1perf-c2c(1) 2=========== 3 4NAME 5---- 6perf-c2c - Shared Data C2C/HITM Analyzer. 7 8SYNOPSIS 9-------- 10[verse] 11'perf c2c record' [<options>] <command> 12'perf c2c record' [<options>] \-- [<record command options>] <command> 13'perf c2c report' [<options>] 14 15DESCRIPTION 16----------- 17C2C stands for Cache To Cache. 18 19The perf c2c tool provides means for Shared Data C2C/HITM analysis. It allows 20you to track down the cacheline contentions. 21 22On x86, the tool is based on load latency and precise store facility events 23provided by Intel CPUs. On PowerPC, the tool uses random instruction sampling 24with thresholding feature. 25 26These events provide: 27 - memory address of the access 28 - type of the access (load and store details) 29 - latency (in cycles) of the load access 30 31The c2c tool provide means to record this data and report back access details 32for cachelines with highest contention - highest number of HITM accesses. 33 34The basic workflow with this tool follows the standard record/report phase. 35User uses the record command to record events data and report command to 36display it. 37 38 39RECORD OPTIONS 40-------------- 41-e:: 42--event=:: 43 Select the PMU event. Use 'perf c2c record -e list' 44 to list available events. 45 46-v:: 47--verbose:: 48 Be more verbose (show counter open errors, etc). 49 50-l:: 51--ldlat:: 52 Configure mem-loads latency. (x86 only) 53 54-k:: 55--all-kernel:: 56 Configure all used events to run in kernel space. 57 58-u:: 59--all-user:: 60 Configure all used events to run in user space. 61 62REPORT OPTIONS 63-------------- 64-k:: 65--vmlinux=<file>:: 66 vmlinux pathname 67 68-v:: 69--verbose:: 70 Be more verbose (show counter open errors, etc). 71 72-i:: 73--input:: 74 Specify the input file to process. 75 76-N:: 77--node-info:: 78 Show extra node info in report (see NODE INFO section) 79 80-c:: 81--coalesce:: 82 Specify sorting fields for single cacheline display. 83 Following fields are available: tid,pid,iaddr,dso 84 (see COALESCE) 85 86-g:: 87--call-graph:: 88 Setup callchains parameters. 89 Please refer to perf-report man page for details. 90 91--stdio:: 92 Force the stdio output (see STDIO OUTPUT) 93 94--stats:: 95 Display only statistic tables and force stdio mode. 96 97--full-symbols:: 98 Display full length of symbols. 99 100--no-source:: 101 Do not display Source:Line column. 102 103--show-all:: 104 Show all captured HITM lines, with no regard to HITM % 0.0005 limit. 105 106-f:: 107--force:: 108 Don't do ownership validation. 109 110-d:: 111--display:: 112 Switch to HITM type (rmt, lcl) or peer snooping type (peer) to display 113 and sort on. Total HITMs (tot) as default, except Arm64 uses peer mode 114 as default. 115 116--stitch-lbr:: 117 Show callgraph with stitched LBRs, which may have more complete 118 callgraph. The perf.data file must have been obtained using 119 perf c2c record --call-graph lbr. 120 Disabled by default. In common cases with call stack overflows, 121 it can recreate better call stacks than the default lbr call stack 122 output. But this approach is not full proof. There can be cases 123 where it creates incorrect call stacks from incorrect matches. 124 The known limitations include exception handing such as 125 setjmp/longjmp will have calls/returns not match. 126 127C2C RECORD 128---------- 129The perf c2c record command setup options related to HITM cacheline analysis 130and calls standard perf record command. 131 132Following perf record options are configured by default: 133(check perf record man page for details) 134 135 -W,-d,--phys-data,--sample-cpu 136 137Unless specified otherwise with '-e' option, following events are monitored by 138default on x86: 139 140 cpu/mem-loads,ldlat=30/P 141 cpu/mem-stores/P 142 143and following on PowerPC: 144 145 cpu/mem-loads/ 146 cpu/mem-stores/ 147 148User can pass any 'perf record' option behind '--' mark, like (to enable 149callchains and system wide monitoring): 150 151 $ perf c2c record -- -g -a 152 153Please check RECORD OPTIONS section for specific c2c record options. 154 155C2C REPORT 156---------- 157The perf c2c report command displays shared data analysis. It comes in two 158display modes: stdio and tui (default). 159 160The report command workflow is following: 161 - sort all the data based on the cacheline address 162 - store access details for each cacheline 163 - sort all cachelines based on user settings 164 - display data 165 166In general perf report output consist of 2 basic views: 167 1) most expensive cachelines list 168 2) offsets details for each cacheline 169 170For each cacheline in the 1) list we display following data: 171(Both stdio and TUI modes follow the same fields output) 172 173 Index 174 - zero based index to identify the cacheline 175 176 Cacheline 177 - cacheline address (hex number) 178 179 Rmt/Lcl Hitm (Display with HITM types) 180 - cacheline percentage of all Remote/Local HITM accesses 181 182 Peer Snoop (Display with peer type) 183 - cacheline percentage of all peer accesses 184 185 LLC Load Hitm - Total, LclHitm, RmtHitm (For display with HITM types) 186 - count of Total/Local/Remote load HITMs 187 188 Load Peer - Total, Local, Remote (For display with peer type) 189 - count of Total/Local/Remote load from peer cache or DRAM 190 191 Total records 192 - sum of all cachelines accesses 193 194 Total loads 195 - sum of all load accesses 196 197 Total stores 198 - sum of all store accesses 199 200 Store Reference - L1Hit, L1Miss, N/A 201 L1Hit - store accesses that hit L1 202 L1Miss - store accesses that missed L1 203 N/A - store accesses with memory level is not available 204 205 Core Load Hit - FB, L1, L2 206 - count of load hits in FB (Fill Buffer), L1 and L2 cache 207 208 LLC Load Hit - LlcHit, LclHitm 209 - count of LLC load accesses, includes LLC hits and LLC HITMs 210 211 RMT Load Hit - RmtHit, RmtHitm 212 - count of remote load accesses, includes remote hits and remote HITMs; 213 on Arm neoverse cores, RmtHit is used to account remote accesses, 214 includes remote DRAM or any upward cache level in remote node 215 216 Load Dram - Lcl, Rmt 217 - count of local and remote DRAM accesses 218 219For each offset in the 2) list we display following data: 220 221 HITM - Rmt, Lcl (Display with HITM types) 222 - % of Remote/Local HITM accesses for given offset within cacheline 223 224 Peer Snoop - Rmt, Lcl (Display with peer type) 225 - % of Remote/Local peer accesses for given offset within cacheline 226 227 Store Refs - L1 Hit, L1 Miss, N/A 228 - % of store accesses that hit L1, missed L1 and N/A (no available) memory 229 level for given offset within cacheline 230 231 Data address - Offset 232 - offset address 233 234 Pid 235 - pid of the process responsible for the accesses 236 237 Tid 238 - tid of the process responsible for the accesses 239 240 Code address 241 - code address responsible for the accesses 242 243 cycles - rmt hitm, lcl hitm, load (Display with HITM types) 244 - sum of cycles for given accesses - Remote/Local HITM and generic load 245 246 cycles - rmt peer, lcl peer, load (Display with peer type) 247 - sum of cycles for given accesses - Remote/Local peer load and generic load 248 249 cpu cnt 250 - number of cpus that participated on the access 251 252 Symbol 253 - code symbol related to the 'Code address' value 254 255 Shared Object 256 - shared object name related to the 'Code address' value 257 258 Source:Line 259 - source information related to the 'Code address' value 260 261 Node 262 - nodes participating on the access (see NODE INFO section) 263 264NODE INFO 265--------- 266The 'Node' field displays nodes that accesses given cacheline 267offset. Its output comes in 3 flavors: 268 - node IDs separated by ',' 269 - node IDs with stats for each ID, in following format: 270 Node{cpus %hitms %stores} (Display with HITM types) 271 Node{cpus %peers %stores} (Display with peer type) 272 - node IDs with list of affected CPUs in following format: 273 Node{cpu list} 274 275User can switch between above flavors with -N option or 276use 'n' key to interactively switch in TUI mode. 277 278COALESCE 279-------- 280User can specify how to sort offsets for cacheline. 281 282Following fields are available and governs the final 283output fields set for cacheline offsets output: 284 285 tid - coalesced by process TIDs 286 pid - coalesced by process PIDs 287 iaddr - coalesced by code address, following fields are displayed: 288 Code address, Code symbol, Shared Object, Source line 289 dso - coalesced by shared object 290 291By default the coalescing is setup with 'pid,iaddr'. 292 293STDIO OUTPUT 294------------ 295The stdio output displays data on standard output. 296 297Following tables are displayed: 298 Trace Event Information 299 - overall statistics of memory accesses 300 301 Global Shared Cache Line Event Information 302 - overall statistics on shared cachelines 303 304 Shared Data Cache Line Table 305 - list of most expensive cachelines 306 307 Shared Cache Line Distribution Pareto 308 - list of all accessed offsets for each cacheline 309 310TUI OUTPUT 311---------- 312The TUI output provides interactive interface to navigate 313through cachelines list and to display offset details. 314 315For details please refer to the help window by pressing '?' key. 316 317CREDITS 318------- 319Although Don Zickus, Dick Fowles and Joe Mario worked together 320to get this implemented, we got lots of early help from Arnaldo 321Carvalho de Melo, Stephane Eranian, Jiri Olsa and Andi Kleen. 322 323C2C BLOG 324-------- 325Check Joe's blog on c2c tool for detailed use case explanation: 326 https://joemario.github.io/blog/2016/09/01/c2c-blog/ 327 328SEE ALSO 329-------- 330linkperf:perf-record[1], linkperf:perf-mem[1] 331