Move core config script here from core/platform/common, and adapt to the larger fmc address bus.

4 files changed, 731 insertions, 10 deletions

diff --git a/build/Makefile b/build/Makefile
index aed01df..d5d4871 100644
--- a/build/Makefile
+++ b/build/Makefile
@@ -22,32 +22,36 @@ all:	$(project).bit
 
 # Build the default core_selector if it doesn't already exist.
 
-CONFIG          = $(CORE_TREE)/platform/common/config
+CONFIG          = .
 core_selector.v core_vfiles.mk:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg
+	$(CONFIG)/config.py
 
 # Build some different configurations
 
 bare:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg -s bare
+	$(CONFIG)/config.py -s bare
 	$(MAKE) project=$(project)_bare ucf=$(ucf)
 
 trng:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg -s trng
+	$(CONFIG)/config.py -s trng
 	$(MAKE) project=$(project)_trng ucf=$(ucf)
 
 hash:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg -s hash
+	$(CONFIG)/config.py -s hash
 	$(MAKE) project=$(project)_hash ucf=$(ucf)
 
 rsa:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg -s rsa
+	$(CONFIG)/config.py -s rsa
 	$(MAKE) project=$(project)_rsa ucf=$(ucf)
 
 mkmif:
-	$(CONFIG)/config.py -c $(CONFIG)/config.cfg -s mkmif
+	$(CONFIG)/config.py -s mkmif
 	$(MAKE) project=$(project)_mkmif ucf=$(ucf)
 
+hsm:
+	$(CONFIG)/config.py -s hsm
+	$(MAKE) project=$(project)_hsm ucf=$(ucf)
+
 # Verilog files that always go with builds on this platform.
 
 vfiles = \
diff --git a/build/config.cfg b/build/config.cfg
new file mode 100644
index 0000000..a7e3151
--- /dev/null
+++ b/build/config.cfg
@@ -0,0 +1,144 @@
+# Config file for the Cryptech Novena FPGA framework.
+#
+# Variables used in this file:
+#
+# default-section: Name of the configuration to build if the user
+#   doesn't specify one.  Only meaningful in the default section.
+#
+# cores: A list of cores to build.  Use with the --section option.
+#
+# vfiles: A list of Verilog files to include in the vfiles list when
+#   including a particular core.  All (optional) cores must have a
+#   vfiles option, so that the configuration program knows what to put
+#   into core_vfiles.mk.
+#
+# requires: A list of other cores whose vfiles must be loaded to build
+#   this core.  This has no effect on the generated core_selector.v
+#   file, and has no effect at all if an instance of a core named here
+#   is already included in the build.
+#
+# error_wire: boolean indicating whether the core wants a error wire.
+#
+# block_memory: boolean indicating whether the core uses block memory.
+#   Effect of this is a bit strange: setting it triggers generation of
+#   a one-cycle timing delay for every core in this build that does
+#   *not* use block memory.  When no cores in the build use block
+#   memory, the delay isn't necessary and is therefore omitted.
+
+[default]
+default-section = hsm
+
+# for quick builds to test the bus
+[bare]
+cores =
+
+[hash]
+cores = sha1 sha256 sha512
+
+[trng]
+cores = trng
+
+[modexp]
+cores = modexpa7
+
+[mkmif]
+cores = mkmif
+
+[rsa]
+cores = mkmif sha256 aes trng modexpa7
+
+# make me one with everything
+[hsm]
+cores = mkmif sha1 sha256 sha512 aes trng modexpa7
+
+# include multiple of the same core
+[multi-test]
+cores = sha256 aes aes chacha aes
+
+[sha1]
+vfiles =
+	hash/sha1/src/rtl/sha1.v
+	hash/sha1/src/rtl/sha1_core.v
+	hash/sha1/src/rtl/sha1_w_mem.v
+
+[sha256]
+vfiles =
+	hash/sha256/src/rtl/sha256.v
+	hash/sha256/src/rtl/sha256_core.v
+	hash/sha256/src/rtl/sha256_k_constants.v
+	hash/sha256/src/rtl/sha256_w_mem.v
+
+[sha512]
+vfiles =
+	hash/sha512/src/rtl/sha512.v
+	hash/sha512/src/rtl/sha512_core.v
+	hash/sha512/src/rtl/sha512_h_constants.v
+	hash/sha512/src/rtl/sha512_k_constants.v
+	hash/sha512/src/rtl/sha512_w_mem.v
+
+[trng]
+requires = chacha sha512
+vfiles =
+	rng/avalanche_entropy/src/rtl/avalanche_entropy.v
+	rng/avalanche_entropy/src/rtl/avalanche_entropy_core.v
+	rng/rosc_entropy/src/rtl/rosc.v
+	rng/rosc_entropy/src/rtl/rosc_entropy.v
+	rng/rosc_entropy/src/rtl/rosc_entropy_core.v
+	rng/trng/src/rtl/trng.v
+	rng/trng/src/rtl/trng_csprng.v
+	rng/trng/src/rtl/trng_csprng_fifo.v
+	rng/trng/src/rtl/trng_mixer.v
+
+[aes]
+vfiles =
+	cipher/aes/src/rtl/aes.v
+	cipher/aes/src/rtl/aes_core.v
+	cipher/aes/src/rtl/aes_decipher_block.v
+	cipher/aes/src/rtl/aes_encipher_block.v
+	cipher/aes/src/rtl/aes_inv_sbox.v
+	cipher/aes/src/rtl/aes_key_mem.v
+	cipher/aes/src/rtl/aes_sbox.v
+
+[chacha]
+vfiles =
+	cipher/chacha/src/rtl/chacha.v
+	cipher/chacha/src/rtl/chacha_core.v
+	cipher/chacha/src/rtl/chacha_qr.v
+
+[modexpa7]
+block_memory = yes
+error_wire = no
+vfiles =
+	math/modexpa7/src/rtl/dsp_multiplier_a7.v
+	math/modexpa7/src/rtl/dsp_subtractor_a7.v
+	math/modexpa7/src/rtl/modexpa7_adder64_carry32.v
+	math/modexpa7/src/rtl/modexpa7_buffer_core.v
+	math/modexpa7/src/rtl/modexpa7_buffer_user.v
+	math/modexpa7/src/rtl/modexpa7_modinv32.v
+	math/modexpa7/src/rtl/modexpa7_montgomery_coeff.v
+	math/modexpa7/src/rtl/modexpa7_montgomery_multiplier.v
+	math/modexpa7/src/rtl/modexpa7_top.v
+	math/modexpa7/src/rtl/modexpa7_wrapper.v
+	math/modexpa7/src/rtl/ram_1rw_1ro_readfirst.v
+
+[modexp]
+error_wire = no
+vfiles =
+	math/modexp/src/rtl/adder.v
+	math/modexp/src/rtl/blockmem1r1w.v
+	math/modexp/src/rtl/blockmem2r1wptr.v
+	math/modexp/src/rtl/blockmem2r1w.v
+	math/modexp/src/rtl/blockmem2rptr1w.v
+	math/modexp/src/rtl/modexp.v
+	math/modexp/src/rtl/modexp_core.v
+	math/modexp/src/rtl/montprod.v
+	math/modexp/src/rtl/residue.v
+	math/modexp/src/rtl/shl.v
+	math/modexp/src/rtl/shr.v
+
+[mkmif]
+error_wire = no
+vfiles =
+	util/mkmif/src/rtl/mkmif.v
+	util/mkmif/src/rtl/mkmif_core.v
+	util/mkmif/src/rtl/mkmif_spi.v
diff --git a/build/config.py b/build/config.py
new file mode 100755
index 0000000..e737c92
--- /dev/null
+++ b/build/config.py
@@ -0,0 +1,573 @@
+#!/usr/bin/env python
+
+"""
+Generate core_selector.v and core_vfiles.mk for a set of cores.
+"""
+
+#=======================================================================
+# Copyright (c) 2015-2016, NORDUnet A/S All rights reserved.
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are
+# met:
+# - Redistributions of source code must retain the above copyright notice,
+#   this list of conditions and the following disclaimer.
+#
+# - Redistributions in binary form must reproduce the above copyright
+#   notice, this list of conditions and the following disclaimer in the
+#   documentation and/or other materials provided with the distribution.
+#
+# - Neither the name of the NORDUnet nor the names of its contributors may
+#   be used to endorse or promote products derived from this software
+#   without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS
+# IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+# TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
+# PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+# HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+# SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
+# TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
+# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
+# LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
+# NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
+# SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+#=======================================================================
+
+# The modexpa7 core drags in a one clock cycle delay to other cores,
+# to compensate for the extra clock cycle consumed by the block
+# memories used in the modexpa7 core.  We probably want a general
+# solution for this, because we're going to run into this problem for
+# any core that handles arguments big enough to require block memory.
+
+# To Do:
+#
+# - Consider automating the one-clock-cycle delay stuff by adding
+#   another boolean flag to the config file.  Default would be no
+#   delay, if any included core sets the "I use block memories" flag,
+#   all other cores would get the delay.  Slightly tedious but
+#   something we can calculate easily enough, and probably an
+#   improvement over wiring in the delay when nothing needs it.
+#
+# - Rename script and its config file to something more meaningful.
+
+
+def main():
+    """
+    Parse arguments and config file, generate core list, generate output.
+    """
+
+    from argparse import ArgumentParser, FileType, ArgumentDefaultsHelpFormatter
+    from sys      import exit
+
+    parser = ArgumentParser(description = __doc__, formatter_class = ArgumentDefaultsHelpFormatter)
+    parser.add_argument("-d", "--debug",   help = "enable debugging",   action = "store_true")
+    parser.add_argument("-s", "--section", help = "config file section")
+    parser.add_argument("-c", "--config",  help = "configuration file",   default = "config.cfg",       type = FileType("r"))
+    parser.add_argument("--verilog",       help = "verilog output file",  default = "core_selector.v",  type = FileType("w"))
+    parser.add_argument("--makefile",      help = "output makefile",      default = "core_vfiles.mk",   type = FileType("w"))
+    parser.add_argument("core",            help = "name(s) of core(s)", nargs = "*")
+    args = parser.parse_args()
+
+    try:
+        cfg = RawConfigParser()
+        cfg.readfp(args.config)
+
+        if args.core:
+            cores = args.core
+        else:
+            section = args.section or cfg.get("default", "default-section")
+            cores = cfg.get(section, "cores").split()
+
+        cores.insert(0, "board_regs")
+        cores.insert(1, "comm_regs")
+
+        cores = tuple(Core.new(core) for core in cores)
+
+        core_number = 0
+        for core in cores:
+            core_number = core.assign_core_number(core_number)
+
+        for core in cores:
+            core.configure(cfg)
+
+        if False:
+
+            # For some reason, attempting to optimize out the delay
+            # code entirely results in a non-working bitstream.  Don't
+            # know why, disabling the optimization works, so just do
+            # that for now.
+
+            Core.need_one_cycle_delay = any(core.block_memory for core in cores)
+
+        args.verilog.write(createModule_template.format(
+            addrs = "".join(core.createAddr()     for core in cores),
+            insts = "".join(core.createInstance() for core in cores),
+            muxes = "".join(core.createMux()      for core in cores),
+            ports = "".join(core.createPort()     for core in cores) ))
+
+        args.makefile.write(listVfiles_template.format(
+            vfiles = "".join(core.listVfiles()    for core in cores)))
+
+    except Exception, e:
+        if args.debug:
+            raise
+        exit(str(e))
+
+
+try:
+    import ConfigParser as configparser
+except ImportError:
+    import configparser   
+
+class RawConfigParser(configparser.RawConfigParser):
+    """
+    RawConfigParser with a few extensions.
+    """
+
+    def getboolean(self, section, option, default = False):
+        if self.has_option(section, option):
+            # RawConfigParser is an old-stle class, super() doesn't work, feh.
+            return configparser.RawConfigParser.getboolean(self, section, option)
+        else:
+            return default
+
+    def getvalues(self, section, option):
+        if self.has_option(section, option):
+            for value in self.get(section, option).split():
+                yield value
+
+
+class Core(object):
+    """
+    Data and methods for a generic core.  We can use this directly for
+    most cores, a few are weird and require subclassing to override
+    particular methods.
+    """
+
+    # Class variable tracking how many times a particular core has
+    # been instantiated.  This controls instance numbering.
+
+    _instance_count = {}
+
+    # Class variable mapping core name to subclass for special cases.
+
+    special_class = {}
+
+    # Class variable recording whether we need a one-cycle delay to
+    # compensate for block memories.
+
+    need_one_cycle_delay = True
+
+    def __init__(self, name):
+        self.name = name
+        self.core_number = None
+        self.vfiles = []
+        self.error_wire = True
+        self.block_memory = False
+        self.instance_number = self._instance_count.get(name, 0)
+        self._instance_count[name] = self.instance_number + 1
+
+    @classmethod
+    def new(cls, name):
+        return cls.special_class.get(name, cls)(name)
+
+    def assign_core_number(self, n):
+        self.core_number = n
+        return n + 1
+
+    def configure(self, cfg):
+        if self.instance_number == 0:
+            self.vfiles.extend(cfg.getvalues(self.name, "vfiles"))
+            for required in cfg.getvalues(self.name, "requires"):
+                if required not in self._instance_count:
+                    self.vfiles.extend(cfg.getvalues(required, "vfiles"))
+        self.error_wire = cfg.getboolean(self.name, "error_wire", self.error_wire)
+        self.block_memory = cfg.getboolean(self.name, "block_memory", self.block_memory)
+
+    @property
+    def instance_name(self):
+        if self._instance_count[self.name] > 1:
+            return "{}_{}".format(self.name, self.instance_number)
+        else:
+            return self.name
+
+    @property
+    def upper_instance_name(self):
+        return self.instance_name.upper()
+
+    @property
+    def reset_pin(self):
+        return ".reset_n(sys_rst_n)"
+
+    @property
+    def error_port(self):
+        return ",\n      .error(error_{core.instance_name})".format(core = self) if self.error_wire else ""
+
+    @property
+    def one_cycle_delay(self):
+        return one_cycle_delay_template.format(core = self) if self.need_one_cycle_delay and not self.block_memory else ""
+
+    @property
+    def mux_data_reg(self):
+        return "read_data_" + self.instance_name + ("_reg" if self.need_one_cycle_delay and not self.block_memory else "")
+
+    @property
+    def mux_error_reg(self):
+        return "error_" + self.instance_name if self.error_wire else "0"
+
+    def createInstance(self):
+        return createInstance_template_generic.format(core = self)
+
+    def createAddr(self):
+        return createAddr_template.format(core = self)
+
+    def createMux(self):
+        return createMux_template.format(core = self, core0 = self)
+
+    def createPort(self):
+        return ""
+
+    def listVfiles(self):
+        return "".join(" \\\n\t$(CORE_TREE)/" + vfile for vfile in self.vfiles)
+
+
+class SubCore(Core):
+    """"
+    Override mux handling for TRNG's sub-cores.
+    """
+
+    def __init__(self, name, parent):
+        super(SubCore, self).__init__(name)
+        self.parent = parent
+
+    def createMux(self):
+        return createMux_template.format(core = self, core0 = self.parent)
+
+
+class TRNGCore(Core):
+    """
+    The TRNG core has an internal mux with slots for 15 sub-cores,
+    most of which are empty.  This is a bit of a mess.
+
+    Mostly this means that our method calls have to iterate over all
+    of the subcores after handling the base TRNG core, but we also use
+    different templates, and fiddle with addresses a bit.
+
+    Mux numbers have to be dug out of the TRNG Verilog source.
+    """
+
+    # TRNG subcore name -> internal mux number.
+    subcore_parameters = dict(avalanche_entropy = 0x1,
+                              rosc_entropy      = 0x2,
+                              trng_mixer        = 0x3,
+                              trng_csprng       = 0x4)
+
+    def __init__(self, name):
+        super(TRNGCore, self).__init__(name)
+        self.subcores = tuple(SubCore(name, self)
+                              for name in sorted(self.subcore_parameters,
+                                                 key = lambda x: self.subcore_parameters[x]))
+
+    def assign_core_number(self, n):
+        n = super(TRNGCore, self).assign_core_number(n)
+        for subcore in self.subcores:
+            subcore.assign_core_number(self.core_number + self.subcore_parameters[subcore.name])
+        return n + 15
+
+    @property
+    def last_subcore_upper_instance_name(self):
+        return self.subcores[-1].upper_instance_name
+
+    def createInstance(self):
+        return createInstance_template_TRNG.format(core = self)
+
+    def createAddr(self):
+        return super(TRNGCore, self).createAddr() + "".join(subcore.createAddr() for subcore in self.subcores)
+
+    def createMux(self):
+        return super(TRNGCore, self).createMux() + "".join(subcore.createMux() for subcore in self.subcores)
+
+
+class ModExpA7Core(Core):
+    """
+    ModExpA7 core consumes as much space as four ordinary cores, and
+    uses different templates to handle the differences in timing and
+    addressing.
+    """
+
+    def assign_core_number(self, n):
+        n = super(ModExpA7Core, self).assign_core_number(n)
+        return n + 3
+
+    def createInstance(self):
+        return createInstance_template_ModExpA7.format(core = self)
+
+    def createMux(self):
+        return createMux_modexpa7_template.format(core = self, core0 = self)
+
+class MkmifCore(Core):
+    """
+    MKM interface core has extra ports for the SPI signal lines.
+    """
+
+    def createPort(self):
+        return """ \
+
+   output wire         mkm_sclk,
+   output wire         mkm_cs_n,
+   input wire          mkm_do,
+   output wire         mkm_di,
+        """
+
+    def createInstance(self):
+        return createInstance_template_MKMIF.format(core = self)
+
+# Hook special classes in as handlers for the cores that require them.
+
+Core.special_class.update(
+    trng        = TRNGCore,
+    modexpa7    = ModExpA7Core,
+    mkmif       = MkmifCore)
+
+
+# Templates (format strings), here instead of inline in the functions
+# that use them, both because some of these are shared between
+# multiple functions and because it's easier to read these (and get
+# the indentation right) when the're separate.
+
+# Template used by .createAddr() methods.
+
+createAddr_template = """\
+   localparam   CORE_ADDR_{core.upper_instance_name:21s} = 15'h{core.core_number:02x};
+"""
+
+# Template used by Core.createInstance().
+
+createInstance_template_generic = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num == CORE_ADDR_{core.upper_instance_name});
+   wire [31: 0]         read_data_{core.instance_name};
+   wire                 error_{core.instance_name};
+
+   {core.name} {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_fmc_rd | sys_fmc_wr)),
+      .we(sys_fmc_wr),
+
+      .address(addr_core_reg),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name}){core.error_port}
+      );
+
+{core.one_cycle_delay}
+
+"""
+
+# Template used by ModExpA7Core.createInstance().  This is different
+# enough from the base template that it's easier to make this separate.
+
+createInstance_template_ModExpA7 = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num >= CORE_ADDR_{core.upper_instance_name}) && (addr_core_num <= CORE_ADDR_{core.upper_instance_name} + 15'h03);
+   wire [31: 0]         read_data_{core.instance_name};
+   wire [1:0]           {core.instance_name}_prefix = addr_core_num[1:0] - CORE_ADDR_{core.upper_instance_name};
+
+   {core.name}_wrapper {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_fmc_rd | sys_fmc_wr)),
+      .we(sys_fmc_wr),
+
+      .address({{{core.instance_name}_prefix, addr_core_reg}}),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name})
+      );
+
+
+"""
+
+# Template used by TRNGCore.createInstance(); this is different enough
+# from the generic template that it's (probably) clearer to have this
+# separate.
+
+createInstance_template_TRNG = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num >= CORE_ADDR_{core.upper_instance_name}) && (addr_core_num <= CORE_ADDR_{core.upper_instance_name} + 15'h0f);
+   wire [31: 0]         read_data_{core.instance_name};
+   wire                 error_{core.instance_name};
+   wire [3:0]           {core.instance_name}_prefix = addr_core_num[3:0] - CORE_ADDR_{core.upper_instance_name};
+
+   {core.name} {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .cs(enable_{core.instance_name} & (sys_fmc_rd | sys_fmc_wr)),
+      .we(sys_fmc_wr),
+
+      .address({{{core.instance_name}_prefix, addr_core_reg}}),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name}),
+      .error(error_{core.instance_name}),
+
+      .avalanche_noise(noise),
+      .debug(debug)
+      );
+
+{core.one_cycle_delay}
+
+"""
+
+# Template used by Mkmif.createInstance().  This is different
+# enough from the base template that it's easier to make this separate.
+
+createInstance_template_MKMIF = """\
+   //----------------------------------------------------------------
+   // {core.upper_instance_name}
+   //----------------------------------------------------------------
+   wire                 enable_{core.instance_name} = (addr_core_num == CORE_ADDR_{core.upper_instance_name});
+   wire [31: 0]         read_data_{core.instance_name};
+
+   {core.name} {core.instance_name}_inst
+     (
+      .clk(sys_clk),
+      {core.reset_pin},
+
+      .spi_sclk(mkm_sclk),
+      .spi_cs_n(mkm_cs_n),
+      .spi_do(mkm_do),
+      .spi_di(mkm_di),
+
+      .cs(enable_{core.instance_name} & (sys_fmc_rd | sys_fmc_wr)),
+      .we(sys_fmc_wr),
+
+      .address(addr_core_reg),
+      .write_data(sys_write_data),
+      .read_data(read_data_{core.instance_name}){core.error_port}
+      );
+
+{core.one_cycle_delay}
+
+"""
+
+# Template for one-cycle delay code.
+
+one_cycle_delay_template = """\
+   reg  [31: 0] read_data_{core.instance_name}_reg;
+   always @(posedge sys_clk)
+     read_data_{core.instance_name}_reg <= read_data_{core.instance_name};
+"""
+
+# Template for .createMux() methods.
+
+createMux_template = """\
+       CORE_ADDR_{core.upper_instance_name}:
+         begin
+            sys_read_data_mux = {core0.mux_data_reg};
+            sys_error_mux = {core0.mux_error_reg};
+         end
+"""
+
+# Template for ModExpA7.createMux() method.
+
+createMux_modexpa7_template = """\
+       CORE_ADDR_{core.upper_instance_name} + 0,
+       CORE_ADDR_{core.upper_instance_name} + 1,
+       CORE_ADDR_{core.upper_instance_name} + 2,
+       CORE_ADDR_{core.upper_instance_name} + 3:
+         begin
+            sys_read_data_mux = {core0.mux_data_reg};
+            sys_error_mux = {core0.mux_error_reg};
+         end
+"""
+
+
+# Top-level (createModule) template.
+
+createModule_template = """\
+// NOTE: This file is generated; do not edit by hand.
+
+module core_selector
+  (
+   input wire          sys_clk,
+   input wire          sys_rst_n,
+
+   input wire [23: 0]  sys_fmc_addr,
+   input wire          sys_fmc_wr,
+   input wire          sys_fmc_rd,
+   output wire [31: 0] sys_read_data,
+   input wire [31: 0]  sys_write_data,
+   output wire         sys_error,
+{ports}
+   input wire          noise,
+   output wire [7 : 0] debug
+   );
+
+
+   //----------------------------------------------------------------
+   // Address Decoder
+   //----------------------------------------------------------------
+   // upper 15 bits specify core being addressed
+   wire [14: 0]         addr_core_num   = sys_fmc_addr[23: 8];
+   // lower 8 bits specify register offset in core
+   wire [ 7: 0]         addr_core_reg   = sys_fmc_addr[ 7: 0];
+
+
+   //----------------------------------------------------------------
+   // Core Address Table
+   //----------------------------------------------------------------
+{addrs}
+
+{insts}
+   //----------------------------------------------------------------
+   // Output (Read Data) Multiplexer
+   //----------------------------------------------------------------
+   reg [31: 0]          sys_read_data_mux;
+   assign               sys_read_data = sys_read_data_mux;
+   reg                  sys_error_mux;
+   assign               sys_error = sys_error_mux;
+
+   always @*
+
+     case (addr_core_num)
+{muxes}
+       default:
+         begin
+            sys_read_data_mux = {{32{{1'b0}}}};
+            sys_error_mux = 1;
+         end
+     endcase
+
+
+endmodule
+
+
+//======================================================================
+// EOF core_selector.v
+//======================================================================
+"""
+
+# Template for makefile snippet listing Verilog source files.
+
+listVfiles_template = """\
+# NOTE: This file is generated; do not edit by hand.
+
+vfiles +={vfiles}
+"""
+
+# Run main program.
+
+if __name__ == "__main__":
+    main()
diff --git a/rtl/alpha_fmc_top.v b/rtl/alpha_fmc_top.v
index 6410d44..79bfba8 100644
--- a/rtl/alpha_fmc_top.v
+++ b/rtl/alpha_fmc_top.v
@@ -235,9 +235,9 @@ module alpha_fmc_top
       .sys_clk(sys_clk),
       .sys_rst_n(sys_rst_n),
 
-      .sys_eim_addr(sys_fmc_addr[16:0]),	// XXX parameterize
-      .sys_eim_wr(sys_fmc_wren),
-      .sys_eim_rd(sys_fmc_rden),
+      .sys_fmc_addr(sys_fmc_addr),
+      .sys_fmc_wr(sys_fmc_wren),
+      .sys_fmc_rd(sys_fmc_rden),
       .sys_write_data(sys_fmc_dout),
       .sys_read_data(sys_fmc_din),