wip: phy: add qdr phy

litespi/clkgen.py

@@ -8,7 +8,29 @@
 
 from litex.gen import *
 
-from litex.build.io import SDROutput, DDROutput
+from litex.build.io import SDROutput, DDROutput, QDROutput
+
+class QDRLiteSPIClkGen(LiteXModule):
+    """SPI Clock generator
+
+    The ``QDRLiteSPIClkGen`` class provides a generic SPI clock generator.
+
+    The class can be combined with QDR PHY Core.
+
+    Parameters
+    ----------
+    pads : Object
+        SPI pads description.
+
+    Attributes
+    ----------
+    en : Signal(), in
+        Clock enable input, output clock will be generated if set to 1, 0 resets the core.
+    """
+    def __init__(self, pads, cd_fastclk):
+        self.en         = en         = Signal()
+
+        self.specials += QDROutput(i1=en, i2=0, i3=en, i4=0, o=pads.clk, fastclk=ClockSignal(cd_fastclk))
 
 class DDRLiteSPIClkGen(LiteXModule):
     """SPI Clock generator

litespi/phy/generic.py

@@ -12,6 +12,7 @@
 
 from litespi.phy.generic_sdr import LiteSPISDRPHYCore
 from litespi.phy.generic_ddr import LiteSPIDDRPHYCore
+from litespi.phy.generic_qdr import LiteSPIQDRPHYCore
 
 # LiteSPI PHY --------------------------------------------------------------------------------------
 
@@ -56,13 +57,14 @@ class LiteSPIPHY(LiteXModule):
         Flash CS signal from ``LiteSPIPHYCore``.
     """
 
-    def __init__(self, pads, flash, device="xc7", clock_domain="sys", default_divisor=9, cs_delay=10, rate="1:1", extra_latency=0):
-        assert rate in ["1:1", "1:2"]
+    def __init__(self, pads, flash, device="xc7", clock_domain="sys", default_divisor=9, cs_delay=10, rate="1:1", extra_latency=0, cd_fastclk="sys4x_ps"):
+        assert rate in ["1:1", "1:2", "1:4"]
         if rate == "1:1":
             phy = LiteSPISDRPHYCore(pads, flash, device, clock_domain, default_divisor, cs_delay)
         if rate == "1:2":
             phy = LiteSPIDDRPHYCore(pads, flash, cs_delay, extra_latency)
-
+        if rate == "1:4":
+            phy = LiteSPIQDRPHYCore(pads, flash, cs_delay, cd_fastclk, extra_latency)
         self.flash = flash
 
         self.source = phy.source

litespi/phy/generic_ddr.py

@@ -83,15 +83,15 @@ def __init__(self, pads, flash, cs_delay, extra_latency=0):
         data_bits = 32
 
         dq_o  = Array([Signal(len(pads.dq)) for _ in range(2)])
-        dq_i  = Array([Signal(len(pads.dq)) for _ in range(2)])
-        dq_oe = Array([Signal(len(pads.dq)) for _ in range(2)])
+        dq_i  = Signal(len(pads.dq))
+        dq_oe = Signal(len(pads.dq))
 
         for i in range(len(pads.dq)):
             self.specials += DDRTristate(
                 io  = pads.dq[i],
                 o1  =  dq_o[0][i],  o2 =  dq_o[1][i],
-                oe1 = dq_oe[0][i], oe2 = dq_oe[1][i],
-                i1  =  dq_i[0][i],  i2 =  dq_i[1][i]
+                oe1 =    dq_oe[i],  oe2 =   dq_oe[i],
+                i1  =     dq_i[i],  i2 =    Signal(),
             )
 
         # Data Shift Registers.
@@ -104,7 +104,6 @@ def __init__(self, pads, flash, cs_delay, extra_latency=0):
 
         # Data Out Shift.
         self.comb += [
-            dq_oe[1].eq(sink.mask),
             Case(sink.width, {
                 1:  dq_o[1].eq(sr_out[-1:]),
                 2:  dq_o[1].eq(sr_out[-2:]),
@@ -116,7 +115,7 @@ def __init__(self, pads, flash, cs_delay, extra_latency=0):
             sr_out.eq(sink.data << (len(sink.data) - sink.len))
         )
         self.sync += If(sr_out_shift,
-            dq_oe[0].eq(dq_oe[1]),
+            dq_oe.eq(sink.mask),
             dq_o[0].eq(dq_o[1]),
             Case(sink.width, {
                 1 : sr_out.eq(Cat(Signal(1), sr_out)),
@@ -129,10 +128,10 @@ def __init__(self, pads, flash, cs_delay, extra_latency=0):
         # Data In Shift.
         self.sync += If(sr_in_shift,
             Case(sink.width, {
-                1 : sr_in.eq(Cat(dq_i[0][1],  sr_in)), # 1: pads.miso
-                2 : sr_in.eq(Cat(dq_i[0][:2], sr_in)),
-                4 : sr_in.eq(Cat(dq_i[0][:4], sr_in)),
-                8 : sr_in.eq(Cat(dq_i[0][:8], sr_in)),
+                1 : sr_in.eq(Cat(dq_i[1],  sr_in)), # 1: pads.miso
+                2 : sr_in.eq(Cat(dq_i[:2], sr_in)),
+                4 : sr_in.eq(Cat(dq_i[:4], sr_in)),
+                8 : sr_in.eq(Cat(dq_i[:8], sr_in)),
             })
         )
 
@@ -172,6 +171,7 @@ def __init__(self, pads, flash, cs_delay, extra_latency=0):
         fsm.act("XFER-END",
             # Stop Clk.
             NextValue(clkgen.en, 0),
+            NextValue(dq_oe, 0),
 
             # Data In Shift.
             sr_in_shift.eq(1),

litespi/phy/generic_qdr.py

@@ -0,0 +1,191 @@
+#
+# This file is part of LiteSPI
+#
+# Copyright (c) 2020 Antmicro <www.antmicro.com>
+# Copyright (c) 2024 Fin Maaß <[email protected]>
+# SPDX-License-Identifier: BSD-2-Clause
+
+from migen import *
+
+from litex.gen import *
+
+from litex.gen.genlib.misc import WaitTimer
+
+from litespi.common import *
+from litespi.clkgen import QDRLiteSPIClkGen
+
+from litex.soc.interconnect import stream
+
+from litex.build.io import QDRTristate
+
+# LiteSPI QDR PHY Core -----------------------------------------------------------------------------
+
+class LiteSPIQDRPHYCore(LiteXModule):
+    """LiteSPI PHY QDR instantiator
+
+    The ``QDRLiteSPIPHYCore`` class provides a generic PHY that can be connected to the ``LiteSPICore``.
+
+    It supports single/dual/quad/octal output reads from the flash chips.
+
+    You can use this class only with devices that supports the QDR primitives.
+
+    Parameters
+    ----------
+    pads : Object
+        SPI pads description.
+
+    flash : SpiNorFlashModule
+        SpiNorFlashModule configuration object.
+
+    Attributes
+    ----------
+    source : Endpoint(spi_phy2core_layout), out
+        Data stream.
+
+    sink : Endpoint(spi_core2phy_layout), in
+        Control stream.
+
+    cs : Signal(), in
+        Flash CS signal.
+    """
+    def __init__(self, pads, flash, cs_delay, cd_fastclk, extra_latency=0):
+        self.source = source = stream.Endpoint(spi_phy2core_layout)
+        self.sink   = sink   = stream.Endpoint(spi_core2phy_layout)
+        self.cs     = Signal()
+
+        if hasattr(pads, "miso"):
+            bus_width = 1
+            pads.dq   = [pads.mosi, pads.miso]
+        else:
+            bus_width = len(pads.dq)
+
+        assert bus_width in [1, 2, 4, 8]
+
+        if flash:
+            # Check if number of pads matches configured mode.
+            assert flash.check_bus_width(bus_width)
+            assert not flash.ddr
+
+        # Clock Generator.
+        self.clkgen = clkgen = QDRLiteSPIClkGen(pads, cd_fastclk)
+
+        # CS control.
+        self.cs_timer = cs_timer  = WaitTimer(cs_delay + 1) # Ensure cs_delay cycles between XFers.
+        cs_enable = Signal()
+        self.comb += cs_timer.wait.eq(self.cs)
+        self.comb += cs_enable.eq(cs_timer.done)
+        self.comb += pads.cs_n.eq(~cs_enable)
+
+        # I/Os.
+        data_bits = 32
+
+        dq_o  = Array([Signal(len(pads.dq)) for _ in range(3)])
+        dq_i  = Array([Signal(len(pads.dq)) for _ in range(2)])
+        dq_oe = Signal(len(pads.dq))
+
+        for i in range(len(pads.dq)):
+            self.specials += QDRTristate(
+                io  = pads.dq[i],
+                o1  = dq_o[0][i],  o2 = dq_o[1][i], o3 = dq_o[1][i], o4= dq_o[2][i],
+                oe1 =   dq_oe[i],
+                i1  = dq_i[0][i],  i2 =   Signal(), i3 = dq_i[1][i],  i4 =  Signal(),
+                fastclk = ClockSignal(cd_fastclk),
+            )
+
+        # Data Shift Registers.
+        sr_cnt       = Signal(8, reset_less=True)
+        sr_out_load  = Signal()
+        sr_out_shift = Signal()
+        sr_out       = Signal(len(sink.data), reset_less=True)
+        sr_in_shift  = Signal()
+        sr_in        = Signal(len(sink.data), reset_less=True)
+
+        # Data Out Shift.
+        self.comb += [
+            Case(sink.width, {
+                1:  {dq_o[1].eq(sr_out[-1:]), dq_o[2].eq(sr_out[-2:-1])},
+                2:  {dq_o[1].eq(sr_out[-2:]), dq_o[2].eq(sr_out[-4:-2])},
+                4:  {dq_o[1].eq(sr_out[-4:]), dq_o[2].eq(sr_out[-8:-4])},
+                8:  {dq_o[1].eq(sr_out[-8:]), dq_o[2].eq(sr_out[-16:-8])},
+            })
+        ]
+        self.sync += If(sr_out_load,
+            sr_out.eq(sink.data << (len(sink.data) - sink.len))
+        )
+        self.sync += If(sr_out_shift,
+            dq_oe.eq(sink.mask),
+            dq_o[0].eq(dq_o[2]),
+            Case(sink.width, {
+                1 : sr_out.eq(Cat(Signal(2), sr_out)),
+                2 : sr_out.eq(Cat(Signal(4), sr_out)),
+                4 : sr_out.eq(Cat(Signal(8), sr_out)),
+                8 : sr_out.eq(Cat(Signal(16), sr_out)),
+            })
+        )
+
+        # Data In Shift.
+        self.sync += If(sr_in_shift,
+            Case(sink.width, {
+                1 : sr_in.eq(Cat(dq_i[1][1],  dq_i[0][1],  sr_in)), # 1: pads.miso
+                2 : sr_in.eq(Cat(dq_i[1][:2], dq_i[0][:2], sr_in)),
+                4 : sr_in.eq(Cat(dq_i[1][:4], dq_i[0][:4], sr_in)),
+                8 : sr_in.eq(Cat(dq_i[1][:8], dq_i[0][:8], sr_in)),
+            })
+        )
+
+        # FSM.
+        self.fsm = fsm = FSM(reset_state="WAIT-CMD-DATA")
+        fsm.act("WAIT-CMD-DATA",
+            # Stop Clk.
+            NextValue(clkgen.en, 0),
+            # Wait for CS and a CMD from the Core.
+            If(cs_enable & sink.valid,
+                # Load Shift Register Count/Data Out.
+                NextValue(sr_cnt, sink.len - sink.width),
+                sr_out_load.eq(1),
+                # Start XFER.
+                NextState("XFER")
+            )
+        )
+
+        fsm.act("XFER",
+            # Generate Clk.
+            NextValue(clkgen.en, 1),
+
+            # Data In Shift.
+            sr_in_shift.eq(1),
+
+            # Data Out Shift.
+            sr_out_shift.eq(1),
+
+            # Shift Register Count Update/Check.
+            NextValue(sr_cnt, sr_cnt - (2 * sink.width)),
+            # End XFer.
+            If(sr_cnt == 0,
+                NextValue(sr_cnt, (2 + 2*extra_latency)*(2 * sink.width)), # FIXME: Explain magic numbers.
+                NextState("XFER-END"),
+            ),
+        )
+        fsm.act("XFER-END",
+            # Stop Clk.
+            NextValue(clkgen.en, 0),
+
+            # Data In Shift.
+            sr_in_shift.eq(1),
+
+            # Shift Register Count Update/Check.
+            NextValue(sr_cnt, sr_cnt - (2 * sink.width)),
+            If(sr_cnt == 0,
+                sink.ready.eq(1),
+                NextState("SEND-STATUS-DATA"),
+            ),
+        )
+        self.comb += source.data.eq(sr_in)
+        fsm.act("SEND-STATUS-DATA",
+            # Send Data In to Core and return to WAIT when accepted.
+            source.valid.eq(1),
+            source.last.eq(1),
+            If(source.ready,
+                NextState("WAIT-CMD-DATA"),
+            )
+        )