Hardware
/
litedram
mirror of https://github.com/enjoy-digital/litedram.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

phy/lpddr5: fix write latency

This commit is contained in:
Jędrzej Boczar 2021-07-12 13:14:51 +02:00 committed by Alessandro Comodi
parent e906c82ea2
commit a015b66e4f
2 changed files with 19 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
litedram/phy/lpddr5/basephy.py
View File

@ -177,7 +177,7 @@ class LPDDR5PHY(Module, AutoCSR):
read_latency = read_data_delay + read_des_delay
# Write latency
write_latency = cwl
write_latency = cwl + cmd_latency
# Registers --------------------------------------------------------------------------------
self._rst = CSRStorage()
@ -350,7 +350,7 @@ class LPDDR5PHY(Module, AutoCSR):
self.comb += self.out.wck[byte].eq(wck_out)
# Write Control Path -----------------------------------------------------------------------
wrtap = cwl - 1
wrtap = write_latency - 1
assert wrtap >= 0
wrdata_en = TappedDelayLine(

27
litedram/phy/lpddr5/sim.py
View File

@ -410,14 +410,14 @@ class CommandsSim(Module, AutoCSR):
),
# data latency
If(self.cmd_info.we,
self.data_latency.eq(self.mode_regs.wl - 2),
If(self.mode_regs.wl < 2,
self.log.error("WL < 2 is currently not supported")
self.data_latency.eq(self.mode_regs.wl - 1),
If(self.mode_regs.wl < 1,
self.log.error("WL < 2 is not supported")
),
).Else(
self.data_latency.eq(self.mode_regs.rl - 2),
If(self.mode_regs.rl < 2,
self.log.error("RL < 2 is currently not supported")
self.data_latency.eq(self.mode_regs.rl - 1),
If(self.mode_regs.rl < 1,
self.log.error("RL < 2 is not supported")
),
),
self.data_timer.trigger.eq(1),
@ -533,18 +533,25 @@ class DataSim(Module, AutoCSR):
self.submodules.burst_p = ClockDomainsRenamer("wck")(Burst("wck", dq_i_p))
self.submodules.burst_n = ClockDomainsRenamer("wck_n")(Burst("wck_n", dq_i_n))
def delay(sig, cycles):
if cycles == 0:
return sig
return delayed(self, sig, cycles=cycles)
# After the WL signal arives we require the data to arrive some time later and then we start
# reading it. This would be adjustable on hardware, but in simulation we rather must set this
# so that it matches the delay that PHY introduces.
wr_start = Signal()
rd_start = Signal()
wr_delay = 1
rd_delay = 0
self.comb += [
wr_start.eq(cmd.valid & cmd.we & latency_ready),
rd_start.eq(cmd.valid & ~cmd.we & latency_ready),
self.burst_p.enable_wr.eq(wr_start),
self.burst_p.enable_rd.eq(rd_start),
self.burst_n.enable_wr.eq(delayed(self, wr_start, cycles=1)),
self.burst_n.enable_rd.eq(delayed(self, rd_start, cycles=1)),
self.burst_p.enable_wr.eq(delay(wr_start, wr_delay)),
self.burst_p.enable_rd.eq(delay(rd_start, rd_delay)),
self.burst_n.enable_wr.eq(delay(wr_start, wr_delay + 1)),
self.burst_n.enable_rd.eq(delay(rd_start, rd_delay + 1)),
cmd.ready.eq(self.burst_p.ready),
]