summaryrefslogtreecommitdiff
path: root/src/mem/SimpleDRAM.py
blob: c450f89924bc6fa6bfdadda01ed4ff43cabe2563 (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
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
# Copyright (c) 2012-2013 ARM Limited
# All rights reserved.
#
# The license below extends only to copyright in the software and shall
# not be construed as granting a license to any other intellectual
# property including but not limited to intellectual property relating
# to a hardware implementation of the functionality of the software
# licensed hereunder.  You may use the software subject to the license
# terms below provided that you ensure that this notice is replicated
# unmodified and in its entirety in all distributions of the software,
# modified or unmodified, in source code or in binary form.
#
# Copyright (c) 2013 Amin Farmahini-Farahani
# 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 copyright holders 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
# OWNER 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.
#
# Authors: Andreas Hansson
#          Ani Udipi

from m5.params import *
from AbstractMemory import *

# Enum for memory scheduling algorithms, currently First-Come
# First-Served and a First-Row Hit then First-Come First-Served
class MemSched(Enum): vals = ['fcfs', 'frfcfs']

# Enum for the address mapping. With Ra, Co, Ba and Ch denoting rank,
# column, bank and channel, respectively, and going from MSB to LSB.
# Available are RaBaChCo and RaBaCoCh, that are suitable for an
# open-page policy, optimising for sequential accesses hitting in the
# open row. For a closed-page policy, CoRaBaCh maximises parallelism.
class AddrMap(Enum): vals = ['RaBaChCo', 'RaBaCoCh', 'CoRaBaCh']

# Enum for the page policy, either open or close.
class PageManage(Enum): vals = ['open', 'close']

# SimpleDRAM is a single-channel single-ported DRAM controller model
# that aims to model the most important system-level performance
# effects of a DRAM without getting into too much detail of the DRAM
# itself.
class SimpleDRAM(AbstractMemory):
    type = 'SimpleDRAM'
    cxx_header = "mem/simple_dram.hh"

    # single-ported on the system interface side, instantiate with a
    # bus in front of the controller for multiple ports
    port = SlavePort("Slave port")

    # the basic configuration of the controller architecture
    write_buffer_size = Param.Unsigned(32, "Number of write queue entries")
    read_buffer_size = Param.Unsigned(32, "Number of read queue entries")

    # threshold in percent for when to trigger writes and start
    # emptying the write buffer as it starts to get full
    write_high_thresh_perc = Param.Percent(70, "Threshold to trigger writes")

    # threshold in percentage for when to stop writes if the read
    # queue has an entry. An optimisaton to give reads priority if
    # sufficient number of writes are scheduled and write queue has
    # sufficient number of free entries
    write_low_thresh_perc = Param.Percent(0, "Threshold to stop writes")

    # scheduler, address map and page policy
    mem_sched_policy = Param.MemSched('frfcfs', "Memory scheduling policy")
    addr_mapping = Param.AddrMap('RaBaChCo', "Address mapping policy")
    page_policy = Param.PageManage('open', "Page closure management policy")

    # pipeline latency of the controller and PHY, split into a
    # frontend part and a backend part, with reads and writes serviced
    # by the queues only seeing the frontend contribution, and reads
    # serviced by the memory seeing the sum of the two
    static_frontend_latency = Param.Latency("10ns", "Static frontend latency")
    static_backend_latency = Param.Latency("10ns", "Static backend latency")

    # the physical organisation of the DRAM
    device_bus_width = Param.Unsigned("data bus width in bits for each DRAM "\
                                      "device/chip")
    burst_length = Param.Unsigned("Burst lenght (BL) in beats")
    device_rowbuffer_size = Param.MemorySize("Page (row buffer) size per "\
                                           "device/chip")
    devices_per_rank = Param.Unsigned("Number of devices/chips per rank")
    ranks_per_channel = Param.Unsigned("Number of ranks per channel")
    banks_per_rank = Param.Unsigned("Number of banks per rank")
    # only used for the address mapping as the controller by
    # construction is a single channel and multiple controllers have
    # to be instantiated for a multi-channel configuration
    channels = Param.Unsigned(1, "Number of channels")

    # timing behaviour and constraints - all in nanoseconds

    # the amount of time in nanoseconds from issuing an activate command
    # to the data being available in the row buffer for a read/write
    tRCD = Param.Latency("RAS to CAS delay")

    # the time from issuing a read/write command to seeing the actual data
    tCL = Param.Latency("CAS latency")

    # minimum time between a precharge and subsequent activate
    tRP = Param.Latency("Row precharge time")

    # minimum time between an activate and a precharge to the same row
    tRAS = Param.Latency("ACT to PRE delay")

    # time to complete a burst transfer, typically the burst length
    # divided by two due to the DDR bus, but by making it a parameter
    # it is easier to also evaluate SDR memories like WideIO.
    # This parameter has to account for burst length.
    # Read/Write requests with data size larger than one full burst are broken
    # down into multiple requests in the SimpleDRAM controller
    tBURST = Param.Latency("Burst duration (for DDR burst length / 2 cycles)")

    # time taken to complete one refresh cycle (N rows in all banks)
    tRFC = Param.Latency("Refresh cycle time")

    # refresh command interval, how often a "ref" command needs
    # to be sent. It is 7.8 us for a 64ms refresh requirement
    tREFI = Param.Latency("Refresh command interval")

    # write-to-read turn around penalty, assumed same as read-to-write
    tWTR = Param.Latency("Write to read switching time")

    # minimum row activate to row activate delay time
    tRRD = Param.Latency("ACT to ACT delay")

    # time window in which a maximum number of activates are allowed
    # to take place, set to 0 to disable
    tXAW = Param.Latency("X activation window")
    activation_limit = Param.Unsigned("Max number of activates in window")

    # Currently rolled into other params
    ######################################################################

    # tRC  - assumed to be tRAS + tRP

# A single DDR3 x64 interface (one command and address bus), with
# default timings based on DDR3-1600 4 Gbit parts in an 8x8
# configuration, which would amount to 4 Gbyte of memory.
class DDR3_1600_x64(SimpleDRAM):
    # 8x8 configuration, 8 devices each with an 8-bit interface
    device_bus_width = 8

    # DDR3 is a BL8 device
    burst_length = 8

    # Each device has a page (row buffer) size of 1KB
    # (this depends on the memory density)
    device_rowbuffer_size = '1kB'

    # 8x8 configuration, so 8 devices
    devices_per_rank = 8

    # Use two ranks
    ranks_per_channel = 2

    # DDR3 has 8 banks in all configurations
    banks_per_rank = 8

    # DDR3-1600 11-11-11-28
    tRCD = '13.75ns'
    tCL = '13.75ns'
    tRP = '13.75ns'
    tRAS = '35ns'

    # 8 beats across an x64 interface translates to 4 clocks @ 800 MHz.
    # Note this is a BL8 DDR device.
    tBURST = '5ns'

    # DDR3, 4 Gbit has a tRFC of 240 CK and tCK = 1.25 ns
    tRFC = '300ns'

    # DDR3, <=85C, half for >85C
    tREFI = '7.8us'

    # Greater of 4 CK or 7.5 ns, 4 CK @ 800 MHz = 5 ns
    tWTR = '7.5ns'

    # Assume 5 CK for activate to activate for different banks
    tRRD = '6.25ns'

    # With a 2kbyte page size, DDR3-1600 lands around 40 ns
    tXAW = '40ns'
    activation_limit = 4


# A single LPDDR2-S4 x32 interface (one command/address bus), with
# default timings based on a LPDDR2-1066 4 Gbit part in a 1x32
# configuration.
class LPDDR2_S4_1066_x32(SimpleDRAM):
    # 1x32 configuration, 1 device with a 32-bit interface
    device_bus_width = 32

    # LPDDR2_S4 is a BL4 and BL8 device
    burst_length = 8

    # Each device has a page (row buffer) size of 1KB
    # (this depends on the memory density)
    device_rowbuffer_size = '1kB'

    # 1x32 configuration, so 1 device
    devices_per_rank = 1

    # Use a single rank
    ranks_per_channel = 1

    # LPDDR2-S4 has 8 banks in all configurations
    banks_per_rank = 8

    # Fixed at 15 ns
    tRCD = '15ns'

    # 8 CK read latency, 4 CK write latency @ 533 MHz, 1.876 ns cycle time
    tCL = '15ns'

    # Pre-charge one bank 15 ns (all banks 18 ns)
    tRP = '15ns'

    tRAS = '42ns'

    # 8 beats across an x32 DDR interface translates to 4 clocks @ 533 MHz.
    # Note this is a BL8 DDR device.
    # Requests larger than 32 bytes are broken down into multiple requests
    # in the SimpleDRAM controller
    tBURST = '7.5ns'

    # LPDDR2-S4, 4 Gbit
    tRFC = '130ns'
    tREFI = '3.9us'

    # Irrespective of speed grade, tWTR is 7.5 ns
    tWTR = '7.5ns'

    # Activate to activate irrespective of density and speed grade
    tRRD = '10.0ns'

    # Irrespective of density, tFAW is 50 ns
    tXAW = '50ns'
    activation_limit = 4

# A single WideIO x128 interface (one command and address bus), with
# default timings based on an estimated WIO-200 8 Gbit part.
class WideIO_200_x128(SimpleDRAM):
    # 1x128 configuration, 1 device with a 128-bit interface
    device_bus_width = 128

    # This is a BL4 device
    burst_length = 4

    # Each device has a page (row buffer) size of 4KB
    # (this depends on the memory density)
    device_rowbuffer_size = '4kB'

    # 1x128 configuration, so 1 device
    devices_per_rank = 1

    # Use one rank for a one-high die stack
    ranks_per_channel = 1

    # WideIO has 4 banks in all configurations
    banks_per_rank = 4

    # WIO-200
    tRCD = '18ns'
    tCL = '18ns'
    tRP = '18ns'
    tRAS = '42ns'

    # 4 beats across an x128 SDR interface translates to 4 clocks @ 200 MHz.
    # Note this is a BL4 SDR device.
    tBURST = '20ns'

    # WIO 8 Gb
    tRFC = '210ns'

    # WIO 8 Gb, <=85C, half for >85C
    tREFI = '3.9us'

    # Greater of 2 CK or 15 ns, 2 CK @ 200 MHz = 10 ns
    tWTR = '15ns'

    # Activate to activate irrespective of density and speed grade
    tRRD = '10.0ns'

    # Two instead of four activation window
    tXAW = '50ns'
    activation_limit = 2

# A single LPDDR3 x32 interface (one command/address bus), with
# default timings based on a LPDDR3-1600 4 Gbit part in a 1x32
# configuration
class LPDDR3_1600_x32(SimpleDRAM):
    # 1x32 configuration, 1 device with a 32-bit interface
    device_bus_width = 32

    # LPDDR3 is a BL8 device
    burst_length = 8

    # Each device has a page (row buffer) size of 1KB
    # (this depends on the memory density)
    device_rowbuffer_size = '1kB'

    # 1x32 configuration, so 1 device
    devices_per_rank = 1

    # Use a single rank
    ranks_per_channel = 1

    # LPDDR3 has 8 banks in all configurations
    banks_per_rank = 8

    # Fixed at 15 ns
    tRCD = '15ns'

    # 12 CK read latency, 6 CK write latency @ 800 MHz, 1.25 ns cycle time
    tCL = '15ns'

    tRAS = '42ns'

    # Pre-charge one bank 15 ns (all banks 18 ns)
    tRP = '15ns'

    # 8 beats across a x32 DDR interface translates to 4 clocks @ 800 MHz.
    # Note this is a BL8 DDR device.
    # Requests larger than 32 bytes are broken down into multiple requests
    # in the SimpleDRAM controller
    tBURST = '5ns'

    # LPDDR3, 4 Gb
    tRFC = '130ns'
    tREFI = '3.9us'

    # Irrespective of speed grade, tWTR is 7.5 ns
    tWTR = '7.5ns'

    # Activate to activate irrespective of density and speed grade
    tRRD = '10.0ns'

    # Irrespective of size, tFAW is 50 ns
    tXAW = '50ns'
    activation_limit = 4