CHANGELOG.md 18.7 KB
Newer Older
Gilbert Lee's avatar
Gilbert Lee committed
1
2
# Changelog for TransferBench

3
Documentation for TransferBench is available at
Lisa's avatar
Lisa committed
4
[https://rocm.docs.amd.com/projects/TransferBench](https://rocm.docs.amd.com/projects/TransferBench).
Lisa Delaney's avatar
Lisa Delaney committed
5

6
7
8
9
## v1.56
### Fixed
- Fixed bug when using interactive mode.  Interactive mode now starts prior to all warmup iterations

10
11
12
13
14
## v1.55
### Fixed
- Fixed missing header error when compiling on CentOS
- Fixed issues when using multi-stream mode for GFX executor

15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
## v1.54
### Modified
- Refactored TransferBench into a header-only library combined with a thin client to facilitate the
  use of TransferBench as the backend for other applications
- Optimized how data validation is handled - this should speed up Tests with many parallel transfers as data is only
  generated once
- Preset benchmarks now no longer take in any extra command line arguments.  Preset settings are only accessed via
  environment variables.  Details for each preset are printed
- The a2a preset benchmark now defaults to using fine-grained memory and GFX unroll of 2
- Refactored how Transfers are launched in parallel which has reduced some CPU-side overheads
- CPU and DMA executor timing now use CPU wall clock timing instead of slowest Transfer time
### Added
- New one2all preset which sweeps over all subests of parallel transfers from one GPU to others
- Adding new warnings for DMA execution relating to how HIP will default to using agents from the source memory
### Removed
- CU scaling preset has been removed.  Similar functionality already exists in the schmoo preset benchmark
- Preparation of source data via GFX kernel has been removed (USE_PREP_KERNEL)
- Removed GFX block-reordering (BLOCK_ORDER)
- Removed NUM_CPU_DEVICES and NUM_GPU_DEVICES from common env vars and only into the presets they apply to.
- Removed SHARED_MEM_BYTES option for GFX executor
- Removed USE_PCIE_INDEX, and SHARED_MEM_BYTES
### Fixed
- Fixed a potential timing reporting issue when DMA executed Transfers end up getting serialized.

39
40
41
42
## v1.53
### Added
- Added ability to specify NULL for sweep preset as source or destination memory type

gilbertlee-amd's avatar
gilbertlee-amd committed
43
44
45
46
47
48
49
50
## v1.52
### Added
- Added USE_HSA_DMA env var to switch to using hsa_amd_memory_async_copy instead of hipMemcpyAsync for DMA execution
- Added ability to set USE_GPU_DMA env var for a2a benchmark
- Adding check for large BAR enablement for GPU devices during topology check
### Fixed
- Potential memory leak if HSA reports 0 hops between GPUs and CPUs

51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
## v1.51

## Modified
- CSV output has been modified slightly to match normal terminal output
- Output for non single stream mode has been changed to match single stream mode (results per Executor)

### Added
- Support for sub-iterations via NUM_SUBITERATIONS.  This allows for additional looping during an iteration
  If set to 0, this should infinitely loop (which may be useful for some debug purposes)
- Support for variable number of subexecutors (currently for GPU-GFX executor only).  Setting subExecutors to
  0 will run over a range of CUs to use, and report only the results of the best one found. This can be tuned
  for performance by setting the MIN_VAR_SUBEXEC and MAX_VAR_SUBEXEC environment variables to narrow the
  search space.  The number of CUs used will be identical for all variable subExecutor transfers
- Experimental new "healthcheck" preset config which currently only supports MI300 series.  This preset runs
  through CPU to GPU bandwidth tests and all-to-all XGMI bandwidth tests and compares against expected values
  Pass criteria limits can be modified (due to platform differences) via the environment variables
  LIMIT_UDIR (undirectional), LIMIT_BDIR (bidirectional), and LIMIT_A2A (Per GPU-GPU link bandwidth)

### Fixed
- Fixed out-of-bounds memory access during topology detection that can happen if the number of
  CPUs is less than the number of NUMA domains
- Fixed CU masking functionality on multi-XCD architectures (e.g. MI300)

74
75
76
77
78
79
80
81
82
## v1.50

### Added
- Adding new parallel copy preset benchmark (pcopy)
  - Usage: ./TransferBench pcopy <numBytes=64M> <#CUs=8> <srcGpu=0> <minGpus=1> <maxGpus=#GPU-1>
### Fixed
- Removed non-copies DMA Transfers (this had previously been using hipMemset)
- Fixed CPU executor when operating on null destination

83
84
85
86
87
## v1.49

### Fixes
* Enumerating previously missed DMA engines used only for CPU traffic in topology display

88
89
90
91
92
93
94
95
96
97
98
99
## v1.48

### Fixes
* Various fixes for TransferBenchCuda

### Additions
* Support for targeting specific DMA engines via executor subindex (e.g. D0.1)
* Printing warnings when exeuctors are overcommited

### Modifications
* USE_REMOTE_READ supported for rwrite preset benchmark

100
101
102
103
104
## v1.47

### Fixes
* Fixing CUDA support

105
106
107
108
109
110
111
112
113
## v1.46

### Fixes
* Fixing GFX_UNROLL set to 13 (past 8) on gfx906 cards

### Modifications
* GFX_SINGLE_TEAM=1 by default
* Adding field showing summation of individual Transfer bandwidths for Executors

gilbertlee-amd's avatar
gilbertlee-amd committed
114
115
116
117
118
119
120
121
122
123
## v1.45

### Additions
* Adding A2A_MODE to a2a preset (0 = copy, 1 = read-only, 2 = write-only)
* Adding GFX_UNROLL to modify GFX kernel's unroll factor
* Adding GFX_WAVE_ORDER to modify order in which wavefronts process data

### Modifications
* Rewrote the GFX reduction kernel to support new wave ordering

124
125
126
127
128
129
## v1.44

### Additions
* Adding rwrite preset to benchmark remote parallel writes
 * Usage: ./TransferBench rwrite <numBytes=64M> <#CUs=8> <srcGpu=0> <minGpus=1> <maxGpus=3>

130
131
132
133
134
## v1.43

### Changes
* Modifying a2a to show executor timing, as well as executor min/max bandwidth

135
136
137
138
139
140
## v1.42

### Fixes
* Fixing schmoo maxNumCus optional arg parsing
* Schmoo output modified to be easier to copy

141
142
143
144
145
146
147
148
149
## v1.41

### Additions
* Adding schmoo preset config benchmarks local/remote reads/writes/copies
  * Usage: ./TransferBench schmoo <numBytes=64M> <localIdx=0> <remoteIdx=1> <maxNumCUs=32>

### Fixes
* Fixing some misreported timings when running with non-fixed number of iterations

150
151
152
153
154
## v1.40

### Fixes
* Fixing XCC defaulting to 0 instead of random for preset configs, ignoring XCC_PREF_TABLE

155
156
157
158
159
160
161
## v1.39

### Additions
* (Experimental) Adding support for Executor sub-index
### Fixes
- Remove deprecated gcnArch code.  ROCm version must include support for hipDeviceMallocUncached

162
163
164
165
166
## v1.38

### Fixes
* Adding missing threadfence which could cause non-fine-grained Transfers to report higher speeds

167
168
169
170
171
172
173
174
## v1.37

### Changes
* USE_SINGLE_STREAM is enabled by default now.  (Disable via USE_SINGLE_STREAM=0)

### Fixes
* Fix unrecognized token error when XCC_PREF_TABLE is unspecified

175
176
177
178
179
180
181
## v1.36

### Additions

* (Experimental) Adding XCC filtering - combined with XCC_PREF_TABLE, this tries to select
  specific XCCs to use for specific (SRC->DST) Transfers

182
183
184
185
186
## v1.35

### Additions

* USE_FINE_GRAIN also applies to a2a preset
187

188
## v1.34
189
190
191
192

### Additions

* Set `GPU_KERNEL=3` as default for gfx942
193

194
## v1.33
Lisa Delaney's avatar
Lisa Delaney committed
195
196
197
198
199

### Additions

* Added the `ALWAYS_VALIDATE` environment variable to allow for validation after every iteration, instead
  of only once at the end of all iterations
200

201
## v1.32
Lisa Delaney's avatar
Lisa Delaney committed
202
203
204
205

### Changes

* Increased the line limit from 2048 to 32768
206

gilbertlee-amd's avatar
gilbertlee-amd committed
207
## v1.31
Lisa Delaney's avatar
Lisa Delaney committed
208
209
210
211
212

### Changes

* `SHOW_ITERATIONS` now shows XCC:CU instead of just CU ID
* `SHOW_ITERATIONS` is printed when `USE_SINGLE_STREAM`=1
gilbertlee-amd's avatar
gilbertlee-amd committed
213

214
## v1.30
Lisa Delaney's avatar
Lisa Delaney committed
215
216
217
218
219
220
221
222
223

### Additions

* `BLOCK_SIZE` has been added to control the threadblock size (must be a multiple of 64, up to 512)
* `BLOCK_ORDER` has been added to control how work is ordered for GFX-executors running
  `USE_SINGLE_STREAM`=1
  * 0 - Threadblocks for transfers are ordered sequentially (default)
  * 1 - Threadblocks for transfers are interleaved
  * 2 - Threadblocks for transfers are ordered randomly
224

225
## v1.29
Lisa Delaney's avatar
Lisa Delaney committed
226
227
228
229
230
231
232
233
234
235
236
237
238
239

### Additions

* A2A preset config now responds to `USE_REMOTE_READ`

### Fixes

* Race-condition during wall-clock initialization caused "inf" during single-stream runs
* CU numbering output after CU masking

### Changes

* The default number of warmups has been reverted to 3
* The default unroll factor for gfx940/941 has been set to 6
240

241
## v1.28
Lisa Delaney's avatar
Lisa Delaney committed
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258

### Additions

* Added `A2A_DIRECT`, which only runs all-to-all on directly connected GPUs (now on by default)
* Added average statistics for P2P and A2A benchmarks
* Added `USE_FINE_GRAIN` for P2P benchmark
  * With older devices, P2P performance with default coarse-grain device memory stops timing as soon
    as a request is sent to data fabric, and not actually when it arrives remotely. This can artificially
    inflate bandwidth numbers, especially when sending small amounts of data.

### Changes

* Modified P2P output to help distinguish between CPU and GPU devices

### Fixes

* Fixed Makefile target to prevent unnecessary re-compilation
259

260
## v1.27
Lisa Delaney's avatar
Lisa Delaney committed
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279

### Additions

* Added cmdline preset to allow specification of  simple tests on command line (e.g.,
  `./TransferBench cmdline 64M "1 4 G0->G0->G1"`)
* Adding the `HIDE_ENV` environment variable, which stops environment variable values from printing
* Adding the `CU_MASK` environment variable, which allows you to select the CUs to run on
* `CU_MASK` is specified in CU indices (0-#CUs-1), where ' - ' can be used to denote ranges of values
  (e.g., `CU_MASK`=3-8,16 requests that transfer be run only on CUs 3,4,5,6,7,8,16)
  * Note that this is somewhat experimental and may not work on all hardware
* `SHOW_ITERATIONS` now shows CU usage for that iteration (experimental)

### Changes

* Added extra comments on commonly missing includes with details on how to install them

### Fixes

* CUDA compilation works again (the `wall_clock64` CUDA alias was not defined)
280

281
282
## v1.26

Lisa Delaney's avatar
Lisa Delaney committed
283
284
285
286
287
288
289
290
291
292
293
294
### Additions

* Setting SHOW_ITERATIONS=1 provides additional information about per-iteration timing for file and
  P2P configs
  * For file configs, iterations are sorted from min to max bandwidth and displayed with standard
    deviation
  * For P2P, min/max/standard deviation is shown for each direction

### Changes

* P2P benchmark formatting now reports bidirectional bandwidth in each direction (as well as sum) for
  clarity
295

296
## v1.25
Lisa Delaney's avatar
Lisa Delaney committed
297
298
299
300
301

### Fixes

* Fixed a bug in the P2P bidirectional benchmark that used the incorrect number of `subExecutors` for
  CPU<->GPU tests
302

gilbertlee-amd's avatar
gilbertlee-amd committed
303
## v1.24
Lisa Delaney's avatar
Lisa Delaney committed
304
305
306
307
308

### Additions

* New All-To-All GPU benchmark accessed by preset "A2A"
* Added gfx941 wall clock frequency
gilbertlee-amd's avatar
gilbertlee-amd committed
309

310
## v1.23
Lisa Delaney's avatar
Lisa Delaney committed
311
312
313
314
315

### Additions

* New GPU subexec scaling benchmark accessed by preset "scaling"
  * Tests GPU-GFX copy performance based on # of CUs used
316

317
## v1.22
Lisa Delaney's avatar
Lisa Delaney committed
318
319
320
321

### Changes

* Switched the kernel timing function to `wall_clock64`
322

gilbertlee-amd's avatar
gilbertlee-amd committed
323
## v1.21
Lisa Delaney's avatar
Lisa Delaney committed
324
325
326
327

### Fixes

* Fixed a bug with `SAMPLING_FACTOR`
gilbertlee-amd's avatar
gilbertlee-amd committed
328

329
## v1.20
Lisa Delaney's avatar
Lisa Delaney committed
330
331
332
333
334

### Fixes

* `VALIDATE_DIRECT` can now be used with `USE_PREP_KERNEL`
* Switched to local GPU for validating GPU memory
335

gilbertlee-amd's avatar
gilbertlee-amd committed
336
## v1.19
Lisa Delaney's avatar
Lisa Delaney committed
337
338
339
340
341

### Additions

* `VALIDATE_DIRECT` now also applies to source memory array checking
* Added null memory pointer check prior to deallocation
gilbertlee-amd's avatar
gilbertlee-amd committed
342

343
## v1.18
Lisa Delaney's avatar
Lisa Delaney committed
344
345
346
347
348
349
350
351
352
353
354
355
356

### Additions

* Adding the ability to validate GPU destination memory directly without going through the CPU
  staging buffer (`VALIDATE_DIRECT`)
  * Note that this only works on AMD devices with large-bar access enabled, and may slow things down
    considerably

### Changes

* Refactored how environment variables are displayed
* Mismatch stops after the first detected error within an array instead of listing all mismatched
  elements
357

358
## v1.17
Lisa Delaney's avatar
Lisa Delaney committed
359
360
361
362
363
364
365
366
367
368
369
370
371
372

### Additions

* Allowed switch to GFX kernel for source array initialization (`USE_PREP_KERNEL`)
  * Note that `USE_PREP_KERNEL` can't be used with `FILL_PATTERN`
* Added the ability to compile with nvcc only (`TransferBenchCuda`)

### Changes

* The default pattern was set to [Element i = ((i * 517) modulo 383 + 31) * (srcBufferIdx + 1)]

### Fixes

* Added the `example.cfg` file
373

374
## v1.16
Lisa Delaney's avatar
Lisa Delaney committed
375
376
377
378
379
380
381

### Additions

* Additional src array validation during preparation
* Added a new environment variable (`CONTINUE_ON_ERROR`) to resume tests after a mis-match
  detection
* Initialized GPU memory to 0 during allocation
382

383
## v1.15
Lisa Delaney's avatar
Lisa Delaney committed
384
385
386
387
388
389
390
391
392

### Fixes

* Fixed a bug that prevented single transfers greater than 8 GB

### Changes

* Removed "check for latest ROCm" warning when allocating too much memory
* Off-source memory value is now printed when a mis-match is detected
393

PedramAlizadeh's avatar
PedramAlizadeh committed
394
## v1.14
Lisa Delaney's avatar
Lisa Delaney committed
395
396
397
398
399
400

### Additions

* Added documentation
* Added pthread linking in src/Makefile and CMakeLists.txt
* Added printing off the hex value of the floats for output and reference
401

PedramAlizadeh's avatar
PedramAlizadeh committed
402
403
## v1.13

Lisa Delaney's avatar
Lisa Delaney committed
404
405
406
407
408
409
410
### Additions

* Added support for cmake

### Changes

* Converted to the Pitchfork layout standard
PedramAlizadeh's avatar
PedramAlizadeh committed
411

412
## v1.12
Lisa Delaney's avatar
Lisa Delaney committed
413
414
415
416
417

### Additions

* Added support for TransferBench on NVIDIA platforms (via `HIP_PLATFORM`=nvidia)
  * Note that CPU executors on NVIDIA platform cannot access GPU memory (no large-bar access)
418

gilbertlee-amd's avatar
gilbertlee-amd committed
419
## v1.11
Lisa Delaney's avatar
Lisa Delaney committed
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465

### Additions

* Added multi-input/multi-output (MIMO) support: transfers now can reduce (element-wise summation)
  multiple input memory arrays and write sums to multiple outputs
* Added GPU-DMA executor 'D', which uses `hipMemcpy` for SDMA copies
  * Previously, this was done using `USE_HIP_CALL`, but now GPU-GFX kernel can run in parallel with
    GPU-DMA, instead of applying to all GPU executors globally
  * GPU-DMA executor can only be used for single-input/single-output transfers
  * GPU-DMA executor can only be associated with one SubExecutor
* Added new "Null" memory type 'N', which represents empty memory. This allows for read-only or
  write-only transfers
* Added new `GPU_KERNEL` environment variable that allows switching between various GPU-GFX
  reduction kernels

### Optimizations

* Improved GPU-GFX kernel performance based on hardware architecture when running with
  fewer CUs

### Changes

* Updated the `example.cfg` file to cover new features
* Updated output to support MIMO
* Changed CU and CPU thread naming to SubExecutors for consistency
* Sweep Preset: default sweep preset executors now includes DMA
* P2P benchmarks:
  * Removed `p2p_rr`, `g2g` and `g2g_rr` (now only works via P2P)
    * Setting `NUM_CPU_DEVICES`=0 can only be used to benchmark GPU devices (like `g2g`)
    * The new `USE_REMOTE_READ` environment variable replaces `_rr` presets
  * New environment variable `USE_GPU_DMA`=1 replaces `USE_HIP_CALL`=1 for benchmarking with
    GPU-DMA Executor
  * Number of GPU SubExecutors for benchmark can be specified via `NUM_GPU_SE`
    * Defaults to all CUs for GPU-GFX, 1 for GPU-DMA
  * Number of CPU SubExecutors for benchmark can be specified via `NUM_CPU_SE`
* Psuedo-random input pattern has been slightly adjusted to have different patterns for each input
  array within same transfer

### Removals

* `USE_HIP_CALL`: use `GPU-DMA` executor 'D' or set `USE_GPU_DMA`=1 for P2P
  benchmark presets
  * Currently, a warning will be issued if `USE_HIP_CALL` is set to 1 and the program will stop
* `NUM_CPU_PER_TRANSFER`: the number of CPU SubExecutors will be whatever is specified for the
  transfer
* `USE_MEMSET`: this function can now be done via a transfer using the null memory type
gilbertlee-amd's avatar
gilbertlee-amd committed
466

467
## v1.10
Lisa Delaney's avatar
Lisa Delaney committed
468
469
470
471

### Fixes

* Fixed incorrect bandwidth calculation when using single stream mode and per-transfer data sizes
472

473
## v1.09
Lisa Delaney's avatar
Lisa Delaney committed
474
475
476
477
478
479
480
481

### Additions

* Printing off src/dst memory addresses during interactive mode

### Changes

* Switching to `numa_set_preferred` instead of `set_mempolicy`
482

483
## v1.08
Lisa Delaney's avatar
Lisa Delaney committed
484
485
486
487
488
489

### Changes

* Fixed handling of non-configured NUMA nodes
* Topology detection now shows actual NUMA node indices
* Fixed 'for' issue with `NUM_GPU_DEVICES`
490

491
## v1.07
Lisa Delaney's avatar
Lisa Delaney committed
492
493
494
495

### Fixes

* Fixed bug with allocations involving non-default CPU memory types
496

gilbertlee-amd's avatar
gilbertlee-amd committed
497
## v1.06
Lisa Delaney's avatar
Lisa Delaney committed
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513

### Additions

* Unpinned CPU memory type ('U'), which may require `HSA_XNACK`=1 in order to access via
  GPU executors
* Added sweep configuration logging to `lastSweep.cfg`
* Ability to specify the number of CUs to use for sweep-based presets

### Changes

* Modified advanced configuration file format to accept bytes-per-transfer

### Fixes

* Fixed random sweep repeatability
* Fixed bug with CPU NUMA node memory allocation
gilbertlee-amd's avatar
gilbertlee-amd committed
514

515
## v1.05
Lisa Delaney's avatar
Lisa Delaney committed
516
517
518
519
520
521
522
523
524

### Additions

* Topology output now includes NUMA node information
* Support for NUMA nodes with no CPU cores (e.g., CXL memory)

### Removals

* The `SWEEP_SRC_IS_EXE` environment variable was removed
525

526
## v1.04
Lisa Delaney's avatar
Lisa Delaney committed
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541

### Additions

* There are new environment variables for sweep based presets:
  * `SWEEP_XGMI_MIN`: The minumum number of XGMI hops for transfers
  * `SWEEP_XGMI_MAX`: The maximum number of XGMI hops for transfers
  * `SWEEP_SEED`: Uses a random seed
  * `SWEEP_RAND_BYTES`: Uses a random amount of bytes (up to pre-specified N) for each transfer

### Changes

* CSV output for sweep now includes an environment variables section followed by output
* CSV output no longer lists environment variable parameters in columns
* We changed the default number of warmup iterations from 3 to 1
* Split CSV output of link type to `ExeToSrcLinkType` and `ExeToDstLinkType`
542

Gilbert Lee's avatar
Gilbert Lee committed
543
## v1.03
Lisa Delaney's avatar
Lisa Delaney committed
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566

### Additions

* There are new preset modes stress-test benchmarks: `sweep` and `randomsweep`
  * `sweep` iterates over all possible sets of transfers to test
  * `randomsweep` iterates over random sets of transfers
  * New sweep-only environment variables can modify `sweep`
    * `SWEEP_SRC`: String containing only "B","C","F", or "G" that defines possible source memory types
    * `SWEEP_EXE`: String containing only "C" or "G" that defines possible executors
    * `SWEEP_DST`: String containing only "B","C","F", or "G" that defines possible destination memory types
    * `SWEEP_SRC_IS_EXE`: Restrict the executor to be the same as the source, if non-zero
    * `SWEEP_MIN`: Minimum number of parallel transfers to test
    * `SWEEP_MAX`: Maximum number of parallel transfers to test
    * `SWEEP_COUNT`: Maximum number of tests to run
    * `SWEEP_TIME_LIMIT`: Maximum number of seconds to run tests
* New environment variables to restrict number of available devices to test on (primarily for sweep
  runs)
  * `NUM_CPU_DEVICES`: Number of CPU devices
  * `NUM_GPU_DEVICES`: Number of GPU devices

### Fixes

* Fixed timing display for CPU executors when using single-stream mode
Gilbert Lee's avatar
Gilbert Lee committed
567

Gilbert Lee's avatar
Gilbert Lee committed
568
## v1.02
Lisa Delaney's avatar
Lisa Delaney committed
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585

### Additions

* Setting `NUM_ITERATIONS` to a negative number indicates a run of -`NUM_ITERATIONS` seconds per
  test

### Changes

* Copies are now referred to as 'transfers' instead of 'links'
* Reordered how environment variables are displayed (alphabetically now)

### Removals

* Combined timing is now always on for kernel-based GPU copies; the `COMBINED_TIMING`
  environment variable has been removed
* Single sync is no longer supported for facility variable iterations; the `USE_SINGLE_SYNC`
  environmental variable has been removed
Gilbert Lee's avatar
Gilbert Lee committed
586

Gilbert Lee's avatar
Gilbert Lee committed
587
## v1.01
Lisa Delaney's avatar
Lisa Delaney committed
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606

### Additions

* Added the `USE_SINGLE_STREAM` feature
  * All Links that run on the same GPU device are run with a single kernel launch on a single stream
  * This doesn't work with `USE_HIP_CALL`, and it forces `USE_SINGLE_SYNC` to collect timings
  * Added the ability to request coherent or fine-grained host memory ('B')

### Changes

* Separated the TransferBench repository from the RCCL repository
* Peer-to-peer benchmark mode now works with `OUTPUT_TO_CSV`
* Toplogy display now works with `OUTPUT_TO_CSV`
* Moved the documentation about the config file into `example.cfg`

### Removals

* Removed config file generation
* Removed the 'show pointer address' (`SHOW_ADDR`) environment variable