[EMAIL PROTECTED] wrote: > Eric, > > If you are interested in tying DTLB misses to > specific regions of user source code, please > consider using 'collect' and 'analyzer'. The > latest builds of these tools for Opteron can > be found here: > > /set/mars/dist/intel-S2/bin >
Can it find specific region in the kernel text section? As my workload is quite simple: a typical run of Iperf that does nothing but read() from a socket for network throughput test. Also the TLB misses mainly happen in kernel pages. :( - yxn > An example of using collect: > > collect -p on -h dtlbm,hi a.out > analyzer test.1.er > er_print -pc test.1.er > ... > Excl. Incl. Excl. DTLB Incl. DTLB Name > User CPU User CPU Misses Misses > sec. sec. > 10.337 10.337 1801261 1801261 <Total> > 0.440 0.440 100070 100070 test_loop + 0x000000A8, line 286 > in "thrashfpleak.c" > 0.430 0.430 50035 50035 test_loop + 0x00000080, line 277 > in "thrashfpleak.c" > 0.420 0.420 70049 70049 test_loop + 0x00000076, line 275 > in "thrashfpleak.c" > ... > > Yukon > > > > Yu Xiangning wrote On 10/31/06 22:11,: >> Hi folks, >> >> While investigating a performance problem in my prototype workspace, I >> found that the dtlb-miss rate is about 16% higher than that of the >> baseline, so I suspect that the high tlb miss could be the issue. >> >> So, how can I get the statistics on the PC addresses when the tlb miss >> happens? Can DTrace or some other facilities help here? Thank you in >> advance! >> >> - yxn >> >> The attached in the sample trapstat output: (ignored the itlb column) >> >> Mine: >> cpu m size| dtlb-miss %tim dtsb-miss %tim |%tim >> ----------+-------------------------------+---- >> 0 u 8k| 7399 0.3 0 0.0 | 0.3 >> 0 u 64k| 0 0.0 0 0.0 | 0.0 >> 0 u 512k| 0 0.0 0 0.0 | 0.0 >> 0 u 4m| 0 0.0 0 0.0 | 0.0 >> - - - - - + - - - - - - - - - - - - - - - + - - >> 0 k 8k| 397423 14.5 237 0.0 |14.5 >> 0 k 64k| 0 0.0 0 0.0 | 0.0 >> 0 k 512k| 0 0.0 0 0.0 | 0.0 >> 0 k 4m| 0 0.0 0 0.0 | 0.0 >> ==========+===============================+==== >> ttl | 404822 14.7 237 0.0 |14.8 >> >> Baseline: >> cpu m size| dtlb-miss %tim dtsb-miss %tim |%tim >> 0 u 8k| 8667 0.3 9 0.0 | 0.3 >> 0 u 64k| 0 0.0 0 0.0 | 0.0 >> 0 u 512k| 0 0.0 0 0.0 | 0.0 >> 0 u 4m| 0 0.0 0 0.0 | 0.0 >> - - - - - + - - - - - - - - - - - - - - - + - - >> 0 k 8k| 342333 12.5 233 0.0 |12.6 >> 0 k 64k| 0 0.0 0 0.0 | 0.0 >> 0 k 512k| 0 0.0 0 0.0 | 0.0 >> 0 k 4m| 21 0.0 0 0.0 | 0.0 >> ==========+===============================+==== >> ttl | 351021 12.8 242 0.0 |12.9 >> >> >> _______________________________________________ >> perf-discuss mailing list >> perf-discuss@opensolaris.org > _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
