready
+
+ return i_valid
class NextControl:
* i_ready: input from next stage indicating that it can accept data
* o_data : an output - added by the user of this class
"""
- def __init__(self):
+ def __init__(self, stage_ctl=False):
+ self.stage_ctl = stage_ctl
self.o_valid = Signal(name="n_o_valid") # self out>> next
self.i_ready = Signal(name="n_i_ready") # self <<in next
self.o_data = None # XXX MUST BE ADDED BY USER
+ if self.stage_ctl:
+ self.d_valid = Signal(reset=1) # INTERNAL (data valid)
+
+ @property
+ def i_ready_test(self):
+ if self.stage_ctl:
+ return self.i_ready & self.d_valid
+ return self.i_ready
def connect_to_next(self, nxt):
""" helper function to connect to the next stage data/valid/ready.
do not use to connect stage-to-stage!
"""
return [nxt.o_valid.eq(self.o_valid),
- self.i_ready.eq(nxt.i_ready),
+ self.i_ready.eq(nxt.i_ready_test),
eq(nxt.o_data, self.o_data),
]
python object, enumerate them, find out the list of Signals that way,
and assign them.
"""
+ res = []
+ if isinstance(o, dict):
+ for (k, v) in o.items():
+ print ("d-eq", v, i[k])
+ res.append(v.eq(i[k]))
+ return res
+
if not isinstance(o, Sequence):
o, i = [o], [i]
- res = []
for (ao, ai) in zip(o, i):
#print ("eq", ao, ai)
if isinstance(ao, Record):
if self.specallocate:
ni = self.chain[idx+1].ispec() # new input on next loop
m.d.comb += eq(ni, o) # assign to next input
+ i = ni
else:
i = o
self.o = o # last loop is the output
class ControlBase:
""" Common functions for Pipeline API
"""
- def __init__(self, in_multi=None):
+ def __init__(self, in_multi=None, stage_ctl=False):
""" Base class containing ready/valid/data to previous and next stages
* p: contains ready/valid to the previous stage
* add i_data member to PrevControl (p) and
* add o_data member to NextControl (n)
"""
-
# set up input and output IO ACK (prev/next ready/valid)
- self.p = PrevControl(in_multi)
- self.n = NextControl()
+ self.p = PrevControl(in_multi, stage_ctl)
+ self.n = NextControl(stage_ctl)
def connect_to_next(self, nxt):
""" helper function to connect to the next stage data/valid/ready.
res += self.n.o_data
return res
+ def _elaborate(self, platform):
+ """ handles case where stage has dynamic ready/valid functions
+ """
+ m = Module()
+ if not self.p.stage_ctl:
+ return m
+
+ # intercept the previous (outgoing) "ready", combine with stage ready
+ m.d.comb += self.p.s_o_ready.eq(self.p._o_ready & self.stage.d_ready)
+
+ # intercept the next (incoming) "ready" and combine it with data valid
+ m.d.comb += self.n.d_valid.eq(self.n.i_ready & self.stage.d_valid)
+
+ return m
+
class BufferedPipeline(ControlBase):
""" buffered pipeline stage. data and strobe signals travel in sync.
input may begin to be processed and transferred directly to output.
"""
- def __init__(self, stage):
- ControlBase.__init__(self)
+ def __init__(self, stage, stage_ctl=False):
+ ControlBase.__init__(self, stage_ctl=stage_ctl)
self.stage = stage
# set up the input and output data
self.n.o_data = stage.ospec()
def elaborate(self, platform):
- m = Module()
+
+ self.m = ControlBase._elaborate(self, platform)
result = self.stage.ospec()
r_data = self.stage.ospec()
if hasattr(self.stage, "setup"):
- self.stage.setup(m, self.p.i_data)
+ self.stage.setup(self.m, self.p.i_data)
# establish some combinatorial temporaries
o_n_validn = Signal(reset_less=True)
i_p_valid_o_p_ready = Signal(reset_less=True)
p_i_valid = Signal(reset_less=True)
- m.d.comb += [p_i_valid.eq(self.p.i_valid_logic()),
+ self.m.d.comb += [p_i_valid.eq(self.p.i_valid_test),
o_n_validn.eq(~self.n.o_valid),
i_p_valid_o_p_ready.eq(p_i_valid & self.p.o_ready),
]
# store result of processing in combinatorial temporary
- m.d.comb += eq(result, self.stage.process(self.p.i_data))
+ self.m.d.comb += eq(result, self.stage.process(self.p.i_data))
# if not in stall condition, update the temporary register
- with m.If(self.p.o_ready): # not stalled
- m.d.sync += eq(r_data, result) # update buffer
+ with self.m.If(self.p.o_ready): # not stalled
+ self.m.d.sync += eq(r_data, result) # update buffer
- with m.If(self.n.i_ready): # next stage is ready
- with m.If(self.p.o_ready): # not stalled
+ with self.m.If(self.n.i_ready_test): # next stage is ready
+ with self.m.If(self.p._o_ready): # not stalled
# nothing in buffer: send (processed) input direct to output
- m.d.sync += [self.n.o_valid.eq(p_i_valid),
- eq(self.n.o_data, result), # update output
+ self.m.d.sync += [self.n.o_valid.eq(p_i_valid),
+ eq(self.n.o_data, result), # update output
]
- with m.Else(): # p.o_ready is false, and something is in buffer.
+ with self.m.Else(): # p.o_ready is false, and something in buffer
# Flush the [already processed] buffer to the output port.
- m.d.sync += [self.n.o_valid.eq(1), # declare reg empty
- eq(self.n.o_data, r_data), # flush buffer
- self.p.o_ready.eq(1), # clear stall condition
+ self.m.d.sync += [self.n.o_valid.eq(1), # reg empty
+ eq(self.n.o_data, r_data), # flush buffer
+ self.p._o_ready.eq(1), # clear stall
]
# ignore input, since p.o_ready is also false.
# (n.i_ready) is false here: next stage is ready
- with m.Elif(o_n_validn): # next stage being told "ready"
- m.d.sync += [self.n.o_valid.eq(p_i_valid),
- self.p.o_ready.eq(1), # Keep the buffer empty
- eq(self.n.o_data, result), # set output data
+ with self.m.Elif(o_n_validn): # next stage being told "ready"
+ self.m.d.sync += [self.n.o_valid.eq(p_i_valid),
+ self.p._o_ready.eq(1), # Keep the buffer empty
+ eq(self.n.o_data, result), # set output data
]
# (n.i_ready) false and (n.o_valid) true:
- with m.Elif(i_p_valid_o_p_ready):
+ with self.m.Elif(i_p_valid_o_p_ready):
# If next stage *is* ready, and not stalled yet, accept input
- m.d.sync += self.p.o_ready.eq(~(p_i_valid & self.n.o_valid))
+ self.m.d.sync += self.p._o_ready.eq(~(p_i_valid & self.n.o_valid))
- return m
+ return self.m
class UnbufferedPipeline(ControlBase):
COMBINATORIALLY (no clock dependence).
"""
- def __init__(self, stage):
- ControlBase.__init__(self)
+ def __init__(self, stage, stage_ctl=False):
+ ControlBase.__init__(self, stage_ctl=stage_ctl)
self.stage = stage
# set up the input and output data
self.n.o_data = stage.ospec() # output type
def elaborate(self, platform):
- m = Module()
+ self.m = ControlBase._elaborate(self, platform)
data_valid = Signal() # is data valid or not
r_data = self.stage.ispec() # input type
if hasattr(self.stage, "setup"):
- self.stage.setup(m, r_data)
+ self.stage.setup(self.m, r_data)
- # some temporarie
+ # some temporaries
p_i_valid = Signal(reset_less=True)
pv = Signal(reset_less=True)
- m.d.comb += p_i_valid.eq(self.p.i_valid_logic())
- m.d.comb += pv.eq(self.p.i_valid & self.p.o_ready)
-
- m.d.comb += self.n.o_valid.eq(data_valid)
- m.d.comb += self.p.o_ready.eq(~data_valid | self.n.i_ready)
- m.d.sync += data_valid.eq(p_i_valid | \
- (~self.n.i_ready & data_valid))
- with m.If(pv):
- m.d.sync += eq(r_data, self.p.i_data)
- m.d.comb += eq(self.n.o_data, self.stage.process(r_data))
- return m
+ self.m.d.comb += p_i_valid.eq(self.p.i_valid_test)
+ self.m.d.comb += pv.eq(self.p.i_valid & self.p.o_ready)
+
+ self.m.d.comb += self.n.o_valid.eq(data_valid)
+ self.m.d.comb += self.p._o_ready.eq(~data_valid | self.n.i_ready_test)
+ self.m.d.sync += data_valid.eq(p_i_valid | \
+ (~self.n.i_ready_test & data_valid))
+ with self.m.If(pv):
+ self.m.d.sync += eq(r_data, self.p.i_data)
+ self.m.d.comb += eq(self.n.o_data, self.stage.process(r_data))
+ return self.m
+
+
+class UnbufferedPipeline2(ControlBase):
+ """ A simple pipeline stage with single-clock synchronisation
+ and two-way valid/ready synchronised signalling.
+
+ Note that a stall in one stage will result in the entire pipeline
+ chain stalling.
+
+ Also that unlike BufferedPipeline, the valid/ready signalling does NOT
+ travel synchronously with the data: the valid/ready signalling
+ combines in a *combinatorial* fashion. Therefore, a long pipeline
+ chain will lengthen propagation delays.
+
+ Argument: stage. see Stage API, above
+
+ stage-1 p.i_valid >>in stage n.o_valid out>> stage+1
+ stage-1 p.o_ready <<out stage n.i_ready <<in stage+1
+ stage-1 p.i_data >>in stage n.o_data out>> stage+1
+ | |
+ r_data result
+ | |
+ +--process ->-+
+
+ Attributes:
+ -----------
+ p.i_data : StageInput, shaped according to ispec
+ The pipeline input
+ p.o_data : StageOutput, shaped according to ospec
+ The pipeline output
+ buf : output_shape according to ospec
+ A temporary (buffered) copy of a valid output
+ This is HELD if the output is not ready. It is updated
+ SYNCHRONOUSLY.
+ """
+
+ def __init__(self, stage, stage_ctl=False):
+ ControlBase.__init__(self, stage_ctl=stage_ctl)
+ self.stage = stage
+
+ # set up the input and output data
+ self.p.i_data = stage.ispec() # input type
+ self.n.o_data = stage.ospec() # output type
+
+ def elaborate(self, platform):
+ self.m = ControlBase._elaborate(self, platform)
+
+ buf_full = Signal() # is data valid or not
+ buf = self.stage.ospec() # output type
+ if hasattr(self.stage, "setup"):
+ self.stage.setup(self.m, self.p.i_data)
+
+ # some temporaries
+ p_i_valid = Signal(reset_less=True)
+ self.m.d.comb += p_i_valid.eq(self.p.i_valid_test)
+
+ self.m.d.comb += self.n.o_valid.eq(buf_full | p_i_valid)
+ self.m.d.comb += self.p._o_ready.eq(~buf_full)
+ self.m.d.sync += buf_full.eq(~self.n.i_ready_test & \
+ (p_i_valid | buf_full))
+ with self.m.If(buf_full):
+ self.m.d.comb += eq(self.n.o_data, buf)
+ with self.m.Else():
+ self.m.d.comb += eq(self.n.o_data,
+ self.stage.process(self.p.i_data))
+ self.m.d.sync += eq(buf, self.n.o_data)
+
+ return self.m
class PassThroughStage(StageCls):
projects / ieee754fpu.git / blobdiff