1 """ Pipeline and BufferedHandshake implementation, conforming to the same API.
2 For multi-input and multi-output variants, see multipipe.
4 Associated development bugs:
5 * http://bugs.libre-riscv.org/show_bug.cgi?id=64
6 * http://bugs.libre-riscv.org/show_bug.cgi?id=57
11 a strategically very important function that is identical in function
12 to nmigen's Signal.eq function, except it may take objects, or a list
13 of objects, or a tuple of objects, and where objects may also be
19 stage requires compliance with a strict API that may be
20 implemented in several means, including as a static class.
21 the methods of a stage instance must be as follows:
23 * ispec() - Input data format specification
24 returns an object or a list or tuple of objects, or
25 a Record, each object having an "eq" function which
26 takes responsibility for copying by assignment all
28 * ospec() - Output data format specification
29 requirements as for ospec
30 * process(m, i) - Processes an ispec-formatted object
31 returns a combinatorial block of a result that
32 may be assigned to the output, by way of the "eq"
34 * setup(m, i) - Optional function for setting up submodules
35 may be used for more complex stages, to link
36 the input (i) to submodules. must take responsibility
37 for adding those submodules to the module (m).
38 the submodules must be combinatorial blocks and
39 must have their inputs and output linked combinatorially.
41 Both StageCls (for use with non-static classes) and Stage (for use
42 by static classes) are abstract classes from which, for convenience
43 and as a courtesy to other developers, anything conforming to the
44 Stage API may *choose* to derive.
49 A useful combinatorial wrapper around stages that chains them together
50 and then presents a Stage-API-conformant interface. By presenting
51 the same API as the stages it wraps, it can clearly be used recursively.
56 A convenience class that takes an input shape, output shape, a
57 "processing" function and an optional "setup" function. Honestly
58 though, there's not much more effort to just... create a class
59 that returns a couple of Records (see ExampleAddRecordStage in
65 A convenience class that takes a single function as a parameter,
66 that is chain-called to create the exact same input and output spec.
67 It has a process() function that simply returns its input.
69 Instances of this class are completely redundant if handed to
70 StageChain, however when passed to UnbufferedPipeline they
71 can be used to introduce a single clock delay.
76 The base class for pipelines. Contains previous and next ready/valid/data.
77 Also has an extremely useful "connect" function that can be used to
78 connect a chain of pipelines and present the exact same prev/next
84 A simple stalling clock-synchronised pipeline that has no buffering
85 (unlike BufferedHandshake). Data flows on *every* clock cycle when
86 the conditions are right (this is nominally when the input is valid
87 and the output is ready).
89 A stall anywhere along the line will result in a stall back-propagating
90 down the entire chain. The BufferedHandshake by contrast will buffer
91 incoming data, allowing previous stages one clock cycle's grace before
94 An advantage of the UnbufferedPipeline over the Buffered one is
95 that the amount of logic needed (number of gates) is greatly
96 reduced (no second set of buffers basically)
98 The disadvantage of the UnbufferedPipeline is that the valid/ready
99 logic, if chained together, is *combinatorial*, resulting in
100 progressively larger gate delay.
102 PassThroughHandshake:
105 A Control class that introduces a single clock delay, passing its
106 data through unaltered. Unlike RegisterPipeline (which relies
107 on UnbufferedPipeline and PassThroughStage) it handles ready/valid
113 A convenience class that, because UnbufferedPipeline introduces a single
114 clock delay, when its stage is a PassThroughStage, it results in a Pipeline
115 stage that, duh, delays its (unmodified) input by one clock cycle.
120 nmigen implementation of buffered pipeline stage, based on zipcpu:
121 https://zipcpu.com/blog/2017/08/14/strategies-for-pipelining.html
123 this module requires quite a bit of thought to understand how it works
124 (and why it is needed in the first place). reading the above is
125 *strongly* recommended.
127 unlike john dawson's IEEE754 FPU STB/ACK signalling, which requires
128 the STB / ACK signals to raise and lower (on separate clocks) before
129 data may proceeed (thus only allowing one piece of data to proceed
130 on *ALTERNATE* cycles), the signalling here is a true pipeline
131 where data will flow on *every* clock when the conditions are right.
133 input acceptance conditions are when:
134 * incoming previous-stage strobe (p.i_valid) is HIGH
135 * outgoing previous-stage ready (p.o_ready) is LOW
137 output transmission conditions are when:
138 * outgoing next-stage strobe (n.o_valid) is HIGH
139 * outgoing next-stage ready (n.i_ready) is LOW
141 the tricky bit is when the input has valid data and the output is not
142 ready to accept it. if it wasn't for the clock synchronisation, it
143 would be possible to tell the input "hey don't send that data, we're
144 not ready". unfortunately, it's not possible to "change the past":
145 the previous stage *has no choice* but to pass on its data.
147 therefore, the incoming data *must* be accepted - and stored: that
148 is the responsibility / contract that this stage *must* accept.
149 on the same clock, it's possible to tell the input that it must
150 not send any more data. this is the "stall" condition.
152 we now effectively have *two* possible pieces of data to "choose" from:
153 the buffered data, and the incoming data. the decision as to which
154 to process and output is based on whether we are in "stall" or not.
155 i.e. when the next stage is no longer ready, the output comes from
156 the buffer if a stall had previously occurred, otherwise it comes
157 direct from processing the input.
159 this allows us to respect a synchronous "travelling STB" with what
160 dan calls a "buffered handshake".
162 it's quite a complex state machine!
167 Synchronised pipeline, Based on:
168 https://github.com/ZipCPU/dbgbus/blob/master/hexbus/rtl/hbdeword.v
171 from nmigen
import Signal
, Cat
, Const
, Mux
, Module
, Value
172 from nmigen
.cli
import verilog
, rtlil
173 from nmigen
.lib
.fifo
import SyncFIFO
, SyncFIFOBuffered
174 from nmigen
.hdl
.ast
import ArrayProxy
175 from nmigen
.hdl
.rec
import Record
, Layout
177 from abc
import ABCMeta
, abstractmethod
178 from collections
.abc
import Sequence
179 from queue
import Queue
182 class RecordObject(Record
):
183 def __init__(self
, layout
=None, name
=None):
184 Record
.__init
__(self
, layout
=layout
or [], name
=None)
186 def __setattr__(self
, k
, v
):
187 if k
in dir(Record
) or "fields" not in self
.__dict
__:
188 return object.__setattr
__(self
, k
, v
)
190 if isinstance(v
, Record
):
191 newlayout
= {k
: (k
, v
.layout
)}
193 newlayout
= {k
: (k
, v
.shape())}
194 self
.layout
.fields
.update(newlayout
)
197 for x
in self
.fields
.values():
202 """ contains signals that come *from* the previous stage (both in and out)
203 * i_valid: previous stage indicating all incoming data is valid.
204 may be a multi-bit signal, where all bits are required
205 to be asserted to indicate "valid".
206 * o_ready: output to next stage indicating readiness to accept data
207 * i_data : an input - added by the user of this class
210 def __init__(self
, i_width
=1, stage_ctl
=False):
211 self
.stage_ctl
= stage_ctl
212 self
.i_valid
= Signal(i_width
, name
="p_i_valid") # prev >>in self
213 self
._o
_ready
= Signal(name
="p_o_ready") # prev <<out self
214 self
.i_data
= None # XXX MUST BE ADDED BY USER
216 self
.s_o_ready
= Signal(name
="p_s_o_rdy") # prev <<out self
217 self
.trigger
= Signal(reset_less
=True)
221 """ public-facing API: indicates (externally) that stage is ready
224 return self
.s_o_ready
# set dynamically by stage
225 return self
._o
_ready
# return this when not under dynamic control
227 def _connect_in(self
, prev
, direct
=False, fn
=None):
228 """ internal helper function to connect stage to an input source.
229 do not use to connect stage-to-stage!
231 i_valid
= prev
.i_valid
if direct
else prev
.i_valid_test
232 i_data
= fn(prev
.i_data
) if fn
is not None else prev
.i_data
233 return [self
.i_valid
.eq(i_valid
),
234 prev
.o_ready
.eq(self
.o_ready
),
235 eq(self
.i_data
, i_data
),
239 def i_valid_test(self
):
240 vlen
= len(self
.i_valid
)
242 # multi-bit case: valid only when i_valid is all 1s
243 all1s
= Const(-1, (len(self
.i_valid
), False))
244 i_valid
= (self
.i_valid
== all1s
)
246 # single-bit i_valid case
247 i_valid
= self
.i_valid
249 # when stage indicates not ready, incoming data
250 # must "appear" to be not ready too
252 i_valid
= i_valid
& self
.s_o_ready
256 def elaborate(self
, platform
):
258 m
.d
.comb
+= self
.trigger
.eq(self
.i_valid_test
& self
.o_ready
)
262 return [self
.i_data
.eq(i
.i_data
),
263 self
.o_ready
.eq(i
.o_ready
),
264 self
.i_valid
.eq(i
.i_valid
)]
267 res
= [self
.i_valid
, self
.o_ready
]
268 if hasattr(self
.i_data
, "ports"):
269 res
+= self
.i_data
.ports()
270 elif isinstance(self
.i_data
, Sequence
):
273 res
.append(self
.i_data
)
278 """ contains the signals that go *to* the next stage (both in and out)
279 * o_valid: output indicating to next stage that data is valid
280 * i_ready: input from next stage indicating that it can accept data
281 * o_data : an output - added by the user of this class
283 def __init__(self
, stage_ctl
=False):
284 self
.stage_ctl
= stage_ctl
285 self
.o_valid
= Signal(name
="n_o_valid") # self out>> next
286 self
.i_ready
= Signal(name
="n_i_ready") # self <<in next
287 self
.o_data
= None # XXX MUST BE ADDED BY USER
289 self
.d_valid
= Signal(reset
=1) # INTERNAL (data valid)
290 self
.trigger
= Signal(reset_less
=True)
293 def i_ready_test(self
):
295 return self
.i_ready
& self
.d_valid
298 def connect_to_next(self
, nxt
):
299 """ helper function to connect to the next stage data/valid/ready.
300 data/valid is passed *TO* nxt, and ready comes *IN* from nxt.
301 use this when connecting stage-to-stage
303 return [nxt
.i_valid
.eq(self
.o_valid
),
304 self
.i_ready
.eq(nxt
.o_ready
),
305 eq(nxt
.i_data
, self
.o_data
),
308 def _connect_out(self
, nxt
, direct
=False, fn
=None):
309 """ internal helper function to connect stage to an output source.
310 do not use to connect stage-to-stage!
312 i_ready
= nxt
.i_ready
if direct
else nxt
.i_ready_test
313 o_data
= fn(nxt
.o_data
) if fn
is not None else nxt
.o_data
314 return [nxt
.o_valid
.eq(self
.o_valid
),
315 self
.i_ready
.eq(i_ready
),
316 eq(o_data
, self
.o_data
),
319 def elaborate(self
, platform
):
321 m
.d
.comb
+= self
.trigger
.eq(self
.i_ready_test
& self
.o_valid
)
325 res
= [self
.i_ready
, self
.o_valid
]
326 if hasattr(self
.o_data
, "ports"):
327 res
+= self
.o_data
.ports()
328 elif isinstance(self
.o_data
, Sequence
):
331 res
.append(self
.o_data
)
336 """ a helper class for iterating twin-argument compound data structures.
338 Record is a special (unusual, recursive) case, where the input may be
339 specified as a dictionary (which may contain further dictionaries,
340 recursively), where the field names of the dictionary must match
341 the Record's field spec. Alternatively, an object with the same
342 member names as the Record may be assigned: it does not have to
345 ArrayProxy is also special-cased, it's a bit messy: whilst ArrayProxy
346 has an eq function, the object being assigned to it (e.g. a python
347 object) might not. despite the *input* having an eq function,
348 that doesn't help us, because it's the *ArrayProxy* that's being
349 assigned to. so.... we cheat. use the ports() function of the
350 python object, enumerate them, find out the list of Signals that way,
353 def iterator2(self
, o
, i
):
354 if isinstance(o
, dict):
355 yield from self
.dict_iter2(o
, i
)
357 if not isinstance(o
, Sequence
):
359 for (ao
, ai
) in zip(o
, i
):
360 #print ("visit", fn, ao, ai)
361 if isinstance(ao
, Record
):
362 yield from self
.record_iter2(ao
, ai
)
363 elif isinstance(ao
, ArrayProxy
) and not isinstance(ai
, Value
):
364 yield from self
.arrayproxy_iter2(ao
, ai
)
368 def dict_iter2(self
, o
, i
):
369 for (k
, v
) in o
.items():
370 print ("d-iter", v
, i
[k
])
374 def _not_quite_working_with_all_unit_tests_record_iter2(self
, ao
, ai
):
375 print ("record_iter2", ao
, ai
, type(ao
), type(ai
))
376 if isinstance(ai
, Value
):
377 if isinstance(ao
, Sequence
):
379 for o
, i
in zip(ao
, ai
):
382 for idx
, (field_name
, field_shape
, _
) in enumerate(ao
.layout
):
383 if isinstance(field_shape
, Layout
):
387 if hasattr(val
, field_name
): # check for attribute
388 val
= getattr(val
, field_name
)
390 val
= val
[field_name
] # dictionary-style specification
391 yield from self
.iterator2(ao
.fields
[field_name
], val
)
393 def record_iter2(self
, ao
, ai
):
394 for idx
, (field_name
, field_shape
, _
) in enumerate(ao
.layout
):
395 if isinstance(field_shape
, Layout
):
399 if hasattr(val
, field_name
): # check for attribute
400 val
= getattr(val
, field_name
)
402 val
= val
[field_name
] # dictionary-style specification
403 yield from self
.iterator2(ao
.fields
[field_name
], val
)
405 def arrayproxy_iter2(self
, ao
, ai
):
407 op
= getattr(ao
, p
.name
)
408 print ("arrayproxy - p", p
, p
.name
)
409 yield from self
.iterator2(op
, p
)
413 """ a helper class for iterating single-argument compound data structures.
416 def iterate(self
, i
):
417 """ iterate a compound structure recursively using yield
419 if not isinstance(i
, Sequence
):
422 print ("iterate", ai
)
423 if isinstance(ai
, Record
):
424 print ("record", list(ai
.layout
))
425 yield from self
.record_iter(ai
)
426 elif isinstance(ai
, ArrayProxy
) and not isinstance(ai
, Value
):
427 yield from self
.array_iter(ai
)
431 def record_iter(self
, ai
):
432 for idx
, (field_name
, field_shape
, _
) in enumerate(ai
.layout
):
433 if isinstance(field_shape
, Layout
):
437 if hasattr(val
, field_name
): # check for attribute
438 val
= getattr(val
, field_name
)
440 val
= val
[field_name
] # dictionary-style specification
441 print ("recidx", idx
, field_name
, field_shape
, val
)
442 yield from self
.iterate(val
)
444 def array_iter(self
, ai
):
446 yield from self
.iterate(p
)
450 """ makes signals equal: a helper routine which identifies if it is being
451 passed a list (or tuple) of objects, or signals, or Records, and calls
452 the objects' eq function.
455 for (ao
, ai
) in Visitor2().iterator2(o
, i
):
457 if not isinstance(rres
, Sequence
):
464 """ flattens a compound structure recursively using Cat
466 from nmigen
.tools
import flatten
467 #res = list(flatten(i)) # works (as of nmigen commit f22106e5) HOWEVER...
468 res
= list(Visitor().iterate(i
)) # needed because input may be a sequence
472 class StageCls(metaclass
=ABCMeta
):
473 """ Class-based "Stage" API. requires instantiation (after derivation)
475 see "Stage API" above.. Note: python does *not* require derivation
476 from this class. All that is required is that the pipelines *have*
477 the functions listed in this class. Derivation from this class
478 is therefore merely a "courtesy" to maintainers.
481 def ispec(self
): pass # REQUIRED
483 def ospec(self
): pass # REQUIRED
485 #def setup(self, m, i): pass # OPTIONAL
487 def process(self
, i
): pass # REQUIRED
490 class Stage(metaclass
=ABCMeta
):
491 """ Static "Stage" API. does not require instantiation (after derivation)
493 see "Stage API" above. Note: python does *not* require derivation
494 from this class. All that is required is that the pipelines *have*
495 the functions listed in this class. Derivation from this class
496 is therefore merely a "courtesy" to maintainers.
508 #def setup(m, i): pass
515 class RecordBasedStage(Stage
):
516 """ convenience class which provides a Records-based layout.
517 honestly it's a lot easier just to create a direct Records-based
518 class (see ExampleAddRecordStage)
520 def __init__(self
, in_shape
, out_shape
, processfn
, setupfn
=None):
521 self
.in_shape
= in_shape
522 self
.out_shape
= out_shape
523 self
.__process
= processfn
524 self
.__setup
= setupfn
525 def ispec(self
): return Record(self
.in_shape
)
526 def ospec(self
): return Record(self
.out_shape
)
527 def process(seif
, i
): return self
.__process
(i
)
528 def setup(seif
, m
, i
): return self
.__setup
(m
, i
)
531 class StageChain(StageCls
):
532 """ pass in a list of stages, and they will automatically be
533 chained together via their input and output specs into a
536 the end result basically conforms to the exact same Stage API.
538 * input to this class will be the input of the first stage
539 * output of first stage goes into input of second
540 * output of second goes into input into third (etc. etc.)
541 * the output of this class will be the output of the last stage
543 def __init__(self
, chain
, specallocate
=False):
545 self
.specallocate
= specallocate
548 return self
.chain
[0].ispec()
551 return self
.chain
[-1].ospec()
553 def _specallocate_setup(self
, m
, i
):
554 for (idx
, c
) in enumerate(self
.chain
):
555 if hasattr(c
, "setup"):
556 c
.setup(m
, i
) # stage may have some module stuff
557 o
= self
.chain
[idx
].ospec() # last assignment survives
558 m
.d
.comb
+= eq(o
, c
.process(i
)) # process input into "o"
559 if idx
== len(self
.chain
)-1:
561 i
= self
.chain
[idx
+1].ispec() # new input on next loop
562 m
.d
.comb
+= eq(i
, o
) # assign to next input
563 return o
# last loop is the output
565 def _noallocate_setup(self
, m
, i
):
566 for (idx
, c
) in enumerate(self
.chain
):
567 if hasattr(c
, "setup"):
568 c
.setup(m
, i
) # stage may have some module stuff
569 i
= o
= c
.process(i
) # store input into "o"
570 return o
# last loop is the output
572 def setup(self
, m
, i
):
573 if self
.specallocate
:
574 self
.o
= self
._specallocate
_setup
(m
, i
)
576 self
.o
= self
._noallocate
_setup
(m
, i
)
578 def process(self
, i
):
579 return self
.o
# conform to Stage API: return last-loop output
583 """ Common functions for Pipeline API
585 def __init__(self
, stage
=None, in_multi
=None, stage_ctl
=False):
586 """ Base class containing ready/valid/data to previous and next stages
588 * p: contains ready/valid to the previous stage
589 * n: contains ready/valid to the next stage
591 Except when calling Controlbase.connect(), user must also:
592 * add i_data member to PrevControl (p) and
593 * add o_data member to NextControl (n)
597 # set up input and output IO ACK (prev/next ready/valid)
598 self
.p
= PrevControl(in_multi
, stage_ctl
)
599 self
.n
= NextControl(stage_ctl
)
601 # set up the input and output data
602 if stage
is not None:
603 self
.p
.i_data
= stage
.ispec() # input type
604 self
.n
.o_data
= stage
.ospec()
606 def connect_to_next(self
, nxt
):
607 """ helper function to connect to the next stage data/valid/ready.
609 return self
.n
.connect_to_next(nxt
.p
)
611 def _connect_in(self
, prev
):
612 """ internal helper function to connect stage to an input source.
613 do not use to connect stage-to-stage!
615 return self
.p
._connect
_in
(prev
.p
)
617 def _connect_out(self
, nxt
):
618 """ internal helper function to connect stage to an output source.
619 do not use to connect stage-to-stage!
621 return self
.n
._connect
_out
(nxt
.n
)
623 def connect(self
, pipechain
):
624 """ connects a chain (list) of Pipeline instances together and
625 links them to this ControlBase instance:
627 in <----> self <---> out
630 [pipe1, pipe2, pipe3, pipe4]
633 out---in out--in out---in
635 Also takes care of allocating i_data/o_data, by looking up
636 the data spec for each end of the pipechain. i.e It is NOT
637 necessary to allocate self.p.i_data or self.n.o_data manually:
638 this is handled AUTOMATICALLY, here.
640 Basically this function is the direct equivalent of StageChain,
641 except that unlike StageChain, the Pipeline logic is followed.
643 Just as StageChain presents an object that conforms to the
644 Stage API from a list of objects that also conform to the
645 Stage API, an object that calls this Pipeline connect function
646 has the exact same pipeline API as the list of pipline objects
649 Thus it becomes possible to build up larger chains recursively.
650 More complex chains (multi-input, multi-output) will have to be
653 eqs
= [] # collated list of assignment statements
655 # connect inter-chain
656 for i
in range(len(pipechain
)-1):
658 pipe2
= pipechain
[i
+1]
659 eqs
+= pipe1
.connect_to_next(pipe2
)
661 # connect front of chain to ourselves
663 self
.p
.i_data
= front
.stage
.ispec()
664 eqs
+= front
._connect
_in
(self
)
666 # connect end of chain to ourselves
668 self
.n
.o_data
= end
.stage
.ospec()
669 eqs
+= end
._connect
_out
(self
)
673 def _postprocess(self
, i
): # XXX DISABLED
674 return i
# RETURNS INPUT
675 if hasattr(self
.stage
, "postprocess"):
676 return self
.stage
.postprocess(i
)
679 def set_input(self
, i
):
680 """ helper function to set the input data
682 return eq(self
.p
.i_data
, i
)
685 return self
.p
.ports() + self
.n
.ports()
687 def _elaborate(self
, platform
):
688 """ handles case where stage has dynamic ready/valid functions
692 if self
.stage
is not None and hasattr(self
.stage
, "setup"):
693 self
.stage
.setup(m
, self
.p
.i_data
)
695 if not self
.p
.stage_ctl
:
698 # intercept the previous (outgoing) "ready", combine with stage ready
699 m
.d
.comb
+= self
.p
.s_o_ready
.eq(self
.p
._o
_ready
& self
.stage
.d_ready
)
701 # intercept the next (incoming) "ready" and combine it with data valid
702 sdv
= self
.stage
.d_valid(self
.n
.i_ready
)
703 m
.d
.comb
+= self
.n
.d_valid
.eq(self
.n
.i_ready
& sdv
)
708 class BufferedHandshake(ControlBase
):
709 """ buffered pipeline stage. data and strobe signals travel in sync.
710 if ever the input is ready and the output is not, processed data
711 is shunted in a temporary register.
713 Argument: stage. see Stage API above
715 stage-1 p.i_valid >>in stage n.o_valid out>> stage+1
716 stage-1 p.o_ready <<out stage n.i_ready <<in stage+1
717 stage-1 p.i_data >>in stage n.o_data out>> stage+1
723 input data p.i_data is read (only), is processed and goes into an
724 intermediate result store [process()]. this is updated combinatorially.
726 in a non-stall condition, the intermediate result will go into the
727 output (update_output). however if ever there is a stall, it goes
728 into r_data instead [update_buffer()].
730 when the non-stall condition is released, r_data is the first
731 to be transferred to the output [flush_buffer()], and the stall
734 on the next cycle (as long as stall is not raised again) the
735 input may begin to be processed and transferred directly to output.
738 def elaborate(self
, platform
):
739 self
.m
= ControlBase
._elaborate
(self
, platform
)
741 result
= self
.stage
.ospec()
742 r_data
= self
.stage
.ospec()
744 # establish some combinatorial temporaries
745 o_n_validn
= Signal(reset_less
=True)
746 n_i_ready
= Signal(reset_less
=True, name
="n_i_rdy_data")
747 nir_por
= Signal(reset_less
=True)
748 nir_por_n
= Signal(reset_less
=True)
749 p_i_valid
= Signal(reset_less
=True)
750 nir_novn
= Signal(reset_less
=True)
751 nirn_novn
= Signal(reset_less
=True)
752 por_pivn
= Signal(reset_less
=True)
753 npnn
= Signal(reset_less
=True)
754 self
.m
.d
.comb
+= [p_i_valid
.eq(self
.p
.i_valid_test
),
755 o_n_validn
.eq(~self
.n
.o_valid
),
756 n_i_ready
.eq(self
.n
.i_ready_test
),
757 nir_por
.eq(n_i_ready
& self
.p
._o
_ready
),
758 nir_por_n
.eq(n_i_ready
& ~self
.p
._o
_ready
),
759 nir_novn
.eq(n_i_ready | o_n_validn
),
760 nirn_novn
.eq(~n_i_ready
& o_n_validn
),
761 npnn
.eq(nir_por | nirn_novn
),
762 por_pivn
.eq(self
.p
._o
_ready
& ~p_i_valid
)
765 # store result of processing in combinatorial temporary
766 self
.m
.d
.comb
+= eq(result
, self
.stage
.process(self
.p
.i_data
))
768 # if not in stall condition, update the temporary register
769 with self
.m
.If(self
.p
.o_ready
): # not stalled
770 self
.m
.d
.sync
+= eq(r_data
, result
) # update buffer
772 # data pass-through conditions
773 with self
.m
.If(npnn
):
774 o_data
= self
._postprocess
(result
)
775 self
.m
.d
.sync
+= [self
.n
.o_valid
.eq(p_i_valid
), # valid if p_valid
776 eq(self
.n
.o_data
, o_data
), # update output
778 # buffer flush conditions (NOTE: can override data passthru conditions)
779 with self
.m
.If(nir_por_n
): # not stalled
780 # Flush the [already processed] buffer to the output port.
781 o_data
= self
._postprocess
(r_data
)
782 self
.m
.d
.sync
+= [self
.n
.o_valid
.eq(1), # reg empty
783 eq(self
.n
.o_data
, o_data
), # flush buffer
785 # output ready conditions
786 self
.m
.d
.sync
+= self
.p
._o
_ready
.eq(nir_novn | por_pivn
)
791 class SimpleHandshake(ControlBase
):
792 """ simple handshake control. data and strobe signals travel in sync.
793 implements the protocol used by Wishbone and AXI4.
795 Argument: stage. see Stage API above
797 stage-1 p.i_valid >>in stage n.o_valid out>> stage+1
798 stage-1 p.o_ready <<out stage n.i_ready <<in stage+1
799 stage-1 p.i_data >>in stage n.o_data out>> stage+1
804 Inputs Temporary Output Data
805 ------- ---------- ----- ----
806 P P N N PiV& ~NiR& N P
813 0 0 1 0 0 0 0 1 process(i_data)
814 0 0 1 1 0 0 0 1 process(i_data)
818 0 1 1 0 0 0 0 1 process(i_data)
819 0 1 1 1 0 0 0 1 process(i_data)
823 1 0 1 0 0 0 0 1 process(i_data)
824 1 0 1 1 0 0 0 1 process(i_data)
826 1 1 0 0 1 0 1 0 process(i_data)
827 1 1 0 1 1 1 1 0 process(i_data)
828 1 1 1 0 1 0 1 1 process(i_data)
829 1 1 1 1 1 0 1 1 process(i_data)
833 def elaborate(self
, platform
):
834 self
.m
= m
= ControlBase
._elaborate
(self
, platform
)
837 result
= self
.stage
.ospec()
839 # establish some combinatorial temporaries
840 n_i_ready
= Signal(reset_less
=True, name
="n_i_rdy_data")
841 p_i_valid_p_o_ready
= Signal(reset_less
=True)
842 p_i_valid
= Signal(reset_less
=True)
843 m
.d
.comb
+= [p_i_valid
.eq(self
.p
.i_valid_test
),
844 n_i_ready
.eq(self
.n
.i_ready_test
),
845 p_i_valid_p_o_ready
.eq(p_i_valid
& self
.p
.o_ready
),
848 # store result of processing in combinatorial temporary
849 m
.d
.comb
+= eq(result
, self
.stage
.process(self
.p
.i_data
))
851 # previous valid and ready
852 with m
.If(p_i_valid_p_o_ready
):
853 o_data
= self
._postprocess
(result
)
854 m
.d
.sync
+= [r_busy
.eq(1), # output valid
855 eq(self
.n
.o_data
, o_data
), # update output
857 # previous invalid or not ready, however next is accepting
858 with m
.Elif(n_i_ready
):
859 o_data
= self
._postprocess
(result
)
860 m
.d
.sync
+= [eq(self
.n
.o_data
, o_data
)]
861 # TODO: could still send data here (if there was any)
862 #m.d.sync += self.n.o_valid.eq(0) # ...so set output invalid
863 m
.d
.sync
+= r_busy
.eq(0) # ...so set output invalid
865 m
.d
.comb
+= self
.n
.o_valid
.eq(r_busy
)
866 # if next is ready, so is previous
867 m
.d
.comb
+= self
.p
._o
_ready
.eq(n_i_ready
)
872 class UnbufferedPipeline(ControlBase
):
873 """ A simple pipeline stage with single-clock synchronisation
874 and two-way valid/ready synchronised signalling.
876 Note that a stall in one stage will result in the entire pipeline
879 Also that unlike BufferedHandshake, the valid/ready signalling does NOT
880 travel synchronously with the data: the valid/ready signalling
881 combines in a *combinatorial* fashion. Therefore, a long pipeline
882 chain will lengthen propagation delays.
884 Argument: stage. see Stage API, above
886 stage-1 p.i_valid >>in stage n.o_valid out>> stage+1
887 stage-1 p.o_ready <<out stage n.i_ready <<in stage+1
888 stage-1 p.i_data >>in stage n.o_data out>> stage+1
896 p.i_data : StageInput, shaped according to ispec
898 p.o_data : StageOutput, shaped according to ospec
900 r_data : input_shape according to ispec
901 A temporary (buffered) copy of a prior (valid) input.
902 This is HELD if the output is not ready. It is updated
904 result: output_shape according to ospec
905 The output of the combinatorial logic. it is updated
906 COMBINATORIALLY (no clock dependence).
910 Inputs Temp Output Data
932 1 1 0 0 0 1 1 process(i_data)
933 1 1 0 1 1 1 0 process(i_data)
934 1 1 1 0 0 1 1 process(i_data)
935 1 1 1 1 0 1 1 process(i_data)
938 Note: PoR is *NOT* involved in the above decision-making.
941 def elaborate(self
, platform
):
942 self
.m
= m
= ControlBase
._elaborate
(self
, platform
)
944 data_valid
= Signal() # is data valid or not
945 r_data
= self
.stage
.ospec() # output type
948 p_i_valid
= Signal(reset_less
=True)
949 pv
= Signal(reset_less
=True)
950 buf_full
= Signal(reset_less
=True)
951 m
.d
.comb
+= p_i_valid
.eq(self
.p
.i_valid_test
)
952 m
.d
.comb
+= pv
.eq(self
.p
.i_valid
& self
.p
.o_ready
)
953 m
.d
.comb
+= buf_full
.eq(~self
.n
.i_ready_test
& data_valid
)
955 m
.d
.comb
+= self
.n
.o_valid
.eq(data_valid
)
956 m
.d
.comb
+= self
.p
._o
_ready
.eq(~data_valid | self
.n
.i_ready_test
)
957 m
.d
.sync
+= data_valid
.eq(p_i_valid | buf_full
)
960 m
.d
.sync
+= eq(r_data
, self
.stage
.process(self
.p
.i_data
))
961 o_data
= self
._postprocess
(r_data
)
962 m
.d
.comb
+= eq(self
.n
.o_data
, o_data
)
966 class UnbufferedPipeline2(ControlBase
):
967 """ A simple pipeline stage with single-clock synchronisation
968 and two-way valid/ready synchronised signalling.
970 Note that a stall in one stage will result in the entire pipeline
973 Also that unlike BufferedHandshake, the valid/ready signalling does NOT
974 travel synchronously with the data: the valid/ready signalling
975 combines in a *combinatorial* fashion. Therefore, a long pipeline
976 chain will lengthen propagation delays.
978 Argument: stage. see Stage API, above
980 stage-1 p.i_valid >>in stage n.o_valid out>> stage+1
981 stage-1 p.o_ready <<out stage n.i_ready <<in stage+1
982 stage-1 p.i_data >>in stage n.o_data out>> stage+1
987 p.i_data : StageInput, shaped according to ispec
989 p.o_data : StageOutput, shaped according to ospec
991 buf : output_shape according to ospec
992 A temporary (buffered) copy of a valid output
993 This is HELD if the output is not ready. It is updated
996 Inputs Temp Output Data
998 P P N N ~NiR& N P (buf_full)
1003 0 0 0 0 0 0 1 process(i_data)
1004 0 0 0 1 1 1 0 reg (odata, unchanged)
1005 0 0 1 0 0 0 1 process(i_data)
1006 0 0 1 1 0 0 1 process(i_data)
1008 0 1 0 0 0 0 1 process(i_data)
1009 0 1 0 1 1 1 0 reg (odata, unchanged)
1010 0 1 1 0 0 0 1 process(i_data)
1011 0 1 1 1 0 0 1 process(i_data)
1013 1 0 0 0 0 1 1 process(i_data)
1014 1 0 0 1 1 1 0 reg (odata, unchanged)
1015 1 0 1 0 0 1 1 process(i_data)
1016 1 0 1 1 0 1 1 process(i_data)
1018 1 1 0 0 0 1 1 process(i_data)
1019 1 1 0 1 1 1 0 reg (odata, unchanged)
1020 1 1 1 0 0 1 1 process(i_data)
1021 1 1 1 1 0 1 1 process(i_data)
1024 Note: PoR is *NOT* involved in the above decision-making.
1027 def elaborate(self
, platform
):
1028 self
.m
= m
= ControlBase
._elaborate
(self
, platform
)
1030 buf_full
= Signal() # is data valid or not
1031 buf
= self
.stage
.ospec() # output type
1034 p_i_valid
= Signal(reset_less
=True)
1035 m
.d
.comb
+= p_i_valid
.eq(self
.p
.i_valid_test
)
1037 m
.d
.comb
+= self
.n
.o_valid
.eq(buf_full | p_i_valid
)
1038 m
.d
.comb
+= self
.p
._o
_ready
.eq(~buf_full
)
1039 m
.d
.sync
+= buf_full
.eq(~self
.n
.i_ready_test
& self
.n
.o_valid
)
1041 o_data
= Mux(buf_full
, buf
, self
.stage
.process(self
.p
.i_data
))
1042 o_data
= self
._postprocess
(o_data
)
1043 m
.d
.comb
+= eq(self
.n
.o_data
, o_data
)
1044 m
.d
.sync
+= eq(buf
, self
.n
.o_data
)
1049 class PassThroughStage(StageCls
):
1050 """ a pass-through stage which has its input data spec equal to its output,
1051 and "passes through" its data from input to output.
1053 def __init__(self
, iospecfn
):
1054 self
.iospecfn
= iospecfn
1055 def ispec(self
): return self
.iospecfn()
1056 def ospec(self
): return self
.iospecfn()
1057 def process(self
, i
): return i
1060 class PassThroughHandshake(ControlBase
):
1061 """ A control block that delays by one clock cycle.
1063 Inputs Temporary Output Data
1064 ------- ------------------ ----- ----
1065 P P N N PiV& PiV| NiR| pvr N P (pvr)
1066 i o i o PoR ~PoR ~NoV o o
1070 0 0 0 0 0 1 1 0 1 1 odata (unchanged)
1071 0 0 0 1 0 1 0 0 1 0 odata (unchanged)
1072 0 0 1 0 0 1 1 0 1 1 odata (unchanged)
1073 0 0 1 1 0 1 1 0 1 1 odata (unchanged)
1075 0 1 0 0 0 0 1 0 0 1 odata (unchanged)
1076 0 1 0 1 0 0 0 0 0 0 odata (unchanged)
1077 0 1 1 0 0 0 1 0 0 1 odata (unchanged)
1078 0 1 1 1 0 0 1 0 0 1 odata (unchanged)
1080 1 0 0 0 0 1 1 1 1 1 process(in)
1081 1 0 0 1 0 1 0 0 1 0 odata (unchanged)
1082 1 0 1 0 0 1 1 1 1 1 process(in)
1083 1 0 1 1 0 1 1 1 1 1 process(in)
1085 1 1 0 0 1 1 1 1 1 1 process(in)
1086 1 1 0 1 1 1 0 0 1 0 odata (unchanged)
1087 1 1 1 0 1 1 1 1 1 1 process(in)
1088 1 1 1 1 1 1 1 1 1 1 process(in)
1093 def elaborate(self
, platform
):
1094 self
.m
= m
= ControlBase
._elaborate
(self
, platform
)
1096 r_data
= self
.stage
.ospec() # output type
1099 p_i_valid
= Signal(reset_less
=True)
1100 pvr
= Signal(reset_less
=True)
1101 m
.d
.comb
+= p_i_valid
.eq(self
.p
.i_valid_test
)
1102 m
.d
.comb
+= pvr
.eq(p_i_valid
& self
.p
.o_ready
)
1104 m
.d
.comb
+= self
.p
.o_ready
.eq(~self
.n
.o_valid | self
.n
.i_ready_test
)
1105 m
.d
.sync
+= self
.n
.o_valid
.eq(p_i_valid | ~self
.p
.o_ready
)
1107 odata
= Mux(pvr
, self
.stage
.process(self
.p
.i_data
), r_data
)
1108 m
.d
.sync
+= eq(r_data
, odata
)
1109 r_data
= self
._postprocess
(r_data
)
1110 m
.d
.comb
+= eq(self
.n
.o_data
, r_data
)
1115 class RegisterPipeline(UnbufferedPipeline
):
1116 """ A pipeline stage that delays by one clock cycle, creating a
1117 sync'd latch out of o_data and o_valid as an indirect byproduct
1118 of using PassThroughStage
1120 def __init__(self
, iospecfn
):
1121 UnbufferedPipeline
.__init
__(self
, PassThroughStage(iospecfn
))
1124 class FIFOControl(ControlBase
):
1125 """ FIFO Control. Uses SyncFIFO to store data, coincidentally
1126 happens to have same valid/ready signalling as Stage API.
1128 i_data -> fifo.din -> FIFO -> fifo.dout -> o_data
1131 def __init__(self
, depth
, stage
, in_multi
=None, stage_ctl
=False,
1132 fwft
=True, buffered
=False, pipe
=False):
1135 * depth: number of entries in the FIFO
1136 * stage: data processing block
1137 * fwft : first word fall-thru mode (non-fwft introduces delay)
1138 * buffered: use buffered FIFO (introduces extra cycle delay)
1140 NOTE 1: FPGAs may have trouble with the defaults for SyncFIFO
1141 (fwft=True, buffered=False)
1143 NOTE 2: i_data *must* have a shape function. it can therefore
1144 be a Signal, or a Record, or a RecordObject.
1146 data is processed (and located) as follows:
1148 self.p self.stage temp fn temp fn temp fp self.n
1149 i_data->process()->result->cat->din.FIFO.dout->cat(o_data)
1151 yes, really: cat produces a Cat() which can be assigned to.
1152 this is how the FIFO gets de-catted without needing a de-cat
1156 assert not (fwft
and buffered
), "buffered cannot do fwft"
1160 self
.buffered
= buffered
1163 ControlBase
.__init
__(self
, stage
, in_multi
, stage_ctl
)
1165 def elaborate(self
, platform
):
1166 self
.m
= m
= ControlBase
._elaborate
(self
, platform
)
1168 # make a FIFO with a signal of equal width to the o_data.
1169 (fwidth
, _
) = self
.n
.o_data
.shape()
1171 fifo
= SyncFIFOBuffered(fwidth
, self
.fdepth
)
1173 fifo
= Queue(fwidth
, self
.fdepth
, fwft
=self
.fwft
, pipe
=self
.pipe
)
1174 m
.submodules
.fifo
= fifo
1176 # store result of processing in combinatorial temporary
1177 result
= self
.stage
.ospec()
1178 m
.d
.comb
+= eq(result
, self
.stage
.process(self
.p
.i_data
))
1180 # connect previous rdy/valid/data - do cat on i_data
1181 # NOTE: cannot do the PrevControl-looking trick because
1182 # of need to process the data. shaaaame....
1183 m
.d
.comb
+= [fifo
.we
.eq(self
.p
.i_valid_test
),
1184 self
.p
.o_ready
.eq(fifo
.writable
),
1185 eq(fifo
.din
, cat(result
)),
1188 # connect next rdy/valid/data - do cat on o_data
1189 connections
= [self
.n
.o_valid
.eq(fifo
.readable
),
1190 fifo
.re
.eq(self
.n
.i_ready_test
),
1192 if self
.fwft
or self
.buffered
:
1193 m
.d
.comb
+= connections
1195 m
.d
.sync
+= connections
# unbuffered fwft mode needs sync
1196 o_data
= cat(self
.n
.o_data
).eq(fifo
.dout
)
1197 o_data
= self
._postprocess
(o_data
)
1204 class UnbufferedPipeline(FIFOControl
):
1205 def __init__(self
, stage
, in_multi
=None, stage_ctl
=False):
1206 FIFOControl
.__init
__(self
, 1, stage
, in_multi
, stage_ctl
,
1207 fwft
=True, pipe
=False)
1209 # aka "BreakReadyStage" XXX had to set fwft=True to get it to work
1210 class PassThroughHandshake(FIFOControl
):
1211 def __init__(self
, stage
, in_multi
=None, stage_ctl
=False):
1212 FIFOControl
.__init
__(self
, 1, stage
, in_multi
, stage_ctl
,
1213 fwft
=True, pipe
=True)
1215 # this is *probably* BufferedHandshake, although test #997 now succeeds.
1216 class BufferedHandshake(FIFOControl
):
1217 def __init__(self
, stage
, in_multi
=None, stage_ctl
=False):
1218 FIFOControl
.__init
__(self
, 2, stage
, in_multi
, stage_ctl
,
1219 fwft
=True, pipe
=False)
1223 # this is *probably* SimpleHandshake (note: memory cell size=0)
1224 class SimpleHandshake(FIFOControl):
1225 def __init__(self, stage, in_multi=None, stage_ctl=False):
1226 FIFOControl.__init__(self, 0, stage, in_multi, stage_ctl,
1227 fwft=True, pipe=False)