Date: | Fri, 26 Feb 2010 00:00:42 -0500 |
---|---|
From: | Carole-Jean Wu <carolewu@xxxxxxxxxxxxx> |
Subject: | [Gems-users] ruby periodic statistic file |
Hello list, I am dumping ruby statistics periodically (period = 10000000) for mcf running on core1 in a 4-core CMP setup. And, I am showing the last entry in my periodic statistic file below. The number instructions executed is 9,691,680,772 and the number of "total_misses" is 132,986,026. Then, I inserted one print statement in the beginning of makeRequest(...) in system/Sequencer.C to get the memory trace for all memory references for core1, where mcf is pinned to. If I count the number of print statements in the trace file, there are only 253,260,832 entries. The ratio of #memory ref. and #inst. executed does not make sense. It is way too smaller than what's published. Does anyone have an idea what causes this discrepancy? Are there other "built-in" counters in ruby, which I can use to count the number of memory references issued? Thanks in advance, Carole Wu ruby_cycles: 47430000001 total_misses: 170926472 [ 7015840 132986026 13365208 17559398 ] instruction_executed: 536087641076 [ 183497245022 9691680772 174930542212 167968173070 ] simics_cycles_executed: 758880000088 [ 189720000005 189720000001 189720000001 189720000081 ] mbytes_resident: 3435.67 mbytes_total: 3576.32 Total_misses: 170926472 total_misses: 170926472 [ 7015840 132986026 13365208 17559398 ] user_misses: 117489130 [ 189887 113459585 218853 3620805 ] supervisor_misses: 53437342 [ 6825953 19526441 13146355 13938593 ] instruction_executed: 536087641076 [ 183497245022 9691680772 174930542212 167968173070 ] simics_cycles_executed: 758880000088 [ 189720000005 189720000001 189720000001 189720000081 ] cycles_per_instruction: 0.353897 [ 0.258478 4.89389 0.271136 0.282375 ] misses_per_thousand_instructions: 0.318841 [ 1.32608e-312 2.31786e-310 8.48798e-314 3.74348e+150 ] Simics Driver Transaction Stats ---------------------------------- Insn requests: 0 Data requests: 7715315597 Memory mapped IO register accesses: 0 Device initiated accesses: 52977 Other initiated accesses: 0 Atomic load accesses: 374534 Exceptions: 5906478 Non stallable accesses: 225508306 Prefetches: 135612838 Cache Flush: 0 . . . Simics Driver Transaction Results Stats ------------------------------------------ Fast path: 7348192646 Request missed: 231457136 Sequencer not ready: 0 Duplicate instruction fetches: 0 Hit return: 228552370 Atomic last accesses: 5663149 L1D_cache cache stats: L1D_cache_total_misses: 231457136 L1D_cache_total_demand_misses: 231457136 . . . L2_cache cache stats: L2_cache_total_misses: 170926472 L2_cache_total_demand_misses: 170926472 . . . |
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | Re: [Gems-users] Removing Supervisor instructions from Opal Simulation, ubaid001 |
---|---|
Next by Date: | [Gems-users] MEMBAR implementation in opal+ruby, Byn Choi |
Previous by Thread: | Re: [Gems-users] Regarding leaking memory, Javi Merino |
Next by Thread: | [Gems-users] simics get unstalled without ruby asking error, sparsh mittal ISU |
Indexes: | [Date] [Thread] |