Hardware/litedram
1
0
Fork 0
mirror of https://github.com/enjoy-digital/litedram.git synced 2025-01-04 09:52:25 -05:00
Code Issues Projects Releases Packages Wiki Activity

phy/s7ddrphy: use our own bitslip module in fabric

Browse source 
we could probably reduce added latency to 2 or 1 in the future.
This commit is contained in:
Florent Kermarrec 2018-10-18 13:37:16 +02:00
parent 20d767532d
commit 8d24163a86
1 changed files with 28 additions and 16 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

44
litedram/phy/s7ddrphy.py
View file

@ -6,6 +6,7 @@ import math
from collections import OrderedDict from collections import OrderedDict
from migen import * from migen import *
from migen.genlib.misc import BitSlip
from litex.soc.interconnect.csr import * from litex.soc.interconnect.csr import *
@ -41,9 +42,9 @@ def get_sys_phases(nphases, sys_latency, cas_latency):
cmd_phase = (dat_phase - 1)%nphases cmd_phase = (dat_phase - 1)%nphases
return cmd_phase, dat_phase return cmd_phase, dat_phase
class S7DDRPHY(Module, AutoCSR): class S7DDRPHY(Module, AutoCSR):
def __init__(self, pads, with_odelay, memtype="DDR3", nphases=4, sys_clk_freq=100e6, iodelay_clk_freq=200e6, def __init__(self, pads, with_odelay, memtype="DDR3", nphases=4, sys_clk_freq=100e6, iodelay_clk_freq=200e6):
additional_read_latency=0):
assert not (memtype == "DDR3" and nphases == 2) # FIXME: Needs BL8 support for nphases=2 assert not (memtype == "DDR3" and nphases == 2) # FIXME: Needs BL8 support for nphases=2
tck = 2/(2*nphases*sys_clk_freq) tck = 2/(2*nphases*sys_clk_freq)
addressbits = len(pads.a) addressbits = len(pads.a)
@ -52,7 +53,6 @@ class S7DDRPHY(Module, AutoCSR):
databits = len(pads.dq) databits = len(pads.dq)
nphases = nphases nphases = nphases
iodelay_tap_average = { iodelay_tap_average = {
200e6: 78e-12, 200e6: 78e-12,
300e6: 52e-12, 300e6: 52e-12,
@ -97,7 +97,7 @@ class S7DDRPHY(Module, AutoCSR):
wrcmdphase=wrcmdphase, wrcmdphase=wrcmdphase,
cl=cl, cl=cl,
cwl=cwl, cwl=cwl,
read_latency=2 + cl_sys_latency + 2 + additional_read_latency, read_latency=2 + cl_sys_latency + 2 + 3,
write_latency=cwl_sys_latency write_latency=cwl_sys_latency
) )
@ -318,25 +318,36 @@ class S7DDRPHY(Module, AutoCSR):
i_DDLY=dq_i_delayed, i_DDLY=dq_i_delayed,
i_CE1=1, i_CE1=1,
i_RST=ResetSignal() | (self._dly_sel.storage[i//8] & self._rdly_dq_bitslip_rst.re), i_RST=ResetSignal(),
i_CLK=ClockSignal(ddr_clk), i_CLKB=~ClockSignal(ddr_clk), i_CLKDIV=ClockSignal(), i_CLK=ClockSignal(ddr_clk), i_CLKB=~ClockSignal(ddr_clk), i_CLKDIV=ClockSignal(),
i_BITSLIP=self._dly_sel.storage[i//8] & self._rdly_dq_bitslip.re, i_BITSLIP=0,
o_Q8=dq_i_data[7], o_Q7=dq_i_data[6], o_Q8=dq_i_data[0], o_Q7=dq_i_data[1],
o_Q6=dq_i_data[5], o_Q5=dq_i_data[4], o_Q6=dq_i_data[2], o_Q5=dq_i_data[3],
o_Q4=dq_i_data[3], o_Q3=dq_i_data[2], o_Q4=dq_i_data[4], o_Q3=dq_i_data[5],
o_Q2=dq_i_data[1], o_Q1=dq_i_data[0] o_Q2=dq_i_data[6], o_Q1=dq_i_data[7]
) )
dq_bitslip = BitSlip(8)
self.comb += dq_bitslip.i.eq(dq_i_data)
self.sync += \
If(self._dly_sel.storage[i//8],
If(self._rdly_dq_bitslip_rst.re,
dq_bitslip.value.eq(0)
).Elif(self._rdly_dq_bitslip.re,
dq_bitslip.value.eq(dq_bitslip.value + 1)
)
)
self.submodules += dq_bitslip
if nphases == 2: if nphases == 2:
self.comb += [ self.comb += [
self.dfi.phases[0].rddata[i].eq(dq_i_data[3]), self.dfi.phases[0].rddata[databits+i].eq(dq_i_data[2]), self.dfi.phases[0].rddata[i].eq(dq_bitslip.o[7]), self.dfi.phases[0].rddata[databits+i].eq(dq_bitslip.o[5]),
self.dfi.phases[1].rddata[i].eq(dq_i_data[1]), self.dfi.phases[1].rddata[databits+i].eq(dq_i_data[0]) self.dfi.phases[1].rddata[i].eq(dq_bitslip.o[6]), self.dfi.phases[1].rddata[databits+i].eq(dq_bitslip.o[7])
] ]
else: else:
self.comb += [ self.comb += [
self.dfi.phases[0].rddata[i].eq(dq_i_data[7]), self.dfi.phases[0].rddata[databits+i].eq(dq_i_data[6]), self.dfi.phases[0].rddata[i].eq(dq_bitslip.o[0]), self.dfi.phases[0].rddata[databits+i].eq(dq_bitslip.o[1]),
self.dfi.phases[1].rddata[i].eq(dq_i_data[5]), self.dfi.phases[1].rddata[databits+i].eq(dq_i_data[4]), self.dfi.phases[1].rddata[i].eq(dq_bitslip.o[2]), self.dfi.phases[1].rddata[databits+i].eq(dq_bitslip.o[3]),
self.dfi.phases[2].rddata[i].eq(dq_i_data[3]), self.dfi.phases[2].rddata[databits+i].eq(dq_i_data[2]), self.dfi.phases[2].rddata[i].eq(dq_bitslip.o[4]), self.dfi.phases[2].rddata[databits+i].eq(dq_bitslip.o[5]),
self.dfi.phases[3].rddata[i].eq(dq_i_data[1]), self.dfi.phases[3].rddata[databits+i].eq(dq_i_data[0]) self.dfi.phases[3].rddata[i].eq(dq_bitslip.o[6]), self.dfi.phases[3].rddata[databits+i].eq(dq_bitslip.o[7])
] ]
if with_odelay: if with_odelay:
@ -378,6 +389,7 @@ class S7DDRPHY(Module, AutoCSR):
# 2 cycles through OSERDESE2 # 2 cycles through OSERDESE2
# cl_sys_latency cycles CAS # cl_sys_latency cycles CAS
# 2 cycles through ISERDESE2 # 2 cycles through ISERDESE2
# 3 cycles through Bitslip
rddata_en = self.dfi.phases[self.settings.rdphase].rddata_en rddata_en = self.dfi.phases[self.settings.rdphase].rddata_en
for i in range(self.settings.read_latency-1): for i in range(self.settings.read_latency-1):
n_rddata_en = Signal() n_rddata_en = Signal()