forked from M-Labs/artiq
sawg: wrap limits init values
This commit is contained in:
parent
cccd01e81e
commit
bd1438d28e
|
@ -89,10 +89,12 @@ class Config(Module):
|
|||
self.limits = [[Signal((width, True)), Signal((width, True))]
|
||||
for i in range(2)]
|
||||
limit = 1 << width - 1
|
||||
self.limits[0][0].reset = -limit
|
||||
self.limits[0][1].reset = limit - 1
|
||||
self.limits[1][0].reset = int(-limit/cordic_gain)
|
||||
self.limits[1][1].reset = int((limit - 1)/cordic_gain)
|
||||
self.limits[0][0].reset = Constant(-limit, (width, True))
|
||||
self.limits[0][1].reset = Constant(limit - 1, (width, True))
|
||||
self.limits[1][0].reset = Constant(int(-limit/cordic_gain),
|
||||
(width, True))
|
||||
self.limits[1][1].reset = Constant(int((limit - 1)/cordic_gain),
|
||||
(width, True))
|
||||
# TODO make persistent, add read-out/notification/clear
|
||||
self.clipped = [Signal(2) for i in range(2)]
|
||||
self.i = Endpoint([("addr", bits_for(4 + 2*len(self.limits) - 1)),
|
||||
|
@ -139,7 +141,7 @@ class Channel(Module, SatAddMixin):
|
|||
coeff = halfgen4_cascade(parallelism, width=.4, order=8)
|
||||
hbf = [ParallelHBFUpsampler(coeff, width=width + 1) for i in range(2)]
|
||||
self.submodules.b = b = SplineParallelDUC(
|
||||
widths._replace(a=width + 1, f=widths.f - width), orders,
|
||||
widths._replace(a=len(hbf[0].o[0]), f=widths.f - width), orders,
|
||||
parallelism=parallelism)
|
||||
cfg = Config(width, b.gain)
|
||||
u = Spline(width=widths.a, order=orders.a)
|
||||
|
@ -178,6 +180,8 @@ class Channel(Module, SatAddMixin):
|
|||
b.ce.eq(cfg.ce),
|
||||
u.o.ack.eq(cfg.ce),
|
||||
Cat(b.clr, a1.clr, a2.clr).eq(cfg.clr),
|
||||
Cat(b.xi).eq(Cat(hbf[0].o)),
|
||||
Cat(b.yi).eq(Cat(hbf[1].o)),
|
||||
]
|
||||
self.sync += [
|
||||
hbf[0].i.eq(self.sat_add(a1.xo[0], a2.xo[0],
|
||||
|
@ -185,11 +189,6 @@ class Channel(Module, SatAddMixin):
|
|||
hbf[1].i.eq(self.sat_add(a1.yo[0], a2.yo[0],
|
||||
limits=cfg.limits[1], clipped=None)), # ignore Q clip data
|
||||
]
|
||||
for i in range(parallelism):
|
||||
self.comb += [
|
||||
b.xi[i].eq(hbf[0].o[i]),
|
||||
b.yi[i].eq(hbf[1].o[i]),
|
||||
]
|
||||
# wire up outputs and q_{i,o} exchange
|
||||
for o, x, y in zip(self.o, b.xo, self.y_in):
|
||||
self.sync += [
|
||||
|
|
Loading…
Reference in New Issue