tracing/hist: Call hist functions directly via a switch statement
Due to retpolines, indirect calls are much more expensive than direct calls. The histograms have a select set of functions it uses for the histograms, instead of using function pointers to call them, create a hist_fn_call() function that uses a switch statement to call the histogram functions directly. This gives a 13% speedup to the histogram logic. Using the histogram benchmark: Before: # event histogram # # trigger info: hist:keys=delta:vals=hitcount:sort=delta:size=2048 if delta > 0 [active] # { delta: 129 } hitcount: 2213 { delta: 130 } hitcount: 285965 { delta: 131 } hitcount: 1146545 { delta: 132 } hitcount: 51854322 { delta: 133 } hitcount: 19896215 { delta: 134 } hitcount: 53118616 { delta: 135 } hitcount: 83816709 { delta: 136 } hitcount: 68329562 { delta: 137 } hitcount: 41859349 { delta: 138 } hitcount: 46257797 { delta: 139 } hitcount: 54400831 { delta: 140 } hitcount: 72875007 { delta: 141 } hitcount: 76193272 { delta: 142 } hitcount: 49504263 { delta: 143 } hitcount: 38821072 { delta: 144 } hitcount: 47702679 { delta: 145 } hitcount: 41357297 { delta: 146 } hitcount: 22058238 { delta: 147 } hitcount: 9720002 { delta: 148 } hitcount: 3193542 { delta: 149 } hitcount: 927030 { delta: 150 } hitcount: 850772 { delta: 151 } hitcount: 1477380 { delta: 152 } hitcount: 2687977 { delta: 153 } hitcount: 2865985 { delta: 154 } hitcount: 1977492 { delta: 155 } hitcount: 2475607 { delta: 156 } hitcount: 3403612 After: # event histogram # # trigger info: hist:keys=delta:vals=hitcount:sort=delta:size=2048 if delta > 0 [active] # { delta: 113 } hitcount: 272 { delta: 114 } hitcount: 840 { delta: 118 } hitcount: 344 { delta: 119 } hitcount: 25428 { delta: 120 } hitcount: 350590 { delta: 121 } hitcount: 1892484 { delta: 122 } hitcount: 6205004 { delta: 123 } hitcount: 11583521 { delta: 124 } hitcount: 37590979 { delta: 125 } hitcount: 108308504 { delta: 126 } hitcount: 131672461 { delta: 127 } hitcount: 88700598 { delta: 128 } hitcount: 65939870 { delta: 129 } hitcount: 45055004 { delta: 130 } hitcount: 33174464 { delta: 131 } hitcount: 31813493 { delta: 132 } hitcount: 29011676 { delta: 133 } hitcount: 22798782 { delta: 134 } hitcount: 22072486 { delta: 135 } hitcount: 17034113 { delta: 136 } hitcount: 8982490 { delta: 137 } hitcount: 2865908 { delta: 138 } hitcount: 980382 { delta: 139 } hitcount: 1651944 { delta: 140 } hitcount: 4112073 { delta: 141 } hitcount: 3963269 { delta: 142 } hitcount: 1712508 { delta: 143 } hitcount: 575941 { delta: 144 } hitcount: 351427 { delta: 145 } hitcount: 218077 { delta: 146 } hitcount: 167297 { delta: 147 } hitcount: 146198 { delta: 148 } hitcount: 116122 { delta: 149 } hitcount: 58993 { delta: 150 } hitcount: 40228 The delta above is in nanoseconds. It brings the fastest time down from 129ns to 113ns, and the peak from 141ns to 126ns. Link: https://lkml.kernel.org/r/20220906225529.411545333@goodmis.org Cc: Ingo Molnar <mingo@kernel.org> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Masami Hiramatsu <mhiramat@kernel.org> Cc: Tom Zanussi <zanussi@kernel.org> Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Showing
Please register or sign in to comment