aboutsummaryrefslogtreecommitdiffstats
path: root/fpga/usrp3/tools/utils/rfnoc-system-sim/ni_hw_models.py
blob: 815003c5f6df06de7109ed087f714c74a3603bbe (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
#!/usr/bin/env python
#
# Copyright 2016 Ettus Research
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#

import rfnocsim
import math

class UsrpX310(rfnocsim.SimComp):
    # Hardware specific constants
    RADIO_LATENCY = 1e-6
    IO_LATENCY    = 1e-6
    MAX_SAMP_RATE = 300e6   # Limited by 10GbE
    BPI = 4                 # Bytes per sample (item)

    """
    Simulation model for the USRP X310
    - Has two producers and consumers of FFT data
    - Computes bandwidth and latency using FFT size and overlap
    """
    def __init__(self, sim_core, index, app_settings):
        rfnocsim.SimComp.__init__(self, sim_core, name='USRP_%03d' % (index), ctype=rfnocsim.comptype.hardware)
        # USRP i carries data for radio 2i and 2i+1 interleaved into one stream
        self.index = index
        items = [rfnocsim.DataStream.submatrix_gen('rx', [2*index]),
                 rfnocsim.DataStream.submatrix_gen('rx', [2*index+1])]
        # Samples are 4 bytes I and Q
        latency = (self.RADIO_LATENCY + self.IO_LATENCY/2) * self.get_tick_rate()
        if app_settings['domain'] == 'frequency':
            # Max latency per direction depends on the FFT size and sample rate
            latency += self.__get_fft_latency(
                app_settings['fft_size'], app_settings['samp_rate'], self.get_tick_rate())
        # An X310 Radio has two producers (RX data) and consumers (TX data) (i.e. two ethernet ports)
        # Both ports can carry data from both radio frontends
        self.sources = ([
            rfnocsim.Producer(sim_core, self.name + '/TX0', self.BPI, items, self.MAX_SAMP_RATE, latency),
            rfnocsim.Producer(sim_core, self.name + '/TX1', self.BPI, items, self.MAX_SAMP_RATE, latency)])
        self.sinks = ([
            rfnocsim.Consumer(sim_core, self.name + '/RX0', self.BPI * self.MAX_SAMP_RATE, latency),
            rfnocsim.Consumer(sim_core, self.name + '/RX1', self.BPI * self.MAX_SAMP_RATE, latency)])
        # The actual sample rate depends over the wire depends on the radio sample rate,
        # the FFT size and FFT overlap
        for src in self.sources:
            if app_settings['domain'] == 'frequency':
                src.set_rate(app_settings['samp_rate'] *
                    (1.0 + (float(app_settings['fft_overlap'])/app_settings['fft_size'])))
            else:
                src.set_rate(app_settings['samp_rate'])

    def inputs(self, i, bind=False):
        return self.sinks[i].inputs(0, bind)

    def connect(self, i, dest):
        self.sources[i].connect(0, dest)

    def get_utilization(self, what):
        return 0.0

    def get_util_attrs(self):
        return []

    def validate(self, chan):
        recvd = self.sinks[chan].get_items()
        idxs = []
        for i in recvd:
            (str_id, idx) = rfnocsim.DataStream.submatrix_parse(i)
            if str_id != 'tx':
                raise RuntimeError(self.name + ' received incorrect TX data on channel ' + str(chan))
            idxs.append(idx[0][0])
        if sorted(idxs) != [self.index*2, self.index*2 + 1]:
            raise RuntimeError(self.name + ' received incorrect TX data. Got: ' + str(sorted(idxs)))

    def __get_fft_latency(self, fft_size, samp_rate, tick_rate):
        FFT_CLK_RATE = 200e6
        fft_cycles = {128:349, 256:611, 512:1133, 1024:2163, 2048:4221, 4096:8323}
        latency = max(
            fft_cycles[fft_size] / FFT_CLK_RATE,    #Min time to leave FFT
            fft_size / samp_rate)                   #Min time to enter FFT
        return latency * tick_rate


class Bee7Fpga(rfnocsim.SimComp):
    """
    Simulation model for a single Beecube BEE7 FPGA
    - Type = hardware
    - Contains 80 IO lanes per FPGA: 16 each to neighboring
      FPGAs and 32 lanes going outside
    """
    # IO lanes (How the various IO lanes in an FPGA are allocated)
    EW_IO_LANES  = list(range(0,16))
    NS_IO_LANES  = list(range(16,32))
    XX_IO_LANES  = list(range(32,48))
    EXT_IO_LANES = list(range(48,80))
    # External IO lane connections
    FP_BASE  = 0    # Front panel FMC
    FP_LANES = 16
    BP_BASE  = 16   # Backplane RTM
    BP_LANES = 16

    # Hardware specific constants
    IO_LN_LATENCY = 1.5e-6
    IO_LN_BW = 10e9/8
    ELASTIC_BUFF_FULLNESS = 0.5
    BRAM_BYTES = 18e3/8

    def __init__(self, sim_core, name):
        self.sim_core = sim_core
        rfnocsim.SimComp.__init__(self, sim_core, name, rfnocsim.comptype.hardware)
        # Max resources from Virtex7 datasheet
        self.max_resources = rfnocsim.HwRsrcs()
        self.max_resources.add('DSP', 3600)
        self.max_resources.add('BRAM_18kb', 2940)
        self.resources = rfnocsim.HwRsrcs()
        # Each FPGA has 80 SERDES lanes
        self.max_io = 80
        self.serdes_i = dict()
        self.serdes_o = dict()
        # Each lane can carry at most 10GB/s
        # Each SERDES needs to have some buffering. We assume elastic buffering (50% full on avg).
        io_buff_size = (self.IO_LN_BW * self.IO_LN_LATENCY) / self.ELASTIC_BUFF_FULLNESS
        # Worst case lane latency
        lane_latency = self.IO_LN_LATENCY * self.get_tick_rate()
        for i in range(self.max_io):
            self.serdes_i[i] = rfnocsim.Channel(sim_core, self.__ioln_name(i)+'/I', self.IO_LN_BW, lane_latency / 2)
            self.serdes_o[i] = rfnocsim.Channel(sim_core, self.__ioln_name(i)+'/O', self.IO_LN_BW, lane_latency / 2)
            self.resources.add('BRAM_18kb', 1 + math.ceil(io_buff_size / self.BRAM_BYTES))  #input buffering per lane
            self.resources.add('BRAM_18kb', 1)                                          #output buffering per lane
        # Other resources
        self.resources.add('BRAM_18kb', 72)     # BPS infrastructure + microblaze
        self.resources.add('BRAM_18kb', 128)    # 2 MIGs

        self.functions = dict()

    def inputs(self, i, bind=False):
        return self.serdes_i[i].inputs(0, bind)

    def connect(self, i, dest):
        self.serdes_o[i].connect(0, dest)

    def get_utilization(self, what):
        if self.max_resources.get(what) != 0:
            return self.resources.get(what) / self.max_resources.get(what)
        else:
            return 0.0

    def get_util_attrs(self):
        return ['DSP', 'BRAM_18kb']

    def rename(self, name):
        self.name = name

    def add_function(self, func):
        if func.name not in self.functions:
            self.functions[func.name] = func
        else:
            raise RuntimeError('Function ' + self.name + ' already defined in ' + self.name)
        self.resources.merge(func.get_rsrcs())

    def __ioln_name(self, i):
        if i in self.EW_IO_LANES:
            return '%s/SER_EW_%02d'%(self.name,i-self.EW_IO_LANES[0])
        elif i in self.NS_IO_LANES:
            return '%s/SER_NS_%02d'%(self.name,i-self.NS_IO_LANES[0])
        elif i in self.XX_IO_LANES:
            return '%s/SER_XX_%02d'%(self.name,i-self.XX_IO_LANES[0])
        else:
            return '%s/SER_EXT_%02d'%(self.name,i-self.EXT_IO_LANES[0])

class Bee7Blade(rfnocsim.SimComp):
    """
    Simulation model for a single Beecube BEE7
    - Contains 4 FPGAs (fully connected with 16 lanes)
    """
    NUM_FPGAS = 4
    # FPGA positions in the blade
    NW_FPGA = 0
    NE_FPGA = 1
    SW_FPGA = 2
    SE_FPGA = 3

    def __init__(self, sim_core, index):
        self.sim_core = sim_core
        self.name = name='BEE7_%03d' % (index)
        # Add FPGAs
        names = ['FPGA_NW', 'FPGA_NE', 'FPGA_SW', 'FPGA_SE']
        self.fpgas = []
        for i in range(self.NUM_FPGAS):
            self.fpgas.append(Bee7Fpga(sim_core, name + '/' + names[i]))
        # Build a fully connected network of FPGA
        # 4 FPGAs x 3 Links x 2 directions = 12 connections
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.NW_FPGA], Bee7Fpga.EW_IO_LANES, self.fpgas[self.NE_FPGA], Bee7Fpga.EW_IO_LANES)
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.NW_FPGA], Bee7Fpga.NS_IO_LANES, self.fpgas[self.SW_FPGA], Bee7Fpga.NS_IO_LANES)
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.NW_FPGA], Bee7Fpga.XX_IO_LANES, self.fpgas[self.SE_FPGA], Bee7Fpga.XX_IO_LANES)
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.NE_FPGA], Bee7Fpga.XX_IO_LANES, self.fpgas[self.SW_FPGA], Bee7Fpga.XX_IO_LANES)
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.NE_FPGA], Bee7Fpga.NS_IO_LANES, self.fpgas[self.SE_FPGA], Bee7Fpga.NS_IO_LANES)
        self.sim_core.connect_multi_bidir(
            self.fpgas[self.SW_FPGA], Bee7Fpga.EW_IO_LANES, self.fpgas[self.SE_FPGA], Bee7Fpga.EW_IO_LANES)

    def inputs(self, i, bind=False):
        IO_PER_FPGA = len(Bee7Fpga.EXT_IO_LANES)
        return self.fpgas[int(i/IO_PER_FPGA)].inputs(Bee7Fpga.EXT_IO_LANES[i%IO_PER_FPGA], bind)

    def connect(self, i, dest):
        IO_PER_FPGA = len(Bee7Fpga.EXT_IO_LANES)
        self.fpgas[int(i/IO_PER_FPGA)].connect(Bee7Fpga.EXT_IO_LANES[i%IO_PER_FPGA], dest)

    @staticmethod
    def io_lane(fpga, fpga_lane):
        IO_PER_FPGA = len(Bee7Fpga.EXT_IO_LANES)
        return (fpga_lane - Bee7Fpga.EXT_IO_LANES[0]) + (fpga * IO_PER_FPGA)

class ManagementHostandSwitch(rfnocsim.SimComp):
    """
    Simulation model for a management host computer
    - Sources channel coefficients
    - Configures radio
    """
    def __init__(self, sim_core, index, num_coeffs, switch_ports, app_settings):
        rfnocsim.SimComp.__init__(self, sim_core, name='MGMT_HOST_%03d'%(index), ctype=rfnocsim.comptype.other)
        if app_settings['domain'] == 'frequency':
            k = app_settings['fft_size']
        else:
            k = app_settings['fir_taps']

        self.sources = dict()
        self.sinks = dict()
        for l in range(switch_ports):
            self.sources[l] = rfnocsim.Producer(
                sim_core, '%s/COEFF_%d'%(self.name,l), 4, ['coeff_%03d[%d]'%(index,l)], (10e9/8)/switch_ports, 0)
            self.sinks[l] = rfnocsim.Consumer(sim_core, self.name + '%s/ACK%d'%(self.name,l))
            self.sources[l].set_rate(k*num_coeffs*app_settings['coherence_rate'])

    def inputs(self, i, bind=False):
        return self.sinks[i].inputs(0, bind)

    def connect(self, i, dest):
        self.sources[i].connect(0, dest)

    def get_utilization(self, what):
        return 0.0

    def get_util_attrs(self):
        return []